WO2018122809A1 - 一种基于静态红外热像图处理的地下管廊渗漏检测方法 - Google Patents

一种基于静态红外热像图处理的地下管廊渗漏检测方法 Download PDF

Info

Publication number
WO2018122809A1
WO2018122809A1 PCT/IB2017/058539 IB2017058539W WO2018122809A1 WO 2018122809 A1 WO2018122809 A1 WO 2018122809A1 IB 2017058539 W IB2017058539 W IB 2017058539W WO 2018122809 A1 WO2018122809 A1 WO 2018122809A1
Authority
WO
WIPO (PCT)
Prior art keywords
crack
temperature difference
area
image
leakage
Prior art date
Application number
PCT/IB2017/058539
Other languages
English (en)
French (fr)
Inventor
杜豫川
潘宁
张晓明
刘成龙
常光照
Original Assignee
同济大学
许军
上海智能交通有限公司
杜豫川
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 同济大学, 许军, 上海智能交通有限公司, 杜豫川 filed Critical 同济大学
Priority to GB1905909.6A priority Critical patent/GB2569751B/en
Priority to CN201780053170.8A priority patent/CN110268190B/zh
Publication of WO2018122809A1 publication Critical patent/WO2018122809A1/zh

Links

Classifications

    • FMECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F17STORING OR DISTRIBUTING GASES OR LIQUIDS
    • F17DPIPE-LINE SYSTEMS; PIPE-LINES
    • F17D5/00Protection or supervision of installations
    • F17D5/02Preventing, monitoring, or locating loss
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01NINVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N3/00Investigating strength properties of solid materials by application of mechanical stress
    • G01N3/08Investigating strength properties of solid materials by application of mechanical stress by applying steady tensile or compressive forces
    • EFIXED CONSTRUCTIONS
    • E01CONSTRUCTION OF ROADS, RAILWAYS, OR BRIDGES
    • E01CCONSTRUCTION OF, OR SURFACES FOR, ROADS, SPORTS GROUNDS, OR THE LIKE; MACHINES OR AUXILIARY TOOLS FOR CONSTRUCTION OR REPAIR
    • E01C23/00Auxiliary devices or arrangements for constructing, repairing, reconditioning, or taking-up road or like surfaces
    • E01C23/01Devices or auxiliary means for setting-out or checking the configuration of new surfacing, e.g. templates, screed or reference line supports; Applications of apparatus for measuring, indicating, or recording the surface configuration of existing surfacing, e.g. profilographs
    • EFIXED CONSTRUCTIONS
    • E01CONSTRUCTION OF ROADS, RAILWAYS, OR BRIDGES
    • E01CCONSTRUCTION OF, OR SURFACES FOR, ROADS, SPORTS GROUNDS, OR THE LIKE; MACHINES OR AUXILIARY TOOLS FOR CONSTRUCTION OR REPAIR
    • E01C7/00Coherent pavings made in situ
    • E01C7/08Coherent pavings made in situ made of road-metal and binders
    • E01C7/18Coherent pavings made in situ made of road-metal and binders of road-metal and bituminous binders
    • FMECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F17STORING OR DISTRIBUTING GASES OR LIQUIDS
    • F17DPIPE-LINE SYSTEMS; PIPE-LINES
    • F17D5/00Protection or supervision of installations
    • F17D5/02Preventing, monitoring, or locating loss
    • F17D5/06Preventing, monitoring, or locating loss using electric or acoustic means
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01MTESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M3/00Investigating fluid-tightness of structures
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01MTESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M3/00Investigating fluid-tightness of structures
    • G01M3/002Investigating fluid-tightness of structures by using thermal means
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01MTESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M3/00Investigating fluid-tightness of structures
    • G01M3/38Investigating fluid-tightness of structures by using light
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01MTESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M5/00Investigating the elasticity of structures, e.g. deflection of bridges or air-craft wings
    • G01M5/0033Investigating the elasticity of structures, e.g. deflection of bridges or air-craft wings by determining damage, crack or wear
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01MTESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M5/00Investigating the elasticity of structures, e.g. deflection of bridges or air-craft wings
    • G01M5/0066Investigating the elasticity of structures, e.g. deflection of bridges or air-craft wings by exciting or detecting vibration or acceleration
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01NINVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84Systems specially adapted for particular applications
    • G01N21/88Investigating the presence of flaws or contamination
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01NINVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84Systems specially adapted for particular applications
    • G01N21/88Investigating the presence of flaws or contamination
    • G01N21/95Investigating the presence of flaws or contamination characterised by the material or shape of the object to be examined
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01NINVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N33/00Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
    • G01N33/42Road-making materials
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01NINVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84Systems specially adapted for particular applications
    • G01N21/88Investigating the presence of flaws or contamination
    • G01N21/8806Specially adapted optical and illumination features
    • G01N2021/8845Multiple wavelengths of illumination or detection
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01NINVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N25/00Investigating or analyzing materials by the use of thermal means
    • G01N25/72Investigating presence of flaws

Definitions

  • the invention belongs to the technical field of image depth processing and leakage detection, and particularly relates to a method for detecting leakage of underground pipe corridor based on infrared thermal image processing.
  • infrared thermal image By introducing infrared thermal image, the gray scale information and temperature information of the internal environment of the underground pipe gallery can be obtained.
  • the gray scale information can realize the conventional target of pipeline line state identification inside the pipe gallery, and the temperature information can be used to detect the pipeline leakage and the later stage.
  • Degree of development The degree of development mentioned in the invention characterizes the extent of damage to the pipeline by the leakage of the pipeline and the severity of the damage recently. It covers the traditional classification of crack severity, and the length, width and area of the crack. It may be related to factors that may increase the severity of crack damage, including these factors, and characterize the level of development of cracks from the beginning to the present.
  • Support vector machine is a machine learning method based on statistical learning theory developed in the mid-1990s. It seeks to minimize the structural risk and improve the generalization ability of learning machine, and realize the minimum of empirical risk and confidence range. In order to achieve a good statistical rule in the case of a small sample size.
  • support vector machines SVMs, which also support vector networks
  • SVMs are supervised learning models related to related learning algorithms that can analyze data, identify patterns, and use for classification and regression analysis. Given a set of training samples, each marked as belonging to two categories, an SVM training algorithm builds a model, assigning new instances to one class or other classes, making it a non-probabilistic binary linear classification. Generally speaking, it is a two-class classification model.
  • the basic model is defined as the linear classifier with the largest interval in the feature space. That is, the learning strategy of the support vector machine is to maximize the interval and finally transform into a convex quadratic. Solving the problem of planning.
  • Classification function The temperature difference data is linearly classified according to the degree of crack development by the support vector machine, and the development degree is divided into 1. 2, 3 three levels, 3 is the most serious, then there will be a line between 1, 2 and 2, 3 as the dividing line, this line expression is the classification function.
  • Crack zone A zone of a pipeline that includes not only the fracture zone itself, but also a range of pipelines around it, including the extent of the pipeline zone that meets image processing and crack identification requirements.
  • Reference temperature difference data Temperature difference data obtained by bringing the measured ambient temperature into two classification functions of the crack development degree detection model.
  • Measured temperature difference data Temperature difference data of the crack area and the pipeline surface area in the crack area obtained after image processing of the acquired infrared image of the crack area.
  • Distortion The meaning of the data distortion degree in the present invention is a reasonable deviation of the unreasonable value distance after the data generates an unreasonable value.
  • Development level A number in 1, 2 or 3 whose size reflects the degree of development of the fracture. The larger the number, the more serious the degree of crack development.
  • Developmental index between 0 and 3, including a number of 0 and 3, the size of which reflects the degree of development of the fracture. The larger the number, the more serious the development of the fracture. The description is indicated by the letter m.
  • Leakage severity index in the present invention means the average degree of severity of pipe cracks in three dimensions of length, width and area. The value ranges from [0-9. 9], retaining 1 decimal place, the size of which reflects the severity of the leak. The larger the number, the more serious the leak. Summary of the invention
  • the internal space of the pipe gallery is photographed by a temperature-measuring infrared camera, and an infrared thermal image near the leaking point of the pipeline is obtained, and a relationship model between the leakage crack and the temperature difference between the leaking portion and the leaking portion of the pipeline is established, and then the machine is combined.
  • the computer vision technology of learning establishes a discriminant model for pipeline leakage.
  • the experimental principle is based on PCT / IB2016/058109.
  • BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a technical route for identifying leaks in the interior of a pipe gallery using infrared thermography.
  • the detection method of obtaining the crack width by the identification of the crack pixel width to determine the crack severity is unreliable, because the crack width of a few millimeters may have only a few pixels in the image, and is subjected to image processing such as noise reduction. The result is more unstable, so it is difficult to discriminate the crack only by the gray scale information of the crack obtained by the ordinary image.
  • Infrared temperature measurement technology uses infrared optical system to make the infrared thermal image of the target to be measured on the infrared focal plane detector. After processing, the infrared image of the target to be measured is obtained, and then according to the gray value of the image and the calibration data, correlation The parameters analyze the temperature field distribution of the object.
  • the temperature measurement technology based on full-field analysis can measure the temperature of a large area, and can simultaneously obtain the temperature of multiple objects and the temperature of multiple objects, which is of great significance for analyzing the state of the object. It has become a hot topic in recent years.
  • the infrared thermal image obtained by the infrared camera can not only obtain the gray scale information of the inner space and crack of the pipe gallery, but also obtain their temperature information, as shown in Figure 2, which is the gray scale information and temperature of the inner space and crack of the pipe gallery. information.
  • the temperature information can be used to detect the leakage of pipelines and provide reference for the later maintenance and repair.
  • the discriminant model of pipeline leakage is established mainly through experiments and computer vision technology combined with machine learning.
  • the space of the underground pipe gallery is dark and humid.
  • the traditional method of detecting the leakage of the internal pipe will be subject to various restrictions due to the dim light and low temperature, resulting in large errors or even being undetectable. Therefore, the present invention clearly recognizes the obvious difference between the leakage area and the corridor environment through the infrared thermal image, can effectively identify the leakage point, and can use image processing technology to display parameters such as the size of the leakage damage.
  • Leakage in underground integrated pipe gallery is mainly divided into leakage caused by small crack damage, leakage caused by local damage and leakage caused by loose pipe joint.
  • the characteristics of the leak point are different, so the image taken by the infrared camera will show different characteristics. Therefore, the present invention needs to set a scientific infrared data information screening standard to distinguish different kinds of leakage conditions, thereby ensuring the applicability of the detection results, and providing a possibility for targeted maintenance according to a specific classification in the future.
  • the invention provides an accurate and efficient positioning technology, and realizes the positioning of the leakage point by screening and matching the interest domain graphic and the template graphic.
  • the object of the present invention is to reflect the degree of crack development by the temperature difference between the surface of the pipe and the crack under certain conditions, and their corresponding relationship is the detection model. Therefore, in order to solve the above problems, the technical solutions adopted by the present invention include:
  • the basis of the leakage analysis model based on infrared thermal imaging is the temperature field information measurement function, which relies on the infrared image temperature field recognition algorithm to realize the analysis of the infrared image collected by the acquisition device or the infrared image data stored on the computer.
  • the corresponding temperature value is calculated according to a certain mathematical model, and the temperature is measured by the set temperature measurement method.
  • the temperature measurement method is a kind of temperature measurement and regional temperature measurement.
  • the point temperature measurement is to measure the temperature of a single pixel, and the area temperature measurement is to measure the average value of a region. Therefore, the global, local area and point measurement of the target scene temperature can be achieved based on the image data.
  • the temperature difference will occur in the vicinity of the leakage point, and the infrared camera can visually detect the temperature difference of the surface above the leaking pipe, and according to the above The temperature measurement method realizes the perception of temperature difference.
  • the temperature of the surface of the pipeline and the interior will be different. Therefore, the dielectric material inside the pipeline will exchange heat with the surface of the pipeline and the ambient air through cracks or other damaged areas, while the surface and cracks of the pipeline There will also be a temperature difference.
  • the greater the length and width of the pipe crack the more intense the air heat exchange between the material in the pipe and the surface of the pipe. The temperature difference will be bigger.
  • the greater the length and width of the crack the more likely it is to cause future water damage, and the leakage is also reflected by the severity index. Therefore, the more severe the crack, the greater the temperature difference between the surface of the pipe and the crack, and we can use the thermal imager to detect the temperature difference and then detect the severity of the leak.
  • the severity of leakage can be measured in a variety of ways.
  • the present invention is described in terms of length, width, and area, respectively, and defines G as the index of severity of leakage.
  • G 1 [ ⁇ x l00%x l0] ( 1 ) where ⁇ is the severity index in the crack length dimension; Z is the length of the crack, the unit is the number of pixels; the outer diameter of the pipe, the unit is mm; The resolution of each pixel in length, in mm/pixel.
  • G 2 [2 ⁇ x l00%X l0] (2) where ⁇ is the severity index in the crack width dimension; S is the crack area, the unit is the number of pixels; Z is the length of the crack, and the unit is the number of pixels; r 2 is the resolution of the width of each pixel of the camera in mm/pixel.
  • is the severity index in the dimension of the crack area
  • S is the area of the crack
  • the unit is the number of pixels
  • is the length of the crack
  • the unit is the number of pixels
  • ⁇ , r 2 is the same as the outer diameter of the pipe, the unit is mm .
  • G Gl +Gz +G3 (4)
  • G is the severity index of the crack, and the result retains 1 decimal place.
  • the relationship between ambient temperature and temperature difference is shown in Figure 4.
  • the severity of the leakage is mainly measured by the damage it has caused to the pipeline, that is, the potential damage in the near future.
  • the severity of the leak is not only related to the width, length and area of the crack, but also to the depth of the crack.
  • the corrosion of the material around the crack is also one of the factors affecting the severity of the crack. Therefore, the severity of the leak needs to be considered comprehensively.
  • image analysis can detect the length, width and area of the crack, and the depth of the crack can be detected by laser radar. In order to estimate the corrosion of the material around the crack, how to accurately and simply reflect the development degree of the crack is an urgent problem to be solved in the project. Image acquisition
  • the uncooled focal plane temperature measuring type infrared thermal imager uses an infrared detector and an optical imaging objective lens to receive the infrared radiation energy distribution pattern of the measured object and reflect it on the photosensitive element of the infrared detector, thereby obtaining an infrared thermal image, which is obtained.
  • the image corresponds to the heat distribution field on the surface of the object.
  • an infrared camera converts invisible infrared energy emitted by an object into a visible thermal image.
  • the different colors above the thermal image represent the different temperatures of the object being measured.
  • Uncooled focal plane temperature measuring infrared camera Uncooled focal plane temperature measuring infrared camera.
  • the height of the distance from the surface of the pipe is 0. 5m, 0. 6m, 0. 7m, 0. 8m, 0. 9m, 1. 0m, it is required to capture the requirements for subsequent image processing for crack recognition.
  • the resolution of the image taken with the camera is at least 384 X 288; at 30 °C, the thermal sensitivity is at least 0. 06 °C; the frame rate is at least 50 Hz, used to collect the experimental water pipe at ambient temperature.
  • Infrared image The main imaging parameters of the DM60-S on-line thermal imaging camera, the infrared image acquired can meet the resolution requirements, and can be used as a reference for obtaining infrared image equipment.
  • the ordinary image acquisition device and the infrared camera constitute a binocular camera, and the angle of view of the two pictures is required to be consistent.
  • the ordinary image and the infrared image collected at the same time have the highest repeatability, and the data acquisition end is designed to be full. Area image acquisition.
  • the thermal imaging camera can be used with a fixed camera. Due to the detection of fixed equipment, the data is irreversible, continuous and trending due to the characteristics of leakage.
  • the irreversible pointer refers to the picture data of the suspicious leakage point collected by the infrared camera. If there is a leakage crack at the previous moment, the next moment should also exist.
  • the continuity refers to the fact that the leakage velocity of the leakage crack is relatively slow, usually 1 hour is the detection period, and the parameters of the leakage crack at the latter moment should only be slightly changed based on the parameters of the leakage crack at the previous moment.
  • the color image of the infrared image and the dimensional parameter curve of the possible leaking crack should be smooth with minimal variation.
  • the trend refers to the fact that the leakage crack is irreversible, so the crack at the latter moment must be greater than or equal to the severity of the previous moment.
  • the color of the infrared image will not change or deepen, and the size parameter of the leaking crack may change or become larger.
  • a temperature measuring fiber sensor is disposed in the pipe gallery, and the detection result is transmitted to the distributed fiber temperature sensing host through the temperature measuring fiber sensor, and the sensing host analyzes and processes the collected signal in real time to realize the monitoring work of the heating pipe.
  • the distributed or robotic inspection type infrared temperature sensing system monitors the temperature field in the pipe corridor environment in real time, and assists the sensing devices such as cameras and sensors configured in various areas of the pipe gallery through the transmission link. The information is processed by the sub-control center and collected in the monitoring center.
  • the temperature abnormality detection data of the optical fiber sensor is used as a control for the detection of the infrared temperature sensing system, and the trade-offs and corrections of the two results are as follows: (1) When the results of the two are the same, the test is considered correct;
  • the infrared thermal imaging image acquisition system is composed of DM60-S model infrared thermal imager and JVS-C300Q model data acquisition card.
  • the infrared thermal imager is used to capture the infrared image of the monitoring area, and the acquisition card assists in collecting thermal imaging image storage. It is easy to call and process at any time on the computer hard disk. At the same time, it assists in patrolling the robot, and installs 360 camera equipment.
  • the machine is equipped with WIFI equipment to realize real-time transmission of data information, which can realize regular fast and accurate leak inspection at any time.
  • the collected leaky crack infrared image is a color image containing brightness and color information. It is necessary to grayscale the crack image, that is, convert the originally collected color image into a grayscale image, and remove the color information in the image.
  • the color of each pixel in a color image is determined by three components: 1?, G, and B.
  • a pixel can have a variation range of more than 16 million (255*255*255) colors.
  • the grayscale image is a special color image with the same components of R, G, and B. The range of one pixel varies from 255. Therefore, in digital image processing, images of various formats are generally converted into gray. Degree image to reduce the amount of subsequent image calculations.
  • the description of the grayscale image like the color image, still reflects the distribution and characteristics of the overall and local chromaticity and brightness levels of the entire image.
  • the grayscale processing of an image can be implemented in two ways.
  • the first method makes the average of the three components of R, G, and B for each pixel, and then assigns this average to the three components of the pixel.
  • the second method is based on the color space of the YUV.
  • the physical meaning of the component of Y is the brightness of the point.
  • the value reflects the brightness level.
  • the brightness can be established with ⁇ 1, G, B.
  • the correspondence of the three color components: Y 0. 3R+0. 59G+0. 11B, the gray value of the image is expressed by this brightness value.
  • the brightness of the leaky crack image collected by the traditional image-based method is not uniform, and the gray value of the crack part and the background part of the image are greatly different. This large difference will bring certain processing to the subsequent processing. Difficulties, such as the selection of thresholds in image segmentation.
  • the present invention introduces an infrared image by using an ordinary image for crack recognition, and the image thereof is only caused by the difference in temperature. In the experimental environment described above, the image is collected without substantially having a temperature difference. Therefore, the present invention does not. Need to consider the impact of gray unevenness. Image noise reduction is suitable for ordinary images and infrared images. In reality, digital images are often subjected to digitization and transmission. Effects such as noise interference between the device and the external environment are called noisy images or noisy images.
  • noise is an important cause of image interference.
  • An image may have a variety of noises in practical applications, which may be generated in transmission or may be generated in processing such as quantization. According to the relationship between noise and signal, it can be divided into three forms: (f (x, y) represents a given original image, g (x, y) represents an image signal, and ⁇ ( ⁇ , y) represents noise.
  • f (x, y) g ( X , y) + n ( X , y), channel noise and light guide tube
  • f (x, y) g (x, y) +n (x, y) g (x, y), flying point
  • Quantization noise which is independent of the input image signal, is the quantization error in the quantization process, and is reflected to the receiver.
  • Noise can generally be defined as unpredictable in theory, and random errors can only be recognized by probabilistic methods. Therefore, it is more appropriate to regard the noise in the image as a multi-dimensional random process.
  • the random process can be used to describe the noise, that is, the probability distribution function and the probability density distribution function are used to represent, and the more mature noise reduction algorithm can be used.
  • the methods of image noise reduction mainly include the following types:
  • the mean value filter using the neighborhood averaging method is very suitable for removing particle noise in an image obtained by scanning.
  • the field averaging method strongly suppresses the noise, and at the same time causes the ambiguity due to the averaging, and the degree of ambiguity is proportional to the radius of the neighborhood.
  • the smoothness achieved by the geometric mean filter can be compared to an arithmetic mean filter, but less image detail is lost during the filtering process.
  • Harmonic averaging filters work better for "salt” noise, but not for "pepper” noise. It is good at dealing with other noises like Gaussian noise.
  • the inverse harmonic mean filter is more suitable for processing impulse noise, but it has the disadvantage that it must be known whether the noise is dark noise or bright noise, in order to select the appropriate filter order symbol, if the symbol of the order is wrong. May cause catastrophic consequences.
  • Combining on and off can be used to filter out noise.
  • the noisy image is turned on.
  • the structure element matrix can be selected to be larger than the noise, so the result of the turn-on is to remove the noise on the background.
  • the image obtained in the previous step is closed to remove the noise on the image.
  • the image type to which this method is applicable is that the size of the object in the image is relatively large, and there is no fine detail, and the effect of denoising the image of this type is better.
  • Edge detection is a fundamental problem in image processing and computer vision. The purpose of edge detection is to identify points in the digital image where the brightness changes significantly. Significant changes in image properties often reflect important events and changes in attributes. These include (i) discontinuities in depth, (i i ) surface discontinuities, (i i i) material property changes, and (iv) scene illumination changes. Edge detection is a research area in image processing and computer vision, especially feature extraction.
  • the search-based edge detection method first calculates the edge intensity, usually expressed by a first derivative, such as a gradient mode, and then uses the calculation to estimate the local direction of the edge, usually using the direction of the gradient, and using this direction to find the maximum value of the local gradient mode.
  • the zero-crossing based method finds the zero crossing of the second derivative derived from the image to locate the edge. Zero crossings are usually used for Laplacian or nonlinear differential equations. Filtering is usually necessary as a pre-processing of edge detection, usually using Gaussian filtering.
  • the published edge detection method applies a measure of the calculated boundary strength, which is fundamentally different from smoothing filtering.
  • edge detection methods use different kinds of filters to estimate the gradients in the X- and y_ directions.
  • Commonly used edge detection templates include Laplacian operator, Roberts operator, Sobel operator, log (Laplacian- Gauss) operator, Kirsch operator and Prewitt operator.
  • the infrared thermogram is used to extract the temperature difference point in combination with the threshold segmentation. The temperature difference between the leak point and the surrounding area is the difference between the RGB values reflected in the gray scale. Therefore, when the appropriate threshold is set, the temperature difference point can be extracted.
  • the crack target to be identified in the leak crack image is compared with the background, and the amount of information is relatively small, and in the process of image acquisition and transmission, the resolution and contrast of the image are reduced due to interference of many factors. Therefore, after filtering and denoising the image, the graphics are further enhanced, so that the crack target we are interested in is more prominent, providing a basis for the subsequent segmentation recognition algorithm.
  • Image threshold segmentation is a widely used segmentation technique. Using the difference between the target region and the background in the image, the image is treated as two types of regions (target region and background region) with different gray levels. The combination of a reasonable threshold is selected to determine whether each pixel in the image should belong to the target area or the background area, thereby generating a corresponding binary image. The purpose of this step of the invention is to find the cracked area and the non-cracked area.
  • the characteristics of the threshold segmentation method are: For the case where the target has a strong contrast with the background grayscale, it is important that the grayscale of the background or object is relatively simple, and the boundary of the closed and connected regions can always be obtained.
  • grayscale or color of cracks and pipes have obvious differences, and it is suitable to use image threshold segmentation algorithm.
  • image threshold segmentation algorithm The advantages of threshold segmentation are simple calculation, high computational efficiency, and fast speed. In applications where operational efficiency is important (such as for hardware implementation).
  • the process of image threshold segmentation algorithm can be represented by Figure 5.
  • a model is built and the original signal is characterized by the features of the model.
  • the model can be a one-dimensional histogram or a two-dimensional histogram. Etc., whether the model is reasonable or not is directly related to the results of subsequent processing. The more information is considered, the larger the calculation amount.
  • the image segmentation method based on two-dimensional histogram is more computational than the one-dimensional histogram based segmentation. The method is much bigger.
  • the second step of threshold segmentation is to determine the criterion for obtaining the threshold. In the case of a certain model, the criterion for obtaining the threshold determines the final segmentation threshold.
  • the third step of threshold segmentation is to obtain the segmentation threshold.
  • the exhaustive method can achieve good results; if the model established in the first step is more complicated, use poor
  • the method is time-consuming and not conducive to practical application. At this time, you can consider using the group intelligence algorithm to obtain the threshold, such as the particle swarm algorithm to find the threshold.
  • the image threshold segmentation method can be divided into single threshold segmentation method and multi-threshold segmentation method. If an image has only two types of target and background, a threshold can separate the target from the background. This method is called single. Threshold segmentation, if the image needs to be divided into multiple classes, that is, there are multiple different regions in the image, multiple thresholds are needed to separate them. This is called multi-threshold segmentation. Let the original image be /( ⁇ , 3, the result of the segmentation is ⁇ , , ⁇ is the segmentation threshold obtained, and the single threshold segmentation method can be defined as
  • the original image is /0,
  • the segmented result is ( , 3:)
  • T k is a series of segmentation thresholds, and the multi-threshold segmentation method can be defined as
  • Table 1 The accuracy of extracting temperature difference points by image threshold segmentation method Actual area of crack area The area extracted by segmentation algorithm
  • the edge extraction method is used to accurately extract the cracks and classify and locate them.
  • the traditional crack analysis technology based on image analysis only pays attention to the crack itself. It is necessary to detect the presence or absence of cracks, location and geometry, and the transition between the crack and the pipeline is basically not considered.
  • the invention mainly needs to obtain the temperature difference between the pipeline area and the crack area. It is necessary to pay attention to the two areas of the crack and the pipeline. Therefore, the influence of the transition area can be considered. After the crack and the pipeline area are obtained by image segmentation, the transition of the two regions is required. The area is cut, because the pipe temperature to the crack temperature is gradual, the temperature of the transition zone between the crack and the pipe is between the crack center temperature and the pipe temperature, and the transition zone is the interference zone for the pipe temperature, as shown in Figure 6. The following measures can be taken to eliminate the interference zone.
  • the transition interference zone is small relative to the area of the pipe area. Therefore, the width of the pipe area can be reduced by a fixed width.
  • the width can be set to w width (0. 5mm ⁇ w ⁇ 2. 5mm), in practice, it is also possible to adopt a method based on the maximum width ratio of the crack region, that is, after detecting the crack region, the crack region is doubled up and down, and the remaining region is defined as a pipeline region without interference zone. Assuming that the surface area of the pipe segmented by the image is 3 ⁇ 4, and finally the determined width is taken, then the following formula (7) is used. This method can better eliminate the influence of the transition interference zone.
  • the range of the transition interference zone is determined, how to delete it specifically can refer to the following:
  • For the pipeline area it is equivalent to moving the image segmentation boundary along the segment boundary to the pipe zone direction to obtain a new pipe zone boundary.
  • For the crack zone The image segmentation boundary is moved along the segmentation boundary radially toward the crack region to obtain a new crack region boundary.
  • the edge is the main part of the local intensity change of the image, mainly between the target and the background. By detecting the edges, the image information to be processed is greatly reduced, and the shape information of the objects in the image is retained.
  • Edge detection is a matrix convolution operation on an image using a template. The convolution operation is to multiply each pixel in the image region used by each element of the template (weight matrix), and the sum of all products is the new center pixel value for that region.
  • the sobel operator consisting of gradient and difference principles. It is a weighted average operator that highlights the edges by weighting them to the center point.
  • the image segmentation algorithm with adjustable threshold is used to distinguish the crack from the background, and the crack region and impurity with lower gray value are converted into black, and the background with higher gray value is converted to white.
  • the area threshold and the area-circumference fractal law are introduced to remove non-cracked areas such as impurities, and only the crack area is maximized.
  • Open and close operations first need to understand corrosion and expansion, corrosion: is a process to eliminate the boundary point and shrink the boundary to the inside. Can be used to eliminate small and meaningless objects; Expansion: is the process of merging all background points in contact with an object into the object, expanding the boundary to the outside. Can be used to fill holes in objects.
  • Open operation The process of first eroding and then expanding is calculated. It is used to eliminate small objects, separate objects at slender points, and smooth the boundaries of larger objects without significantly changing their area. The open operation is usually used when it is necessary to remove small particle noise and to break the adhesion between the targets. Its main function is similar to corrosion, and it has the advantage of keeping the original size of the target unchanged compared with the corrosion operation. Closed operation: The process of first eroding and then eroding is called a closed operation. It is used to fill small voids in objects, connect adjacent objects, and smooth their boundaries without significantly changing their area.
  • Morphological transformation of binary images in mathematical morphology is a process for collections.
  • the essence of the morphological operator is to express the interaction between the set of objects or shapes and the structural elements.
  • the shape of the structural elements determines the shape information of the signals extracted by this operation.
  • Morphological image processing is to move a structural element in the image, then intersect the structural element with the binary image below, and then perform the set operation.
  • the calculation method of fractal dimension is adopted. Not all irregular graphics have fractal features, and only graphics that satisfy self-similarity within a certain scale range satisfy the fractal feature.
  • the key is to calculate whether the fractal dimension values corresponding to the surface cracks of the structure meet the requirements of non-European space.
  • the fractal dimension can quantitatively and qualitatively reflect the degree of damage on the surface of the pipeline.
  • the irregular figure has a relationship between its circumference and the area: ⁇ ⁇ ⁇ 2 , where the circumference is expressed, ⁇ is the area, and the fractal dimension is obtained, and the logarithm of the two sides is obtained:
  • ⁇ ogP 0.5D ⁇ ogA + C (9) where: C is a constant.
  • the linear fitting accuracy is calculated by the following formula. When the value of ? 2 is close to 0, it means that the fitting accuracy is very poor; when the value of ? 2 is close to 1, it means that the fitting precision is very high.
  • the clear, complete and undisturbed pipe crack image is converted into a binary image by adaptive threshold segmentation, and the crack is extracted as a black region; then the boundary tracking method is used to calculate the crack area and perimeter of the formula, ie The number of pixels in the Unicom area is the area of the crack. The number of pixels on the statistical boundary is the perimeter of the crack. Finally, the area and perimeter values are substituted into the former formula, and the least squares method is used to obtain the pair. The line equation of the area-circumference relationship in numerical coordinates can be used to find the fractal dimension of the crack. Calculate by selecting 10 clear, complete cracks.
  • the crack region and the pipeline surface region can be obtained by using the existing mature algorithm, and then the RGB average value of the image in each region is obtained, and finally the average RGB value is matched with the colorbar, from left to right. Perform, step by two pixel width to ensure the accuracy of the temperature value, calculate the temperature value according to the matched position and color range.
  • Image area feature analysis is the ability of a computer to recognize or recognize an image, that is, image recognition.
  • Feature selection is a key issue in image recognition. The basic task of feature selection and extraction is how to find the most effective features from among the many features.
  • a set of original features is generated by calculation, which is called feature formation.
  • the number of original features is large, or the original sample is in a high-dimensional space.
  • the feature description in high-dimensional space can be described by the feature of low-dimensional space through mapping or transformation.
  • This process is called feature extraction.
  • Some of the most effective features are selected from a set of features to achieve the goal of reducing the dimensionality of the feature space. This process is called feature selection.
  • Concavity and convexity is one of the basic features of a region.
  • the concavity and convexity of a region can be determined by a method: a line connecting any two pixels in a region passes through a pixel outside the region, and this region is concave. Instead, connect a line segment of any two pixels in the image, if you don't pass a pixel other than this graphic, Then this figure is called convex.
  • the convex closure of this graphic After removing the portion of the original graphic from the convex closure, the position and shape of the resulting graphic will be an important clue for shape feature analysis.
  • the type of leakage of a known leak point Based on the regional characteristics of the leak point, such as roundness, density and other parameters, determine the type of leakage of a known leak point. When a certain characteristic of the leakage area satisfies the following conditions, it can be considered that the leakage area at this point is such a type of leakage.
  • This section mainly considers the perimeter/area of a leaking area and the area of the circumscribed rectangle of the leaking area & / the area S 2 of the area.
  • Table 2 shows the determined temperatures for leaks and undamaged points at different ambient temperatures. Table 2 Determining the temperature of the leaking point and the undamaged point at different ambient temperatures
  • the leakage temperature change values at each temperature are averaged to obtain the average temperature change curve of the leak, which is the non-destructive area.
  • the angle bisector of the two curves is taken as the boundary of the temperature change trend of the leak and the temperature of the undamaged portion, as indicated by the thick black dotted line in Fig. 10.
  • the data is linearly classified using a support vector machine.
  • the ambient temperature is plotted on the abscissa, and the temperature difference between the crack and the surface of the pipe is plotted on the ordinate. There are three levels of 1, 2, and 3. The greater the number, the more serious the development is.
  • the classification function graph can be obtained.
  • test results are judged according to the following:
  • the development degree of ⁇ ⁇ ⁇ 2 is 1; ⁇ 12 ⁇ ⁇ ⁇ ⁇ 23 , the development degree is 2, ⁇ ⁇ ⁇ 23
  • the degree of development is 3.
  • the temperature difference between the surface of the pipe and the crack is mainly related to the temperature, and the temperature difference between the crack and the surface of the pipe and the degree of development of the crack are related, so the crack and the pipe can be detected by the infrared camera.
  • the temperature difference of the surface and then use the above classification function and 2 , to detect the degree of crack development.
  • the detection environment meets the requirements, that is, the dark and humid underground pipe gallery space and the environmental temperature in the pipe gallery space are uniformly increased by more than 4 degrees Celsius. Then using the thermal imaging data acquisition, recording and when the ambient air temperature at each image capture, the post-treatment to give the pipe surface temperature cracks, degree of development of classification threshold values calculated from the temperature and eight 72 eight.
  • the pipe structure is a belt structure for transporting liquid.
  • the environmental factors and loads are the main causes of the structural damage of the pipeline.
  • the physical characteristics of the pipeline area are relatively consistent. Therefore, in the infrared image, the temperature values of the pipeline area are basically the same. It is convenient to handle, and does not need to be separately considered, that is, after obtaining the pipeline area in the image segmentation, it returns to the infrared image to calculate the RGB average value of the whole region.
  • the crack area is usually of a slender shape, such as a two-meter-long crack, which may be only a few millimeters wide.
  • the two-meter-long crack has a half-meter-long area, and the development is more serious. According to the average RGB value, it may need to be repaired, but for another three. Decimeter-long cracks, which also have a half-meter-long area, are more developed. The development degree index obtained from the average RGB value is lower than the previous one, and even shows no repair. It is obviously unscientific.
  • the effective crack region is divided into the P segments by the straight line along the y-axis direction, and each segment length is arbitrary; for the longitudinal crack, the effective crack region is divided into the p segments by the straight line along the X-axis direction of the image, each length Arbitrarily; for other types of cracks, not segmented or divided into segments according to the geometric center of the crack, each segment corresponds to the central angle.
  • the crack region divided by the image can be divided into p(p ⁇ 2) segments for consideration.
  • Each segment can be processed according to the technical route described above, that is, image graying and noise reduction are first performed.
  • image segmentation is performed, and the crack region is obtained and then divided into w segments for subsequent processing, including performing average RGB value calculation for each segment, and then matching the average RGB value with the color value in the legend to determine the segment.
  • the temperature value finally, the length of the p-segment crack region/the temperature difference ⁇ between the crack region and the pipe temperature is obtained, that is, the following array is obtained, as shown in the formula (17): ⁇ ⁇ 2 ⁇ ⁇ ⁇ ) ( 17)
  • ( ⁇ , ⁇ 2 ... ⁇ ⁇ ) ( 18) where ⁇ is the temperature difference between the segment/section and the pipe after the fracture zone is divided into sections.
  • a weighted average calculation can be performed according to the following formula (19) to obtain a final measured temperature difference ⁇ of the effective pipeline region and the effective crack region.
  • index of the development degree of each segment of the fracture region obtained by the array is as follows (20):
  • the invention divides the degree of crack development into three levels of 1, 2 and 3 through machine learning.
  • SVM supports vector machine.
  • the linear inseparable sample in low-dimensional space is mapped to the kernel function.
  • a linearly separable sample space in a high-dimensional space is calculated by a kernel function to obtain a linear classifier.
  • kernel functions such as linear kernel functions, polynomial kernel functions, radial basis kernel functions, Sigmoid kernel functions, and composite kernel functions.
  • the linear kernel function is first used to classify the 1 and 2 grades, and the classification function / 12 is obtained , and then the 2, 3 grades are classified according to the data to obtain the classification function / 23 , and there is a certain error in the machine learning classification.
  • the classification data some of the cracks belonging to the lighter development degree are assigned to the category of more serious development degree, and the cracks with the heavier development degree are assigned to the lighter category.
  • the average development index of the weights there will be a small number of fracture development index, because there is a certain error in the classification of these, although the development index of the decimals does not necessarily make the fracture development index more Precise, but can reflect the relative size, 2.
  • the development index of the 6 is divided into cracks with a developmental index of 3, and 2. 4 is divided into cracks with a developmental index of 2, which is the previous classification.
  • the method but the 2.6 developmental degree index and the 2.4 developmental degree index calculated by the weighting are actually not much different, which has a considerable influence on the manager's decision-making. Therefore, in the calculation of the fracture development degree index, the invention can take a decimal point to indicate the calculated relative relationship.
  • the developmental index of 2.4 is not necessarily more serious than the developmental index of 2.3, because as mentioned above, there is an error in the classification itself. 2 This integer part may already be inaccurate. .
  • the invention determines an error term according to the variance of the weighted average calculation method in the calculation process, that is, the development degree index can take one decimal place and is expressed as: m ⁇ form.
  • ( ° C ) is the ambient temperature
  • ( ° C ) is the temperature difference between the surface of the pipe and the crack
  • ⁇ 23 is a linear classification function coefficient
  • the value range is 0. 02-0. 03
  • b 23 linear classification function constant term The value ranges from 1.80-2. 55.
  • test results are judged according to the following
  • the result calculated by the formula can take one decimal place.
  • m' is the result of multiple tests on the same crack, and then the standard deviation is obtained from the above formula.
  • the development index of the crack can be written to one decimal place, and the error is defined by a fixed error term.
  • the fixed error term is obtained experimentally, and the new developmental degree index is written as follows (26):
  • the detection system of the pipeline crack development degree based on infrared thermography is used to construct the detection system.
  • the actual detection system needs to use infrared double
  • the eye camera captures the image of the pipe.
  • the binocular camera includes a normal camera and an infrared camera to simultaneously capture the image of the pipe and process it for identification of the presence, size, size and development of the crack. Data fusion can improve the robustness of the detection method proposed by the invention.
  • the crack detection system is modular in design and can be combined with other pipeline disease detection modules.
  • the crack detection module includes a camera fixing device, a binocular camera, a data transmission line, a vehicle terminal, a GPS receiving device, and inertial navigation.
  • the camera fixture can be customized with a fixed iron frame to ensure that the binocular camera can be reliably fixed on the engineering vehicle.
  • the patrol car head can be used to control the binocular camera angle while fixing the camera.
  • the pan/tilt has at least two degrees of freedom. , that is, it can be rotated in the horizontal direction and the vertical direction respectively, and more degrees of freedom will bring higher operation of the gimbal.
  • the minimum configuration of the infrared camera is: 320 X 240 resolution, with replaceable lens, maintenance-free uncooled microbolometer, microscopy and close-up measurement, data transfer speeds up to 60 Hz.
  • the normal image camera supports real-time output of 1080P HD images.
  • the data transmission line includes at least two, which respectively support high-frequency transmission of high-definition ordinary images and infrared images, and supports at least 100 Hz. 1080p video transmission.
  • the inspection vehicle terminal includes two schemes: one is front-end processing, the embedded PC is used for real-time video stream processing, and the processed data is transmitted through the 3g ⁇ 4g network.
  • the embedded PC can be compatible with the data of the ordinary image and the infrared image.
  • the processor supports real-time processing of video stream; the other is back-end processing, the inspection vehicle terminal as the front end of data acquisition, only responsible for data acquisition and storage, lower configuration and development requirements, requires an interface for collection
  • the three-party device is connected for data processing.
  • the module uses the video image splicing technology to restore the longitudinal image of the pipeline during the entire acquisition process, and then separately cuts along the length direction to identify the accuracy and stability of the pipeline crack detection. This system idea requires high image processing algorithms and relatively few devices.
  • the GPS accepts the device and the inertial navigation, which together locate the acquisition device to ensure accuracy and real-time. The accuracy of positioning needs to be within 10m.
  • Another type of crack detection module includes a camera fixing device, a binocular camera, a data transmission line, a patrol vehicle terminal, a GPS receiving device, an inertial navigation, a photoelectric encoder, and a synchronization controller (refer to CN104749187).
  • the photoelectric encoder is mounted on the wheel central axis of the inspection vehicle moving platform for measuring the running speed and distance of the inspection vehicle moving platform;
  • the GPS receiver is installed on the inspection vehicle mobile platform for the inspection The high-precision positioning and timing of the in-vehicle mobile platform;
  • the inertial navigation is installed on the platform of the inspection vehicle, and is used for measuring the in-vehicle mobile platform when the GPS receiver does not receive the GPS signal in the tunnel.
  • Position and attitude data realizing high-precision position estimation inside the tunnel;
  • Synchronous controller installed on the inspection vehicle mobile platform, used to synchronize the image acquisition time of ordinary cameras and infrared cameras to ensure uniform time and space between the two Benchmark. This method accurately measures the vehicle speed through the photoelectric encoder.
  • the synchronous controller automatically controls the binocular image acquisition time according to the vehicle speed and the field of view of the binocular camera, ensuring that the pipeline images of two adjacent effective clips have good continuity.
  • the collected pipes are completely covered and do not overlap each other.
  • the front-end processing or storage function is executed. This system has higher requirements on the device and lower image processing algorithms.
  • BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a technical road map for identifying the leakage of the internal space of the pipe gallery by infrared thermal image.
  • Figure 2 shows the grayscale information and temperature information of the interior space and cracks of the pipe gallery.
  • Figure 3 is a graph of illuminance versus temperature difference.
  • Figure 4 is a graph showing the relationship between the increase in ambient temperature and the temperature difference.
  • Figure 5 shows the image threshold segmentation model.
  • Figure 6 shows the crack area on the surface of the pipe, the pipe area and the transition interference zone.
  • Figure 7 is a flow chart of the crack extraction algorithm combining the skeleton and fractal features.
  • Figure 8 shows the fractal dimension of crack and impurity regions.
  • Figure 9 shows the threshold for looseness leakage at the interface.
  • Figure 10 shows the temperature change at the leak and undamaged after the ambient temperature rise.
  • Figure 11 is a linear classification diagram of the support vector machine.
  • Figure 12 is a schematic diagram showing different temperature differences obtained for different crack sections. detailed description (1) Environmental determination
  • the inspection model needs to be used under certain environmental conditions to ensure accuracy. First, it is necessary to ensure that the environmental conditions during data collection are satisfied: The dark and humid underground pipe gallery space and the ambient temperature in the pipe corridor space are uniformly increased by 4 degrees Celsius or more. It is necessary to ensure that the environmental conditions are stable when the data is collected, that is, the ambient temperature inside the pipe gallery is uniform, and after a certain temperature is raised, no drastic changes will occur.
  • the inner space of the pipe gallery was photographed by a temperature-measuring infrared camera, and the infrared thermal image of the crack region was analyzed.
  • the thermal imager is photographed at a horizontal distance of 1 m from the pipeline, and the machine position is kept at a constant speed from the crack area.
  • the infrared camera uses an infrared detector and an optical imaging objective to receive the infrared radiation energy distribution pattern of the target to be reflected on the photosensitive element of the infrared detector, thereby obtaining an infrared thermal image, the thermal image and the surface heat of the object.
  • the distribution field corresponds.
  • an infrared camera converts invisible infrared energy emitted by an object into a visible thermal image.
  • the different colors above the thermal image represent the different temperatures of the object being measured.
  • Invention of the current mainstream infrared imaging device Uncooled focal plane micro-thermal infrared camera.
  • the ambient temperature is directly measured by a thermometer, and the degree of crack development is manually measured according to the crack width.
  • the crack is divided into three levels of light, medium and heavy, and then consider the factors such as the humidity and depth of the crack.
  • the developmental degree index is scored, and the detection model is established based on the actual data.
  • the image pre-processing is performed to perform grayscale processing on the infrared image; the image is subjected to wavelet denoising and median filtering processing of different rectangles, while filtering noise, try not to blur the edge; using image gray enhancement algorithm, enhancing The contrast between the crack and the background area facilitates the extraction of the crack; the image segmentation algorithm with an adjustable threshold is used to segment the image, the crack region with lower gray value and the impurity are converted to black, and the background with higher gray value is converted to white, for the impurity
  • the non-cracked area can be removed by the area threshold and the area-circumference fractal rule, leaving only the crack area.
  • the RGB values are matched, and the temperature represented by the most consistent position is the temperature of the region.
  • the crack region and the pipe surface region are respectively matched with the legend to obtain respective temperatures, thereby obtaining the temperature difference between the crack and the pipe surface.
  • the image segmentation technology in image processing can identify the crack area directly after the temperature identification process of the crack area and the pipe surface area, and can also analyze the transition interference zone directly between the crack and the pipe surface according to the above description. .
  • the final temperature difference data can be obtained directly to calculate the temperature difference between the entire crack area and the surface area of the pipeline. It is also possible to divide the crack area into sections, and the average distribution method can be adopted, and then the section with higher development index is given.
  • the high weight according to the method of giving different weights, calculates the temperature difference between the final crack area and the surface area of the pipeline. Considering the different conditions of different pipe corridor environments, it is possible to try to establish a model of the relationship between the temperature difference of the cracks in the pipe corridor environment and the ambient temperature conditions.
  • the current data is corrected and discarded according to historical data; the temperature difference obtained by image analysis, and the temperature information collected by other instruments and the temperature difference data collected by the optical fiber temperature measurement are used to correct the relationship model.
  • the temperature difference obtained by image analysis, and the acquired image are collected by other instruments.
  • the temperature information is established to be related to the degree of crack development, that is, the specific values of the correlation coefficients of the linear classification functions / 12 and / 23 are mainly determined.
  • the above detection model can be used to detect the actual degree of crack development, and it is necessary to collect the infrared image of the cracked road surface and the current ambient temperature.
  • the index of crack development can be roughly divided into three levels of 1, 2, and 3. Only one of the three numbers can be taken. The larger the value, the more serious the development.
  • the calculation can also use the developmental index with a decimal number as mentioned above, the value range is 0-3, and the specific development degree index can be expressed in the form of ⁇ .

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Immunology (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • Pathology (AREA)
  • Mechanical Engineering (AREA)
  • Civil Engineering (AREA)
  • Architecture (AREA)
  • General Engineering & Computer Science (AREA)
  • Structural Engineering (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Medicinal Chemistry (AREA)
  • Food Science & Technology (AREA)
  • Acoustics & Sound (AREA)
  • Image Processing (AREA)
  • Investigating Or Analyzing Materials Using Thermal Means (AREA)
  • Radiation Pyrometers (AREA)
  • Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)

Abstract

一种基于静态红外热像图处理的地下管廊渗漏检测方法,其属于图像深度处理和渗漏检测技术领域。通过引入红外热图像可以获取地下管廊内部环境的灰度信息和温度信息,灰度信息可以实现管廊内部管道线路状态识别的常规目标,而利用温度信息可以进行管道渗漏的检测以及后期检修维护提供参考。

Description

一种基于静态红外热像图处理的地下管廊渗漏检测方法
技术领域
本发明属于图像深度处理和渗漏检测技术领域,具体涉及一种基于红外热像图处理的地下 管廊渗漏检测方法。 通过引入红外热图像可以获取地下管廊内部环境的灰度信息和温度信息, 灰度信息可以实现管廊内部管道线路状态识别的常规目标,而利用温度信息可以进行管道渗漏 的检测以及为后期检修维护提供参考,主要通过实验和结合机器学习的技术建立起管道渗漏的 判别模型。 背景技术
近年来, 我国地下综合管廊空间建设迅速发展, 其管理和养护问题也愈发凸显, 其中渗漏 检测就成为相关管养部门的工作重点之一。 传统的现场式排査的检测效率取决于管道的尺寸, 材料和深度, 检测过程需要手动干预, 检测条件依赖于天气、 管道表面状态以及水压等条件。 用于规模式检测渗漏的现有技术仍以闭路电视监控技术为主, 效率低下且需要人工排査, 时间 和人力成本较高。得益于数字图像处理技术的飞速发展,地下管廊智能检测手段也越来越多样 化。 最新最有效的方法是光纤技术, 无线网络传感器检测技术, 超声导波技术, 管道内微渗漏 检测技术和热成像技术。 现有技术
CN206573258U; CN101070947A; CN102155628A o
术语解释
发育程度: 发明中所提及的发育程度表征管道渗漏裂缝对管道的已有损害程度以及近期发 生损害的严重程度, 它涵盖了传统的裂缝严重程度分级, 与裂缝的长度、 宽度和面积等等可能 会使裂缝损害严重程度加剧的因素有关, 囊括了这些因素,表征了裂缝从开始出现到现在的发 展水平。
支持向量机: 支持向量机 (SVM)是 90年代中期发展起来的基于统计学习理论的一种机器 学习方法,通过寻求结构化风险最小来提高学习机泛化能力, 实现经验风险和置信范围的最小 化, 从而达到在统计样本量较少的情况下, 亦能获得良好统计规律的目的。 在机器学习中, 支 持向量机(SVM,还支持矢量网络)是与相关的学习算法有关的监督学习模型,可以分析数据, 识别模式, 用于分类和回归分析。 给定一组训练样本, 每个标记为属于两类, 一个 SVM训练算 法建立了一个模型,分配新的实例为一类或其他类,使其成为非概率二元线性分类。通俗来讲, 它是一种二类分类模型, 其基本模型定义为特征空间上的间隔最大的线性分类器, 即支持向量 机的学习策略便是间隔最大化, 最终可转化为一个凸二次规划问题的求解。
分类函数:通过支持向量机将温差数据按照裂缝发育程度进行线性分类,发育程度分为 1、 2、 3三个等级, 3最严重, 那么 1、 2之间和 2、 3之间就会有一直线作为分界线, 此直线表达 式即为分类函数。
裂缝区: 管道某区域, 该区域不仅包括裂缝区域本身, 还包括其周围一定范围内的管道区 域, 包括管道区域范围满足图像处理和裂缝识别的要求。
参照温差数据: 将实测环境温度带入裂缝发育程度检测模型两个分类函数后得到的温差数 据。
实测温差数据: 对采集到的裂缝区红外图像进行图像处理后得到的该裂缝区内裂缝区域和 管道表面区域的温差数据。
失真度:本发明中数据失真度的含义是数据产生不合理值后,该不合理值距离合理的偏差。 发育程度等级: 1、 2或 3中的一个数, 其大小反映了裂缝的发育程度, 数字越大, 裂缝发 育程度越严重。
发育程度指数: 介于 0-3之间, 包括 0和 3的一个数, 其大小反映了裂缝的发育程度, 数 字越大, 裂缝发育程度越严重。 说明中采用字母 m表示。
渗漏严重程度指数: 本发明中渗漏严重程度指数的含义是指管道裂缝在长度、 宽度和面积 三个维度上的严重程度的平均。 取值范围在 [0-9. 9]之间, 保留 1位小数, 其大小反映了渗漏 的严重程度, 数字越大, 渗漏程度越严重。 发明内容
本发明的目的在于, 提供一种基于红外热像图处理的地下管廊渗漏检测方法。通过测温型 红外热像仪拍摄管廊内部空间, 获取管道渗漏处附近的红外热像图, 建立起渗漏裂缝与管道未 渗漏处和渗漏处温差的关系模型,再通过结合机器学习的计算机视觉技术建立起管道渗漏的判 别模型。 实验原理参照 PCT / IB2016/058109。 附图 1为利用红外热像图识别管廊内部空间渗 漏的技术路线。
普通图像只能获得地下管廊环境空间, 管道线路和渗漏裂缝的灰度信息, 灰度信息可以实 现管廊内部管道线路状态识别的常规目标,普通图像难以较好识别渗漏裂缝的原因是因为管道 表面材质具有颗粒性, 灰度的杂乱往往会影响裂缝区域的识别。我们通常通过识别裂缝的宽度 来识别渗漏裂缝, 而裂缝的宽度是通过识别裂缝区域短边的像素点个数, 并结合摄像机机位拍 摄裂缝时的高度来确定。然而,通过裂缝像素点宽度的识别来得到裂缝宽度进而确定裂缝严重 程度的检测方法很不可靠, 因为几毫米的的裂缝宽度在图像中可能就只有几个像素点,在经过 降噪等图像处理后结果更不稳定,所以, 只由普通图像获取的裂缝的灰度信息进行裂缝判别是 很困难的。
红外温度测量技术是利用红外光学***将待测目标的红外热图像成在红外焦平面探测器 上, 经过处理后得到待测目标的红外图像, 然后根据图像的灰度值及定标数据、相关参数对物 体的温度场分布进行分析。基于全场分析的温度测量技术可以对大面积的区域进行测温, 可以 同时得到一个物体多区域的温度和多个物体的温度,对分析物体的状态具有重要的意义, 因此 成为了近些年研究的热点。 由红外热像仪获取的红外热图像, 不仅可以获取管廊内部空间和裂 缝的灰度信息还可以得到他们的温度信息, 如附图 2, 为管廊内部空间和裂缝的灰度信息和温 度信息。利用温度信息可以进行管道渗漏的检测以及为后期检修维护提供参考, 主要通过实验 和结合机器学习的计算机视觉技术建立起管道渗漏的判别模型。 使用红外热像仪检测渗漏时, 本发明主要解决以下三个问题:
( 1 ) 地下管廊空间较为阴暗潮湿,利用传统方式检测内部管道的渗漏问题,会由于光线昏暗、 温度较低等因素受到多种限制, 导致误差较大, 甚至无法被检测到。 所以, 本发明通过 红外热图像清晰地发觉渗漏区域与管廊环境的明显区别, 可以有效识别渗漏点, 并且利 用图像处理技术可以显示渗漏破坏的大小等参数。
( 2 ) 地下综合管廊内的渗漏主要分为小型裂缝破坏引起的渗漏、局部区域损坏引起的渗漏和 管道接口松动引起的渗漏。 以上三种情况, 渗漏点的特征是有区别的, 所以红外热像仪 拍摄得到的图像就会呈现出不同特征。所以,本发明需要设定科学的红外数据信息筛选 标准来区别不同种类的渗漏情况, 从而保证检测结果的适用性, 为日后根据具体分类进 行定向的检修提供可能。
( 3 ) 成功识别地下综合管廊内的渗漏破坏后, 需要及时地提取该渗漏点的位置信息, 并将其 反馈给检修维护相关部门,方便对管廊内危险渗漏点的即时修复。本发明给出一个准确 高效的定位技术, 通过对兴趣域图形和模板图形的筛选匹配, 实现渗漏点的定位。 由前文所述, 裂缝发育程度越严重, 管道表面以下温度就会越多的体现出来, 而管道表面 以下的温度与其表面温度是有差距的, 故裂缝发育程度越严重, 其在相同条件下所表现出来与 管道表面温度的差异性就会越大。因此,本发明目的就是通过在一定条件下管道表面与裂缝的 温度差异来体现裂缝的发育程度, 它们的对应关系就是检测模型。 故为解决上述问题, 本发明 采用的技术方案包括:
基于红外热成像的渗漏分析模型的基础是温度场信息测量功能,该功能依赖红外图像温度 场识别算法,可实现对采集设备采集到的红外图像或者保存在计算机上的红外图像数据进行分 析,根据图像上各像素点的灰度值按照一定的数学模型计算出对应的温度值, 并以设置的测温 方式进行测温。 测温方式有点测温、 区域测温两种, 点测温是对单个像素点进行测温, 区域测 温是对一块区域的平均值进行测温。因此可根据图像数据实现目标场景温度全局、局部区域和 点的测量。
同时, 由于水和水管材料的比热容不同, 当管道发生漏水, 会导致渗漏点附近区域产生温 度差异, 红外热像仪可以可视化地检测到渗漏点管道上方的表面的温度差异, 并且根据上述温 度测量方式实现温度差异的感知。
由管道表面温度滞后性和积累性的特点, 管道表面和内部的温度会不同, 因此管道内部的 介质材料会通过裂缝或其他破损区域与管道表面和环境空气发生热交换,而管道表面和裂缝处 也会有温度差。管道裂缝长度和宽度越大,管内介质和管道表面的材料空气热交换就会越剧烈, 温差就会越大。裂缝的长度和宽度越大, 将来的水损害也会更容易发生, 渗漏也通过严重程度 指数反映其程度。 所以, 裂缝越严重, 管道表面与裂缝的温差就会越大, 而我们可以利用热像 仪来检测温差, 进而检测渗漏的严重程度。
通过研究发现,照度会影响管道表面的温度,但是此因素会同时影响管道表面和裂缝温度, 对于他们的温度差异影响不大, 只会对管道表面和环境温度差异产生较大的影响, 如附图 3 所示。 其他条件相同的情况下, 气温越高, 裂缝与管道表面的温差越大。 而我们需要通过温差 来反映裂缝的发育严重程度, 因此需要确定严重程度指数就需要有确定的温度差标准, 因此, 需要对不同气温下的裂缝与管道表面温度差进行修正。
渗漏的严重程度, 可以有多种衡量方式, 本发明分别从长度、 宽度、 面积三个角度加以描 述, 定义 G为渗漏严重程度指数。
G1 = [^ x l00%x l0] ( 1 ) 其中, ^为裂缝长度维度上的严重程度指数; Z为裂缝的长度,单位为像素点数; 为管道 的外直径, 单位为 mm; 为相机的每个像素在长度上的分辨率, 单位为 mm/像素点。
s
G2 = [2^ x l00%X l0] (2) 其中, ^为裂缝宽度维度上的严重程度指数; S为裂缝的面积, 单位为像素点数; Z为裂缝 的长度, 单位为像素点数; r2为相机的每个像素在宽度上的分辨率, 单位为 mm/像素点。
G3 = [¾^ 100%X 10] (3) τχ
其中, ^为裂缝面积维度上的严重程度指数; S为裂缝的面积, 单位为像素点数; Ζ为裂缝 的长度, 单位为像素点数; ^, r2同上, 为管道的外直径, 单位为 mm。
G = Gl +Gz +G3 (4)
3
G为裂缝的严重程度指数, 结果保留 1位小数。
在发布层次, 将管道渗漏的严重程度指数分成了以下轻、 中和重三个程度等级: 轻度: G = [0 - 2.9];
中等: G = [3— 6.9];
严重: G = [7— 9.9]。
常规的严重程度指数在 [0-9. 9]之间,但是不排除特别严重的裂缝导致该指数达到 10及以 上的情况, 此时取 G=9. 9。 环境升温与温差关系图如附图 4所示。
渗漏的严重程度主要衡量依据是它对管道已经造成的危害即近期的潜在损害,其严重程度 的根本不仅与裂缝的宽度、 长度和面积有关, 还应该与裂缝的深度有关, 深度越大其对管道所 造成的损害就越大, 进一步引发水损害的可能性也越大, 同时, 裂缝周围材料的腐蚀情况也是 影响裂缝严重程度的因素之一, 因此, 渗漏严重程度需要我们综合考虑。 在现有技术中, 图像 分析已经可以探测裂缝的长度、 宽度和面积, 通过激光雷达可以检测裂缝的深度, 肉眼观察可 以估计裂缝周围材料的腐蚀情况,但是如何准确而简单的反映出裂缝的发育程度, 是工程中所 亟待解决的问题。 图像获取
获取稳定的、 足够质量的红外图像是进行后续处理的基础。
非制冷焦平面测温型红外热像仪是利用红外探测器和光学成像物镜接受被测目标的红外 辐射能量分布图形反映到红外探测器的光敏元件上, 从而获得红外热像图,这种热像图与物体 表面的热分布场相对应。通俗地讲红外热像仪就是将物体发出的不可见红外能量转变为可见的 热图像。热图像的上面的不同颜色代表被测物体的不同温度。发明现在主流红外成像装置 非制冷焦平面测温型红外热像仪。
利用红外热像仪采集裂缝区时, 设备距离管道表面的高度为 0. 5-lm, 具体地高度可采用 0. 5m, 0. 6m, 0. 7m, 0. 8m, 0. 9m, 1. 0m, 要求拍摄到能够满足后续图像处理进行裂缝识别的要 求即可。 使用热像仪拍摄图像分辨率至少为 384 X 288 ; 在温度为 30 °C时, 热灵敏度至少保证 在 0. 06 °C ; 拍摄帧频至少为 50Hz, 用于采集实验水管在环境温度下的红外图像。 DM60-S型在 线式红外热像仪的主要成像参数, 其采集到的红外图像可以达到分辨率要求, 可作为获取红外 图像设备的参考。
同理,普通图像采集设备与红外热像仪构成双目摄像机, 要求两者拍摄图片的视角保持一 致, 同一时刻所采集的普通图像与红外图像具有最高的可重复性,数据采集端设计为全区域图 像采集。
红外热像仪可采用固定摄像机进行拍摄。 由于固定设备检测, 则因为渗漏发生的特点, 其 数据具有不可逆性、 连续性以及趋势性。
其中不可逆性指针对红外热像仪采集到的可疑渗漏点的图片数据,若前一时刻有渗漏裂缝 存在, 则下一时刻应当也存在。
其中连续性指, 因为渗漏裂缝扩散的速度较为缓慢, 通常以 1小时为检测周期, 后一时刻 的渗漏裂缝的参数应当只在前一时刻的渗漏裂缝的参数基础上发生微小改变。表现在红外图像 上, 红外图像的颜色和可能渗漏裂缝的尺寸参数变化曲线应当是平滑的, 变化极小的。
其中趋势性指, 由于渗漏裂缝发生具有不可逆性, 故后一时刻的裂缝一定会大于等于前一 时刻的严重程度。表现在红外图像上, 红外图像的颜色会不变或加深, 同时可能渗漏裂缝的尺 寸参数会不变或变大。
管廊内敷设测温光纤传感器,通过测温光纤传感器将检测结果传输给分布式光纤温度传感 主机, 传感主机对采集信号实时分析处理, 实现对供热管道的监测工作。 同时, 分布式或机器 人巡检式的红外测温传感***, 实时对管廊环境内的温度场进行监测,辅助以管廊内各区域配 置的摄像机、传感器等传感设备, 通过传输链路将信息经分控中心处理后汇集到监控中心。其 中, 光纤传感器的温度异常检测数据, 作为红外测温传感***检测的对照, 二者结果的取舍与 修正如下: ( 1 ) 当二者结果相同时, 认为检测正确;
( 2 ) 当二者结果不同时,若光纤传感器结果显示发生渗漏, 而红外测温传感***显示未 发生渗漏, 认为发生渗漏; 若相反, 则认为没发生渗漏。
( 3 ) 当某一仪器没有反馈数据时, 取另一台数据的结果;
( 4) 当二者均没有反馈数据时, 认为***出现故障, 及时检修。
其中, 红外热成像图像采集***由 DM60-S型号的红外热像仪和 JVS-C300Q型号的数据采 集卡组成, 红外热像仪用以拍摄监测区域的红外图像,采集卡辅助采集热成像图片存储于计算 机硬盘内, 方便随时的调用和处理。 同时, 辅助以巡检机器人, 并且安装 360摄像头设备, 机 上配备有 WIFI设备, 实现数据信息的实时传输, 随时可以实现定期快速准确的渗漏巡检。 图像处理
在实际应用中,由于各种环境干扰将严重影响裂缝图像质量,对后期的裂缝检测造成困难。 所以图像的预处理是很有必要的,通过预处理可以把图像中的一些冗余信息去掉, 突出我们感 兴趣的目标, 从而达到减少图像信息量和改善图像质量的目的。 为此, 需要对拍摄的图像进行 预处理, 主要包括图像灰度化、 降噪、 边缘检测或者阈值分割、 开闭运算、 区域分割、 裂缝几 何信息识别几大步骤。 采集到的渗漏裂缝红外图像是包含有亮度和色彩信息等的彩色图像。需要对裂缝图像灰度 化, 即将原来采集到的彩色图像转换为灰度图像, 将图像中的色彩信息去掉。彩色图像中的每 个像素的颜色有1?、 G、 B三个分量决定, 一个像素点可以有 1600多万 (255*255*255 ) 的颜色 的变化范围。 而灰度图像是 R、 G、 B三个分量相同的一种特殊的彩色图像, 其一个像素点的变 化范围为 255种,所以在数字图像处理中一般先将各种格式的图像转变成灰度图像以减少后续 的图像计算量。灰度图像的描述与彩色图像一样仍然反映了整幅图像的整体和局部的色度和亮 度等级的分布和特征。 图像的灰度化处理可用两种方法来实现。
第一种方法使求出每个像素点的 R、 G、 B三个分量的平均值, 然后将这个平均值赋予给这 个像素的三个分量。
第二种方法是根据 YUV的颜色空间中, Y的分量的物理意义是点的亮度, 由该值反映亮度 等级, 根据 RGB和 YUV颜色空间的变化关系可建立亮度 ¥与 1?、 G、 B三个颜色分量的对应: Y=0. 3R+0. 59G+0. 11B, 以这个亮度值表达图像的灰度值。
传统基于普通图像的方法采集到得渗漏裂缝图像的亮度是不均匀的,图像中裂缝部分和背 景部分的灰度值有较大的差异,这种比较大的差异会给后续处理带来一定的困难, 比如图像分 割中阈值的选取等。而本发明在利用普通图像进行裂缝识别的同时引入红外图像进行处理, 其 图像只会因为温度的不同造成差异,在前文所述实验环境下采集图像基本不会有温度差,因此, 本发明不需要考虑灰度不均匀所造成的的影响。 图像降噪适用于普通图像和红外图像,现实中的数字图像在数字化和传输过程中常受到成 像设备与外部环境噪声干扰等影响,称为含噪图像或噪声图像。减少数字图像中噪声的过程称 为图像降噪,有时候又称为图像去噪。 噪声是图像干扰的重要原因。 一幅图像在实际应用中可 能存在各种各样的噪声,这些噪声可能在传输中产生,也可能在量化等处理中产生。根据噪声和 信号的关系可将其分为三种形式: (f (x, y)表示给定原始图像, g (x, y)表示图像信号, η (χ, y)表 示噪声。
(1) 加性噪声,此类噪声与输入图像信号无关,含噪图像可表示为 f (x,y) =g (X,y) +n (X,y), 信道噪声及光导摄像管的摄像机扫描图像时产生的噪声就属这类噪声;
(2) 乘 性 噪 声 , 此 类 噪 声 与 图 像 信 号 有 关 , 含 噪 图 像 可 表 示 为 f (x, y) =g (x, y) +n (x, y) g (x, y),飞点扫描器扫描图像时的噪声,电视图像中的相干噪声,胶片中 的颗粒噪声就属于此类噪声。
(3) 量化噪声,此类噪声与输入图像信号无关,是量化过程存在量化误差,再反映到接收端 而产生。
噪声一般在理论上可定义为不可预测的, 只能用概率统计方法来认识的随机误差。因此可 以将图像中的噪声看成多维随机过程是比较恰当的, 可以借助随机过程来描述噪声, 即用概率 分布函数和概率密度分布函数去表示,使用现在已有较成熟的降噪算法即可, 图像降噪的方法 主要有以下几类:
(1)均值滤波器
采用邻域平均法的均值滤波器非常适用于去除通过扫描得到的图象中的颗粒噪声。领域平 均法有力地抑制了噪声,同时也由于平均而引起了模糊现象,模糊程度与邻域半径成正比。几何 均值滤波器所达到的平滑度可以与算术均值滤波器相比,但在滤波过程中会丢失更少的图象细 节。 谐波均值滤波器对 "盐"噪声效果更好,但是不适用于 "胡椒"噪声。 它善于处理像高斯 噪声那样的其他噪声。逆谐波均值滤波器更适合于处理脉冲噪声,但它有个缺点,就是必须要知 道噪声是暗噪声还是亮噪声,以便于选择合适的滤波器阶数符号,如果阶数的符号选择错了可 能会引起灾难性的后果。
(2)自适应维纳滤波器
它能根据图象的局部方差来调整滤波器的输出,局部方差越大,滤波器的平滑作用越强。它 的最终目标是使恢复图像 f' (x, y)与原始图像 f (x, y)的均方误差 e2=E [ (f (x, y) -f' (x, y) 2]最 小。该方法的滤波效果比均值滤波器效果要好,对保留图像的边缘和其他高频部分很有用,不过 计算量较大。 维纳滤波器对具有白噪声的图象滤波效果最佳。
(3)中值滤波器
它是一种常用的非线性平滑滤波器,其基本原理是把数字图像或数字序列中一点的值用该 点的一个领域中各点值的中值代换其主要功能是让周围象素灰度值的差比较大的像素改取与 周围的像素值接近的值,从而可以消除孤立的噪声点,所以中值滤波对于滤除图像的椒盐噪声 非常有效。中值滤波器可以做到既去除噪声又能保护图像的边缘,从而获得较满意的复原效果, 而且,在实际运算过程中不需要图象的统计特性,这也带来不少方便,但对一些细节多,特别是 点、 线、 尖顶细节较多的图象不宜采用中值滤波的方法。
(4)形态学噪声滤除器
将开启和闭合结合起来可用来滤除噪声,首先对有噪声图象进行开启操作,可选择结构要 素矩阵比噪声的尺寸大,因而开启的结果是将背景上的噪声去除。 最后是对前一步得到的图象 进行闭合操作,将图象上的噪声去掉。根据此方法的特点可以知道,此方法适用的图像类型是图 象中的对象尺寸都比较大,且没有细小的细节,对这种类型的图像除噪的效果会比较好。
(5)小波去噪
这种方法保留了大部分包含信号的小波系数,因此可以较好地保持图象细节。 小波分析进 行图像去噪主要有 3 个步骤:对图象信号进行小波分解;对经过层次分解后的高频系数进行阈 值量化;利用二维小波重构图象信号。 边缘检测是图像处理和计算机视觉中的基本问题,边缘检测的目的是标识数字图像中亮度 变化明显的点。 图像属性中的显著变化通常反映了属性的重要事件和变化。 这些包括(i )深 度上的不连续、 (i i )表面方向不连续、 (i i i )物质属性变化和 (iv)场景照明变化。 边缘检 测是图像处理和计算机视觉中, 尤其是特征提取中的一个研究领域。
有许多用于边缘检测的方法, 他们大致可分为两类: 基于搜索和基于零交叉。基于搜索的 边缘检测方法首先计算边缘强度, 通常用一阶导数表示, 例如梯度模, 然后, 用计算估计边 缘的局部方向, 通常采用梯度的方向, 并利用此方向找到局部梯度模的最大值。 基于零交叉 的方法找到由图像得到的二阶导数的零交叉点来定位边缘。 通常用拉普拉斯算子或非线性微 分方程的零交叉点。滤波作为边缘检测的预处理通常是必要的, 通常采用高斯滤波。 已发表的 边缘检测方法应用计算边界强度的度量, 这与平滑滤波有本质的不同。 正如许多边缘检测方 法依赖于图像梯度的计算, 他们用不同种类的滤波器来估计 X-方向和 y_方向的梯度。 常用的 边缘检测模板有 Laplacian算子、 Roberts算子、 Sobel算子、 log ( Laplacian- Gauss )算子、 Kirsch算子和 Prewitt算子等。 其次, 利用红外热像图, 结合阈值分割提取温度差异点。漏点和周围区域之间的温度差是 灰度中反映的 RGB值之间的差异。 因此, 当设置适当的阈值时, 可以提取温度差异点。
渗漏裂缝图像中需要识别的裂缝目标和背景相比较,信息量比较少, 而且在图片的采集和 传输过程中, 由于很多因素的干扰也会使得图像的清晰度和对比度有所降低。所以在对图像进 行滤波去噪之后, 还要对图形进一步增强, 使得我们感兴趣的裂缝目标更加突出, 为后面的分 割识别算法提供基础。
图像阈值分割是一种广泛应用的分割技术,利用图像中要提取的目标区域与其背景在灰度 特性上的差异,把图像看作具有不同灰度级的两类区域(目标区域和背景区域)的组合,选取一 个比较合理的阈值, 以确定图像中每个像素点应该属于目标区域还是背景区域, 从而产生相应 的二值图像。本发明此步骤目的就是要找出裂缝区域和非裂缝区域。 阈值分割法的特点是: 适 用于目标与背景灰度有较强对比的情况, 重要的是背景或物体的灰度比较单一, 而且总可以得 到封闭且连通区域的边界。无论是在灰度图像还是红外图像中,裂缝与管道的灰度或色彩都具 有较明显的差异性,适合使用图像阈值分割算法。阈值分割的优点是计算简单、运算效率较高、 速度快。 在重视运算效率的应用场合 (如用于硬件实现)。
图像阈值分割算法的过程可以用附图 5来表示, 先建立一个模型, 用这个模型的特征去表 征原有的信号, 在图像阈值分割领域, 这个模型可以是一维直方图、 二维直方图等, 模型的合 理与否直接关系到后续处理的结果, 考虑的信息越多, 计算量也就越大, 如基于二维直方图的 图像分割方法的计算量就比基于一维直方图的分割方法要大很多。阈值分割的第二步是确定求 取阈值的准则, 在模型一定的情况下, 求取阈值的准则决定了最终的分割阈值, 求取阈值的准 则有很多, 如最大熵法, 最大类间方差法等。 阈值分割的第三步是求取分割阈值, 在阈值分割 第一步建立的模型不复杂的情况下, 用穷举法可以取得好的效果; 如果第一步建立的模型的较 为复杂, 用穷举法耗时大, 不利于实际应用, 这个时候可以考虑用群智能算法求取阈值, 如粒 子群算法求阈值。
根据图像需要分类的多少, 图像阈值分割法可以分为单阈值分割法和多阈值分割法, 如果 一幅图像只有目标和背景两类,一个阈值就可以将目标和背景分开,这种方法叫做单阈值分割 法,如果图像需要分为多个类,即图像中有多个不同的区域,则需要多个阈值才能将他们分开, 这叫做多阈值分割方法。设原始图像为 /(χ, 3 ,分割后的结果为^^, , Τ为求取的分割阈值, 单阈值分割方法可以定义为
Figure imgf000011_0001
在多阈值分割方法中, 设原始图像为 /0,) , 分割后的结果为 ( , 3:), T0,T\... Tk为一系 分割阈值, 多阈值分割方法可以定义为
g( ,j = t,当; < f(x,y) < Tk+l,k = 0,\,2n. (6) 上式中, k为分割后图像中各个不同区域的标号。 提取温度差异点的准确率如表 1。
表 1 图像阈值分割法提取温度差异点的准确率 裂缝区域实际面积 分割算法提取的面积
环境升温 (°C) 正确率 (%)
(像素) (像素)
0 335.00 90.05
2 339.00 91.13
4 352.00 94.62
6 356.00 95.70
372.00
8 347.00 93.28
10 349.00 93.82
12 351.00 94.35
14 352.00 94.62 16 364.00 97.85 边缘提取与图像分割
之后, 运用边缘提取的手段, 将裂缝准确提取, 并进行分类和定位。
传统基于图像分析的裂缝识别技术仅关注裂缝本身, 需要检测出裂缝的有无、位置和几何 形态等因素,裂缝与管道之间的过渡区域基本不作考虑。发明主要是要获取管道区域和裂缝区 域的温度差, 是需要关注裂缝和管道两个区域的, 因此可以考虑过渡区域的影响, 通过图像分 割获取裂缝和管道区域后, 需要对两个区域边界过渡区域进行删减, 因为管道温度到裂缝温度 是渐变的,裂缝和管道的过渡区温度介于裂缝中心温度和管道温度之间,过渡区对于管道温度 的获取就是干扰区, 如附图 6所示, 可以采取下列措施对干扰区进行排除。
对于管道区域,过渡干扰区相对管道区域面积而言很小, 因此在进行排除的时候可以采取 固定的宽度进行管道区域的删减, 此处的宽度可以设定为 w宽(0. 5mm〈w〈2. 5mm), 实际还可以 采取基于裂缝区域最大宽度比值的方法, 即检测出裂缝区域后,将裂缝区域上下各增加一倍宽 度后, 剩下的区域定为无干扰区的管道区域, 假设图像分割出来的管道表面区域宽度为 ¾, 最终采取确定的宽度为 , 则有以下公式 (7), 此方法可以较好的排除过渡干扰区的影响。
D = D0 - 2d0 (7) 对于裂缝区域, 因为其宽度较小, 因此过渡干扰区的排除需要更细致, 可以采用基于裂缝 区域宽度比例的方法来进行排除,对图像分割的裂缝区域需要进行瘦身,假设裂缝的宽度为 图像分割出来的裂缝区域宽度为 αϋ, 则满足以下条件如式 (8 ) 所示- dx = d^ - 2ad0 (8) a为为了排除过渡干扰区而对裂缝宽度上下部分分别删减的比例, 0.1≤ «≤ 0.2.
过渡干扰区范围确定后, 如何具体对其进行删除可以参考以下: 对于管道区域, 相当于将 图像分割边界沿着分割边界径向往管道区域方向移动 距离得到新的管道区域边界, 对于裂 缝区域, 相当于将图像分割边界沿着分割边界径向往裂缝区域方向移动 距离, 得到新的裂 缝区域边界。 在数字图像中, 边缘是图像局部强度变化的主要部分, 主要是在目标和背景之间。通过检 测边缘, 大大减少了要处理的图像信息, 保留了图像中物体的形状信息。边缘检测是使用模板 对图像进行矩阵卷积运算。卷积运算是使用的图像区域中的每个像素乘以模板的每个元素(权 重矩阵), 并且所有乘积之和是该区域的新的中心像素值。
Figure imgf000012_0001
新的中心像素值可表示为, R5 (centerpixel) = + R2G2 + R3G3 + R4G4 + R,G5 + Rfi6 + R7G7 + ¾ 8 + ϋβ9 , 之后, 在数字图像检测方法中使用由梯度和差分原理组成的 sobel算子。 它是一个加权 平均算子, 通过加权到中心点突出显示边缘。 同时, 利用阈值可调的图像分割算法对裂缝和背 景进行区分, 将灰度值较低的裂缝区域和杂质转化为黑色, 将灰度值较高的背景转换为白色。 此外, 引入面积阈值和面积 -周长分形定律去除杂质等非裂纹区域, 仅使裂纹面积达到最大。
开闭运算, 首先需要了解腐蚀和膨胀, 腐蚀: 是一种消除边界点, 使边界向内部收缩的过 程。 可以用来消除小且无意义的物体; 膨胀: 是将与物体接触的所有背景点合并到该物体中, 使边界向外部扩张的过程。 可以用来填补物体中的空洞。 开运算: 先腐蚀后膨胀的过程开运 算。用来消除小物体、在纤细点处分离物体、平滑较大物体的边界的同时并不明显改变其面积。 开运算通常是在需要去除小颗粒噪声, 以及断开目标物之间粘连时使用。其主要作用与腐蚀相 似, 与腐蚀操作相比, 具有可以基本保持目标原有大小不变的优点。 闭运算: 先膨胀后腐蚀 的过程称为闭运算。用来填充物体内细小空洞、 连接邻近物体、 平滑其边界的同时并不明显改 变其面积。
虽然腐蚀处理可以将粘连的目标物进行分离,膨胀处理可以将断开的目标物进行接续,但 同时都存在一个问题, 就是经过腐蚀处理后, 目标物的面积小于原有面积, 而经过膨胀处理之 后, 目标物的面积大于原有面积。 开、 闭运算就是为了解决这个问题而被提出的。 数学形态 学中二值图像的形态变换是一种针对集合的处理过程。其形态算子的实质是表达物体或形状的 集合与结构元素间的相互作用, 结构元素的形状就决定了这种运算所提取的信号的形状信息。 形态学图像处理是在图像中移动一个结构元素,然后将结构元素与下面的二值图像进行交、并 等集合运算。 对于裂缝形态的准确提取和处理, 采用分形维数的计算方法。 不是所有不规则图形都具有 分形特征, 只有在一定标度范围内满足自相似性的图形才满足分形特征。研究管道裂缝的分形 特征, 关键在于计算结构表面裂缝对应的分形维数值是否满足非欧式空间的要求。分形维数能 够定量和定性地反映管道表面损伤程度。现行求解分形维数的方法有很多: 1) 差分盒子维法; 2) 面积-周长法; 3) 毯子覆盖法; 4) 分形布朗运动自相似模型法; 5) 基于多尺度的分数维 法, 这里采用面积-周长法计算分形维数, 研究管道裂缝的分形规律。
根据分形几何理论, 不规则图形其周长与面积满足关系: Ρ Αυ2, 式中 表示周长, ^表示面积, 表示分形维数, 两边取对数得到式:
\ogP = 0.5D\ogA + C (9) 式中: C为常数。 为了衡量上式的拟合精度, 采用下式计算线性拟合精度。 当?2的值接近 于 0, 意味着拟合精度很差; 当?2的值接近于 1时, 则表示拟合精度很高。
R2 = ("∑ -∑ ∑ )2
("∑ 2— (∑ )2Χ"Σ 2— (Σ ) 2 ) ( )) 首先,将拍摄的清晰、完整和没有干扰的管道裂缝图像通过自适应阈值分割转换成二值图 像, 裂缝被提取出来, 为黑色区域; 然后利用边界跟踪方法计算公式中裂缝面积和周长, 即统 计 8 联通区域内像素个数为该裂缝的面积, 统计边界上像素点的个数为该裂缝的周长; 最后 将面积和周长数值代入到前式中, 利用最小二乘法获得在双对数坐标下面积-周长关系的直线 方程, 可求出裂缝的分形维数£>。 通过选取 10条清晰、 完整的裂缝进行计算。
分析数据, 可得裂缝的面积 -周长关系公式为:
log = 1.0289- log^ -0.4211 ( 11) 其拟合度为: R2 = 0.9957,表明管道表面裂缝直线符合分形规律,具有较高的拟合精度。 为了进行比较, 在图像中选取了不同尺寸、 形状的 10个杂质, 进行了同样的拟合计算, 得到 其面积和周长的关系为:
\ogP = \.024- \ogA - 0.27 ( 12) 其拟合度为: ?2 = 0.9165。
定义以下函数: ^A^^ logi5— 1.0289' log^ + 0.4211 对于管道裂缝而言, 在一定的误差范围内, 应满足下列公式: (Λ^^ ο
而在实际的管道表面图像中,由于多种干扰、污染的存在,上式一般不会自动满足。为此, 提出一种结合骨架提取和分形特征的裂缝检测方法,即对图像分割后获得的裂缝边缘位置进行 优化,使其满足式的要求。本文提出的结合骨架和分形特征的裂缝提取算法流程如附图 7所示。 裂缝与杂质区域的分形维数规律如附图 8。 获取到红外图像后, 需要对图像进行处理得到管道裂缝和管道的各自温度数据,然后再进 行进一步分析。
由红外热图像的灰度信息,利用现有的成熟算法可以得到裂缝区域和管道表面区域,然后 求得各自区域内图像 RGB均值, 最后将平均的 RGB值与 colorbar进行匹配, 从左向右依次进 行, 步进取两个像素宽度以保证温度值精度, 根据匹配到的位置和颜色范围计算温度值。 图像区域特征分析是让计算机具有认识或者识别图像的能力, 即图像识别。特征选择是图 像识别的一个关键问题。 特征选择和提取的基本任务是如何从众多特征中找出最有效的特征。
根据待识别的图像,通过计算产生一组原始特征,称之为特征形成。原始特征的数量很大, 或者说原始样本处于一个高维空间中,通过映射或变换的方法可以将高维空间中的特征描述用 低维空间的特征来描述,这个过程就叫做特征提取。从一组特征中挑选出一些最有效的特征以 达到降低特征空间维数的目的, 这个过程叫做特征选择。
除了颜色特征外, 形状特征也是用于区分图像特征的一个重要方面。凹凸性是区域的基本 特征之一,区域凹凸性可通过以方法进行判别:区域内任意两像素间的连线穿过区域外的像素, 则此区域为凹形。 相反, 连接图像内任意两个像素的线段, 如果不通过这个图形以外的像素, 则这个图形称为凸的。任何一个图形, 把包含它的最小的凸图形叫这个图形的凸闭包。从凸闭 包除去原始图形的部分后, 所产生的图形的位置和形状将成为形状特征分析的重要线索。
基于渗漏点的区域特征, 如圆度, 密集度等参数, 判断某一已知渗漏点的渗漏类型。 当渗 漏区域某一个特征满足如下的条件, 则可以认为该处渗漏区域为该类渗漏。本节主要是考虑了 某渗漏区域周长 /面积和渗漏区域外接矩形的面积 & /该区域的面积 S2两个特征。
若某渗漏区域周长 /面积〉 0. 5, 则认为该渗漏为裂缝类;
如附图 9 ( a) 为极端假设某渗漏区域为正圆形时, 设圆形半径为 r, 裂缝宽度为 r / 5, 则 该渗漏区域外接矩形的面积 该区域的面积 &可表示如下式:
S S2 = 2r - r/ ^r - - = 3.18 ( 13) 如附图 9 (b), 为极端假设某渗漏区域为绝对裂缝, 裂缝宽度为 r / 5, 长度为 ^, 则该渗 漏区域外接矩形的面积 /该区域的面积 &可表示如下式:
Sx /S2 = 7θΊ TO- = \ ( 14) 故若某渗漏区域外接矩形的面积 该区域的面积 S2 >3. 18, 则认为该渗漏为接口松动类 渗漏; 若 1〈外接矩形的面积 该区域的面积 S2〈3. 18, 则需要更细致的判断。 运用边缘提取和轮廓追踪技术, 在红外图像中, 准确提取红外图像中可能的渗漏发生处, 并且可以根据边缘分形优化的方式, 避免大面积杂质或表皮脱落等因素造成的误差。 模型建立
首先, 研究裂缝与渗漏处和环境温差的关系。
研究利用红外热像仪检测管道渗漏裂缝时,在提取出温度差异点后,确定渗漏裂缝与管道 未渗漏处和渗漏处温差的关系模型, 再通过结合机器学习技术建立起管道渗漏的判别模型。
由于管道和容积水的比热容量不同, 当环境温度超过给定值时, 水温和管道温度将有显着 差异。 表 2显示了不同环境温度下泄漏点和未损坏点的确定温度。 表 2 不同环境温度下泄漏点和未损坏点的确定温度
测试组 1 测试组 2 测试组 3
渗漏处 未损坏点 渗漏处 未损坏点 渗漏处 未损坏点
0 24 24 24 24 24 24
2 24.4 24.6 24.2 26.4 23.8 25.6
4 25.2 26.2 24.4 27.9 23.4 26.2
6 24.3 25.5 24.4 28 23.3 25.8
8 25.3 27.1 24.4 28.2 23.2 25.2
10 25.5 27.5 24.4 28 23.1 26.1
12 26.1 27.5 24.4 27.9 23 27 14 25.5 28.1 24.6 28.1 ―
If- 27 28.6 25 28.2 ― ― 根据三次泄漏的温度变化曲线,将各温度下的泄漏温度变化值进行平均,得出泄漏的平均 温度变化曲线, 即为无损区域。将两条曲线的角平分线作为泄漏处温度变化趋势的边界和未损 伤部位的温度, 如附图 10中黑色粗虚线所示。 采用支持向量机对数据进行线性分类。 以环境温度为横坐标、裂缝与管道表面温度差为纵 坐标绘制点, 共有 1、 2、 3三个等级, 数字越大发育程度越严重, 如附图 11所示为分类函数 图, 可以得到如下两个分类函数, 如式 (15) (16)。
3、 2分类函数: Δ¾= :Τ+ 3 (15) 其中 (°C) 为环境温度, (°C) 为管道与裂缝的温度差; α23为线性分类函数系数, 取值范围为 0.02-0.03, b23线性分类函数常数项, 取值范围为 1.80-2.55.
2、 1分类函数: ^u aj+bu (16) 其中 (°C) 为环境温度, (°C) 为管道与裂缝的温度差; 《12为线性分类函数系数, 取值范围为 0.0075-0.0100, b12线性分类函数常数项, 取值范围为 1.2-1.95.
检测结果根据以下判断:
首先根据环境温度计算出 472和八 , 然后将测量出的 与 ΔΓ12、 Δ 进行对比, Δ ≤Δ 2则发育程度为 1; Δ 12≤Δ ≤Δ 23则发育程度为 2, Δ ≥Δ 23的发育程度为 3.
现在在其它管道上进行数据采集,验证上述检测模型的精度。由上文分析,在一般情况下, 管道表面与裂缝的温差主要与温度有关系,并且裂缝与管道表面的温度差和裂缝的发育程度是 相关的,因此可以由红外热像仪检测裂缝与管道表面的温度差,再利用上文中分类函数 和 2, 进行裂缝发育程度的检测。
首先确定检测环境满足要求, 即阴暗潮湿的地下管廊空间、管廊空间内环境温度均匀升高 4摄氏度以上。 然后利用红外热像仪进行数据采集, 并记录下采集每张图片时的环境气温, 处 理后得到裂缝与管道表面的温差, 根据温度计算发育程度分级阈值八72和八 。
在目前大多裂缝检测都只关注数量的背景下,可以根据检测模型结果赋予不同的权值给不 同发育程度的裂缝, 为管道养护提供更精准的参考, 提高社会效益。 区域 值计算
管道结构是用于运输液体的带状结构物, 环境因素和荷载的作用是造成管道结构破坏的主 要原因, 管道区域物理特性比较一致, 因此在红外图像中, 管道区域的温度值基本一样, 因此 处理起来比较方便, 不需要进行分别考虑, 即在图像分割中得到管道区域后就返回到红外图像 中计算该区域整体的 RGB平均值。
但是, 裂缝区域通常为细长的形状, 譬如一条两米长的裂缝, 可能只有几个毫米宽, 这种 情况下, 同一裂缝不同位置处的严重程度可能会有较大的差异,两米长的裂缝有半米长的区域 发育程度比较严重,根据平均 RGB值计算出来可能需要修补,但是对于另外一条三分米长的裂 缝, 其也有半米长的区域发育程度比较严重, 由平均 RGB值计算出来的温差得到的发育程度指 数就会比之前一条要低, 甚至显示为不修补, 显然是不科学的, 因此, 对于裂缝区域需要分区 域进行 RGB值得计算, 并对发育程度较严重的区域给予较高的权重,这样计算出来一条裂缝区 域与管道的温差就会是一个数组, 如附图 12所示, 不同区段裂缝可以得到不同温差值。
对于横向裂缝,利用图像沿 y轴方向直线将有效裂缝区域分成所述 P段,每一段长度任意; 对于纵向裂缝, 利用图像沿 X轴方向直线将有效裂缝区域分成所述 p段, 每一段长度任意; 对 于其他种类裂缝, 不分段或者根据裂缝几何中心按角度分成所述 P段,每一段对应中心角度任 思。
根据精度要求, 可以将图像分割出来的裂缝区域分成 p(p≥ 2)段来进行考虑, 每一小段均 可以按照前文所述的技术路线进行处理, 即首先进行图像灰度化、 降噪, 然后进行图像增强后 再进行图像分割, 得到裂缝区域后再分成 w段进行后续处理, 包括对每一段进行平均 RGB值得 计算, 再由平均 RGB值与图例中的颜色值进行匹配, 确定该段的温度值, 最后可以得到 p段裂 缝区域的长度 /和裂缝区域与管道温度的温度差 Δ , 即得到如下数组, 如式 (17 ) 所示: ί υ 2 ··· ίρ) ( 17)
其中 /,即为裂缝区域划分区段后第 段的长度, 计算温度差数组如式 (18 ) 所示。
ΔΓ = (Δ , ΔΓ2… ΔΓρ) ( 18) 其中 Δ 即为裂缝区域划分区段后第 /段与管道的温度差。
然后可以根据下式(19 )进行加权平均计算, 得到所述有效管道区域和所述有效裂缝区域 的最终实测温差 Δ .
ΔΤ = +〜 + I ( 十…+ ^ ) ( 19)
最后将此 与参照温差数据进行对比, 得到裂缝发育程度相关参数。
或者由 数组得到各段裂缝区域的发育程度指数如下式 (20 ):
m = (m1, m2■■· mp) (20) 其中 即为裂缝区域划分区段后第 /段的裂缝发育程度指数。
根据裂缝的发育程度赋予不同区段的裂缝权重,因此可以得到裂缝的最终发育程度指数如 下式 (21 ):
m' = (m^ + m2Z2 +… + mnZp ) I ( ^ +…+ ^ ) (21 ) 计算结果四舍五入取 1、 2或 3, 即裂缝发育程度的三个等级。
发明中可以采取平均分配的方法, 那么就有 = 2 =— = ^, 公式就可以化成下列形式 ( 22):
m' = (7 + m2 +— h mp)/n (22) 数据处理 红外图像数据可能会有失真度, 即泄漏点和未损坏点的确定温度超出合理范围。考虑到渗 漏发生具有不可逆性、连续性以及趋势性, 即当前一时段有可疑渗漏裂缝反馈时, 后一时段的 数据必须也有裂缝反馈, 且渗漏处和未渗漏处温度差异只能不变或缓慢变大,经过边缘提取后 也可以体现在渗漏裂缝尺寸参数不变或缓慢变大, 故当数据违背该规律时, 认为失真。此时处 理方式有两种:
( 1 ) 直接抛弃失真数据, 将前一时段的数据(渗漏裂缝的尺寸参数)迁移到该时段。 例 如, 若以 1小时为检测周期, 当前一时段内 (前 lh) 红外图像上存在温度差异, 则将改组数据原样迁移到该失真数据时段 (该 lh)。
( 2 ) 等待后续数据再对该失真数据时段进行修正。 若某一时段数据 (时段 B)相比于前 一时段数据 (时段 A)失真, 等待随后的一个时段的数据 (时段 C), 若时段 C数据 与时段 A数据符合不可逆性、 连续性以及趋势性, 则将时段 A数据迁移到时段 B 上, 若时段 C数据与时段 A数据不符合上述三条特性, 则认为时段 A数据有误, 予 以剔除, 同时时段 A的结果迁移给时段 B。 此时, 时段 A之前的时段的数据不予考 虑修正。
在模型建立过程中, 发明通过机器学***均发育程度指数的计算中会出现 有小数的裂缝发育程度指数, 因为本身的分级就存在一定误差,这些有小数的发育程度指数虽 然不一定可以使裂缝发育程度指数更精确, 但是却可以反映出相对大小, 2. 6的发育程度指数 四舍五入被划分为发育程度指数为 3的裂缝, 2. 4的就被划分为发育程度指数为 2的裂缝, 这 是之前的分类方法,但是由赋予权重计算出来的 2. 6发育程度指数和 2. 4发育程度指数实际上 是差别不大的,这对于管理者的决策有相当重要的影响。因此,在裂缝发育程度指数的计算中, 发明可以取一位小数来表明计算出来的相对关系。但是, 需要注意的是 2. 4的发育程度指数不 一定比 2. 3的发育程度指数要更严重, 因为前文说过, 本身分类就存在误差, 2这个整数部分 可能就已经是不准确的了。 因此, 为了避免这种问题, 发明在计算过程中根据加权平均计算方 法的方差确定了一个误差项, 即发育程度指数可以取一位小数并表示为: m ± 的形式。 采用支持向量机对数据进行线性分类。 以环境温度为横坐标、裂缝与管道表面温度差为纵 坐标绘制点,共有 1、 2、 3三个等级,数字越大发育程度越严重,可以得到如下两个分类函数。 3、 2分类函数: Δ¾ = :Τ+ 3 (23)
其中 ( °C )为环境温度, ( °C )为管道表面与裂缝的温度差; α23为线性分类函数系 数, 取值范围为 0. 02-0. 03, b23线性分类函数常数项, 取值范围为 1. 80-2. 55.
2、 1分类函数: ^u aj+bu (24)
其中 ( °C )为环境温度, ( °C )为管道表面与裂缝的温度差; 《12为线性分类函数系 数, 取值范围为 0. 0075-0. 0100, b12线性分类函数常数项, 取值范围为 1. 2-1. 95.
检测结果根据以下判断
首先根据环境温度计算出 4 和 Δ , 然后将测量出的 与 ΔΓ12、 ΑΤ13进行对比, Δ = Δ 2则发育程度为 1 ; ΔΓ = ΔΓ23则发育程度为 2, 介于之间的采用线性插值法计算发育程 度指数, 指数大于 3的取 3.
若要对裂缝区域进行区段划分, 并根据裂缝的发育程度赋予不同的权重, 那么由公式计算出 来的结果可以取一位小数。
误差项 的取值可以由以下 (25 ) 得到:
Figure imgf000019_0001
m' 为对同一裂缝进行多次检测的结果, 然后由以上公式得出标准差, 据此可以将裂缝的 发育程度指数写到小数点后一位, 并通过一个固定的误差项来进行误差限定, 固定的误差项由 实验得到, 新的发育程度指数写法如下式 (26 ) :
τη = ιη ± ασ (26) 其中 "为误差项系数, 取值范围为 1-3. 采用基于红外热像图的管道裂缝发育程度检测方法进行检测***的构建,实际的检测*** 需要利用红外双目摄像机进行管道图像的采集。双目摄像机包括一个普通相机和一个红外热像 仪, 同步进行管道的图像采集, 并进行处理进行裂缝的有无、 位置、 大小和发育程度的识别, 两种图像数据融合, 可以提高发明提出的检测方法的鲁棒性。
裂缝检测***采用模块化设计, 可以很好的与其它管道病害检测模块进行结合。裂缝检测 模块包括摄像机固定装置、 双目摄像机、 数据传输线、 车载终端、 GPS 接受装置、 惯性导航。 摄像机固定装置可以采用定制铁架,保证可以在工程车上可靠固定双目摄像机; 可以采用巡检 车云台, 固定摄像机的同时可以进行双目摄像机视角的控制, 云台至少拥有两个自由度, 即可 以分别在水平方向和竖直方向旋转, 更多自由度将带来云台的更高操作性。红外摄像头的最低 配置为: 分辨率为 320 X 240, 带可更换的镜头, 免维护的非制冷微测辐射热计, 显微技术 和特写镜头测量功能, 数据传输速度高达 60 Hz。 普通图像摄像头支持 1080P高清图像实时输 出。数据传输线包括至少两根,分别支持高清普通图像和红外图像的高频传输,支持至少 100Hz 的 1080p视频传输。 巡检车载终端包括两种方案: 一种是前端处理, 采用嵌入式 PC进行实时 视频流的处理, 通过 3g\4g网络进行处理后的数据传输, 嵌入式 PC可以兼容普通图像和红外 图像的数据接口, 处理器支持视频流的实时处理; 另外一种是后端处理, 巡检车载终端作为数 据采集的前端, 只负责数据采集和存储, 配置和开发要求更低, 需要有接口供采集后第三方设 备接入进行数据处理。 此种模块通过视频图像拼接技术, 还原整个采集过程的管道纵向图像, 然后再沿长度方向进行切割后分别识别,保证管道裂缝检测的准确性和稳定性。此***思路图 像处理算法要求高, 设备相对要求更少。 GPS接受装置和惯性导航, 两者共同为采集设备进行 定位, 确保精度和实时性。 定位的精度需要在 10m内。
另外一种裂缝检测模块包括摄像机固定装置、 双目摄像机、 数据传输线、 巡检车载终端、 GPS接受装置、 惯性导航、 光电编码器和同步控制器(参照 CN104749187)。 光电编码器安装在 巡检车载移动平台的车轮中心轴上, 用以测量巡检车载移动平台的运行速度和距离; 所述 GPS 接收机, 安装在巡检车载移动平台上, 用于所述巡检车载移动平台的高精度定位及授时; 所述 惯性导航, 安装在巡检车载移动那平台上, 用于在隧道内 GPS接受机接受不到 GPS信号的情况 下, 测量巡检车载移动平台的位置、 姿态数据, 实现在隧道内部高精度的位置推算; 同步控制 器, 安装在巡检车载移动平台上, 用于同步普通摄像机和红外摄像机的图像采集时间, 保证两 者具有统一的时间和空间基准。此种方法通过光电编码器精确测量车速, 同步控制器根据车速 和双目摄像机的视野大小, 自动控制双目图像采集时间,确保相邻两张有效剪辑的管道图像具 有很好的连续性, 可以完整的覆盖所采集的管道, 并且相互不重叠。 图像传输到巡检车载终端 后执行前端处理或者存储功能, 此***思路对于设备要求较高, 图像处理算法要求较低。 附图筒要说明 图 1为红外热像图识别管廊内部空间渗漏的技术路线图。
图 2为管廊内部空间和裂缝的灰度信息和温度信息。
图 3为照度与温差关系图。
图 4为环境温度升高与温差关系图。
图 5为图像阈值分割模型。
图 6为管道表面裂缝区域、 管道区域和过渡干扰区。
图 7为结合骨架和分形特征的裂缝提取算法流程图。
图 8为裂缝与杂质区域的分形维数规律。
图 9为接口处松动类渗漏判断临界值。
图 10环境升温后渗漏处与未损坏处温度变化图。
图 11为支持向量机线性分类图示。
图 12为不同裂缝区段获得不同温差示意图。 具体实施方式 ( 1 ) 环境确定
检测模型需要在一定环境条件下使用才能保证精确性, 首先需要保证数据采集时环境条件 满足: 阴暗潮湿的地下管廊空间、 管廊空间内环境温度均匀升高 4摄氏度以上。 需要保证数据 采集时环境条件稳定, 即管廊内部环境温度均匀, 且升高一定温度后, 不会发生剧烈变化。
( 2) 图像采集
采用测温型红外热像仪对管廊内部空间进行拍摄, 分析得到裂缝区域红外热像图。 热像仪 拍摄距离管道水平距离为 1 m, 机位离裂缝区域保持距离不变匀速连续拍摄。 红外热像仪是利 用红外探测器和光学成像物镜接受被测目标的红外辐射能量分布图形反映到红外探测器的光 敏元件上, 从而获得红外热像图, 这种热像图与物体表面的热分布场相对应。通俗地讲红外热 像仪就是将物体发出的不可见红外能量转变为可见的热图像。热图像的上面的不同颜色代表被 测物体的不同温度。 发明现在主流红外成像装置 非制冷焦平面微热型红外热像仪。 同时, 需要对采集裂缝图像时的气温和裂缝发育程度进行记录, 环境温度直接使用温度计进行测量, 裂缝的发育程度根据裂缝宽度进行人工衡量。
同时, 还需要对裂缝的真实发育程度指数进行打分, 主要参考传统分级方式, 即将裂缝分 为轻、 中、 重三个等级的分类方式, 再考虑裂缝的湿度和深度等因素, 请专家对裂缝发育程度 指数进行打分, 据此实际数据建立检测模型。
( 3) 图像分析
首先进行图像预处理,对红外图像进行灰度化处理; 对图像使用小波降噪和不同矩形的中 值滤波处理, 在滤除噪声的同时, 尽量不模糊边缘; 采用图像灰度增强算法, 增强裂缝与背景 区域的对比度, 便于裂缝的提取; 采用可调阈值的图像分割算法分割图像, 灰度值较低的裂缝 区域和杂质转换为黑色, 灰度值较高的背景转换为白色,对于杂质等非裂缝区域可采用面积阈 值和面积 -周长分形规律进行剔除, 只保留裂缝区域。 然后得到裂缝区域和管道表面非裂缝区 域; 再根据两种区域位置在最初的红外图像中进行区域定位; 最后计算红外图像裂缝区域和管 道表面区域的 RGB平均值, 将 RGB平均值依次与 colorbar图例的 RGB值进行匹配, 最相符的 位置所代表的的温度即为该区域的温度,将裂缝区域和管道表面区域分别与图例匹配得到各自 的温度, 进而得到裂缝与管道表面的温度差异。
图像处理中的图像分割技术,可以识别出裂缝区域后直接对裂缝区域和管道表面区域进行 温度识别处理,也可以根据前文所述来对裂缝与管道表面直接的过渡干扰区进行分析后做排除 处理。
而最终温差数据的获得可以直接计算整条裂缝区域与管道表面区域的温差值,也可以将裂 缝区域进行区段划分, 可以采取平均分配的方法,然后对于发育程度指数较高的区段给予较高 的权重, 按照给予不同权重的方法来计算最终整条裂缝区域与管道表面区域的温差值。 考虑到不同管廊环境的条件差异较大,可以尝试建立本区域管廊环境的裂缝温差与环境温 度条件关系模型。 根据历史数据对当前数据进行修正和舍弃; 利用图像分析得到的温度差异, 以及采集图像同时利用其他仪器采集到的温度信息, 以及光纤测温采集到的温差数据,矫正关 系模型。经过前三个步骤后, 我们具有了裂缝与路面温度差异数据、裂缝发育程度的数据和采 集每个样本时的环境温度数据,利用图像分析得到的温度差异, 以及采集图像同时利用其他仪 器采集到的温度信息, 建立与裂缝发育程度的关系模型, 即主要确定线性分类函数 /12和/23的 相关系数的具体数值。
( 5 ) 检测验证
利用以上检测模型可以进行实际裂缝发育程度的检测, 需要采集有裂缝路面的红外图像和 当下的环境温度。
检测最后计算裂缝的发育程度指标时, 可以直接粗略的将裂缝发育程度指标分为 1、 2、 3 三个等级, 只取三个数中的某个即可, 数值越大代表发育程度越严重; 计算也可以采用前文提 出的有一位小数的发育程度指数, 取值范围为 0-3, 具体的发育程度指标可以表示为 士 Ωσ的 形式。
以上所述, 仅为本发明典型的实施方式, 但本发明的保护范围并不局限于此, 任何熟悉本 技术领域的技术人员在本发明揭露的技术范围内, 可轻易想到的变化或替换, 都应涵盖在本发 明的保护范围之内。 因此, 本发明的保护范围应该以权力要求的保护范围为准。

Claims

杈利要求书
1. 一种基于静态红外热像图处理的地下管廊渗漏检测方法, 包括如下步骤: a) 建立裂缝发育程度检测模型, 具体包含:
al) 釆集管道表面至少 10个样本裂缝区的红外热像图, 同时记录样本环境 温度及所述样本裂缝区内的裂缝发育程度, 根据传统的裂缝发育程度分类方法, 对裂缝发育程度进行估计,将所釆集裂缝的裂缝发育程度分为 1、2、3三个等级; a2) 对步骤 al)中所述的裂缝区的红外热像图分别进行处理, 得到对应的裂 缝区内的裂缝和裂缝区内的管道表面的温差数据;
a3) 利用步骤 a2)中所得到的温差数据, 以及步骤 al)中所记录的样本环境 温度和裂缝区内的样本裂缝发育程度进行支持向量机分类, 得到两个分类函数 ΔΓ12 = ¾2 + 612和 7^ = a23T + b23 , 所述的分类函数自变量为所记录的样本 环境温度, 因变量为参照温差数据;
b) 数据釆集: 釆集管道裂缝区的红外热像图, 同时记录环境温度; c) 数据预处理: 将原始红外热像图进行降噪和灰度化操作;
d) 图像数据处理: 对经过降噪和灰度化操作的红外热像图进行数据处理, 包 括边缘检测及增强、 边缘提取、 杂质剔除和计算实测温差数据;
e) 温差数据处理: 包括数据合理性判定、实测温差数据一次修正和实测温差 数据二次修正;
f) 渗漏分析:包括计算参照温差数据、 确定渗漏严重程度和确定渗漏类型。
2. 如权利要求 1所述的基于静态红外热像图处理的地下管廊渗漏检测方法, 其 特征在于, 步骤 b)包括:
bl) 固定设备釆集: 用非制冷焦平面测温型红外热像仪设备在地下综合管廊 内,设备距离管道表面的高度为 0. 5-lm, 釆集管廊内部环境及管道表面状况, 得 到红外热像图;
b2) 记录环境温度: 利用温度计在所述的管道裂缝区同时记录环境温度。
3. 如权利要求 1所述的基于静态红外热像图处理的地下管廊渗漏检测方法, 其 特征在于, 步骤 c)包括:
cl) 降噪: 将原始红外热像图, 进行降噪处理, 去除其在数字化和传输过程 中受到成像设备与外部环境噪声干扰等的影响, 得到降噪后的红外热像图; c2) 灰度化: 将降噪后的红外热像图中的色彩信息去掉, 转换为灰度图像, 得到灰度化红外热像图。
4. 如权利要求 3所述的基于静态红外热像图处理的地下管廊渗漏检测方法, 其 特征在于, 步骤 cl)所述的降噪方法至少包括以下五种之一: 均值滤波器、 自适 应维纳滤波器、 中值滤波器、 形态学噪声滤除器和小波去噪。
5. 如权利要求 1至 4之一所述的基于静态红外热像图处理的地下管廊渗漏检测 方法, 其特征在于, 步骤 d)包括:
dl) 边缘检测及增强: 用搜索或零交叉对图像属性的显著变化进行识别, 得 到温度差异点;
d2) 边缘提取: 对所述温度差异点, 运用单值或多值阈值分割, 得到裂缝区 域和管道表面区域;
d3) 杂质剔除: 分别对所述裂缝区域和管道表面区域进行杂质干扰剔除, 得 到有效裂缝区域和有效管道表面区域;
d4) 计算实测温差数据: 处理步骤 d3)中的有效裂缝区域和有效管道表面区 域的红外热像图,得到对应的裂缝区内裂缝和裂缝区内的管道表面之间的实测温 差数据 ΔΤ。
6. 如权利要求 5所述的基于静态红外热像图处理的地下管廊渗漏检测方法, 其 特征在于, 步骤 dl)所述的边缘检测及增强, 釆用的边缘检测模板至少包括以下 五种算子之一: Laplacian算子、 Roberts算子、 Sobel算子、 log (Laplacian- Gauss)算子、 Kirsch算子禾口 Prewitt算子。
7. 如权利要求 5所述的基于静态红外热像图处理的地下管廊渗漏检测方法, 其 特征在于, 步骤 d4)包括:
d41)对于所述有效管道表面区域进行图例匹配, 得到实测温度 Γ。; d42) 将所述有效裂缝区域划分为 p段, p 2, 每段长度为 Zf Zp, 分别对每 一段有效裂缝区域进行图例匹配, 得到实测温度 7ν·· Γρ
d43) 将所述有效裂缝区域各段实测温度减去所述有效管道表面实测温度, 得到实测温差 Δ7^·· ΔΓρ ;
d44) 根据下式进行加权平均计算, 得到所述有效管道表面区域和所述有效 裂缝区域的最终实测温差数据 ΔΓ,
ΔΓ = + 〜 + I +…十/" ) 。
8. 如权利要求 1至 4之一所述的基于静态红外热像图处理的地下管廊渗漏检测 方法, 其特征在于, 步骤 e)包括:
el) 数据合理性判定:根据渗漏发生的不可逆性、 连续性以及趋势性三点特 性, 依据历史实测温差数据对实测温差数据进行合理性判定;
e2) 实测温差数据一次修正:根据历史实测温差数据及渗漏发生的三点特性, 对具有失真度的实测温差数据进行取舍和修正,得到一次修正后的实测温差数据; e3) 实测温差数据二次修正:根据光纤测温得到的温差数据, 对一次修正后 的实测温差数据进行取舍和修正, 如有光纤测温数据, 则用该光纤测温数据替换 原实测温差数据,得到二次修正后的实测温差数据。
9. 如权利要求 1至 4之一所述的基于静态红外热像图处理的地下管廊渗漏检测 方法, 其特征在于, 步骤 f)包括:
fl) 计算参照温差数据: 将步骤 b2)所记录的环境温度数据带入步骤 a)中所 述的裂缝发育程度检测模型, 得到参照温差数据; 将所述参照温差数据及二次修 正后的实测温差数据带入步骤 a)所述检测模型, 推算裂缝发育程度;
f2) 确定渗漏严重程度: 从长度、 宽度和面积三个维度, 对渗漏严重程度进 行定义, 得到 G= [0_9. 9],G为渗漏严重程度指数, 结果保留 1位小数;
f3) 确定渗漏类型: 釆用区域特征分析的手段, 考虑圆度, 密集度参数, 对 裂缝区的特征进行判别, 判断所述管道裂缝区渗漏点的渗漏类型。
10. 如权利要求 9所述的基于静态红外热像图处理的地下管廊渗漏检测方法, 其特征在于, 步骤 f3)所述的渗漏类型判别方法包括:
f31) 若某渗漏区域周长 /面积 >0. 5, 则认为该渗漏为裂缝类;
f32) 若某渗漏区域周长 /面积≤0. 5,且 > 3.18,则认为该渗漏为接口松 动类渗漏;
f33) 若某渗漏区域周长 /面积≤0. 5,且 1 ≤ 3.18, 则认为该渗漏为 其他类渗漏;
上述步骤 f32)和 f33)中, 为某渗漏区域外接矩形的面积, 为该区域的面 积。
PCT/IB2017/058539 2016-12-30 2017-12-30 一种基于静态红外热像图处理的地下管廊渗漏检测方法 WO2018122809A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
GB1905909.6A GB2569751B (en) 2016-12-30 2017-12-30 A method for leakage detection of underground corridor based on static infrared thermal image processing
CN201780053170.8A CN110268190B (zh) 2016-12-30 2017-12-30 一种基于静态红外热像图处理的地下管廊渗漏检测方法

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
IBPCT/IB2016/058109 2016-12-30
PCT/IB2016/058109 WO2018122589A1 (zh) 2016-12-30 2016-12-30 一种基于红外热像图分析的沥青路面裂缝发育程度检测方法

Publications (1)

Publication Number Publication Date
WO2018122809A1 true WO2018122809A1 (zh) 2018-07-05

Family

ID=59713564

Family Applications (5)

Application Number Title Priority Date Filing Date
PCT/IB2016/058109 WO2018122589A1 (zh) 2016-12-30 2016-12-30 一种基于红外热像图分析的沥青路面裂缝发育程度检测方法
PCT/IB2017/058548 WO2018122818A1 (zh) 2016-12-30 2017-12-30 一种基于红外热像图分析的沥青路面裂缝发育程度检测方法
PCT/IB2017/058549 WO2018122819A1 (zh) 2016-12-30 2017-12-30 一种基于双目图像分析的沥青路面病害检测***
PCT/IB2017/058540 WO2018122810A1 (zh) 2016-12-30 2017-12-30 一种基于动态红外热像图处理的地下管廊渗漏检测方法
PCT/IB2017/058539 WO2018122809A1 (zh) 2016-12-30 2017-12-30 一种基于静态红外热像图处理的地下管廊渗漏检测方法

Family Applications Before (4)

Application Number Title Priority Date Filing Date
PCT/IB2016/058109 WO2018122589A1 (zh) 2016-12-30 2016-12-30 一种基于红外热像图分析的沥青路面裂缝发育程度检测方法
PCT/IB2017/058548 WO2018122818A1 (zh) 2016-12-30 2017-12-30 一种基于红外热像图分析的沥青路面裂缝发育程度检测方法
PCT/IB2017/058549 WO2018122819A1 (zh) 2016-12-30 2017-12-30 一种基于双目图像分析的沥青路面病害检测***
PCT/IB2017/058540 WO2018122810A1 (zh) 2016-12-30 2017-12-30 一种基于动态红外热像图处理的地下管廊渗漏检测方法

Country Status (4)

Country Link
US (2) US11486548B2 (zh)
CN (5) CN109804119B (zh)
GB (6) GB201711412D0 (zh)
WO (5) WO2018122589A1 (zh)

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109816656A (zh) * 2019-02-01 2019-05-28 辽宁工程技术大学 一种火电厂负压侧***漏点精确定位方法
CN109859125A (zh) * 2019-01-14 2019-06-07 广东工业大学 基于形态学检测与小波变换的图像高光修复方法
WO2020015354A1 (zh) * 2018-07-17 2020-01-23 北京讯腾智慧科技股份有限公司 一种基于北斗差分定位的燃气管道的修补方法及***
CN111242123A (zh) * 2020-01-07 2020-06-05 西安交通大学 一种基于红外图像的电力设备故障诊断方法
CN112927223A (zh) * 2021-03-29 2021-06-08 南通大学 一种基于红外热成像仪的玻璃幕墙检测方法
CN113777028A (zh) * 2021-11-11 2021-12-10 成都理工大学 测量凝胶类堵漏材料与岩石壁面粘附强度的装置和方法
CN114034405A (zh) * 2021-11-08 2022-02-11 北京航空航天大学 一种非接触式测温方法及***
CN114511469A (zh) * 2022-04-06 2022-05-17 江苏游隼微电子有限公司 一种图像智能降噪先验检测方法
CN116823839A (zh) * 2023-08-31 2023-09-29 梁山中维热力有限公司 基于热红外图像的管道泄漏检测方法
CN117392140A (zh) * 2023-12-13 2024-01-12 宏发建设有限公司 一种基于图像处理的建筑幕墙玻璃破裂检测方法及***
CN118154594A (zh) * 2024-05-10 2024-06-07 山东力乐包装股份有限公司 一种基于电数字数据处理的木质托盘缺陷分析***

Families Citing this family (153)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10970590B2 (en) * 2015-06-05 2021-04-06 Schlumberger Technology Corporation Image-based wellsite equipment health monitoring
US10551297B2 (en) * 2017-09-22 2020-02-04 Saudi Arabian Oil Company Thermography image processing with neural networks to identify corrosion under insulation (CUI)
CN107767377B (zh) * 2017-11-09 2024-02-06 高视科技(苏州)股份有限公司 基于双目视觉***的液晶屏缺陷与灰尘区分法及检测装置
CN108364280B (zh) * 2018-01-03 2022-04-15 东南大学 结构裂缝自动化描绘及宽度精准测量方法与设备
US10643324B2 (en) 2018-08-30 2020-05-05 Saudi Arabian Oil Company Machine learning system and data fusion for optimization of deployment conditions for detection of corrosion under insulation
US10533937B1 (en) 2018-08-30 2020-01-14 Saudi Arabian Oil Company Cloud-based machine learning system and data fusion for the prediction and detection of corrosion under insulation
JP7056804B2 (ja) * 2018-09-28 2022-04-19 日本電気株式会社 経験損失推定システム、経験損失推定方法および経験損失推定プログラム
CN109488888B (zh) * 2018-11-06 2020-07-17 沈阳天眼智云信息科技有限公司 基于对红外温度场多元分析的金属管道泄漏监测方法
CN109358060B (zh) * 2018-11-14 2023-08-25 浙江工业大学 一种用于检测并标记道路地砖松动情况的***及方法
US10878596B2 (en) 2018-11-29 2020-12-29 International Business Machines Corporation Object oriented image normalization
US11143599B2 (en) 2018-12-03 2021-10-12 Mistras Group, Inc. Systems and methods for inspecting pipelines using a pipeline inspection robot
US10783623B2 (en) 2018-12-03 2020-09-22 Mistras Group, Inc. Systems and methods for inspecting pipelines using a robotic imaging system
CN109632963B (zh) * 2019-01-11 2020-10-20 南京航空航天大学 一种基于时不变特征信号构建的结构损伤四维成像方法
CN109801282A (zh) * 2019-01-24 2019-05-24 湖北大学 路面状况检测方法、处理方法、装置及***
CN109800739B (zh) * 2019-02-21 2023-04-25 四川中天炬矿业有限公司 电加热旋转窑温度检测装置
CN109858570A (zh) * 2019-03-08 2019-06-07 京东方科技集团股份有限公司 图像分类方法及***、计算机设备及介质
CN109919139B (zh) * 2019-04-01 2021-02-09 杭州晶一智能科技有限公司 基于双目立体视觉的路面状况快速检测方法
CN109903325B (zh) * 2019-04-03 2021-05-11 杭州晶一智能科技有限公司 基于立体视觉深度信息的地面精确描述方法
CN110232666B (zh) * 2019-06-17 2020-04-28 中国矿业大学(北京) 基于暗原色先验的地下管道图像快速去雾方法
CN110349134B (zh) * 2019-06-27 2022-12-09 广东技术师范大学天河学院 一种基于多标签卷积神经网络的管道病害图像分类方法
CN110205910A (zh) * 2019-07-08 2019-09-06 河北工程大学 一种铺面结构表面平整性智能检测与分析装置
CN110320236B (zh) * 2019-07-19 2021-09-14 沈阳工业大学 大型风力机叶片内部缺陷深度的红外测量方法
CN110490862B (zh) * 2019-08-22 2022-08-09 联峰钢铁(张家港)有限公司 一种提高连铸探伤合格率的方法及装置
JP7285174B2 (ja) * 2019-09-04 2023-06-01 株式会社トプコン 壁面のひび割れ測定機および測定方法
CN110674754A (zh) * 2019-09-25 2020-01-10 武汉易视维科技有限公司 一种在线间歇中空滤棒视觉缺陷检测识别***
CN111025285B (zh) * 2019-11-01 2021-08-24 长安大学 一种基于图谱灰度自适应选取的沥青路面水损害检测方法
CN110793722B (zh) * 2019-11-08 2021-11-30 国家计算机网络与信息安全管理中心 基于机器学习的铅酸蓄电池非接触式漏液检测装置及方法
CN111062959B (zh) * 2019-11-28 2022-04-12 重庆大学 一种航空薄壁微构零件切出底边毛刺特征的提取及表征方法
CN111209794A (zh) * 2019-12-11 2020-05-29 浙江省交通运输科学研究院 一种基于探地雷达图像的地下管道识别方法
CN111191540A (zh) * 2019-12-20 2020-05-22 数海信息技术有限公司 一种基于温度梯度的对象状态分析方法及***
US11651278B2 (en) * 2019-12-23 2023-05-16 Saudi Arabian Oil Company Pipeline sensor integration for product mapping
CN111145205B (zh) * 2019-12-24 2022-04-12 天津农学院 基于红外图像的多猪只场景下猪体温检测方法
CN111210405A (zh) * 2019-12-27 2020-05-29 山东省特种设备检验研究院有限公司 一种基于红外成像的挥发性有机物泄露检测方法
US11774990B2 (en) 2019-12-30 2023-10-03 Marathon Petroleum Company Lp Methods and systems for inline mixing of hydrocarbon liquids based on density or gravity
CA3104319C (en) 2019-12-30 2023-01-24 Marathon Petroleum Company Lp Methods and systems for spillback control of in-line mixing of hydrocarbon liquids
US11607654B2 (en) 2019-12-30 2023-03-21 Marathon Petroleum Company Lp Methods and systems for in-line mixing of hydrocarbon liquids
CN111209876B (zh) * 2020-01-10 2023-04-07 汕头大学 一种漏油缺陷检测方法及***
US11054538B1 (en) * 2020-01-27 2021-07-06 S-L Ip, Llc Modification and assessment
CN111400891B (zh) * 2020-03-11 2023-08-11 中煤航测遥感集团有限公司 管道巡检点偏差程度获取方法、装置、设备及存储介质
CN111583413B (zh) * 2020-03-24 2023-06-02 交通运输部科学研究院 一种路面病害bim参数化建模与增强现实移动巡检方法
CN111445462A (zh) * 2020-03-30 2020-07-24 国家计算机网络与信息安全管理中心 一种基于神经网络和热像图的蓄电池漏液检测方法
CN111524146B (zh) * 2020-04-13 2023-05-05 同济大学 基于红外图像的暖通空调热水管道故障诊断方法及***
CN111489352B (zh) * 2020-04-29 2023-06-02 安徽国钜工程机械科技有限公司 一种基于数字图像处理的隧道缝隙检测与测量方法及装置
CN111583362A (zh) * 2020-04-30 2020-08-25 东南大学 一种沥青路面病害状况检测的可视化记录方法
US11645766B2 (en) 2020-05-04 2023-05-09 International Business Machines Corporation Dynamic sampling for object recognition
CN111754455B (zh) * 2020-05-15 2024-03-19 华能国际电力股份有限公司海门电厂 一种火电厂漏水检测方法和***
CN111640117B (zh) * 2020-06-03 2024-03-05 四川正大新材料科技有限公司 一种寻找建筑物渗漏源位置的方法
CN111738990A (zh) * 2020-06-03 2020-10-02 东北林业大学 基于log算法的损伤水果温度场检测方法
CN111784645B (zh) * 2020-06-15 2023-08-22 北京科技大学 一种充填管道裂纹检测方法
CN111767815A (zh) * 2020-06-22 2020-10-13 浙江省机电设计研究院有限公司 一种隧道渗漏水识别方法
CN111746537B (zh) * 2020-06-22 2022-05-17 重庆长安汽车股份有限公司 基于路面平整度的自适应巡航车速控制***、方法及车辆
CN111693536A (zh) * 2020-06-24 2020-09-22 河南高建工程管理有限公司 一种基于红外热成像的地下综合管廊施工防水检测方法
CN111750283A (zh) * 2020-06-26 2020-10-09 西北工业大学 基于深度学习的强背景噪声环境下的气体管道泄漏识别方法
CN111795883B (zh) * 2020-07-06 2023-04-25 中国石油天然气集团有限公司 一种含裂纹金相试样的浸蚀及图像处理方法和设备
CN111767874B (zh) * 2020-07-06 2024-02-13 中兴飞流信息科技有限公司 一种基于深度学习的路面病害检测方法
CN112036073A (zh) * 2020-07-16 2020-12-04 成都飞机工业(集团)有限责任公司 一种3d打印零件测量结果矫正方法
CN111812215B (zh) * 2020-07-22 2021-06-29 南京航空航天大学 一种飞行器结构损伤的监测方法
CN114002222B (zh) * 2020-07-28 2024-05-14 宝山钢铁股份有限公司 一种用于在役管线的智能探伤装置、***及其方法
CN111811933B (zh) * 2020-07-31 2022-03-11 中国矿业大学 一种承载煤岩损伤破裂过程中的红外辐射信息去噪方法
CN112032568A (zh) * 2020-08-14 2020-12-04 常州机电职业技术学院 一种埋地燃气管道泄漏危险度预测算法及预测方法
CN112032567A (zh) * 2020-08-14 2020-12-04 常州机电职业技术学院 一种埋地燃气管道泄漏危险度预测***
CN112004093B (zh) * 2020-09-02 2022-07-12 烟台艾睿光电科技有限公司 一种红外数据压缩方法、装置及设备
CN111999001B (zh) * 2020-09-09 2022-07-08 中国南方电网有限责任公司超高压输电公司大理局 一种基于图像处理的换流站空冷器泄漏检测方法及***
CN112085037B (zh) * 2020-09-21 2024-04-09 国网吉林省电力有限公司电力科学研究院 一种变电设备红外热故障特征提取及数字化表达方法
CN112183516A (zh) * 2020-09-22 2021-01-05 广东中鹏热能科技有限公司 基于图形图像识别技术对单照片多区域最高温的读取方法
CN112367461B (zh) * 2020-10-27 2021-12-17 京东方科技集团股份有限公司 仪表图像样本制作方法及***、存储介质、电子设备
CN112270663B (zh) * 2020-10-27 2023-11-24 北京京能东方建设工程有限公司 基于蜂巢网络环境的沥青路面过筛修复***
CN112345084B (zh) * 2020-11-05 2021-09-28 北京易达恩能科技有限公司 基于数字孪生环境的三维温度场构建方法及装置
CN112347919B (zh) * 2020-11-06 2023-07-25 中国矿业大学(北京) 一种地下天然气微泄漏点的遥感探测方法
CN112381796B (zh) * 2020-11-16 2021-08-03 广东电网有限责任公司肇庆供电局 一种基于红外数据的导线缺陷识别及检测方法
CN112464779B (zh) * 2020-11-23 2022-10-11 武汉舜陈技术有限公司 一种基于柔性材料的图形识别定位方法
CN112485329B (zh) * 2020-11-27 2024-01-26 重庆商勤科技有限公司 基于热成像与超声相结合检测排污口的方法、装置及***
CN112541887B (zh) * 2020-12-02 2024-05-03 中国华能集团有限公司南方分公司 一种火电厂多管道设备运行现场漏水缺陷检测方法
CN112698015B (zh) * 2020-12-08 2024-04-02 温州鼎玛建筑技术有限公司 一种道路桥梁裂缝检测***
CN112417378A (zh) * 2020-12-10 2021-02-26 常州大学 一种基于无人机图像处理的中华绒螯蟹质量估算方法
CN112609547A (zh) * 2020-12-11 2021-04-06 中山火炬职业技术学院 一种沥青路面各施工阶段层厚的监测方法
CN112560707B (zh) * 2020-12-18 2022-10-21 中国民用航空总局第二研究所 基于激光光源的移动式道面检测方法及***
CN112508944A (zh) * 2020-12-27 2021-03-16 中信重工开诚智能装备有限公司 一种应用于煤矿井下供水管路的泄漏检测方法
CN112766251B (zh) * 2020-12-30 2022-06-14 广东电网有限责任公司佛山供电局 变电设备红外检测方法、***、储存介质及计算机设备
CN112767322B (zh) * 2021-01-05 2023-06-13 成都圭目机器人有限公司 一种机场水泥道面fod风险评估方法和装置
CN112881432B (zh) * 2021-01-12 2022-11-29 成都泓睿科技有限责任公司 一种带液玻璃瓶瓶口裂纹检测方法
CN112763349B (zh) * 2021-01-21 2021-11-26 北京航空航天大学 一种复合材料结构冲击损伤的监测方法
CN112834457B (zh) * 2021-01-23 2022-06-03 中北大学 基于反射式激光热成像的金属微裂纹三维表征***及方法
CN112669316B (zh) * 2021-01-29 2023-05-30 南方电网调峰调频发电有限公司 电力生产异常监控方法、装置、计算机设备和存储介质
CA3211954A1 (en) * 2021-03-15 2022-09-22 Emil Sylvester RAMOS System and method for automatic monitoring of pavement condition
US11578836B2 (en) 2021-03-16 2023-02-14 Marathon Petroleum Company Lp Scalable greenhouse gas capture systems and methods
US12012883B2 (en) 2021-03-16 2024-06-18 Marathon Petroleum Company Lp Systems and methods for backhaul transportation of liquefied gas and CO2 using liquefied gas carriers
US11655940B2 (en) 2021-03-16 2023-05-23 Marathon Petroleum Company Lp Systems and methods for transporting fuel and carbon dioxide in a dual fluid vessel
CN113077562B (zh) * 2021-04-09 2021-12-14 北京市燃气集团有限责任公司 一种燃气管网智能巡检方法与***
CN113203743B (zh) * 2021-05-20 2023-12-12 中铁二十一局集团第四工程有限公司 一种基于红外热成像分析的路基裂缝检测识别及修复方法
CN113252724B (zh) * 2021-05-21 2022-05-31 山东中坚工程质量检测有限公司 一种外墙保温性能的检测方法
CN113177611B (zh) * 2021-05-24 2022-11-01 河北工业大学 基于力学指标和人工神经网络的路面病害快速巡检方法
EP4357746A1 (en) * 2021-06-16 2024-04-24 Konica Minolta, Inc. Gas concentration feature quantity estimation device, gas concentration feature quantity estimation method, program, and gas concentration feature quantity inference model generation device
CN113375065B (zh) * 2021-07-01 2022-05-24 北京化工大学 管道泄漏监测中趋势信号的消除方法及装置
DE102021207204A1 (de) * 2021-07-08 2023-01-12 Zf Friedrichshafen Ag System und Verfahren zum Schätzen der Tiefe mindestens eines zumindest zum Teil mit Wasser gefüllten Schlaglochs und entsprechendes Fahrerassistenzsystem
CN113628164A (zh) * 2021-07-12 2021-11-09 北京科技大学 一种基于深度学习与web端定位的路面裂缝检测方法
CN113606502B (zh) * 2021-07-16 2023-03-24 青岛新奥燃气设施开发有限公司 一种基于机器视觉判断操作人员执行管道漏气检测的方法
CN113592798B (zh) * 2021-07-21 2023-08-15 山东理工大学 一种道路病害智能辨识方法、***、终端及介质
CN113849901B (zh) * 2021-07-28 2024-05-03 上海机电工程研究所 针对接触换热系数辨识的改进自适应优化方法及***
CN113591721B (zh) * 2021-08-02 2022-01-25 山东省交通科学研究院 一种利用无人机确定新摊铺沥青路面取芯点位置的方法
US11447877B1 (en) 2021-08-26 2022-09-20 Marathon Petroleum Company Lp Assemblies and methods for monitoring cathodic protection of structures
CN113838078B (zh) * 2021-09-06 2023-06-30 中国矿业大学(北京) 采煤塌陷地裂缝的识别与提取方法、装置及存储介质
CN113503974B (zh) * 2021-09-09 2021-11-23 江苏沃泰冶金设备有限公司 基于pid的热成像检测***、方法及瓦斯灰输送装置
CN113963285B (zh) * 2021-09-09 2022-06-10 山东金宇信息科技集团有限公司 一种基于5g的道路养护方法及设备
CN113884464B (zh) * 2021-09-27 2024-04-26 西安空天能源动力智能制造研究院有限公司 一种基于红外热像仪的涂层波段发射率外场测量方法
CN114239194A (zh) * 2021-10-20 2022-03-25 中州水务控股有限公司 一种大水量输供水管网漏损分析和漏点定位方法
US11598689B1 (en) 2021-10-24 2023-03-07 Philip Becerra Method of detecting and identifying underground leaking pipes
CN114113217B (zh) * 2021-11-15 2024-07-12 中国矿业大学 一种煤岩体损伤程度的红外辐射量化评价方法
CN114037633B (zh) * 2021-11-18 2022-07-15 南京智谱科技有限公司 一种红外图像处理的方法及装置
CN114049336B (zh) * 2021-11-18 2024-06-14 国网重庆市电力公司电力科学研究院 Gis套管温度异常检测方法、装置、设备及可读存储介质
CN114384073B (zh) * 2021-11-30 2023-08-04 杭州申昊科技股份有限公司 一种地铁隧道裂纹检测方法及***
CN114487012B (zh) * 2021-12-29 2023-11-03 南京大学 一种土体表面裂隙发育预判方法
CN114383789B (zh) * 2022-01-05 2024-06-18 中国科学院合肥物质科学研究院 基于激励源与容性阻尼模型的金属容器气密性红外检测方法
CN114363520A (zh) * 2022-01-11 2022-04-15 河北德冠隆电子科技有限公司 一种自动巡视快速定位目标的方法
CN114119614B (zh) * 2022-01-27 2022-07-26 天津风霖物联网科技有限公司 一种远程检测建筑物的裂缝的方法
CN114459708A (zh) * 2022-02-10 2022-05-10 合肥永信科翔智能技术有限公司 一种基于气体智能感知的泄漏气体监测***
TWI833168B (zh) * 2022-02-23 2024-02-21 南亞科技股份有限公司 異常診斷方法
CN114565793B (zh) * 2022-02-28 2023-05-23 湖南北斗微芯产业发展有限公司 一种道路交通裂缝监测方法以及***
CN114577405A (zh) * 2022-02-28 2022-06-03 辽宁石油化工大学 一种基于红外成像技术的城市地下污水管道缺陷检测装置
CN114663408B (zh) * 2022-03-30 2023-04-28 江苏天晶智能装备有限公司 一种基于人工智能的门窗水密性能检测方法
CN114913316B (zh) * 2022-04-02 2023-04-07 淮沪电力有限公司田集第二发电厂 工业设备表计识别的图像分类方法、装置、电子设备和存储介质
CN114459728B (zh) * 2022-04-13 2022-06-24 中国空气动力研究与发展中心高速空气动力研究所 一种低温温敏漆转捩测量试验方法
CN114838297B (zh) * 2022-04-14 2024-03-15 七腾机器人有限公司 一种原油管道泄漏检测方法、装置、存储介质及***
US11686070B1 (en) 2022-05-04 2023-06-27 Marathon Petroleum Company Lp Systems, methods, and controllers to enhance heavy equipment warning
CN114972757A (zh) * 2022-05-31 2022-08-30 山东大学 一种隧道渗漏水区域识别方法及***
CN114878796B (zh) * 2022-07-12 2022-09-16 唐山陆达公路养护有限公司 基于道路养护的评估监测平台
CN115146513B (zh) * 2022-07-26 2023-04-07 北京科技大学 一种管线渗漏致塌模拟仿真预警方法和***
CN115100209B (zh) * 2022-08-28 2022-11-08 电子科技大学 一种基于摄像机的图像质量的修正方法及修正***
CN115184378B (zh) * 2022-09-15 2024-03-29 北京思莫特科技有限公司 一种基于移动设备的混凝土结构病害检测***及方法
CN115290211B (zh) * 2022-10-10 2023-01-20 广东电网有限责任公司中山供电局 一种基于光纤传感技术的输电线环境温度测量方法
CN115762155B (zh) * 2022-11-14 2024-03-22 东南大学 一种高速公路路面异常监测方法和***
CN115899598B (zh) * 2022-11-17 2024-06-21 浙江英集动力科技有限公司 一种融合听觉和视觉特征的供热管网状态监测方法及***
US12012082B1 (en) 2022-12-30 2024-06-18 Marathon Petroleum Company Lp Systems and methods for a hydraulic vent interlock
CN115684272B (zh) * 2023-01-03 2023-03-21 广州市市政工程试验检测有限公司 一种基于红外摄像机的钢结构裂纹检测方法
US12006014B1 (en) 2023-02-18 2024-06-11 Marathon Petroleum Company Lp Exhaust vent hoods for marine vessels and related methods
CN115861825B (zh) * 2023-02-27 2023-04-25 中铁电气化局集团有限公司 一种基于图像识别的2c检测方法
CN115855070B (zh) * 2023-03-01 2023-04-28 东莞先知大数据有限公司 一种水管漏水检测方法、装置、电子设备和存储介质
CN116416270B (zh) * 2023-03-24 2024-01-23 北京城市轨道交通咨询有限公司 一种监测地下作业面渗漏水的方法及装置
CN116095914B (zh) * 2023-04-10 2023-08-25 同方德诚(山东)科技股份公司 一种基于大数据的智慧建筑照明调节方法及***
CN116152231B (zh) * 2023-04-17 2023-07-14 卡松科技股份有限公司 基于图像处理的润滑油内杂质检测方法
US11953161B1 (en) 2023-04-18 2024-04-09 Intelcon System C.A. Monitoring and detecting pipeline leaks and spills
CN116308217B (zh) * 2023-05-19 2023-08-01 中交第四航务工程勘察设计院有限公司 一种基于物联网的混凝土监控平台管理方法及***
CN116403057B (zh) * 2023-06-09 2023-08-18 山东瑞盈智能科技有限公司 一种基于多源图像融合的输电线路巡检方法及***
CN116433668B (zh) * 2023-06-14 2023-09-12 东营孚瑞特能源设备有限公司 一种智能液压油管漏油检测方法
CN116758085B (zh) * 2023-08-21 2023-11-03 山东昆仲信息科技有限公司 一种气体污染红外图像视觉辅助检测方法
CN116862918B (zh) * 2023-09-05 2023-11-24 浙江奔月电气科技有限公司 基于人工智能的环网柜内部凝露实时检测方法
CN117169086A (zh) * 2023-09-08 2023-12-05 中建安装集团有限公司 一种建筑物地下防水层施工质量检测方法、介质及***
CN117309247B (zh) * 2023-11-28 2024-03-01 北京市农林科学院信息技术研究中心 鸡舍密闭性的判别方法、装置、***、设备及介质
CN117830300B (zh) * 2024-03-04 2024-05-14 新奥新能源工程技术有限公司 一种基于视觉的燃气管道外观质量检测方法
CN117949143B (zh) * 2024-03-26 2024-07-02 四川名人居门窗有限公司 一种门窗渗漏检测及反馈***及方法
CN117974633B (zh) * 2024-03-28 2024-06-07 潍坊科技学院 基于图像处理的番茄病虫害智能检测方法
CN118037730B (zh) * 2024-04-12 2024-06-18 广州航海学院 一种基于计算机图像处理的裂缝长度检测***
CN118038283B (zh) * 2024-04-15 2024-06-21 贵州黔通工程技术有限公司 一种沥青路面隐伏病害检测方法及设备

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5272646A (en) * 1991-04-11 1993-12-21 Farmer Edward J Method for locating leaks in a fluid pipeline and apparatus therefore
CN1936414A (zh) * 2006-08-12 2007-03-28 陈宜中 非导电材料供水管道检漏法
CN102135234A (zh) * 2010-01-27 2011-07-27 捷达世软件(深圳)有限公司 水管泄漏监控***及方法
CN102155628A (zh) * 2010-12-01 2011-08-17 广西大学 地下排水管道渗漏检测方法及装置
CN102374385A (zh) * 2011-07-21 2012-03-14 王斌 一种管道漏水检测装置及方法
CN102537667A (zh) * 2011-12-29 2012-07-04 杭州翰平电子技术有限公司 一种地下水管渗漏检测定位***及方法
CN104776318A (zh) * 2014-05-19 2015-07-15 白运福 一种地下水管漏水检测处理装置
CN105465613A (zh) * 2015-11-19 2016-04-06 中建七局第二建筑有限公司 城市地下排水管道渗漏定位***及其施工方法
CN205786366U (zh) * 2016-05-26 2016-12-07 国网浙江省电力公司宁波供电公司 一种电缆隧道水泥管片缺陷渗水特征红外热像实验装置

Family Cites Families (60)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
SU894372A1 (ru) * 1978-06-30 1981-12-30 Центральный Ордена Трудового Красного Знамени Научно-Исследовательский Институт Черной Металлургии Им.И.П.Бардина Измеритель скорости распространени трещины в металле
JPS5629129A (en) * 1979-08-17 1981-03-23 Sharp Corp Measuring system of road surface temperature
JPS62172249A (ja) * 1986-01-25 1987-07-29 Kajima Corp 煙突の劣化診断方法及び装置
US4899296A (en) * 1987-11-13 1990-02-06 Khattak Anwar S Pavement distress survey system
JPH08184398A (ja) * 1995-01-05 1996-07-16 Mitsubishi Denki Bill Techno Service Kk 埋設配管漏水箇所特定方法
JPH0961138A (ja) * 1995-08-24 1997-03-07 Mitsubishi Heavy Ind Ltd ひび割れ抽出装置
JP3460896B2 (ja) 1995-09-14 2003-10-27 ローランド株式会社 電子楽器の楽音生成装置
JPH1010064A (ja) * 1996-06-19 1998-01-16 Constec:Kk モルタル吹き付け法面の点検方法
AUPP107597A0 (en) * 1997-12-22 1998-01-22 Commonwealth Scientific And Industrial Research Organisation Road pavement deterioration inspection system
JP2000035372A (ja) * 1998-07-16 2000-02-02 Ishikawajima Inspection & Instrumentation Co 赤外線を用いた発泡検査方法
JP3205806B2 (ja) * 1999-04-02 2001-09-04 鹿島建設株式会社 アスファルト表面層内部の水探知方法および装置
JP2001215164A (ja) * 2000-02-02 2001-08-10 Kansai Electric Power Co Inc:The 霧状水滴を利用した真空・ガス漏れ検知装置
JP2004191258A (ja) * 2002-12-12 2004-07-08 Yoshitake Eda 建築物の漏水路検知方法
US6874932B2 (en) * 2003-06-30 2005-04-05 General Electric Company Methods for determining the depth of defects
US7073979B2 (en) * 2003-11-26 2006-07-11 Aries Industries Incorporated Method and apparatus for performing sewer maintenance with a thermal sensor
CN100402753C (zh) * 2003-12-10 2008-07-16 刘世俊 高级路面裂纹处理工艺方法
US7358860B2 (en) * 2005-03-31 2008-04-15 American Air Liquide, Inc. Method and apparatus to monitor and detect cryogenic liquefied gas leaks
CN101070947A (zh) 2006-04-28 2007-11-14 王明根 管道接缝渗漏检测***
US8288726B2 (en) * 2006-12-19 2012-10-16 Weil Gary J Remote sensing of subsurface artifacts by use of visual and thermal imagery
WO2010106639A1 (ja) * 2009-03-17 2010-09-23 西日本高速道路エンジニアリング四国株式会社 構造物の損傷深さ判定方法とその装置及び構造物の損傷処置判定方法とその装置
CN201449248U (zh) * 2009-03-18 2010-05-05 河海大学 一种土体裂隙发育监测仪
CN101701919B (zh) * 2009-11-20 2011-05-11 长安大学 一种基于图像的路面裂缝检测***及检测方法
US20110221906A1 (en) * 2010-03-12 2011-09-15 Board Of Regents, The University Of Texas System Multiple Camera System for Automated Surface Distress Measurement
CN101845787A (zh) * 2010-04-09 2010-09-29 同济大学 基于双目视觉的水泥混凝土路面错台检测装置及方法
WO2012096530A2 (en) * 2011-01-13 2012-07-19 Samsung Electronics Co., Ltd. Multi-view rendering apparatus and method using background pixel expansion and background-first patch matching
CN102108666B (zh) * 2011-01-17 2012-05-30 长安大学 一种沥青路面施工质量实时控制方法
CN102182137A (zh) * 2011-02-25 2011-09-14 广州飒特电力红外技术有限公司 路面缺陷检测***及方法
CN102621419A (zh) * 2012-03-28 2012-08-01 山东省电力学校 基于激光和双目视觉图像对线路电气设备自动识别和监测方法
CN102636313B (zh) * 2012-04-11 2014-12-03 浙江工业大学 基于红外热成像图像处理的渗漏源检测装置
RU2511275C2 (ru) * 2012-07-16 2014-04-10 Федеральное государственное унитарное предприятие "Научно-исследовательский институт физических проблем им. Ф.В. Лукина" Наноструктурный ик-приемник (болометр) с большой поверхностью поглощения
CN103308521A (zh) * 2012-08-29 2013-09-18 中国人民解放军第二炮兵工程大学 一种增强红外热波检测图像缺陷对比度的方法
CN102927448B (zh) * 2012-09-25 2016-12-21 北京声迅电子股份有限公司 管道无损检测方法
CN103217256A (zh) * 2013-03-20 2013-07-24 北京理工大学 基于红外图像的局部灰度-熵差的泄漏检测定位方法
CN103321129A (zh) * 2013-06-18 2013-09-25 中山市拓维电子科技有限公司 基于3g网络的红外热像的远程路面施工诊断***及方法
CN103808760B (zh) * 2013-12-12 2017-04-26 交通运输部公路科学研究所 混凝土结构红外热成像无损检测用热激励装置
CN103882891B (zh) * 2014-01-16 2016-06-08 同济大学 利用红外热场快速预测地下连续墙侧壁渗漏的方法
CN103912791B (zh) * 2014-01-26 2016-05-04 清华大学深圳研究生院 地下管网泄漏探测方法
US9857228B2 (en) * 2014-03-25 2018-01-02 Rosemount Inc. Process conduit anomaly detection using thermal imaging
US10576907B2 (en) * 2014-05-13 2020-03-03 Gse Technologies, Llc Remote scanning and detection apparatus and method
CN103983514B (zh) * 2014-05-22 2016-06-01 中国矿业大学 一种煤岩裂隙发育红外辐射监测试验方法
CN103983513B (zh) * 2014-05-22 2016-03-02 中国矿业大学 一种采用红外辐射观测煤岩裂隙发育过程的装置及方法
CN104048969A (zh) * 2014-06-19 2014-09-17 樊晓东 一种隧道病害的识别方法
CN104034733A (zh) * 2014-07-02 2014-09-10 中国人民解放军国防科学技术大学 基于双目视觉监测与表面裂纹图像识别的寿命预测方法
CN104574393B (zh) * 2014-12-30 2017-08-11 北京恒达锦程图像技术有限公司 一种三维路面裂缝图像生成***和方法
CN104713885B (zh) * 2015-03-04 2017-06-30 中国人民解放军国防科学技术大学 一种用于pcb板在线检测的结构光辅助双目测量方法
CN104749187A (zh) * 2015-03-25 2015-07-01 武汉武大卓越科技有限责任公司 基于红外温度场和灰度图像的隧道衬砌病害检测装置
CN104764528B (zh) * 2015-04-03 2018-01-12 中国矿业大学 一种煤岩裂隙发育过程中的热红外信息去噪方法
CN105113375B (zh) * 2015-05-15 2017-04-19 南京航空航天大学 一种基于线结构光的路面裂缝检测***及其检测方法
SE539312C2 (en) * 2015-06-10 2017-06-27 Conny Andersson Med Firma Ca Konsult A method of determining the quality of a newly produced asphalt pavement
CN205115977U (zh) * 2015-11-23 2016-03-30 张先 道路桥梁沥青路面裂缝检测装置
CN105719283A (zh) * 2016-01-18 2016-06-29 苏州科技学院 一种基于Hessian矩阵多尺度滤波的路面裂缝图像检测方法
CN105717163A (zh) * 2016-01-29 2016-06-29 中国商用飞机有限责任公司 红外热像检测缺陷的方法
CN105527165B (zh) * 2016-02-02 2018-07-24 山东省交通科学研究院 一种沥青路面裂缝荷载响应相对位移测试方法及测试装置
CN105938620B (zh) * 2016-04-14 2018-12-25 北京工业大学 一种小口径管内焊缝表面缺陷识别装置
CN205748654U (zh) * 2016-06-21 2016-11-30 国家电网公司 基于红外热成像的变压器实时监测***
CN106018096A (zh) * 2016-07-20 2016-10-12 中国矿业大学 煤岩破裂过程中裂隙发育区的红外辐射监测定位方法
CN106124949B (zh) * 2016-08-30 2019-08-13 国网山东省电力公司济南供电公司 一种基于热红外成像技术对绝缘子故障在线监测方法
CN206573258U (zh) 2017-02-20 2017-10-20 广东工业大学 一种管道渗漏检测装置
CN206629279U (zh) 2017-03-30 2017-11-10 中建地下空间有限公司 一种应用于综合管廊的移动巡检***
WO2018216629A1 (ja) * 2017-05-22 2018-11-29 キヤノン株式会社 情報処理装置、情報処理方法、及びプログラム

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5272646A (en) * 1991-04-11 1993-12-21 Farmer Edward J Method for locating leaks in a fluid pipeline and apparatus therefore
CN1936414A (zh) * 2006-08-12 2007-03-28 陈宜中 非导电材料供水管道检漏法
CN102135234A (zh) * 2010-01-27 2011-07-27 捷达世软件(深圳)有限公司 水管泄漏监控***及方法
CN102155628A (zh) * 2010-12-01 2011-08-17 广西大学 地下排水管道渗漏检测方法及装置
CN102374385A (zh) * 2011-07-21 2012-03-14 王斌 一种管道漏水检测装置及方法
CN102537667A (zh) * 2011-12-29 2012-07-04 杭州翰平电子技术有限公司 一种地下水管渗漏检测定位***及方法
CN104776318A (zh) * 2014-05-19 2015-07-15 白运福 一种地下水管漏水检测处理装置
CN105465613A (zh) * 2015-11-19 2016-04-06 中建七局第二建筑有限公司 城市地下排水管道渗漏定位***及其施工方法
CN205786366U (zh) * 2016-05-26 2016-12-07 国网浙江省电力公司宁波供电公司 一种电缆隧道水泥管片缺陷渗水特征红外热像实验装置

Cited By (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020015354A1 (zh) * 2018-07-17 2020-01-23 北京讯腾智慧科技股份有限公司 一种基于北斗差分定位的燃气管道的修补方法及***
CN109859125B (zh) * 2019-01-14 2022-10-21 广东工业大学 基于形态学检测与小波变换的图像高光修复方法
CN109859125A (zh) * 2019-01-14 2019-06-07 广东工业大学 基于形态学检测与小波变换的图像高光修复方法
CN109816656A (zh) * 2019-02-01 2019-05-28 辽宁工程技术大学 一种火电厂负压侧***漏点精确定位方法
CN109816656B (zh) * 2019-02-01 2023-06-20 辽宁工程技术大学 一种火电厂负压侧***漏点精确定位方法
CN111242123A (zh) * 2020-01-07 2020-06-05 西安交通大学 一种基于红外图像的电力设备故障诊断方法
CN111242123B (zh) * 2020-01-07 2022-10-28 西安交通大学 一种基于红外图像的电力设备故障诊断方法
CN112927223A (zh) * 2021-03-29 2021-06-08 南通大学 一种基于红外热成像仪的玻璃幕墙检测方法
CN114034405A (zh) * 2021-11-08 2022-02-11 北京航空航天大学 一种非接触式测温方法及***
CN113777028B (zh) * 2021-11-11 2022-01-18 成都理工大学 测量凝胶类堵漏材料与岩石壁面粘附强度的装置和方法
CN113777028A (zh) * 2021-11-11 2021-12-10 成都理工大学 测量凝胶类堵漏材料与岩石壁面粘附强度的装置和方法
CN114511469A (zh) * 2022-04-06 2022-05-17 江苏游隼微电子有限公司 一种图像智能降噪先验检测方法
CN114511469B (zh) * 2022-04-06 2022-06-21 江苏游隼微电子有限公司 一种图像智能降噪先验检测方法
CN116823839A (zh) * 2023-08-31 2023-09-29 梁山中维热力有限公司 基于热红外图像的管道泄漏检测方法
CN116823839B (zh) * 2023-08-31 2023-12-01 梁山中维热力有限公司 基于热红外图像的管道泄漏检测方法
CN117392140A (zh) * 2023-12-13 2024-01-12 宏发建设有限公司 一种基于图像处理的建筑幕墙玻璃破裂检测方法及***
CN117392140B (zh) * 2023-12-13 2024-02-06 宏发建设有限公司 一种基于图像处理的建筑幕墙玻璃破裂检测方法及***
CN118154594A (zh) * 2024-05-10 2024-06-07 山东力乐包装股份有限公司 一种基于电数字数据处理的木质托盘缺陷分析***

Also Published As

Publication number Publication date
GB2569751A (en) 2019-06-26
CN109804232A (zh) 2019-05-24
CN109804232B (zh) 2021-06-01
GB2581294A (en) 2020-08-12
GB2581293A (en) 2020-08-12
GB2571016B (en) 2020-09-02
CN110268190B (zh) 2021-04-30
US11486548B2 (en) 2022-11-01
GB2573429A (en) 2019-11-06
GB2569751B (en) 2020-04-08
CN109804119A (zh) 2019-05-24
GB201905912D0 (en) 2019-06-12
CN109716108B (zh) 2021-07-13
WO2018122589A1 (zh) 2018-07-05
GB201711412D0 (en) 2017-08-30
CN110268190A (zh) 2019-09-20
WO2018122810A1 (zh) 2018-07-05
GB201905909D0 (en) 2019-06-12
GB2573429B (en) 2020-07-15
CN109743879B (zh) 2021-02-12
CN109804119B (zh) 2021-03-19
GB2571016A (en) 2019-08-14
GB202006107D0 (en) 2020-06-10
GB202006108D0 (en) 2020-06-10
US20190331301A1 (en) 2019-10-31
CN109743879A (zh) 2019-05-10
WO2018122818A1 (zh) 2018-07-05
CN109716108A (zh) 2019-05-03
US11221107B2 (en) 2022-01-11
GB2581294B (en) 2021-01-06
GB201909416D0 (en) 2019-08-14
GB2581293B (en) 2021-01-06
US20190339209A1 (en) 2019-11-07
WO2018122819A1 (zh) 2018-07-05

Similar Documents

Publication Publication Date Title
US11221107B2 (en) Method for leakage detection of underground pipeline corridor based on dynamic infrared thermal image processing
Iyer et al. Segmentation of pipe images for crack detection in buried sewers
Yu et al. Efficient crack detection method for tunnel lining surface cracks based on infrared images
CN109544535B (zh) 一种基于红外截止滤镜光学滤波特性的窥视摄像头检测方法及***
CN111667470B (zh) 一种基于数字图像的工业管道探伤内壁检测方法
CN112862744B (zh) 一种基于超声图像的电容内部缺陷智能检测方法
CN107481233A (zh) 一种应用于fod异物检测雷达中的图像识别方法
CN113313107B (zh) 一种斜拉桥缆索表面多类型病害智能检测和识别方法
CN105787912B (zh) 一种基于分类的阶跃型边缘亚像素定位方法
CN105787870A (zh) 一种图形图像拼接融合***
Li et al. Detection algorithm of defects on polyethylene gas pipe using image recognition
Hu et al. Rail surface spalling detection based on visual saliency
Qi et al. Micro-concrete crack detection of underwater structures based on convolutional neural network
KR102355997B1 (ko) 스마트글래스 기반의 콘크리트 구조물 건전도 모니터링 방법
Kröhnert et al. Segmentation of environmental time lapse image sequences for the determination of shore lines captured by hand-held smartphone cameras
KR101557271B1 (ko) 영상 내 원 형상 검출에 따른 검출 원 형상의 근사화 방법
Gao et al. Intelligent crack damage detection system in shield tunnel using combination of retinanet and optimal adaptive selection
CN112927223A (zh) 一种基于红外热成像仪的玻璃幕墙检测方法
CN117036259A (zh) 一种基于深度学习的金属板材表面缺陷检测方法
Singh et al. Segmentation technique for the detection of Micro cracks in solar cell using support vector machine
Sun et al. Research on image segmentation and extraction algorithm for bicolor water level gauge
CN112734745B (zh) 一种融合gis数据的无人机热红外影像供暖管道泄漏探测方法
CN109299655A (zh) 一种基于无人机的海上溢油在线快速识别方法
Li et al. Graph network refining for pavement crack detection based on multiscale curvilinear structure filter
CN112329572B (zh) 一种基于边框和闪光点的快速静态活体检测方法及装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17888959

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 201905909

Country of ref document: GB

Kind code of ref document: A

Free format text: PCT FILING DATE = 20171230

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 17888959

Country of ref document: EP

Kind code of ref document: A1