CN115690062A - Rail surface damage state detection method and device and electronic equipment - Google Patents
Rail surface damage state detection method and device and electronic equipment Download PDFInfo
- Publication number
- CN115690062A CN115690062A CN202211394590.1A CN202211394590A CN115690062A CN 115690062 A CN115690062 A CN 115690062A CN 202211394590 A CN202211394590 A CN 202211394590A CN 115690062 A CN115690062 A CN 115690062A
- Authority
- CN
- China
- Prior art keywords
- rail surface
- damage
- dimensional point
- dimensional
- grid
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention provides a rail surface damage state detection method, a rail surface damage state detection device and electronic equipment, wherein the method comprises the following steps: acquiring a plurality of three-dimensional point clouds corresponding to a rail surface to be detected; determining a plurality of depth images corresponding to the rail surface to be detected based on the plurality of three-dimensional point clouds; determining a damage area image of the rail surface to be detected based on the plurality of depth images and a pre-trained rail surface damage identification model; according to the damage area image, determining the damage state of the rail surface to be detected, wherein the damage state comprises the following steps: lesion length, lesion width, and lesion depth. Determining a damage area image of a rail surface to be detected through a plurality of depth images and a pre-trained rail surface damage identification model; and determining the damage length, the damage width and the damage depth of the rail surface to be detected according to the damage area image, and accurately and quickly detecting the damage length, the damage width and the damage depth of the rail surface.
Description
Technical Field
The invention relates to the technical field of rail surface damage state detection, in particular to a rail surface damage state detection method and device and electronic equipment.
Background
The rail surface damage state of the steel rail is an important factor influencing safe running of the railway and comfortable riding of people. The traditional steel rail surface damage state detection method is divided into two types, namely manual inspection and area-array camera image detection, but the two methods have the defects to a certain extent: for manual detection, the detection efficiency and precision of the method can not meet the detection requirements of the existing high-speed railway, and for area array camera image detection, although the detection efficiency and precision of the method are greatly improved, the method can not detect the important detection item of the surface damage depth of the steel rail through an area array image, and only can detect the length and the width of the surface damage of the steel rail.
Disclosure of Invention
The invention aims to provide a rail surface damage state detection method, a rail surface damage state detection device and electronic equipment, so that the length, width and depth information of rail surface damage can be accurately and quickly detected.
The invention provides a rail surface damage state detection method, which comprises the following steps:
acquiring a plurality of three-dimensional point clouds corresponding to a rail surface to be detected;
determining a plurality of depth images corresponding to the rail surface to be detected based on the plurality of three-dimensional point clouds;
determining a damage area image of the rail surface to be detected based on the plurality of depth images and a pre-trained rail surface damage identification model;
according to the damage area image, determining the damage state of the rail surface to be detected, wherein the damage state comprises the following steps: lesion length, lesion width and lesion depth.
Further, the step of acquiring a plurality of three-dimensional point clouds corresponding to the rail surface to be detected comprises the following steps;
a plurality of three-dimensional point clouds corresponding to the rail surface to be detected are collected through a collecting device according to a preset step length, and a plurality of three-dimensional point clouds are obtained.
Further, after the step of obtaining a plurality of three-dimensional point clouds corresponding to the rail surface to be detected, each three-dimensional point cloud comprises a plurality of first three-dimensional points and corresponding X, Y and Z coordinate values, the method further comprises the following steps:
aiming at each first three-dimensional point, taking the current first three-dimensional point as a center, and calculating the number of second three-dimensional points within a filtering radius range based on a preset filtering radius;
if the number of the second three-dimensional points is smaller than a first preset threshold value, filtering the current first three-dimensional points to obtain a plurality of filtered third three-dimensional points;
selecting a third three-dimensional point corresponding to the maximum Z coordinate value from the plurality of third three-dimensional points;
and searching fourth three-dimensional points respectively corresponding to the Z coordinate values not greater than the second preset threshold value based on the maximum Z coordinate value to obtain a plurality of rail surface point clouds corresponding to the plurality of three-dimensional point clouds.
Further, the step of determining a plurality of depth images corresponding to the rail surface to be detected based on the plurality of three-dimensional point clouds comprises the steps of;
aiming at each rail surface point cloud, acquiring X, Y and Z coordinate values of each fourth three-dimensional point in the rail surface point cloud;
selecting the maximum X coordinate value, the maximum Y coordinate value, the minimum X coordinate value and the minimum Y coordinate value from the X, Y and Z coordinate values of each fourth three-dimensional point, and determining the length value and the width value of the depth image corresponding to the rail surface point cloud;
and calculating to obtain the grid line number and the grid column number corresponding to each fourth three-dimensional point based on the preset grid size of the depth image and the X and Y coordinate values of each fourth three-dimensional point.
And acquiring the number of the fourth three-dimensional points in each grid based on the row number and the column number of the grid.
And aiming at each grid, calculating the corresponding position intensity value of the grid based on the number of the fourth three-dimensional points in the grid and the Z coordinate value of each fourth three-dimensional point in the grid.
And generating a plurality of depth images corresponding to the rail surface to be detected based on the grid line number and the grid column number of each grid and the position strength value.
Further, according to the damaged area image, determining a damaged state of the rail surface to be detected, wherein the damaged state comprises: the steps of lesion length, lesion width and lesion depth include;
determining the damage length and the damage width of the rail surface to be detected based on the grid row number and the grid column number of each grid and the grid size;
calculating an X coordinate value of a rail surface central line of the rail surface to be detected based on a first preset algorithm;
determining a plurality of target three-dimensional points in the damaged area image based on the X coordinate value of the rail surface central line and a third preset threshold;
and acquiring the damage depth of the rail surface to be detected based on the plurality of target three-dimensional points.
Further, the step of acquiring the damage depth of the rail surface to be detected based on the plurality of target three-dimensional points comprises the following steps;
fitting a plurality of target three-dimensional points into a three-dimensional plane based on a second preset algorithm;
calculating the distance value between each target three-dimensional point in the point cloud and the three-dimensional plane;
and determining the maximum distance value as the damage depth of the rail surface to be detected.
The invention provides a rail surface damage state detection device, comprising: the acquisition module is used for acquiring a plurality of three-dimensional point clouds corresponding to the rail surface to be detected; the first determining module is used for determining a plurality of depth images corresponding to the rail surface to be detected based on a plurality of three-dimensional point clouds; the second determining module is used for determining the damage position information of the rail surface to be detected based on the plurality of depth images and the rail surface damage identification model trained in advance; the third determining module is used for determining the damage state of the rail surface to be detected according to the damage position information; wherein, the state of injury includes: lesion length, lesion width and lesion depth.
Further, the obtaining module is further configured to: and acquiring a plurality of three-dimensional point clouds corresponding to the rail surface to be detected according to a preset step length by using an acquisition device to obtain the plurality of three-dimensional point clouds.
The electronic device provided by the invention comprises a processor and a memory, wherein the memory stores computer executable instructions capable of being executed by the processor, and the processor executes the computer executable instructions to realize the rail surface damage state detection method.
The invention provides a computer-readable storage medium, which stores computer-executable instructions, and when the computer-executable instructions are called and executed by a processor, the computer-executable instructions cause the processor to implement any one of the rail surface damage state detection methods.
The invention provides a rail surface damage state detection method, a rail surface damage state detection device and electronic equipment, wherein the method comprises the steps of obtaining a plurality of three-dimensional point clouds corresponding to a rail surface to be detected; determining a plurality of depth images corresponding to the rail surface to be detected based on the plurality of three-dimensional point clouds; determining the damage position information of the rail surface to be detected based on the plurality of depth images and a pre-trained rail surface damage identification model; determining the damage state of the rail surface to be detected according to the damage position information; wherein the damage state comprises: length, width and depth of the lesion site. Determining the damage position information of the rail surface to be detected through a plurality of depth images and a pre-trained rail surface damage identification model; the length, the width and the depth are determined according to the damage position information, and the length, the width and the depth information of the rail surface damage position can be accurately and quickly detected.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a flowchart of a rail surface damage state detection method according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a rail surface damage identification model according to an embodiment of the present invention;
fig. 3 is a flowchart of another rail surface damage state detection method according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of an acquisition device according to an embodiment of the present invention;
fig. 5 is a schematic diagram of a three-dimensional point cloud of a rail surface damage and drop block according to an embodiment of the present invention;
fig. 6 is a flowchart illustrating a detection process of a damage depth of a rail surface to be detected according to an embodiment of the present invention;
fig. 7 is a schematic structural diagram of a rail surface damage state detection device according to an embodiment of the present invention;
fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
The technical solutions of the present invention are clearly and completely described in conjunction with the embodiments, and it is obvious that the described embodiments are some, not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In recent years, high-speed rails have become a business card displayed externally in China, and are praised as 'four new inventions'. However, with the increase of railway operation mileage and the improvement of running speed in China, the safe service state of the railway track is more challenged than the prior art, and the rail surface damage state of the steel rail is an important factor influencing safe running of the railway and comfortable riding of people. Rail surface damages such as the chipping and the cracking of the surface of the steel rail can seriously limit the use performance of the steel rail and the safe running of the train in the running process of the train, so that how to timely and effectively detect the rail surface damage state of the steel rail becomes a problem to be solved urgently.
The traditional steel rail surface damage state detection method comprises two types, namely manual inspection and area-array camera image detection, wherein the manual inspection mainly refers to that a railway worker inspects the rail surface damage state along a line by using lighting equipment and human eyes at a limited skylight point time; the area-array camera image detection mainly refers to detecting an abnormal rail surface state image in an image through an image post-processing method by using an image acquired by an area-array camera, and then obtaining the rail surface damage of an actual railway through mileage association.
However, both of the above methods have some disadvantages: for manual detection, the detection efficiency and precision of the detection device cannot meet the detection requirements of the existing high-speed railway, and for two steel rails on the left and right tracks, the manual detection needs two workers to perform detection at walking speed within limited skylight time, so that the efficiency is low. And the damage states such as a plurality of tiny rail surface cracks and the like cannot be rapidly identified by naked eyes. For the image detection of the area-array camera, although the detection efficiency and the detection precision are improved to a great extent, the important detection item of the surface damage depth of the steel rail cannot be detected through the area-array image, and only the length and the width of the surface damage of the steel rail can be detected.
To facilitate understanding of the present embodiment, first, a rail surface damage state detection method disclosed in the embodiment of the present invention is described in detail, and as shown in fig. 1, the method includes the following steps:
and S102, acquiring a plurality of three-dimensional point clouds corresponding to the rail surface to be detected.
The rail surface to be detected can be generally understood as the top surface of the rail, also called a rail head; the three-dimensional point cloud may be understood as a data set of points (i.e., a set of points in a three-dimensional space), which is a set of points representing the surface characteristics and spatial distribution of the target object under the same reference coordinate system. The three-dimensional point cloud has corresponding data (including the spatial coordinates X, Y, Z, etc. corresponding to each point in the three-dimensional point cloud).
In practical implementation, the rail has a certain length, so that the rail surface to be detected also has a certain length, the length can be divided into a plurality of distance sections, and each distance is short and corresponds to a three-dimensional point cloud; specifically, the three-dimensional point cloud and the corresponding data thereof can be obtained through a laser radar or a depth camera and other acquisition devices.
And step S104, determining a plurality of depth images corresponding to the rail surface to be detected based on the plurality of three-dimensional point clouds.
The depth image is also referred to as a range image, and refers to an image in which a distance (depth) from the acquisition device to each point in the target to be measured is used as a pixel value, the depth image can be calculated as data corresponding to a three-dimensional point cloud through coordinate conversion, and the depth image can be obtained through coordinate conversion of the point cloud data.
And S106, determining a damage area image of the rail surface to be detected based on the plurality of depth images and the rail surface damage identification model trained in advance.
The rail surface damage identification model is an improvement based on a U-net model, and can automatically identify rail surface damage. The U-net model is divided into an encoder and a decoder and has a symmetrical structure, the encoder performs an image down-sampling process, and characteristic information of an image is extracted through a typical convolutional neural network structure. The decoder performs an up-sampling process, deconvolution is used to combine information of down-sampling layers to restore detail information, the decoder and the encoder perform four times of sampling, and the difference from the encoder is that the number of up-sampling channels of the decoder is halved each time, the size of a feature map is enlarged, and the picture precision is gradually restored. According to the method, an improved U-net model (rail surface damage identification model) is adopted, as shown in fig. 2, a global module is added between a feature extraction part and an up-sampling part in the whole network, the calculation amount of the whole network is greatly reduced due to the addition of the module, the network performance is improved on the basis of realizing network lightweight, meanwhile, effective modeling on global context is kept, and finally, rapid and accurate rail surface damage identification is realized.
In a specific implementation process, the trained rail surface damage identification model can be obtained by training an initial detection model constructed based on a machine learning algorithm by using a training set, wherein the training set can be historical depth images of a damaged rail surface and image data corresponding to the historical depth images. The trained rail surface damage identification model may also be obtained by using a deep learning target positioning algorithm, specifically, a large number of depth images corresponding to the damaged rail surface and image data corresponding to the depth images (for example, row-column coordinates, gray values, and the like of each pixel) are collected first, the rail surface damage identification model is trained based on the image data, then the trained rail surface damage identification model is used to detect a damage area image in the depth images, and the depth images are specifically input into the trained rail surface damage identification model to obtain the damage area image.
The method is used for rapidly, accurately and automatically identifying the rail surface damage based on the U-net improved model, and experiments prove that: the improved U-net is better than 98.43% of the traditional U-net model in global recognition accuracy rate of 99.19%, and the improved U-net is greatly improved in recognition accuracy rate by 83.53 compared with 80.43 of the U-net; the improved U-net was nearly 10% higher than U-net by 80.72% in recall, demonstrating that the improved U-net is better overall for multiple disease identification.
Step S108, determining the damage state of the rail surface to be detected according to the damage position information; wherein the damage state comprises: length, width and depth of the lesion site. According to the damage area image, determining the damage state of the rail surface to be detected, wherein the damage state comprises the following steps: lesion length, lesion width and lesion depth.
The above-mentioned lesion region image may be understood as a part of the depth image; the depth image may include a plurality of region images; if the rail surface damage position exactly corresponds to one or more adjacent area images, determining the one or more adjacent area images as a damage area image; in practical implementation, when the depth image is input into the trained rail surface damage identification model, a damage area image in a plurality of area images can be acquired.
The rail surface damage state detection method comprises the following steps: acquiring a plurality of three-dimensional point clouds corresponding to a rail surface to be detected; determining a plurality of depth images corresponding to the rail surface to be detected based on the plurality of three-dimensional point clouds; determining a damage area image of the rail surface to be detected based on the plurality of depth images and a pre-trained rail surface damage identification model; according to the damage area image, determining the damage state of the rail surface to be detected, wherein the damage state comprises the following steps: lesion length, lesion width and lesion depth. Determining a damage area image of a rail surface to be detected through a plurality of depth images and a pre-trained rail surface damage identification model; and determining the damage length, the damage width and the damage depth of the rail surface to be detected according to the damage area image, and accurately and quickly detecting the damage length, the damage width and the damage depth of the rail surface.
The embodiment of the invention also provides another rail surface damage state detection method which is realized on the basis of the method of the embodiment; as shown in fig. 3, the method comprises the steps of:
step S202, a plurality of three-dimensional point clouds corresponding to the rail surface to be detected are collected through a collecting device according to a preset step length, and a plurality of three-dimensional point clouds are obtained.
The above-mentioned collecting device can be understood as a movable trolley comprising: the system comprises a laser acquisition unit, a navigation positioning unit, a synchronous control unit and a power supply unit; the navigation positioning unit also comprises an inertial measurement unit and a milemeter, and the laser acquisition unit comprises a linear sensor 1 and a linear sensor 2; specifically, referring to the schematic structural diagram of the acquisition device shown in fig. 4, the laser acquisition unit and the navigation positioning unit are respectively connected to the synchronous control unit, and the laser acquisition unit, the navigation positioning unit and the synchronous control unit are respectively connected to the power supply unit; the collection device is connected with the microelectronic unit. The linear sensor 1 and the linear sensor 2 are used as three-dimensional sensing sensors and can acquire three-dimensional point clouds corresponding to rail surfaces of a left rail and a right rail; the inertial measurement unit is used as an attitude sensor to acquire instantaneous speed and attitude data of the acquisition device. The odometer is used as a position sensor to acquire accumulated running mileage data of the acquisition device and the like. The power supply unit adopts a lithium battery pack to supply power, and an internal integrated overload overvoltage and undervoltage protection circuit guarantees the electrical safety of the internal sensor and the power supply unit.
In a specific implementation, the microelectronic unit may set an acquisition step size, such as 5m, 10m, etc., according to an actual requirement. Taking 10m as an example, the acquisition device corresponds to one three-dimensional point cloud every time the acquisition device runs for 10m, and if the total length of the rail surface to be detected is 500m, 50 three-dimensional point clouds can be acquired. The microelectronic unit may also set a trigger time for the odometer; setting the laser intensity to a laser acquisition unit; after receiving the trigger time and the acquisition step length, the odometer starts to move; and synchronous time information is sent to a synchronous control unit, and the synchronous control unit carries out time-space synchronous trigger control on all the sensors to ensure the synchronism of data. The linear sensor 1 and the linear sensor 2 transmit the collected three-dimensional point cloud to a plurality of sets of integrated embedded computers (microelectronic units) for storage.
In actual implementation, the acquisition device can simultaneously acquire three-dimensional point clouds corresponding to the rail surfaces of the left and right rails at the speed of 5-10 km/h; the rail surface to be detected is only the rail surface of one of the rails, and the method for detecting the rail surface damage state of the other rail is the same as that of the rail surface to be detected.
Step S204, aiming at each first three-dimensional point, taking the current first three-dimensional point as the center, and calculating the number of second three-dimensional points belonging to the range of the filtering radius based on the preset filtering radius.
Step S206, if the number of the second three-dimensional points is less than the first preset threshold, filtering the current first three-dimensional point to obtain a plurality of filtered third three-dimensional points.
Each three-dimensional point cloud comprises a plurality of first three-dimensional points and corresponding X, Y and Z coordinate values thereof.
In actual implementation, all sensors are noisy and have several types of noise, including point cloud disturbance and abnormal values, which means that a first three-dimensional point has a certain probability of being located in a certain radius range near a sampled place (disturbance), or the first three-dimensional point may appear in any position (abnormal value) in space, so that the noise point can be filtered through a radius filtering method, then the highest point of a Z value in the point cloud is found out to be the highest point of a rail surface to be detected, and finally all points within 16mm below the Z value are extracted to be the point cloud of the rail surface.
Specifically, a filtering radius may be set, for example, 1m, and then the number of other points (second three-dimensional points) of each first three-dimensional point within the filtering radius range may be calculated, taking one of the first three-dimensional points as an example, if the number of other points within the radius range is less than a certain set threshold (a first preset threshold), the first three-dimensional point is filtered. After each first three-dimensional point is subjected to the operation, a part of the first three-dimensional points are filtered, a part of the first three-dimensional points are reserved, and the reserved first three-dimensional points are determined to be a plurality of third three-dimensional points.
Step S208, selecting a third three-dimensional point corresponding to the maximum Z coordinate value from the plurality of third three-dimensional points.
Step S210, based on the maximum Z coordinate value, searching fourth three-dimensional points respectively corresponding to the Z coordinate value not greater than a second preset threshold value, and obtaining a plurality of rail surface point clouds corresponding to the three-dimensional point clouds.
The maximum Z coordinate value (Z coordinate value) can be found from the X, Y and Z coordinate values respectively corresponding to the reserved first three-dimensional points (the filtered third three-dimensional points) max )。
Then finding out Z coordinate value less than or equal to Z from the reserved first three-dimensional points max -16 (second preset threshold) fourth three-dimensional points, obtaining a trajectory surface point cloud. After each three-dimensional point cloud is subjected to the operation, a corresponding rail surface point cloud can be obtained.
Step S212, aiming at each rail surface point cloud, obtaining the X, Y and Z coordinate values of each fourth three-dimensional point in the rail surface point cloud.
Step S214, selecting the maximum X coordinate value, the maximum Y coordinate value, the minimum X coordinate value and the minimum Y coordinate value from the X, Y and Z coordinate values of each fourth three-dimensional point, and determining the length value and the width value of the depth image corresponding to the rail surface point cloud.
During specific implementation, the size of the rail surface point cloud projection depth image can be determined firstly, wherein the size comprises a length value and a width value, and the length value can correspond to a calculation result obtained by subtracting the minimum Y coordinate value from the maximum Y coordinate value; the width value may correspond to the calculation of the maximum X coordinate value minus the minimum X coordinate value.
Step S216, based on the preset grid size of the depth image and the X and Y coordinate values of each fourth three-dimensional point, a grid row number and a grid column number corresponding to each fourth three-dimensional point are calculated.
The pixel grid size q of the depth image may then be set, for example 10mm x 10mm, etc.
Then, calculation is performed according to formulas 2.1 and 2.2, and corresponding grid row and column numbers (R, C), (int) after the projection of each point (fourth three-dimensional point) can be obtained to represent that the calculation result is subjected to integer taking.
Wherein m is a Y coordinate value of each fourth three-dimensional point; m is in The minimum Y coordinate value.
Wherein n is the X coordinate value of each fourth three-dimensional point; n is a radical of an alkyl radical in The minimum X coordinate value.
Step S218, based on the grid row number and the grid column number, the number of the fourth three-dimensional points in each grid is obtained.
In particular, the corresponding grid row and column numbers (R, C) may be the same after the projection of the different fourth three-dimensional points, and therefore, one grid may include a plurality of fourth three-dimensional points.
Step S220, for each grid, calculating a position strength value corresponding to the grid based on the number of the fourth three-dimensional points in the grid and the Z-coordinate value of each fourth three-dimensional point in the grid.
The depth image may include a plurality of meshes (pixels), each of which is the same size, and then the position intensity value (gray value) of each mesh may be calculated according to equation 2.3.
Wherein, I (R,C) As positional intensity values of the grid, I i Representing the elevation (Z coordinate) value of the ith fourth three-dimensional point prior to projection, and k represents the total number of fourth three-dimensional points within the grid column (R, C) position.
Step S222, based on the grid row number and the grid column number of each grid and the position intensity value, generating a plurality of depth images corresponding to the rail surface to be detected.
Each grid in the depth image has a corresponding row-column number (R, C) and a grid size q, and grid row-column coordinates of each grid can be calculated according to (R, C) and the grid size q, for example, when q is 10mm × 10mm, and (R, C) is (3, 2), grid coordinates of the grid are 30mm × 20mm.
Finally, the grid coordinates and the position intensity values of each grid can be normalized to generate a depth image of the rail surface point cloud;
after the operation is performed on the plurality of rail surface point clouds, a plurality of depth images (a plurality of depth images corresponding to the rail surface to be detected) can be obtained.
And S224, determining a damage area image of the rail surface to be detected based on the plurality of depth images and the rail surface damage identification model trained in advance.
In actual implementation, a plurality of depth images can be input into a rail surface damage recognition model trained in advance to be recognized to obtain a depth image corresponding to rail surface damage; (image of damaged area). Specifically, if abnormal depth images exist in the multiple depth images after the multiple depth images are input into a rail surface damage recognition model trained in advance, a damage area image can be further recognized; the lesion area image may be understood as one or more adjacent meshes in the anomalous depth image.
Step S226, determining the damage length and the damage width of the rail surface to be detected based on the grid row number and the grid column number of each grid and the grid size.
After determining the damaged area image of the rail surface to be detected, the row and column numbers of the grids corresponding to the damaged area image can be acquired, assuming that two adjacent grids corresponding to the damaged area image have row numbers (3, 2) and column numbers (4, 2), and the grid size is 10mm × 10mm, that is, the damaged area image occupies 2 rows and 1 column in total, and the corresponding damaged length of the rail surface to be detected is 10 multiplied by 2 and equal to 20mm, and the damaged width is 10 multiplied by 1 and equal to 10mm.
In actual implementation, assuming that the 8 th depth image corresponding to the rail surface to be detected is an abnormal depth image, the grid row and column coordinates of the two grids corresponding to the damaged area image are 5mm × 5mm, and the sampling step length is 10m, the damaged position of the rail surface to be detected is 75 (70 +5= 75) rail length.
Step S228, based on the first preset algorithm, calculating an X coordinate value of the rail surface center line of the rail surface to be detected.
The first preset algorithm can be a multi-stage edge detection algorithm, and when the first preset algorithm is specifically implemented, firstly, the X coordinates X of the left and right edge points of the rail surface to be detected with obvious features can be extracted from the damaged area image through the multi-stage edge detection algorithm Left And X Right Then X is added Left And X Right The average value of (d) is taken as the X coordinate X of the center line of the rail surface mid 。
And step S230, determining a plurality of target three-dimensional points in the damaged area image based on the X coordinate value of the central line of the rail surface and a third preset threshold value.
In practical implementation, the third preset threshold is usually set to 15mm, and the grid column coordinates in the image of the damaged area may be determined at first in X mid -15mm to X mid A target grid within +15mm, wherein points in the target grid are the target three-dimensional points.
And step S232, acquiring the damage depth of the rail surface to be detected based on the plurality of target three-dimensional points.
Specifically, the step S232 can be implemented by the following steps one to three:
the method comprises the following steps: and fitting the target three-dimensional points into a three-dimensional plane based on a second preset algorithm.
Step two: and calculating the distance value of each target three-dimensional point to the three-dimensional plane.
Step three: and determining the maximum distance value as the damage depth of the rail surface to be detected.
The second preset algorithm may be a random sampling consensus algorithm, specifically, the target three-dimensional point cloud (including a plurality of target three-dimensional points) is fitted to a three-dimensional plane, and finally, the distance from each target three-dimensional point in the target three-dimensional point cloud to the three-dimensional plane and the maximum distance D are calculated max That is, the damage depth of the rail surface is obtained, see a three-dimensional point cloud schematic diagram of a rail surface damage block as shown in fig. 5, and the application is in the followingBesides the plane information of rail surface damage, three-dimensional depth information can be extracted.
In order to better understand the above embodiments, the present application provides a flowchart for detecting the damage depth of the rail surface to be detected, as shown in fig. 6: firstly, rail surface point cloud can be extracted, then the rail surface point cloud is projected into a depth image, and then rail surface damage automatic identification based on a U-net improved model (a rail surface damage identification model trained in advance) is executed, so that a damage area image can be obtained; and then converting the damaged area image into a point cloud (target three-dimensional point cloud), and extracting the depth of the disease (damage) after the point cloud is processed. Specifically, the three-dimensional point cloud (including a plurality of target three-dimensional points) may be inversely calculated according to the target grid points of the damaged area image, wherein an X coordinate of each target three-dimensional point corresponds to a grid column coordinate of the grid where the target three-dimensional point is located, a Y coordinate corresponds to a grid row coordinate, and a Z coordinate corresponds to a position intensity value of the grid where the target three-dimensional point is located.
According to the rail surface damage state detection method, firstly, a collection device is used for collecting rail surface three-dimensional point cloud data, then a normal rail surface state image area (a damage area image) is found through a machine learning or image processing method, the length and width information of rail surface damage is determined, finally, corresponding original point cloud data is found through the area image, damage depth information is calculated after three-dimensional point cloud processing, and rail surface damage length, width and depth information can be found quickly, accurately and comprehensively.
An embodiment of the present invention further provides a rail surface damage state detection device, as shown in fig. 7, the device includes: the acquisition module 70 is used for acquiring a plurality of three-dimensional point clouds corresponding to the rail surface to be detected; the first determining module 71 is configured to determine, based on the multiple three-dimensional point clouds, multiple depth images corresponding to a rail surface to be detected; the second determining module 72 is configured to determine damage position information of the rail surface to be detected based on the multiple depth images and the rail surface damage identification model trained in advance; the third determining module 73 is configured to determine a damage state of the rail surface to be detected according to the damage position information; wherein the damage state comprises: lesion length, lesion width, and lesion depth.
Above-mentioned rail surface damage state detection device includes: acquiring a plurality of three-dimensional point clouds corresponding to a rail surface to be detected; determining a plurality of depth images corresponding to the rail surface to be detected based on the plurality of three-dimensional point clouds; determining a damage area image of the rail surface to be detected based on the plurality of depth images and a pre-trained rail surface damage identification model; determining the damage state of the rail surface to be detected according to the damage area image, wherein the damage state comprises the following steps: lesion length, lesion width and lesion depth. The device determines a damage area image of a rail surface to be detected through a plurality of depth images and a pre-trained rail surface damage identification model; and determining the damage length, the damage width and the damage depth of the rail surface to be detected according to the damage area image, and accurately and quickly detecting the damage length, the damage width and the damage depth of the rail surface.
Further, the obtaining module is further configured to: and acquiring a plurality of three-dimensional point clouds corresponding to the rail surface to be detected according to a preset step length by using an acquisition device to obtain the plurality of three-dimensional point clouds.
Furthermore, each three-dimensional point cloud comprises a plurality of first three-dimensional points and corresponding X, Y and Z coordinate values, and the device further comprises: aiming at each first three-dimensional point, taking the current first three-dimensional point as a center, and calculating the number of second three-dimensional points belonging to a filtering radius range based on a preset filtering radius; if the number of the second three-dimensional points is smaller than a first preset threshold value, filtering the current first three-dimensional points to obtain a plurality of filtered third three-dimensional points; selecting a third three-dimensional point corresponding to the maximum Z coordinate value from the plurality of third three-dimensional points; and searching fourth three-dimensional points respectively corresponding to the Z coordinate values not greater than the second preset threshold value based on the maximum Z coordinate value to obtain a plurality of rail surface point clouds corresponding to the plurality of three-dimensional point clouds.
Further, the first determining module is further configured to; aiming at each rail surface point cloud, acquiring X, Y and Z coordinate values of each fourth three-dimensional point in the rail surface point cloud; selecting the maximum X coordinate value, the maximum Y coordinate value, the minimum X coordinate value and the minimum Y coordinate value from the X, Y and Z coordinate values of each fourth three-dimensional point, and determining the length value and the width value of the depth image corresponding to the rail surface point cloud; and calculating to obtain the grid row number and the grid column number corresponding to each fourth three-dimensional point based on the preset grid size of the depth image and the X and Y coordinate values of each fourth three-dimensional point. And acquiring the number of the fourth three-dimensional points in each grid based on the row number and the column number of the grid. And aiming at each grid, calculating the corresponding position intensity value of the grid based on the number of the fourth three-dimensional points in the grid and the Z coordinate value of each fourth three-dimensional point in the grid. And generating a plurality of depth images corresponding to the rail surface to be detected based on the grid line number and the grid column number of each grid and the position strength value.
Further, the third determining module is further configured to: determining the damage length and the damage width of the rail surface to be detected based on the grid row number and the grid column number of each grid and the grid size; calculating an X coordinate value of a rail surface central line of the rail surface to be detected based on a first preset algorithm; determining a plurality of target three-dimensional points in the damaged area image based on the X coordinate value of the rail surface central line and a third preset threshold; and acquiring the damage depth of the rail surface to be detected based on the plurality of target three-dimensional points.
Further, the third determining module is further configured to: fitting a plurality of target three-dimensional points into a three-dimensional plane based on a second preset algorithm; calculating the distance value between each target three-dimensional point in the point cloud and the three-dimensional plane; and determining the maximum distance value as the damage depth of the rail surface to be detected.
The rail surface damage state detection device provided by the embodiment of the invention has the same realization principle and technical effect as the rail surface damage state detection method embodiment, and the rail surface damage state detection device embodiment can refer to the corresponding contents in the rail surface damage state detection method embodiment.
An embodiment of the present invention further provides an electronic device, which is shown in fig. 8, and the electronic device includes a processor 130 and a memory 131, where the memory 131 stores machine executable instructions that can be executed by the processor 130, and the processor 130 executes the machine executable instructions to implement the rail surface damage state detection method.
Further, the electronic device shown in fig. 8 further includes a bus 132 and a communication interface 133, and the processor 130, the communication interface 133, and the memory 131 are connected through the bus 132.
The Memory 131 may include a high-speed Random Access Memory (RAM) and may also include a non-volatile Memory (non-volatile Memory), such as at least one disk Memory. The communication connection between the network element of the system and at least one other network element is realized through at least one communication interface 133 (which may be wired or wireless), and the internet, a wide area network, a local network, a metropolitan area network, and the like can be used. The bus 132 may be an ISA bus, PCI bus, EISA bus, or the like. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one double-headed arrow is shown in FIG. 8, but that does not indicate only one bus or one type of bus.
The processor 130 may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method may be performed by integrated logic circuits of hardware or instructions in the form of software in the processor 130. The Processor 130 may be a general-purpose Processor, and includes a Central Processing Unit (CPU), a Network Processor (NP), and the like; the device can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, a discrete Gate or transistor logic device, or a discrete hardware component. The various methods, steps and logic blocks disclosed in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in connection with the embodiments of the present invention may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software module may be located in ram, flash memory, rom, prom, or eprom, registers, etc. storage media as is well known in the art. The storage medium is located in the memory 131, and the processor 130 reads the information in the memory 131 and completes the steps of the method of the foregoing embodiment in combination with the hardware thereof.
The embodiment of the present invention further provides a computer-readable storage medium, where the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are called and executed by a processor, the computer-executable instructions cause the processor to implement the rail surface damage state detection method, and specific implementation may refer to method embodiments, and is not described herein again.
The rail surface damage state detection method, the rail surface damage state detection device and the electronic equipment provided by the embodiment of the invention comprise a computer-readable storage medium storing program codes, wherein instructions included in the program codes can be used for executing the method in the previous method embodiment, and specific implementation can refer to the method embodiment and is not described herein again.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. A rail surface damage state detection method is characterized by comprising the following steps:
acquiring a plurality of three-dimensional point clouds corresponding to a rail surface to be detected;
determining a plurality of depth images corresponding to the rail surface to be detected based on the plurality of three-dimensional point clouds;
determining a damage area image of the rail surface to be detected based on the plurality of depth images and a pre-trained rail surface damage identification model;
determining the damage state of the rail surface to be detected according to the damage area image, wherein the damage state comprises the following steps: lesion length, lesion width and lesion depth.
2. The method according to claim 1, wherein the step of obtaining a plurality of three-dimensional point clouds corresponding to the rail surface to be detected comprises;
and acquiring a plurality of three-dimensional point clouds corresponding to the rail surface to be detected according to a preset step length by using an acquisition device to obtain a plurality of three-dimensional point clouds.
3. The method according to claim 2, wherein each three-dimensional point cloud comprises a plurality of first three-dimensional points and corresponding X, Y, and Z coordinate values, and after the step of obtaining a plurality of three-dimensional point clouds corresponding to the rail surface to be detected, the method further comprises:
aiming at each first three-dimensional point, taking the current first three-dimensional point as a center, and calculating the number of second three-dimensional points within the filtering radius range based on a preset filtering radius;
if the number of the second three-dimensional points is smaller than a first preset threshold value, filtering the current first three-dimensional points to obtain a plurality of filtered third three-dimensional points;
selecting a third three-dimensional point corresponding to the maximum Z coordinate value from the plurality of third three-dimensional points;
and searching fourth three-dimensional points respectively corresponding to the Z coordinate values not greater than a second preset threshold value based on the maximum Z coordinate value to obtain a plurality of rail surface point clouds corresponding to the three-dimensional point clouds.
4. The method according to claim 3, wherein the step of determining a plurality of depth images corresponding to the rail surface to be detected based on the plurality of three-dimensional point clouds comprises;
aiming at each rail surface point cloud, acquiring X, Y and Z coordinate values of each fourth three-dimensional point in the rail surface point cloud;
selecting the maximum X coordinate value, the maximum Y coordinate value, the minimum X coordinate value and the minimum Y coordinate value from the X coordinate value, the Y coordinate value and the Z coordinate value of each fourth three-dimensional point, and determining the length value and the width value of the depth image corresponding to the point cloud of the rail surface;
calculating to obtain a grid row number and a grid column number corresponding to each fourth three-dimensional point based on the preset grid size of the depth image and the X and Y coordinate values of each fourth three-dimensional point;
acquiring the number of the fourth three-dimensional points in each grid based on the grid row number and the grid column number;
for each grid, calculating a position intensity value corresponding to the grid based on the number of the fourth three-dimensional points in the grid and the Z coordinate value of each fourth three-dimensional point in the grid;
and generating a plurality of depth images corresponding to the rail surface to be detected based on the grid line number and the grid column number of each grid and the position intensity value.
5. The method according to claim 4, characterized in that the damage state of the rail surface to be detected is determined according to the damage area image, wherein the damage state comprises: the steps of lesion length, lesion width and lesion depth include;
determining the damage length and the damage width of the rail surface to be detected based on the grid row number and the grid column number of each grid and the grid size;
calculating an X coordinate value of the rail surface central line of the rail surface to be detected based on a first preset algorithm;
determining a plurality of target three-dimensional points in the damaged area image based on the X coordinate value of the rail surface central line and a third preset threshold;
and acquiring the damage depth of the rail surface to be detected based on the plurality of target three-dimensional points.
6. The method according to claim 5, wherein the step of obtaining the damage depth of the rail surface to be detected based on the plurality of target three-dimensional points comprises;
fitting a plurality of target three-dimensional points into a three-dimensional plane based on a second preset algorithm;
calculating a distance value from each target three-dimensional point to the three-dimensional plane;
and determining the maximum distance value as the damage depth of the rail surface to be detected.
7. A rail surface damage state detection device, characterized in that the device includes:
the acquisition module is used for acquiring a plurality of three-dimensional point clouds corresponding to the rail surface to be detected;
the first determining module is used for determining a plurality of depth images corresponding to the rail surface to be detected based on the plurality of three-dimensional point clouds;
the second determining module is used for determining a damage area image of the rail surface to be detected based on the plurality of depth images and a pre-trained rail surface damage identification model;
the second determining module is configured to determine a damage state of the rail surface to be detected according to the damage region image, where the damage state includes: lesion length, lesion width and lesion depth.
8. The apparatus of claim 7, wherein the obtaining module is further configured to:
and collecting a plurality of three-dimensional point clouds corresponding to the rail surface to be detected according to a preset step length through a collecting device to obtain a plurality of three-dimensional point clouds.
9. An electronic device, comprising a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement the rail face flaw state detection method of any one of claims 1 to 6.
10. A computer-readable storage medium having stored thereon computer-executable instructions that, when invoked and executed by a processor, cause the processor to implement the rail face damage state detection method of any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211394590.1A CN115690062A (en) | 2022-11-08 | 2022-11-08 | Rail surface damage state detection method and device and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211394590.1A CN115690062A (en) | 2022-11-08 | 2022-11-08 | Rail surface damage state detection method and device and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115690062A true CN115690062A (en) | 2023-02-03 |
Family
ID=85050373
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211394590.1A Pending CN115690062A (en) | 2022-11-08 | 2022-11-08 | Rail surface damage state detection method and device and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115690062A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115830019A (en) * | 2023-02-14 | 2023-03-21 | 南京慧然科技有限公司 | Three-dimensional point cloud calibration processing method and device for steel rail detection |
| CN117197136A (en) * | 2023-11-06 | 2023-12-08 | 中数智科(杭州)科技有限公司 | Straddle type monorail track beam damage detection positioning system, method and storage medium |
-
2022
- 2022-11-08 CN CN202211394590.1A patent/CN115690062A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115830019A (en) * | 2023-02-14 | 2023-03-21 | 南京慧然科技有限公司 | Three-dimensional point cloud calibration processing method and device for steel rail detection |
| CN117197136A (en) * | 2023-11-06 | 2023-12-08 | 中数智科(杭州)科技有限公司 | Straddle type monorail track beam damage detection positioning system, method and storage medium |
| CN117197136B (en) * | 2023-11-06 | 2024-01-26 | 中数智科(杭州)科技有限公司 | Straddle type monorail track beam damage detection positioning system, method and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115690062A (en) | Rail surface damage state detection method and device and electronic equipment | |
| US20210319561A1 (en) | Image segmentation method and system for pavement disease based on deep learning | |
| CN110084116B (en) | Road surface detection method, road surface detection device, computer equipment and storage medium | |
| CN110493816B (en) | Real-time prediction method for rail transit subway station passenger flow volume | |
| CN108986465B (en) | Method, system and terminal equipment for detecting traffic flow | |
| EP3023913A1 (en) | Crack data collection method and crack data collection program | |
| EP3023912A1 (en) | Crack data collection apparatus and server apparatus to collect crack data | |
| CN109489584B (en) | Tunnel clearance detection system and tunnel clearance identification method based on 3D technology | |
| EP2927871A1 (en) | Method and device for calculating number of pedestrians and crowd movement directions | |
| CN107341508B (en) | Fast food picture identification method and system | |
| CN116644373A (en) | Automobile flow data analysis management system based on artificial intelligence | |
| CN113470374A (en) | Vehicle overspeed monitoring method and device, computer equipment and storage medium | |
| CN115511836B (en) | Bridge crack grade assessment method and system based on reinforcement learning algorithm | |
| Zhao et al. | Image-based comprehensive maintenance and inspection method for bridges using deep learning | |
| CN111325292B (en) | Object behavior recognition method and device | |
| CN112036269A (en) | Fall detection method and device, computer equipment and storage medium | |
| CN108596968B (en) | Sleeper counting method based on track 3D depth image | |
| CN106952049A (en) | The state evaluating method and device of transmission line of electricity | |
| CN113643232A (en) | Pavement pit automatic detection method based on binocular camera and convolutional neural network | |
| CN113033443A (en) | Unmanned aerial vehicle-based automatic pedestrian crossing facility whole road network checking method | |
| CN117173647A (en) | Insulator abnormality detection method and device, electronic equipment and storage medium | |
| CN110135382B (en) | Human body detection method and device | |
| CN108846824B (en) | Linear array scanning image sleeper positioning and counting method based on gradient projection | |
| CN103996266A (en) | Tailing reservoir safety monitoring device based on video analysis | |
| CN117058912A (en) | Method and device for detecting abnormal parking of inspection vehicle, storage medium and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |