CN117214172A - Method and device for detecting defects of inner wall of long barrel cylinder and storage medium - Google Patents
Method and device for detecting defects of inner wall of long barrel cylinder and storage medium Download PDFInfo
- Publication number
- CN117214172A CN117214172A CN202310912895.5A CN202310912895A CN117214172A CN 117214172 A CN117214172 A CN 117214172A CN 202310912895 A CN202310912895 A CN 202310912895A CN 117214172 A CN117214172 A CN 117214172A
- Authority
- CN
- China
- Prior art keywords
- defect
- wall
- barrel
- image data
- image
- 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
- 230000007547 defect Effects 0.000 title claims abstract description 172
- 238000000034 method Methods 0.000 title claims abstract description 36
- 238000003860 storage Methods 0.000 title claims abstract description 15
- 238000001514 detection method Methods 0.000 claims abstract description 43
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 30
- 238000012545 processing Methods 0.000 claims abstract description 17
- 238000007781 pre-processing Methods 0.000 claims abstract description 16
- 230000002950 deficient Effects 0.000 claims description 28
- 238000012937 correction Methods 0.000 claims description 14
- 238000012549 training Methods 0.000 claims description 13
- 238000003708 edge detection Methods 0.000 claims description 12
- 238000013507 mapping Methods 0.000 claims description 8
- 238000005520 cutting process Methods 0.000 claims description 6
- 230000001502 supplementing effect Effects 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 4
- 238000009826 distribution Methods 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 4
- 208000032544 Cicatrix Diseases 0.000 claims description 3
- 230000007797 corrosion Effects 0.000 claims description 3
- 238000005260 corrosion Methods 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims description 3
- 231100000241 scar Toxicity 0.000 claims description 3
- 230000037387 scars Effects 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 4
- 238000013519 translation Methods 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 229910052751 metal Inorganic materials 0.000 description 3
- 239000002184 metal Substances 0.000 description 3
- 230000000007 visual effect Effects 0.000 description 3
- 229910052782 aluminium Inorganic materials 0.000 description 2
- XAGFODPZIPBFFR-UHFFFAOYSA-N aluminium Chemical compound [Al] XAGFODPZIPBFFR-UHFFFAOYSA-N 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000011521 glass Substances 0.000 description 2
- 238000005286 illumination Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 241001085205 Prenanthella exigua Species 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 206010015037 epilepsy Diseases 0.000 description 1
- 239000006260 foam Substances 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a method, a device and a storage medium for detecting defects of the inner wall of a long barrel cylinder, and the method comprises the following steps: s1, carrying out full coverage shooting on the inner wall of a long barrel cylinder to be detected by a multi-camera combined image acquisition system, and obtaining an image data set corresponding to the inner wall of the long barrel cylinder to be detected; s2, preprocessing image data in the image data set; carrying out three-dimensional reconstruction of the inner wall of the long barrel cylinder to be detected through a three-dimensional reconstruction model of the inner surface of the barrel to obtain a three-dimensional model of the inner wall of the barrel containing position coordinates; s3, inputting the two-dimensional model of the inner wall surface of the barrel into a defect classification recognition algorithm model for defect recognition processing to obtain all the defect areas recognized in the two-dimensional model of the inner wall surface of the barrel, and obtaining defect area recognition result data. The invention can realize the task of three-dimensional reconstruction of the inner wall of the long barrel and the defect detection operation of the two-dimensional image of the inner wall surface, and has higher accuracy of defect identification and higher accuracy of defect positioning.
Description
Technical Field
The invention relates to the technical field of detection of inner walls of pipes, in particular to a method and a device for detecting defects of inner walls of long-barrel cylinders and a storage medium.
Background
The small-caliber metal long barrel cylinder is widely applied in the fields of industry, military and the like. Whether the surface of the inner wall barrel has micro defect or not is also an important judgment standard for whether the metal long barrel is qualified or not. Therefore, the method for accurately, rapidly and automatically measuring the inner wall surface of the metal long barrel has important significance for improving the production efficiency and accuracy of the pipeline quality and accurately estimating the residual service life. However, at present, the inner wall line size detection is manually detected by mechanical contact measurement methods such as gauges, plug gauges, inner diameter dial indicators and the like, and the process has great limitations in measurement speed, repeatability, accuracy and the like. Therefore, an automatic detection system and method are necessary for equipping a production quality detection stage, and the visual conditions of the three-dimensional appearance and defects of the pt inner wall are reflected.
The Chinese patent application No. CN201510689977.3 discloses an inner wall flaw epilepsy detection device, which comprises a panoramic vision sensor part and a detection analysis system part, wherein the detection part consists of a visible light CCD camera and a laser scanning sensor, and the detection analysis system part consists of a laser analysis unit, a three-dimensional modeling unit, a panoramic image unfolding processing unit and the like. The device provides possibility for comprehensively, accurately and quantitatively recognizing the inner surface condition and change rule of the fire bore, and changes the low-accuracy method of visual estimation in the past. However, the main problem that exists in this kind of technique is that laser sensor module volume is great, to the gun inner wall below 100mm bore, is annular laser emitter or some laser emitter all hardly satisfy the installation detection demand, and two-dimensional image acquisition is inconsistent with the illumination condition that laser three-dimensional reconstruction is required in addition, and single position stationary detection time is longer, leads to whole shooting collection flow loaded down with trivial details, and the practicality is not good enough.
Disclosure of Invention
The invention aims to overcome the technical problem of defect detection of the inner wall of the existing small-caliber long barrel, and provides a method, a device and a storage medium for detecting the defect of the inner wall of a long barrel cylinder.
The aim of the invention is achieved by the following technical scheme:
a method for detecting defects of the inner wall of a long barrel cylinder comprises the following steps:
s1, constructing a multi-phase combined image acquisition system, and performing full coverage shooting on the inner wall of a long barrel cylinder to be detected by the multi-phase combined image acquisition system to obtain an image data set corresponding to the inner wall of the long barrel cylinder to be detected;
s2, preprocessing image data in the image data set; carrying out three-dimensional reconstruction of the inner wall of the long barrel to be detected according to the preprocessed image data set through the three-dimensional reconstruction model of the inner surface of the barrel to obtain a three-dimensional model of the inner wall of the barrel containing position coordinates;
s3, constructing a defect classification recognition algorithm model, inputting a defect image data training sample library into the defect classification recognition algorithm model for model training, wherein defect categories are correspondingly marked in defect image data in the defect image data training sample library, and include scratches, corrosion, cracks, skin warping, pits, scars and defects; the three-dimensional model of the inner wall surface of the barrel is split into a two-dimensional model of the inner wall surface of the barrel, the two-dimensional model of the inner wall surface of the barrel is input into a defect classification and identification algorithm model to carry out defect identification treatment to obtain all defect areas, and the defect areas are obtained according to the following method:
wherein f i (n, m) a set of pixel points representing a defective region i, S i The area of the defective region i pixel is represented, n and m represent the rows and columns of the defective region pixel, N, M represent the total number of rows and total columns occupied by the defective region pixel, and the defective region, the defective position, the defective type, and the defective area are stored as defective region identification result data.
In order to better realize the method for detecting the defect of the inner wall of the long barrel cylinder, the method further comprises the following steps after the step S3:
s4, a channel edge detection module is included in the defect classification recognition algorithm model, contour extraction is conducted on each defect area through the channel edge detection module, a closed contour point set is obtained, minimum circumscribed rectangle scribing is conducted on the closed contour point set corresponding to the defect area, the scribed minimum circumscribed rectangle is correspondingly marked on the defect area, and the defect area or the minimum circumscribed rectangle corresponds to recognition result data for the defect area.
Preferably, the method for detecting the defect of the inner wall of the long barrel cylinder further comprises the following steps:
s5, sequentially cutting the three-dimensional model of the inner wall surface of the barrel into continuous image blocks and recording position information corresponding to the image blocks, and carrying out defect recognition processing on each image block by using a defect classification recognition algorithm model to obtain image block defect result data, wherein the image block defect result data comprises defect types and defect position coordinate data; and the three-dimensional model is correspondingly marked on the surface of the inner wall of the barrel according to the position information of the image block.
Preferably, in step S1, the multi-camera combined image acquisition system includes a mobile carrier and a driving device for driving the mobile carrier to move on the inner wall of the long barrel cylinder to be detected, wherein a plurality of cameras are distributed on the mobile carrier in a circumferential distribution manner, each camera corresponds to a camera shooting area on the inner wall section of the long barrel cylinder to be detected, and all cameras completely cover the inner wall section of the long barrel cylinder to be detected for shooting; the movable bearing frame is also provided with a storage battery and a plurality of light supplementing lamps, and all the light supplementing lamps and all the cameras are respectively and electrically connected with the storage battery.
Preferably, in step S2, the preprocessing method of image data in the image dataset includes:
s21, adjusting brightness values of the image data: setting a brightness threshold value, screening out an image area smaller than the brightness threshold value from the image data, amplifying the image area until the pixels of the amplified image area larger than the brightness threshold value are equal to the pixels smaller than the brightness threshold value, and adjusting the brightness value of the pixels of the amplified image area smaller than the brightness threshold value as follows: selecting k pixel points which are nearest to and higher than the brightness threshold value, calculating the average brightness value of the k pixel points, and adjusting the pixel points smaller than the brightness threshold value to the average brightness value;
s22, splicing the image data in the image data set to obtain a fused image data set, wherein the image data set is image data corresponding to the inner wall of the long barrel cylinder to be detected.
The technical scheme is as follows: step S22 further includes the following distortion correction processing:
and constructing checkerboard calibration paper corresponding to the inner wall of the cylinder of the long barrel to be detected, wherein the checkerboard calibration paper is provided with calibration grids, the calibration grids are provided with cross points, and the corresponding mapping correction processing is carried out on the distortion areas in the image data before splicing through the checkerboard calibration paper.
Preferably, the three-dimensional reconstruction in the step S2 is correspondingly constructed by adopting a point cloud data mode, a coordinate system is arranged in the three-dimensional reconstruction model of the inner surface of the barrel, the three-dimensional reconstruction model of the inner surface of the barrel firstly extracts the preprocessed image data set and generates sparse point cloud data, then the preprocessed image data set is deeply extracted to carry out dense point cloud reconstruction on the basis of the sparse point cloud data, and further the three-dimensional model of the inner wall surface of the barrel containing the position coordinates is obtained.
The multi-phase combined image acquisition system comprises a movable bearing frame and a driving device for driving the movable bearing frame to move on the inner wall of the long barrel cylinder to be detected, wherein a plurality of cameras are distributed on the movable bearing frame in a circumferential manner, and the multi-phase combined image acquisition system performs full coverage shooting on the inner wall of the long barrel cylinder to be detected and obtains an image data set corresponding to the inner wall of the long barrel cylinder to be detected; the defect classification recognition system comprises a preprocessing module, a barrel inner surface three-dimensional reconstruction model and a defect classification recognition algorithm model, wherein the preprocessing module is used for preprocessing brightness value adjustment, splicing and fusion of image data in an image data set, and the barrel inner surface three-dimensional reconstruction model is used for carrying out three-dimensional reconstruction of the inner wall of a long barrel cylinder to be detected according to the preprocessed image data set to obtain a barrel inner wall surface three-dimensional model containing position coordinates; the defect classification recognition algorithm model is used for cutting the three-dimensional model of the inner wall surface of the barrel into a two-dimensional model of the inner wall surface of the barrel, performing defect recognition processing according to the two-dimensional model of the inner wall surface of the barrel to obtain all defect areas, calculating to obtain defect areas corresponding to the defect areas, and correspondingly storing the defect areas, defect positions, defect categories and defect areas as defect area recognition result data; the output module is used for outputting data.
Preferably, the defect classification recognition algorithm model includes a canny edge detection module inside, the canny edge detection module is used for extracting the outline of each defect area and obtaining a closed outline point set, and dividing the closed outline point set corresponding to the defect area by a minimum circumscribed rectangle, and the divided minimum circumscribed rectangle is correspondingly marked on the defect area.
A storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the long barrel cylinder inner wall defect detection method of the present invention.
Compared with the prior art, the invention has the following advantages:
(1) The invention is used for solving the problems of more manual links, high detection difficulty, low detection precision, non-visual results and the like in the existing long barrel inner wall surface detection process, researches are conducted around two aspects of three-dimensional reconstruction and defect detection, an image data set is collected by a multi-camera combined image collection system for pretreatment, a barrel inner wall surface three-dimensional model containing position coordinates is further obtained through three-dimensional reconstruction, defect identification processing is conducted through a defect classification identification algorithm model to obtain all defect areas, and the defect area of each defect area is obtained.
(2) The invention can realize the task of three-dimensional reconstruction of the inner wall and the task of defect detection of the two-dimensional image of the inner surface, the working pipe diameter size of the invention covers the pipe diameter of 76-155mm, the accuracy of defect identification and the accuracy of defect positioning are higher, and important technical support is provided for the standard degree and the service state of the inner wall of the bore.
Drawings
FIG. 1 is a flow chart of a method for detecting defects on the inner wall of a long barrel cylinder according to the present invention;
FIG. 2 is a schematic block diagram of a multi-camera combined image acquisition system in an embodiment;
FIG. 3 is a schematic block diagram of a long barrel cylinder inner wall defect detection device of the present invention;
FIG. 4 is a schematic diagram of a multi-camera combined image acquisition system according to an embodiment;
FIG. 5 is a schematic diagram showing the results of a three-dimensional model of the inner wall surface of a barrel in the example;
FIG. 6 is a schematic diagram showing the image data brightness value adjustment before and after comparison in the embodiment;
FIG. 7 is a schematic diagram of an example checkerboard calibration paper in an embodiment;
FIG. 8 is a schematic diagram of a portion of an image captured in a preprocessed image dataset in an embodiment;
fig. 9 is a schematic drawing illustrating a minimum bounding rectangle corresponding to a defective region in an embodiment.
Wherein, the names corresponding to the reference numerals in the drawings are:
1-moving a bearing frame, 2-cameras, 21-camera shooting areas and 3-detecting the section of the inner wall of the long barrel cylinder.
Detailed Description
The invention is further illustrated by the following examples:
examples
As shown in fig. 1 to 9, a method for detecting defects on the inner wall of a long barrel cylinder includes:
s1, constructing a multi-phase combined image acquisition system, and performing full coverage shooting on the inner wall of the long barrel cylinder to be detected by the multi-phase combined image acquisition system to obtain an image data set corresponding to the inner wall of the long barrel cylinder to be detected. As shown in fig. 2, the multi-camera combined image acquisition system of the embodiment includes a movable carrier 1 and a driving device for driving the movable carrier 1 to move on the inner wall of the long barrel cylinder to be detected, a plurality of cameras 2 are distributed on the movable carrier 1 in a circumferential distribution manner, each camera 2 corresponds to a camera shooting area 21 on the inner wall section 3 of the long barrel cylinder to be detected, and all cameras 2 completely cover the inner wall section 3 of the long barrel cylinder to be detected for shooting. The movable bearing frame 1 is also provided with a storage battery and a plurality of light supplementing lamps (preferably LED lamps), and all the light supplementing lamps and all the cameras 2 are respectively and electrically connected with the storage battery; the data communication mode can adopt a mode of wired communication (connection is carried out through a USB data wire, collected image data is stored to be provided with an exchangeable image file format (EXIF), TIFF/JPG (95% compression ratio) is used for data transmission and real-time processing storage) and wireless communication. The fixed angle of each camera meets the image information that the view angle range and the view field area can cover the 360-degree range of the inner wall of the barrel, as shown in fig. 4, the multi-camera combined image acquisition system adopts eight five-megapixel FPC flexible flat cable industrial board camera modules, the optical axis of each camera is kept vertical to the inner surface, and the horizontal collimation of the x and y directions of the images in a focal plane can be ensured at a working distance of 40 mm; the anti-vibration foam can be arranged at the outer edge of each camera lens, and the optical filter and the light-transmitting glass are sequentially overlapped and nested on the camera from bottom to top, so that the camera cover plate is aligned with the cylindrical fixing screw holes to be synthesized into a whole outer cover for fixing the light-transmitting glass. The LED lamp strips are sequentially connected in parallel and encircling the side surfaces of the cover plates corresponding to the eight cameras in a wiring connection mode, are perpendicular to the plane of the camera lens and are fixed on the inner side of the aluminum photo frame; the LED lamp strip power line and the camera USB transmission line share one side to penetrate through staggered wiring holes and externally connect to one side of the development board to supply power; the bright white light emitted from the strip lights the image capture area of each camera lens by diffuse reflection on a black aluminum frame.
S2, preprocessing the image data in the image data set. And carrying out three-dimensional reconstruction of the inner wall of the long barrel to be detected according to the preprocessed image data set through the three-dimensional reconstruction model of the inner surface of the barrel to obtain a three-dimensional model of the inner wall of the barrel containing position coordinates (shown in figure 5). The two-dimensional images acquired by the multi-camera combination image acquisition system come from eight camera lens modules with incomplete acquisition contents, the whole images lack ordering information of corresponding shooting position relations, and particularly, the sequence image position relations corresponding to known camera parameters need to be referred to when the three-dimensional reconstruction of the two-dimensional images is carried out; therefore, the multi-camera combined image acquisition system needs to save a series of long barrel two-dimensional images in a serial naming manner before being transferred to a subsequent three-dimensional reconstruction unit.
In some embodiments, the three-dimensional reconstruction is correspondingly constructed by adopting a point cloud data mode, a coordinate system is arranged in a three-dimensional reconstruction model of the inner surface of the barrel, the three-dimensional reconstruction model of the inner surface of the barrel firstly extracts a preprocessed image data set and generates sparse point cloud data, then the preprocessed image data set is deeply extracted to carry out dense point cloud reconstruction (which also assists in triangularization and texture mapping) on the basis of the sparse point cloud data, and further the three-dimensional model of the inner surface of the barrel with position coordinates (which can reflect the morphological characteristics of the inner surface of the barrel) is obtained.
In some embodiments, the three-dimensional point cloud model reconstruction process is: extracting feature points of any image by using the obtained omnidirectional image sequence, obtaining a description matrix corresponding to any image, determining a matching pair according to the distance relation between the description moments of the images, and obtaining the rotation R and the translation t of the image frame according to the feature points on the matching pair of the images; according to the rotation R and the translation t of the images of the adjacent frames, obtaining a reprojection error of the matched pair, and optimizing the rotation R and the translation t; taking the optimized rotation R and translation t as input, extracting edge points of the image frames and optimizing luminosity errors, so as to match the edge points; and carrying out triangulation on the matched edge points to obtain dense three-dimensional point clouds so as to realize three-dimensional reconstruction.
In some embodiments, corresponding feature point acquisition and tracking is constrained to a local camera range of motion by employing SIFT algorithms. And extracting the characteristic points of the rest cameras in sequence, combining pose information specified by known parameters of the cameras, accurately optimizing the motion process tracks reflected by eight cameras of the whole multi-camera combined image acquisition system, searching and finding matched adjacent characteristic points through matching characteristic points on the overlapping edges of the multi-camera images on the basis of the existing sparse point cloud, and generating seed patches with the size of 5 multiplied by 5 pixels. And taking the connecting line of the searched feature points and the camera center as the normal vector of the seed patch, and calculating the value screening feature points of the normalized cross-correlation principle NCC. The magnitude of the NCC value represents the magnitude of the correlation of the feature points, and if the magnitude of the NCC value is too small, the feature points are removed if the correlation of the feature points is small. The patch is then expanded and a three-dimensional patch is projected onto the image, which is copied to an adjacent location if the depth is continuous but there is no patch projection. And finally, removing some adjacent patches in the projection space and non-adjacent patches in the space adjacent to each other to obtain dense point cloud reconstruction. Finally, the texture features are sampled at intervals by utilizing the correlation similarity of the continuous texture features in the sequence images, so that the interval texture features are obtained; and mapping the texture image by using the interval texture features to obtain the three-dimensional model of the inner surface of the barrel.
S3, constructing a defect classification recognition algorithm model, inputting a defect image data training sample library into the defect classification recognition algorithm model to carry out model training, wherein defect categories are correspondingly marked in defect image data in the defect image data training sample library, and include scratches, corrosion, cracks, skin warping, pits, scars and defects. The three-dimensional model of the inner wall surface of the barrel is split into a two-dimensional model of the inner wall surface of the barrel, the two-dimensional model of the inner wall surface of the barrel is input into a defect classification and identification algorithm model to carry out defect identification treatment to obtain all defect areas, and the defect areas are obtained according to the following method:
wherein f i (n, m) a set of pixel points representing a defective region i, S i The area of the defective region i pixel is represented, n and m represent the rows and columns of the defective region pixel, N, M represent the total number of rows and total columns occupied by the defective region pixel, and the defective region, the defective position, the defective type, and the defective area are stored as defective region identification result data.
The algorithm of the defect classification recognition algorithm model comprises the following steps: target recognition based on strong supervised learning and abnormal sample detection based on unsupervised learning.
As an alternative embodiment, the target recognition step based on the strong supervised learning is as follows: firstly, collecting a certain number of plane defect pictures, and then marking the positions of the plane defects to obtain a training set; secondly, training the training set by using a YOLO target detection network to obtain a detection model; finally, the detection model is used for two-dimensional defect detection of the plane expansion image.
As an alternative embodiment, the abnormal sample detection step based on the unsupervised learning is as follows: firstly, a preset unsupervised self-encoder is utilized to manufacture a pseudo tag for the non-tag barrel inner wall detection data in the initial defect detection data set, and a target detection data set is generated; training a preset classifier according to the existing detection data set to obtain a trained target classifier; and when the data to be detected is received, carrying out anomaly detection on the data to be detected by utilizing the target classifier to obtain a defect detection result of the abnormal sample.
The defect classification recognition algorithm model is used for calculating the type information, the size information, the area of a defect area, the minimum circumscribed rectangular range, the outline size length and the like of the defect of the inner wall of the long barrel, transmitting the information to an image acquisition control processing analysis interface of a software end, and generating a defect detection result.
In some embodiments, a method of preprocessing image data in an image dataset includes:
s21, adjusting brightness value of the image data (because different areas in the image acquired by the camera have obvious brightness difference, and the area with overlarge brightness or low brightness is easy to cause false detection in the process of defect identification, aiming at the problems, the embodiment of the invention also provides an image brightness correction method, which combines the characteristics of gamma correction and linear correction to carry out brightness correction on the input image to obtain an image with balanced brightness so as to repair the phenomenon of unbalanced brightness): setting a brightness threshold value, screening out an image area smaller than the brightness threshold value from the image data, amplifying the image area until the pixels of the amplified image area larger than the brightness threshold value are equal to the pixels smaller than the brightness threshold value, and adjusting the brightness value of the pixels of the amplified image area smaller than the brightness threshold value as follows: and selecting k pixel points which are closest to the brightness threshold value and are higher than the brightness threshold value, calculating the average brightness value of the k pixel points, and adjusting the pixel points which are smaller than the brightness threshold value to the average brightness value. FIG. 6 is a graph showing the contrast of the effect of image brightness correction according to an embodiment of the present invention; the original image on the left side of fig. 6 is an original image on which the entire brightness correction of the image is not performed, and the reference image on the right side of fig. 6 is an effect image using the image brightness correction method of the present embodiment. As can be seen by comparison, the luminance value of the whole image area of the right comparison chart of fig. 6 is significantly improved compared with the luminance value of the left original chart of fig. 6.
In some embodiments, the pixels of each region may be sampled by calculating the average brightness value of each region, obtaining the brightness value of the input image including the over-dark region, the normal illumination region, and the over-bright region, and then calculating the average brightness of the sample, and defining the optimal target brightness required for detection. Specifically, if an image of an area has N (N is greater than or equal to 1 and is an integer), sampling the N pixel points at an interval N (N is greater than or equal to 1 and is an integer), calculating a brightness value of each sample pixel point, and averaging the brightness values to obtain an average brightness value of the area.
S22, splicing the image data (shown in fig. 8) in the image data set to obtain a fused image data set, wherein the image data set is image data corresponding to the inner wall of the long barrel cylinder to be detected.
Further, the preferred distortion correction (the inner wall surface photographed by the camera is a curved object rather than a horizontal plane, and the curved object of the inner surface presents a significant barrel distortion on the image compared with the planar object, so that the internal image needs to be unfolded into the planar image, and according to the mapping relationship, the actual size corresponding to each pixel point in the image is ensured to be unchanged) processing technical scheme is as follows:
and constructing checkerboard calibration paper corresponding to the inner wall of the long barrel cylinder to be detected, wherein the checkerboard calibration paper is provided with calibration grids, the calibration grids are provided with cross points, and the corresponding mapping correction processing is carried out on the distortion areas in the image data before splicing through the checkerboard calibration paper.
Sticking checkerboard calibration paper on the inner surface of the barrel, so that the checkerboard calibration paper covers the range of a camera acquisition view field, the checkerboard calibration paper in a square lattice form is shown in fig. 7, and the side lengths of all black and white lattice points are the same; eight cameras respectively acquire barrel-shaped distortion images of checkerboard calibration paper in the corresponding view field area; calculating the coordinates of each intersection point in the barrel distortion image of each piece of checkerboard calibration paper; calculating the mapping relation between barrel distortion and a plane image according to the coordinates of each intersection point in the barrel distortion image of the checkerboard calibration paper and the coordinates of each intersection point in the checkerboard calibration paper; each camera shoots a barrel-shaped distortion image of the inner surface of the barrel after the checkerboard calibration paper is removed; and converting the barrel-shaped distortion image of the inner surface of the barrel into a plane image of the inner surface of the barrel according to the distortion correction corresponding mapping relation obtained by priori pre-calculation.
In some embodiments, the following method is further included after step S3:
s4, a channel edge detection module is included in the defect classification recognition algorithm model, contour extraction is conducted on each defect area through the channel edge detection module, a closed contour point set is obtained, minimum circumscribed rectangle scribing is conducted on the closed contour point set corresponding to the defect area, the scribed minimum circumscribed rectangle is correspondingly marked on the defect area, and as shown in FIG. 9, the defect area or the minimum circumscribed rectangle corresponds to recognition result data for the defect area.
In some embodiments, the invention further comprises the following method:
s5, sequentially cutting the three-dimensional model of the inner wall surface of the barrel into continuous image blocks, recording position information corresponding to the image blocks, and carrying out defect recognition processing on each image block by using a defect classification recognition algorithm model to obtain image block defect result data, wherein the image block defect result data comprises defect types and defect position coordinate data. And the three-dimensional model is correspondingly marked on the surface of the inner wall of the barrel according to the position information of the image block.
The utility model provides a long barrel cylinder inner wall defect detection device, includes that polyphase unit makes up image acquisition system, defect classification system and output module, polyphase unit makes up image acquisition system including removing carrier 1 and drive and remove the drive arrangement who carries carrier 1 and be detecting long barrel cylinder inner wall motion, is circumference distribution on removing carrier 1 and has laid a plurality of camera 2, polyphase unit makes up image acquisition system and carries out full coverage shooting and obtain the image dataset that the long barrel cylinder inner wall that waits to detect corresponds. The defect classification recognition system comprises a preprocessing module, a barrel inner surface three-dimensional reconstruction model and a defect classification recognition algorithm model, wherein the preprocessing module is used for preprocessing brightness value adjustment, splicing and fusion of image data in an image data set, and the barrel inner surface three-dimensional reconstruction model is used for carrying out three-dimensional reconstruction of the inner wall of the long barrel cylinder to be detected according to the preprocessed image data set to obtain a barrel inner wall surface three-dimensional model containing position coordinates. The defect classification recognition algorithm model is used for cutting the three-dimensional model of the inner wall surface of the barrel into a two-dimensional model of the inner wall surface of the barrel, performing defect recognition processing according to the two-dimensional model of the inner wall surface of the barrel to obtain all defect areas, calculating to obtain defect areas corresponding to the defect areas, and storing the defect areas, defect positions, defect types and defect areas as defect area recognition result data. The output module is used for outputting data.
The defect classification recognition algorithm model comprises a canny edge detection module, wherein the canny edge detection module is used for carrying out contour extraction on each defect region to obtain a closed contour point set, carrying out minimum circumscribed rectangle division on the closed contour point set corresponding to the defect region, and correspondingly marking the divided minimum circumscribed rectangle on the defect region.
A storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the long barrel cylinder inner wall defect detection method of the present invention.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.
Claims (10)
1. A method for detecting defects of the inner wall of a long barrel cylinder is characterized by comprising the following steps of: the method comprises the following steps:
s1, constructing a multi-phase combined image acquisition system, and performing full coverage shooting on the inner wall of a long barrel cylinder to be detected by the multi-phase combined image acquisition system to obtain an image data set corresponding to the inner wall of the long barrel cylinder to be detected;
s2, preprocessing image data in the image data set; carrying out three-dimensional reconstruction of the inner wall of the long barrel to be detected according to the preprocessed image data set through the three-dimensional reconstruction model of the inner surface of the barrel to obtain a three-dimensional model of the inner wall of the barrel containing position coordinates;
s3, constructing a defect classification recognition algorithm model, inputting a defect image data training sample library into the defect classification recognition algorithm model for model training, wherein defect categories are correspondingly marked in defect image data in the defect image data training sample library, and include scratches, corrosion, cracks, skin warping, pits, scars and defects; the three-dimensional model of the inner wall surface of the barrel is split into a two-dimensional model of the inner wall surface of the barrel, the two-dimensional model of the inner wall surface of the barrel is input into a defect classification and identification algorithm model to carry out defect identification treatment to obtain all defect areas, and the defect areas are obtained according to the following method:
wherein f i (n, m) a set of pixel points representing a defective region i, S i The area of the defective region i pixel is represented, n and m represent the rows and columns of the defective region pixel, N, M represent the total number of rows and total columns occupied by the defective region pixel, and the defective region, the defective position, the defective type, and the defective area are stored as defective region identification result data.
2. The long barrel cylinder inner wall defect detection method according to claim 1, wherein: the method further comprises the following steps after the step S3:
s4, a channel edge detection module is included in the defect classification recognition algorithm model, contour extraction is conducted on each defect area through the channel edge detection module, a closed contour point set is obtained, minimum circumscribed rectangle scribing is conducted on the closed contour point set corresponding to the defect area, the scribed minimum circumscribed rectangle is correspondingly marked on the defect area, and the defect area or the minimum circumscribed rectangle corresponds to recognition result data for the defect area.
3. The long barrel cylinder inner wall defect detection method according to claim 1, wherein: the method also comprises the following steps:
s5, sequentially cutting the three-dimensional model of the inner wall surface of the barrel into continuous image blocks and recording position information corresponding to the image blocks, and carrying out defect recognition processing on each image block by using a defect classification recognition algorithm model to obtain image block defect result data, wherein the image block defect result data comprises defect types and defect position coordinate data; and the three-dimensional model is correspondingly marked on the surface of the inner wall of the barrel according to the position information of the image block.
4. The long barrel cylinder inner wall defect detection method according to claim 1, wherein: in the step S1, the multi-camera combined image acquisition system comprises a movable bearing frame (1) and a driving device for driving the movable bearing frame (1) to move on the inner wall of the long barrel cylinder to be detected, wherein a plurality of cameras (2) are distributed on the movable bearing frame (1) in a circumferential manner, each camera (2) corresponds to a camera shooting area (21) on the inner wall section (3) of the long barrel cylinder to be detected, and all cameras (2) completely cover the inner wall section (3) of the long barrel cylinder to be detected for shooting; the movable bearing frame (1) is also provided with a storage battery and a plurality of light supplementing lamps, and all the light supplementing lamps and all the cameras (2) are respectively and electrically connected with the storage battery.
5. The long barrel cylinder inner wall defect detection method according to claim 1, wherein: in step S2, the preprocessing method of image data in the image dataset includes:
s21, adjusting brightness values of the image data: setting a brightness threshold value, screening out an image area smaller than the brightness threshold value from the image data, amplifying the image area until the pixels of the amplified image area larger than the brightness threshold value are equal to the pixels smaller than the brightness threshold value, and adjusting the brightness value of the pixels of the amplified image area smaller than the brightness threshold value as follows: selecting k pixel points which are nearest to and higher than the brightness threshold value, calculating the average brightness value of the k pixel points, and adjusting the pixel points smaller than the brightness threshold value to the average brightness value;
s22, splicing the image data in the image data set to obtain a fused image data set, wherein the image data set is image data corresponding to the inner wall of the long barrel cylinder to be detected.
6. The long barrel cylinder inner wall defect detection method according to claim 5, wherein: step S22 further includes the following distortion correction processing:
and constructing checkerboard calibration paper corresponding to the inner wall of the long barrel cylinder to be detected, wherein the checkerboard calibration paper is provided with calibration grids, the calibration grids are provided with cross points, and the corresponding mapping correction processing is carried out on the distortion areas in the image data before splicing through the checkerboard calibration paper.
7. The long barrel cylinder inner wall defect detection method according to claim 1, wherein: the three-dimensional reconstruction of the step S2 is correspondingly constructed in a point cloud data mode, a coordinate system is arranged in a three-dimensional reconstruction model of the inner surface of the barrel, the three-dimensional reconstruction model of the inner surface of the barrel firstly extracts a preprocessed image data set and generates sparse point cloud data, then the preprocessed image data set is deeply extracted to carry out dense point cloud reconstruction on the basis of the sparse point cloud data, and further the three-dimensional model of the inner surface of the barrel containing position coordinates is obtained.
8. The utility model provides a long barrel cylinder inner wall defect detection device which characterized in that: the system comprises a multiphase combined image acquisition system, a defect classification and identification system and an output module, wherein the multiphase combined image acquisition system comprises a movable bearing frame (1) and a driving device for driving the movable bearing frame (1) to move on the inner wall of a long barrel cylinder to be detected, a plurality of cameras (2) are distributed on the movable bearing frame (1) in a circumferential distribution manner, and the multiphase combined image acquisition system performs full coverage shooting on the inner wall of the long barrel cylinder to be detected and obtains an image data set corresponding to the inner wall of the long barrel cylinder to be detected; the defect classification recognition system comprises a preprocessing module, a barrel inner surface three-dimensional reconstruction model and a defect classification recognition algorithm model, wherein the preprocessing module is used for preprocessing brightness value adjustment, splicing and fusion of image data in an image data set, and the barrel inner surface three-dimensional reconstruction model is used for carrying out three-dimensional reconstruction of the inner wall of a long barrel cylinder to be detected according to the preprocessed image data set to obtain a barrel inner wall surface three-dimensional model containing position coordinates; the defect classification recognition algorithm model is used for cutting the three-dimensional model of the inner wall surface of the barrel into a two-dimensional model of the inner wall surface of the barrel, performing defect recognition processing according to the two-dimensional model of the inner wall surface of the barrel to obtain all defect areas, calculating to obtain defect areas corresponding to the defect areas, and correspondingly storing the defect areas, defect positions, defect categories and defect areas as defect area recognition result data; the output module is used for outputting data.
9. The long barrel cylinder inner wall defect detecting apparatus according to claim 8, wherein: the defect classification recognition algorithm model comprises a canny edge detection module, wherein the canny edge detection module is used for carrying out contour extraction on each defect region to obtain a closed contour point set, carrying out minimum circumscribed rectangle division on the closed contour point set corresponding to the defect region, and correspondingly marking the divided minimum circumscribed rectangle on the defect region.
10. A storage medium having a computer program stored thereon, characterized by: the computer program implementing the steps of the method according to any of claims 1 to 7 when executed by a processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310912895.5A CN117214172A (en) | 2023-07-24 | 2023-07-24 | Method and device for detecting defects of inner wall of long barrel cylinder and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310912895.5A CN117214172A (en) | 2023-07-24 | 2023-07-24 | Method and device for detecting defects of inner wall of long barrel cylinder and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117214172A true CN117214172A (en) | 2023-12-12 |
Family
ID=89039606
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310912895.5A Pending CN117214172A (en) | 2023-07-24 | 2023-07-24 | Method and device for detecting defects of inner wall of long barrel cylinder and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117214172A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117571739A (en) * | 2024-01-16 | 2024-02-20 | 中国人民解放军陆军装甲兵学院 | Pipe wall ablation abrasion degree assessment method based on intelligent algorithm |
| CN118130501A (en) * | 2024-05-10 | 2024-06-04 | 深圳玩智商科技有限公司 | Defect detection method and device applied to inner surface of column |
| CN118130501B (en) * | 2024-05-10 | 2024-07-26 | 深圳玩智商科技有限公司 | Defect detection method and device applied to inner surface of column |
-
2023
- 2023-07-24 CN CN202310912895.5A patent/CN117214172A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117571739A (en) * | 2024-01-16 | 2024-02-20 | 中国人民解放军陆军装甲兵学院 | Pipe wall ablation abrasion degree assessment method based on intelligent algorithm |
| CN117571739B (en) * | 2024-01-16 | 2024-03-12 | 中国人民解放军陆军装甲兵学院 | Pipe wall ablation abrasion degree assessment method based on intelligent algorithm |
| CN118130501A (en) * | 2024-05-10 | 2024-06-04 | 深圳玩智商科技有限公司 | Defect detection method and device applied to inner surface of column |
| CN118130501B (en) * | 2024-05-10 | 2024-07-26 | 深圳玩智商科技有限公司 | Defect detection method and device applied to inner surface of column |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109544679B (en) | Three-dimensional reconstruction method for inner wall of pipeline | |
| JP6289564B2 (en) | Method, apparatus and computer readable medium for detecting changes to structures | |
| CN111047568B (en) | Method and system for detecting and identifying steam leakage defect | |
| US10930013B2 (en) | Method and system for calibrating imaging system | |
| CN104048969A (en) | Tunnel defect recognition method | |
| CN117214172A (en) | Method and device for detecting defects of inner wall of long barrel cylinder and storage medium | |
| US11567207B2 (en) | Apparatus and method for efficient point cloud feature extraction and segmentation framework | |
| CN108665468B (en) | Device and method for extracting tangent tower insulator string | |
| CN112687231B (en) | Brightness and chrominance data extraction method, equipment and computer readable storage medium | |
| US9204130B2 (en) | Method and system for creating a three dimensional representation of an object | |
| CN114280075A (en) | Online visual inspection system and method for surface defects of pipe parts | |
| CN115639248A (en) | System and method for detecting quality of building outer wall | |
| CN114119614A (en) | Method for remotely detecting cracks of building | |
| CN117308881A (en) | Detection device and detection method for inclination angle of tunnel inspection and repair cover plate | |
| CN111260735B (en) | External parameter calibration method for single-shot LIDAR and panoramic camera | |
| CN116363302B (en) | Pipeline three-dimensional reconstruction and pit quantification method based on multi-view geometry | |
| CN116883483A (en) | Fish body measuring method based on laser camera system | |
| CN114441450B (en) | Multi-metal nodule typical characteristic simulation analysis device and method | |
| CN112798608B (en) | Optical detection device and optical detection method for side wall of inner cavity of mobile phone camera support | |
| CN114529808A (en) | Pipeline detection panoramic shooting processing method | |
| CN110991387B (en) | Distributed processing method and system for robot cluster image recognition | |
| CN113655066A (en) | Device, system and method for detecting damage of net cage | |
| CN211086121U (en) | Optical imaging module for tool three-dimensional cutting edge chipping detection device | |
| Du et al. | Mosaicking of mountain tunnel images guided by laser rangefinder | |
| CN113808117A (en) | Lamp detection method, device, equipment and storage medium |
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 |