CN113916892A - Multi-view vision-based brake disc gluing defect detection device and method - Google Patents
Multi-view vision-based brake disc gluing defect detection device and method Download PDFInfo
- Publication number
- CN113916892A CN113916892A CN202111146179.8A CN202111146179A CN113916892A CN 113916892 A CN113916892 A CN 113916892A CN 202111146179 A CN202111146179 A CN 202111146179A CN 113916892 A CN113916892 A CN 113916892A
- Authority
- CN
- China
- Prior art keywords
- defect
- brake disc
- pixel
- pixel value
- gradient
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Image Processing (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention discloses a brake disc gluing defect detection device and method based on multi-view vision, wherein the detection device consists of a checkerboard calibration plate, an illumination module, an imaging module and a main control unit; the checkerboard calibration board is used for solving a transformation matrix among the plurality of cameras; the illumination module is used for providing illumination for detection; the imaging module is used for acquiring a brake disc image required to be detected and filtering the influence of natural light, and is used for adjusting the position and the angle of the camera through the bracket; the main control unit is used for controlling the lighting module and the imaging module, detecting and positioning defects of collected images, communicating with the PLC through the IO card, receiving a signal to be detected, transmitting a detection result and further controlling production on a production line. The invention realizes the online automatic detection of the gluing defect of the brake disc on the production line, makes up the defects of manual inspection, has high defect identification precision, low missing detection and false detection rate and high detection speed, and ensures the gluing quality and the production efficiency of the brake disc.
Description
Technical Field
The invention relates to a brake disc gluing defect detection technology, in particular to a brake disc gluing defect detection device and method based on multi-view vision.
Background
Among various parts of an automobile, the importance of a brake disc is self-evident, and the brake disc is also called a brake disc, which can directly show the good or bad braking effect of the automobile. In the manufacturing process of the brake disc, special colloid needs to be coated on the specific position at present to prepare for adding friction materials subsequently, so that the quality of the coated colloid has a crucial influence on the final quality of the brake disc.
At present, the detection of the gluing quality of a brake disc in a factory mostly depends on human eyes, and the glue adopted in the gluing process is special transparent glue, so that the contrast between the glued part and the non-glued part is extremely low under the condition of natural light, the glue is difficult to observe and consumes energy, and the detection precision is lower along with the extension of the working time of workers; and the brake disc needs to be taken down from the production line for convenient observation, and the process not only consumes time, but also easily causes irreparable damage to the coated colloid. Therefore, the traditional manual detection scheme is not suitable for rapid modern industrial production due to the fact that the labor investment is large, the repeated labor intensity is high, the detection period is long, and the production efficiency is low.
Disclosure of Invention
The invention mainly aims to provide a brake disc gluing defect detection device and method based on multi-view vision, overcomes the defect of manual detection, improves the detection precision and the detection speed, and is suitable for modern assembly line production.
The technical solution for realizing the purpose of the invention is as follows: a brake disc gluing defect detection device based on multi-view vision is composed of a chessboard format calibration plate, an illumination module, an imaging module and a main control unit;
asymmetric circular ring characteristic points are arranged at the corners of the chessboard pattern calibration plate and are used for calculating transformation matrixes among the cameras;
the lighting module is used for providing lighting conditions for the detection device;
the imaging module is used for acquiring a brake disc image to be detected;
the main control unit is used for controlling the lighting module and the imaging module, detecting and positioning defects of the acquired images through the image processing step, communicating with the PLC through the IO card, receiving the signals to be detected and transmitting the detection result.
Furthermore, five asymmetric circular ring characteristic points are distributed on four corners of the chessboard grid calibration plate and are used for judging the direction during calibration, so that the characteristic points extracted at different angles can be arranged according to the same sequence; each of the checkerboards is a square black-white checkerboard, and the number of the grid points is 8x 8.
Further, the illumination module is composed of 16 infrared LED light sources with the wavelength of 850nm and a light source controller and is used for illumination during detection.
Furthermore, the imaging module consists of 4 black-and-white industrial cameras, an infrared filter and a camera bracket; selecting a lens with a focal length of 8mm, and matching with an infrared filter with a wavelength of 850 nm; the camera support is used for installing 4 black and white industrial cameras, the main camera is in the center, and the auxiliary cameras are distributed around the main camera.
Further, the main control unit can operate on a UI interface and is composed of two parts, namely signal control and image processing: the signal control comprises the control of a camera signal, the control of a light source signal and the operation of a communication control assembly line through an IO card and a PLC, wherein the communication comprises the steps that the PLC sends a signal to be detected to a master control program and the master control signal sends a detection result signal to the PLC; the UI interface is used for detecting result access, signal control and image processing algorithm parameter modification.
A brake disc gluing defect detection method based on multi-view vision comprises the following steps:
(1) a calibration process: extracting angular points of calibration plate images acquired by the four cameras; extracting 8x8 feature points according to a certain sequence by taking the corner with two circular feature points as the third corner in the clockwise direction; constructing a standard dot matrix of 8x8 as a standard reference space according to a calibration plate, and solving an imaging perspective transformation matrix H from four cameras to the standard space;
(2) processing the collected images of the brake disc according to the perspective transformation matrix calculated in the step (1), and mapping the images to a standard space;
(3) and (4) performing the same operation as the operation (4) to (6) on the four images to obtain a black and white segmentation image with defects and non-defects on the target detection area, wherein the pixel value 255 represents the defects, and the pixel value 0 represents the non-defects.
(4) Positioning the brake disc according to the inherent circular hole characteristics on the brake disc to obtain an annular region ROI required to be detected;
(5) performing regional adaptive histogram analysis on the ROI to obtain the region where the abnormal pixel value is located;
(6) performing gradient histogram analysis on the ROI to obtain a region contained in the abnormal boundary;
(7) performing phase comparison on the black and white segmentation images obtained in the step (5) and the step (6) to obtain a possible defect area on each image;
(8) and (4) giving different weights to the defect map obtained by the main camera and the defect map obtained by the auxiliary camera in the step (7), performing fusion analysis on the defect area to obtain a final defect map, and marking the final defect map on the original map.
Further, the step (4) is specifically as follows:
(41) performing morphological gradient processing on the image to obtain a gradient image;
(42) searching the contour of the gradient map, and screening out the specific contour on the brake disc according to the perimeter of the contour and the length-width difference of the circumscribed rectangle;
(43) performing circular ring fitting according to the central point group of the screened contour, and simultaneously removing abnormal values by using RANSAC; and obtaining the radius of the center of the brake disc and the radius of the ring where the round hole is located through ring fitting, and obtaining the annular MASK of the target area by the center and the radius and comparing the annular MASK with the original image to obtain the ROI.
Further, the step (5) is specifically as follows:
(51) calculating the angle between each point on the ROI and the center point of the brake disc, and dividing the ROI into specific blocks according to the angles to perform (52) partition histogram analysis;
(52) and performing histogram analysis on the partitions, finding a second peak value area formed in a high pixel value range due to the defect of no gluing according to the gradient change condition of the histogram, and marking pixel points of the pixel values in the area as defects.
Further, the step (6) is specifically as follows:
(61) performing morphological gradient processing on the ROI;
(62) performing histogram analysis on the gradient map to obtain a gradient histogram, obtaining a peak value of the histogram, taking a certain proportion according to the peak value, and finding a first gradient value on the histogram, wherein the number of the first gradient values is smaller than the proportion, and the gradient value is larger than the gradient value corresponding to the peak value;
(63) taking the gradient value found in the step (62) as a threshold value, carrying out binarization on the gradient map, searching a contour, and determining an abnormal boundary;
(64) the inside of the obtained abnormal boundary is filled (63) as a defective region.
Further, the step (8) is specifically as follows:
(81) performing phase comparison on the four black-and-white images to obtain a black-and-white image which is necessary to be defective, wherein the defective pixel value is recorded as 255, and the non-defective pixel value is recorded as 0;
(82) subtracting the defect map marked by the main camera from the image obtained in the step (81) to obtain a suspicious defect mark map, and recording the suspicious defect pixel value as 85; subtracting the defect map marked by the three auxiliary cameras from the image obtained in the step (81) to obtain a reference defect mark map, wherein the pixel value of the marked area is marked as 51;
(83) adding the four images obtained in the step (82) and adding the four images to the determined image obtained in the step (81) to obtain a combined image; the pixel value is 0 to represent non-defective pixel points, the pixel value is 51 to represent a pixel point which is considered as a defect by one auxiliary camera, the pixel value is 102 to represent a pixel point which is considered as a defect by two auxiliary cameras, the pixel value is 153 to represent a pixel point which is considered as a defect by three auxiliary cameras, the pixel value is 85 to represent a suspicious defect pixel point to be determined, the pixel value is 136 to represent a pixel point which is considered as a defect by the main camera and one auxiliary camera, the pixel value is 187 to represent a pixel point which is considered as a defect by the main camera and two auxiliary cameras, and the pixel value is 255 to represent a pixel point which is considered as a defect by the main camera and three auxiliary cameras, namely a pixel point which must be a defect;
(84) and taking the point with the pixel value of 255 as a seed point, optionally taking the value between 69 and 118 as a threshold value, and performing flood filling on the combined graph to obtain a final defect graph, wherein the filled area is a pixel point which is adjacent to the pixel point which must be defective and is considered to be defective by two auxiliary cameras, and in the flood filling process, an iterative process exists, namely the filled area can be also taken as the pixel point which must be defective, and the adjacent pixel point which meets the condition can be continuously filled.
Compared with the prior art, the invention has the beneficial effects that:
(1) according to the invention, the negative influence on detection caused by the difference of the imaging brightness of the brake disc due to uneven illumination is effectively relieved by adopting the partition histogram analysis.
(2) The threshold value selection of the defect area and the gluing area is carried out by adopting a self-adaptive analysis method, and the robustness is provided for various illumination conditions.
(3) By adopting multi-camera detection, the negative influence of reflection on detection is effectively suppressed, and the omission ratio is reduced by combining a multi-camera analysis result weight fusion method.
Drawings
Fig. 1 is a schematic diagram of a device architecture.
Fig. 2 is a schematic view of the apparatus structure.
Fig. 3 is a schematic view of an adjustable mount for a camera.
FIG. 4 is a schematic view of a combined LED illumination source.
FIG. 5 is a drawing of a tailored calibration plate object.
Fig. 6 is an operation interface image of the main control program.
Fig. 7 is a brake disc image captured by four cameras in the embodiment.
FIG. 8 is a brake disc image with calibration corrected to standard space for four cameras in the embodiment.
FIG. 9 is an image of an embodiment in which defects are detected and marked by a defect detection algorithm.
Detailed Description
The invention relates to a brake disc gluing defect detection device and method based on multi-view vision, which are characterized in that an industrial digital camera is used for collecting an image of a brake disc which is glued on a production line, a certain contrast is generated on the collected image by utilizing different reflectivities of infrared light of a colloid and the surface of the brake disc, and an image processing method is used for detecting and marking an area of the brake disc which is not glued.
As shown in fig. 1, the brake disc gluing defect detecting device based on multi-view vision provided by the invention comprises:
the calibration board is used for calculating a transformation matrix among the cameras;
the illumination module consists of 16 infrared LED light sources and a light source controller and is used for illumination during detection;
the main body of the imaging module is provided with four industrial cameras;
and the main control unit comprises an image processing program and a signal control program and is used for controlling the lighting module and the imaging module and communicating with the PLC.
As shown in fig. 2, the whole structure of the device is that the adjustable support is fixed on the production line and located at the top, the light source support is fixed on the production line and located in the middle, 16 light sources are uniformly distributed on the light source support, the calibration plate is located at the bottom, and during the production process, the brake disc replaces the position of the calibration plate.
As shown in fig. 5, the calibration board is a specially customized checkerboard, the thickness of the calibration board is 0.3mm, each check in the checkerboard is a square black-and-white checkerboard, the number of the check points is 8x8, the side length of each check is 20mm, 5 asymmetric annular features are arranged around the check points for determining the direction during calibration, and feature points extracted from different angles can be arranged in the same sequence. The four characteristic points are distributed on the four corners, so that the detection range can be reduced to a certain range, the influence of a complex environment on the subsequent checkerboard corner extraction is prevented, and the detection precision is improved. The definition condition of the fifth feature point is to ensure that the symmetry of the calibration board can be broken, so that the feature points extracted from the four cameras can be arranged in the same order. The thickness of the calibration plate is designed to be extremely thin, so that the feature points extracted from the checkerboard can be in the same plane with the surface of the brake disc to ensure the calibration accuracy.
As shown in fig. 4, the lighting module is composed of 16 infrared LED light sources with a wavelength of 850nm and a dc light source controller. According to the reason that the surfaces of the colloid and the brake disc have different reflectivity to light and certain contrast is generated during imaging, a plurality of different light sources are tested, and finally, compared with other light sources, the 850nm infrared light source can obviously increase the contrast between the colloid and the surface of the brake disc. Aiming at the characteristic that the brake disc is circular in shape, the brake disc is formed by circumferentially arranging and combining 16 small LED infrared light sources, so that the uniformity of light on the brake disc can be ensured; the LED lamp is characterized in that the lighting mode is low-angle lighting, and the influence of light reflection is reduced.
As shown in fig. 3, the imaging module is composed of 4 black and white industrial cameras, the central camera is used as a main camera, the three peripheral cameras are used as auxiliary cameras, and a lens with a focal length of 8mm is selected to be matched with an infrared filter with a wavelength of 850nm, so that the influence of ambient light in production is filtered. The acquisition device can acquire a high-definition image with the resolution of 1292x964, and can realize the detection precision of 0.25 mm. Meanwhile, the adjustable camera support is matched, a groove for the camera to slide is formed in the adjustable camera support, the position and the angle of the camera can be freely adjusted according to actual conditions, and imaging is complete.
The main control unit can operate on a UI interface and consists of two parts of signal control and image processing: the signal control may control the switching and brightness of the lighting module, the switching of the camera, and the communication with the PLC via the IO card. The communication comprises the step that the PLC sends a signal to be detected and a master control signal to the PLC, and sends a detection result signal.
The interactive interface can realize the access of the detection result, the control of the signal and the modification of the image processing algorithm parameter.
The brake disc gluing defect detection method based on multi-view vision mainly comprises four steps of image processing, correction, positioning, detection and fusion.
The correction step comprises the steps of collecting images of a calibration plate through 4 cameras, correspondingly calculating a perspective transformation matrix by using the specific sequence characteristic points extracted from the 4 images and a standard dot matrix, and then mapping the collected brake disc images to a standard space by using the matrix to correct distortion.
The positioning step comprises the steps of filtering, morphological gradient, binaryzation, contour searching, round holes on the brake disc are screened out from the contour, RANSAC circle fitting is carried out on the round holes to determine the center of the brake disc and the radius of a ring where the round holes are located, then the corresponding relation between the actual size and the pixel level is formed, and then the annular ROI area is determined according to the gluing standard.
The detecting step is characterized in that the partition adaptive histogram analysis and the gradient histogram analysis are combined.
The fusion step is characterized in that different weights are given to images shot by different cameras, and the images are combined to analyze defects, so that missing detection and false detection are effectively avoided.
The defect detection algorithm adopts the partition self-adaptive histogram analysis, and can effectively solve the influence on the histogram distribution caused by the problems of processing marks, chromatic aberration, uneven illumination and the like on the surface of the brake disc in the global histogram analysis method. Meanwhile, the adopted method based on multi-view visual fusion can effectively avoid the misjudgment influence of reflection on defect analysis caused by uneven gluing and bubble doping.
The following describes in detail the steps of the detection method:
(1) calibration procedure (after each adjustment of camera position): and extracting corner points of the calibration plate images collected by the four cameras. Because there are five asymmetric ring feature points of distribution on the calibration board four corners, can regard the corner that possess two ring feature points as the third corner of clockwise, therefore no matter how rotatory the calibration board, and what angle is with the camera, all can extract 8x8 feature points according to certain order. Meanwhile, in order to correct the influence of brake disc imaging distortion caused by the fact that the imaging plane of the camera is not parallel to the plane of the brake disc, a standard dot matrix of 8x8 is constructed according to the calibration plate and is used as a standard reference space, and an imaging perspective transformation matrix H from the four cameras to the standard space is obtained.
(2) And (3) processing the acquired images of the brake disc according to the perspective transformation matrix calculated in the step (1), and mapping the images to a standard space.
(3) And (4) performing the same operation as the operation (4) to (6) on the four images to obtain a black and white segmentation image with defects and non-defects on the target detection area, wherein the pixel value 255 represents the defects, and the pixel value 0 represents the non-defects.
(4) And positioning the brake disc according to the inherent circular hole characteristics on the brake disc to obtain the annular region ROI required to be detected.
(5) And carrying out regional adaptive histogram analysis on the ROI to obtain the region where the abnormal pixel value is located.
(6) And performing gradient histogram analysis on the ROI to obtain a region contained in the abnormal boundary.
(7) And (4) performing phase comparison on the black and white segmentation images obtained in the step (5) and the step (6) to obtain possible defect areas on each image.
(8) And (4) giving different weights to the defect map obtained by the main camera and the defect map obtained by the auxiliary camera in the step (7), performing fusion analysis on the defect area to obtain a final defect map, and marking the final defect map on the original map.
Wherein the detailed steps of (4) are as follows:
(41) and performing morphological gradient processing on the image to obtain a gradient map.
(42) And searching the contour of the gradient map, and screening out the specific contour on the brake disc according to the perimeter of the contour and the length-width difference of the circumscribed rectangle.
(43) And performing circular ring fitting according to the central point group of the screened contour, and simultaneously removing some abnormal values by using RANSAC. The radius of the center of the brake disc and the radius of the ring where the round hole is located can be obtained through ring fitting, and the annular MASK of the target area is obtained according to the center and the radius and is compared with the original image, so that the ROI is obtained.
Wherein the detailed steps of (5) are as follows:
(51) and calculating the angle between each point on the ROI and the central point of the brake disc, and dividing the ROI into specific blocks according to the angle to perform partition analysis so as to reduce the influence of the problems of trace on the disc surface of the brake disc, chromatic aberration, uneven illumination and the like on histogram distribution.
(52) And performing histogram analysis on the subareas to find a second peak value area formed in a high pixel value range due to the defect of no gluing, and marking pixel points of the pixel values in the area as defects.
Wherein the detailed steps of (6) are as follows:
(61) the ROI was subjected to morphological gradient processing.
(62) Performing histogram analysis on the gradient map to obtain a gradient histogram, obtaining a peak value of the histogram, taking a certain proportion according to the peak value, and finding a first gradient value on the histogram, wherein the first gradient value is smaller than the proportion and is larger than the gradient value corresponding to the peak value.
(63) And (4) taking the gradient value found in the step (62) as a threshold value, carrying out binarization on the gradient map, searching for a contour, and determining an abnormal boundary.
(64) The inside of the obtained abnormal boundary is filled (63) as a defective region.
Wherein the detailed steps of (8) are as follows:
(81) and performing phase comparison on the four black-and-white images to obtain a black-and-white image which is necessary to be defective, wherein the value of a defective pixel is recorded as 255, and the value of a non-defective pixel is recorded as 0.
(82) And (4) subtracting the defect map marked by the main camera from the image obtained in the step (81) to obtain a suspicious defect mark map, and recording the suspicious defect pixel value as 85. And (4) subtracting the defect map of the three auxiliary camera marks from the image obtained in the step (81) to obtain a reference defect mark map, wherein the mark area pixel value is recorded as 51.
(83) And (4) adding the four images obtained in the step (82) and adding the four images to the determined image obtained in the step (81) to obtain a combined image. The pixel value is 0 to represent non-defective pixel points, the pixel value is 51 to represent pixel points considered as defective by one auxiliary camera, the pixel value is 102 to represent pixel points considered as defective by two auxiliary cameras, the pixel value is 153 to represent pixel points considered as defective by three auxiliary cameras, the pixel value is 85 to represent suspicious defective pixel points to be determined, the pixel value is 136 to represent pixel points considered as defective by the main camera and one auxiliary camera, the pixel value is 187 to represent pixel points considered as defective by the main camera and the two auxiliary cameras, and the pixel value is 255 to represent pixel points considered as defective by the main camera and the three auxiliary cameras, namely pixel points which must be defective.
(84) And taking the point with the pixel value of 255 as a seed point, optionally taking the value between 69 and 118 as a threshold value, and performing flood filling on the combined graph to obtain a final defect graph, wherein the filled area is a pixel point which is adjacent to the pixel point which must be defective and is considered to be defective by two auxiliary cameras, and in the flood filling process, an iterative process exists, namely the filled area can be also taken as the pixel point which must be defective, and the adjacent pixel point which meets the condition can be continuously filled.
The defect detection algorithm adopts the partition self-adaptive histogram analysis, and can effectively solve the influence on the histogram distribution caused by the problems of processing marks, chromatic aberration, uneven illumination and the like on the surface of the brake disc in the global histogram analysis method. Meanwhile, the adopted method based on multi-view visual fusion can effectively avoid the misjudgment influence of reflection on defect analysis caused by uneven gluing and bubble doping.
The brake disc gluing defect detection device and method based on multi-view vision according to the invention are described below with reference to the accompanying drawings.
Firstly, selecting a detection result saving path 101 on the operation interface of fig. 6; selecting a connection port of the light source controller and turning on the light source at 301; opening a camera 201, adjusting camera parameters 202, adjusting the angle of the camera to enable the area where the assembly line brake disc is located to be completely imaged, adjusting the focal length to enable the imaging to be clear, and observing the imaging conditions of the four cameras in real time at 203; placing a calibration plate in a detection area, and obtaining a perspective transformation matrix through a calibration process 402; the brake disc with the typical gluing defect is placed in a detection area, detection is carried out through a detection algorithm 403, algorithm parameters 401 are adjusted according to a detection result and a detection standard so that the detection meets requirements, and then the brake disc can be put into production line production.
The present embodiment is one of the detection results of the present detection device. The inner radius of gluing detection is determined to be 95mm and the outer radius of gluing detection is determined to be 109.6mm mainly according to the gluing standard of the brake disc. Fig. 7 is an original image acquired by four cameras, fig. 8 is an image mapped to a standard space through perspective transformation, and fig. 9 is a detection result.
Claims (10)
1. A brake disc gluing defect detection device based on multi-view vision is characterized by comprising a chessboard pattern calibration plate, an illumination module, an imaging module and a main control unit;
asymmetric circular ring characteristic points are arranged at the corners of the chessboard pattern calibration plate and are used for calculating transformation matrixes among the cameras;
the lighting module is used for providing lighting conditions for the detection device;
the imaging module is used for acquiring a brake disc image to be detected;
the main control unit is used for controlling the lighting module and the imaging module, detecting and positioning defects of the acquired images through the image processing step, communicating with the PLC through the IO card, receiving the signals to be detected and transmitting the detection results.
2. The brake disc gluing defect detecting device based on the multi-view vision as claimed in claim 1, wherein five asymmetric circular ring feature points are distributed at four corners of the chessboard pattern calibration plate and are used for judging the direction during calibration; each of the checkerboards is a square black-white checkerboard, and the number of the grid points is 8x 8.
3. The brake disc gluing defect detecting device based on the multi-view vision as claimed in claim 1, wherein the lighting module is composed of 16 infrared LED light sources with the wavelength of 850nm and a light source controller and is used for lighting during detection.
4. The brake disc gluing defect detecting device based on the multi-view vision as claimed in claim 1, wherein the imaging module is composed of 4 black and white industrial cameras, an infrared filter and a camera bracket; selecting a lens with a focal length of 8mm, and matching with an infrared filter with a wavelength of 850 nm; the camera support is used for installing 4 black and white industrial cameras, the main camera is in the center, and the auxiliary cameras are distributed around the main camera.
5. The brake disc gluing defect detection device based on the multi-view vision as claimed in claim 1, wherein the main control unit provides a UI interface, and the UI interface comprises two parts of signal control and image processing: the signal control comprises the control of a camera signal, the control of a light source signal and the operation of a communication control assembly line through an IO card and a PLC, wherein the communication comprises the steps that the PLC sends a signal to be detected to a master control program and the master control signal sends a detection result signal to the PLC; the UI interface is used for detecting result access, signal control and image processing algorithm parameter modification.
6. The method for detecting the brake disc gluing defect detection device based on the multi-view vision as claimed in any one of claims 1 to 5 is characterized by comprising the following steps:
(1) a calibration process: extracting angular points of calibration plate images acquired by the four cameras; extracting 8x8 feature points according to a certain sequence by taking the corner with two circular feature points as the third corner in the clockwise direction; constructing a standard dot matrix of 8x8 as a standard reference space according to a calibration plate, and solving an imaging perspective transformation matrix H from four cameras to the standard space;
(2) processing the collected images of the brake disc according to the perspective transformation matrix calculated in the step (1), and mapping the images to a standard space;
(3) performing the same operations of (4) to (6) on the four images to obtain a black and white segmentation image with defects and non-defects on a target detection area, wherein a pixel value of 255 represents the defects, and a pixel value of 0 represents the non-defects;
(4) positioning the brake disc according to the inherent circular hole characteristics on the brake disc to obtain an annular region ROI required to be detected;
(5) performing regional adaptive histogram analysis on the ROI to obtain the region where the abnormal pixel value is located;
(6) performing gradient histogram analysis on the ROI to obtain a region contained in the abnormal boundary;
(7) performing phase comparison on the black and white segmentation images obtained in the step (5) and the step (6) to obtain a possible defect area on each image;
(8) and (4) giving different weights to the defect map obtained by the main camera and the defect map obtained by the auxiliary camera in the step (7), performing fusion analysis on the defect area to obtain a final defect map, and marking the final defect map on the original map.
7. The brake disc gluing defect detection method based on the multi-view vision as claimed in claim 6, wherein the step (4) is specifically as follows:
(41) performing morphological gradient processing on the image to obtain a gradient image;
(42) searching the contour of the gradient map, and screening out the specific contour on the brake disc according to the perimeter of the contour and the length-width difference of the circumscribed rectangle;
(43) performing circular ring fitting according to the central point group of the screened contour, and simultaneously removing abnormal values by using RANSAC; and obtaining the radius of the center of the brake disc and the radius of the ring where the round hole is located through ring fitting, and obtaining the annular MASK of the target area by the center and the radius and comparing the annular MASK with the original image to obtain the ROI.
8. The brake disc gluing defect detection method based on the multi-view vision as claimed in claim 6, wherein the step (5) is specifically as follows:
(51) calculating the angle between each point on the ROI and the center point of the brake disc, and dividing the ROI into specific blocks according to the angles to perform (52) partition histogram analysis;
(52) and performing histogram analysis on the partitions, finding a second peak value area formed in a high pixel value range due to the defect of no gluing according to the gradient change condition of the histogram, and marking pixel points of the pixel values in the area as defects.
9. The brake disc gluing defect detection method based on the multi-view vision as claimed in claim 6, wherein the step (6) is specifically as follows:
(61) performing morphological gradient processing on the ROI;
(62) performing histogram analysis on the gradient map to obtain a gradient histogram, obtaining a peak value of the histogram, taking a certain proportion according to the peak value, and finding a first gradient value on the histogram, wherein the number of the first gradient values is smaller than the proportion, and the gradient value is larger than the gradient value corresponding to the peak value;
(63) taking the gradient value found in the step (62) as a threshold value, carrying out binarization on the gradient map, searching a contour, and determining an abnormal boundary;
(64) the inside of the obtained abnormal boundary is filled (63) as a defective region.
10. The brake disc gluing defect detection method based on the multi-view vision as claimed in claim 6, wherein the step (8) is specifically as follows:
(81) performing phase comparison on the four black-and-white images to obtain a black-and-white image which is necessary to be defective, wherein the defective pixel value is recorded as 255, and the non-defective pixel value is recorded as 0;
(82) subtracting the defect map marked by the main camera from the image obtained in the step (81) to obtain a suspicious defect mark map, and recording the suspicious defect pixel value as 85; subtracting the defect map marked by the three auxiliary cameras from the image obtained in the step (81) to obtain a reference defect mark map, wherein the pixel value of the marked area is marked as 51;
(83) adding the four images obtained in the step (82) and adding the four images to the determined image obtained in the step (81) to obtain a combined image; the pixel value is 0 to represent non-defective pixel points, the pixel value is 51 to represent a pixel point which is considered as a defect by one auxiliary camera, the pixel value is 102 to represent a pixel point which is considered as a defect by two auxiliary cameras, the pixel value is 153 to represent a pixel point which is considered as a defect by three auxiliary cameras, the pixel value is 85 to represent a suspicious defect pixel point to be determined, the pixel value is 136 to represent a pixel point which is considered as a defect by the main camera and one auxiliary camera, the pixel value is 187 to represent a pixel point which is considered as a defect by the main camera and two auxiliary cameras, and the pixel value is 255 to represent a pixel point which is considered as a defect by the main camera and three auxiliary cameras, namely a pixel point which must be a defect;
(84) and taking the point with the pixel value of 255 as a seed point, optionally taking the value between 69 and 118 as a threshold value, and performing flood filling on the combined graph to obtain a final defect graph, wherein the filled area is a pixel point which is adjacent to the pixel point which must be defective and is considered to be defective by two auxiliary cameras, and in the flood filling process, an iterative process exists, namely the filled area is also taken as the pixel point which must be defective, and the adjacent pixel point which meets the condition is continuously filled.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111146179.8A CN113916892B (en) | 2021-09-28 | 2021-09-28 | Multi-view vision-based brake disc gluing defect detection device and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111146179.8A CN113916892B (en) | 2021-09-28 | 2021-09-28 | Multi-view vision-based brake disc gluing defect detection device and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113916892A true CN113916892A (en) | 2022-01-11 |
| CN113916892B CN113916892B (en) | 2022-10-28 |
Family
ID=79236918
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111146179.8A Active CN113916892B (en) | 2021-09-28 | 2021-09-28 | Multi-view vision-based brake disc gluing defect detection device and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113916892B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117451728A (en) * | 2023-12-26 | 2024-01-26 | 常州漫舒医疗科技有限公司 | Ostomy product production detection system based on visual detection |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108760767A (en) * | 2018-05-31 | 2018-11-06 | 电子科技大学 | Large-size LCD Screen defect inspection method based on machine vision |
| CN109141232A (en) * | 2018-08-07 | 2019-01-04 | 常州好迪机械有限公司 | A kind of circle plate casting online test method based on machine vision |
| CN109490316A (en) * | 2018-11-30 | 2019-03-19 | 熵智科技(深圳)有限公司 | A kind of surface defects detection algorithm based on machine vision |
| US20190096057A1 (en) * | 2017-05-11 | 2019-03-28 | Jacob Nathaniel Allen | Object inspection system and method for inspecting an object |
| CN109829876A (en) * | 2018-05-30 | 2019-05-31 | 东南大学 | Carrier bar on-line detection device of defects and method based on machine vision |
| KR20200014438A (en) * | 2017-07-07 | 2020-02-10 | 주식회사 고영테크놀러지 | Apparatus and method for optimizing examination outside of the subject |
| CN111784779A (en) * | 2020-06-15 | 2020-10-16 | 南京理工大学 | Checkerboard image recognition and positioning system and method based on convolutional neural network and nested contour recognition |
| CN111830049A (en) * | 2020-07-24 | 2020-10-27 | 深圳市道通科技股份有限公司 | Brake disc wear area identification method and wear identification system |
| CN111929309A (en) * | 2020-09-14 | 2020-11-13 | 成都卓识维景科技有限公司 | Cast part appearance defect detection method and system based on machine vision |
-
2021
- 2021-09-28 CN CN202111146179.8A patent/CN113916892B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190096057A1 (en) * | 2017-05-11 | 2019-03-28 | Jacob Nathaniel Allen | Object inspection system and method for inspecting an object |
| KR20200014438A (en) * | 2017-07-07 | 2020-02-10 | 주식회사 고영테크놀러지 | Apparatus and method for optimizing examination outside of the subject |
| CN109829876A (en) * | 2018-05-30 | 2019-05-31 | 东南大学 | Carrier bar on-line detection device of defects and method based on machine vision |
| CN108760767A (en) * | 2018-05-31 | 2018-11-06 | 电子科技大学 | Large-size LCD Screen defect inspection method based on machine vision |
| CN109141232A (en) * | 2018-08-07 | 2019-01-04 | 常州好迪机械有限公司 | A kind of circle plate casting online test method based on machine vision |
| CN109490316A (en) * | 2018-11-30 | 2019-03-19 | 熵智科技(深圳)有限公司 | A kind of surface defects detection algorithm based on machine vision |
| CN111784779A (en) * | 2020-06-15 | 2020-10-16 | 南京理工大学 | Checkerboard image recognition and positioning system and method based on convolutional neural network and nested contour recognition |
| CN111830049A (en) * | 2020-07-24 | 2020-10-27 | 深圳市道通科技股份有限公司 | Brake disc wear area identification method and wear identification system |
| CN111929309A (en) * | 2020-09-14 | 2020-11-13 | 成都卓识维景科技有限公司 | Cast part appearance defect detection method and system based on machine vision |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117451728A (en) * | 2023-12-26 | 2024-01-26 | 常州漫舒医疗科技有限公司 | Ostomy product production detection system based on visual detection |
| CN117451728B (en) * | 2023-12-26 | 2024-02-20 | 常州漫舒医疗科技有限公司 | Ostomy product production detection system based on visual detection |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113916892B (en) | 2022-10-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108445007B (en) | Detection method and detection device based on image fusion | |
| CN109215063B (en) | Registration method of event trigger camera and three-dimensional laser radar | |
| CN110657785B (en) | Efficient scene depth information acquisition method and system | |
| CN110261410A (en) | A kind of detection device and method of glass lens defect | |
| CN112669394B (en) | Automatic calibration method for vision detection system | |
| CN110570412B (en) | Part error vision judgment system | |
| CN110108711A (en) | The vision detection system of annulus side wall defect | |
| CN107869954B (en) | Binocular vision volume weight measurement system and implementation method thereof | |
| CN105021628A (en) | Detection method for surface defects of optical fiber image inverter | |
| CN113269762B (en) | Screen defect detection method, system and computer storage medium | |
| WO2020019352A1 (en) | Method, device and system for inspecting camera module, and machine readable storage medium | |
| CN112801947A (en) | Visual detection method for dead pixel of LED display terminal | |
| CN109406527B (en) | System and method for detecting fine appearance defects of micro camera module lens | |
| CN113916892B (en) | Multi-view vision-based brake disc gluing defect detection device and method | |
| US20120128232A1 (en) | Method for detecting a bridge connecting failure | |
| KR20160108644A (en) | Device for detecting defect of device | |
| JP2022533848A (en) | Systems and methods for determining if camera components are damaged | |
| CN114820475B (en) | Edge identification method and system, wafer processing device and method for determining concentric state of wafer and processing table | |
| CN114577805A (en) | MiniLED backlight panel defect detection method and device | |
| CN111985435A (en) | Unmanned aerial vehicle water area monitoring and cruising method based on machine vision | |
| KR20180048986A (en) | Measuring the rotational position of a lenticular lens sheet | |
| CN117092113A (en) | Camera mould welding quality detection device and system thereof | |
| CN111784779A (en) | Checkerboard image recognition and positioning system and method based on convolutional neural network and nested contour recognition | |
| CN112184578A (en) | Method for calibrating central coordinates of each microlens imaging in light field imaging | |
| JP4036048B2 (en) | Defect identification method |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |