Dynamic defect monitoring and early warning method and system for machine vision rapid image recognition

Abstract

The invention discloses a dynamic defect monitoring and early warning method and system for machine vision rapid image recognition, and relates to the technical field of vehicle paint spraying processing, and the system comprises a data acquisition module, a preprocessing module, a dynamic monitoring module and a classification analysis module. The imaging unit is used for collecting image data of the surface of a vehicle and transmitting the image data to the image unit, and the image unit is used for synchronously collecting the image data generated by the imaging unit and transmitting the image data to the preprocessing module in real time; high-quality original image data is provided through the data acquisition module, high-reflection interference of vehicle paint is effectively reduced, a complex curved surface and an edge area are covered, the preprocessing module performs optimization processing on an image, defect features are further highlighted, the image quality of a complex area is improved, and the processed image data is transmitted to the dynamic monitoring module.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle painting processing, and more specifically to a dynamic defect monitoring and early warning method and system for rapid machine vision image recognition. Background Art

[0002] With the rapid development of the automobile industry, the appearance quality of vehicles, especially the defect monitoring of paint surface, has become a crucial link in the automobile manufacturing process. The paint not only affects the appearance of the vehicle, but also directly affects the anti-corrosion performance and service life of the vehicle. However, the traditional paint defect detection method mainly relies on manual visual inspection, which is not only inefficient and easily affected by subjective factors, but also difficult to meet the high-efficiency and high-precision production requirements of modern automobile manufacturing. In recent years, the rapid development of machine vision technology has provided a new idea for solving this problem. Through high-resolution cameras, advanced image processing algorithms and automated monitoring equipment, machine vision technology can comprehensively and accurately monitor the paint surface in a short time. High-resolution cameras and customized LED matrix lighting, combined with robot-guided sensors, can achieve rapid monitoring of paint defects. However, in actual use, the paint surface of a car usually has a high reflective property, especially the high-gloss paint surface after multiple layers of spraying. This property will cause the traditional machine vision system to have too many specular reflection components during the imaging process. The high-gloss reflective paint surface may cover up tiny scratches, particles or shrinkage holes during imaging, making defect monitoring difficult; Moreover, automobile bodies usually have complex curved surfaces and ridge structures. Defect monitoring in these areas is a weak link in existing systems. The ridges of the car body cannot form a flat effect, which will reduce the probability of the monitoring system detecting defects such as particles. For the edges and occluded areas of the car body, traditional vision systems may not be able to achieve full coverage, thereby missing potential defects. Summary of the invention

[0003] In order to solve the above problems, the present invention provides a dynamic defect monitoring and early warning method and system for rapid image recognition by machine vision.

[0004] The present invention provides a dynamic defect monitoring and early warning method and system for rapid machine vision image recognition, including a data acquisition module, a preprocessing module, a dynamic monitoring module and a classification analysis module: The data acquisition module includes an imaging unit and an image unit, wherein the imaging unit is used to acquire image data of the vehicle surface and transmit the image data to the image unit, and the image unit is used to synchronously acquire image data generated by the imaging unit and transmit it to the preprocessing module in real time; The preprocessing module is used to receive the image data transmitted by the data acquisition module in real time, and preprocess the acquired image to further highlight the defect characteristics, obtain the preprocessed image data, and transmit it to the dynamic monitoring module; The dynamic monitoring module is used to receive the image data preprocessed by the preprocessing module in real time, monitor it in real time, identify and locate the defects of the vehicle surface paint, and if it is determined that there are defects, generate an alarm signal and transmit it to the mobile phone terminal of the staff, and track the change data of the defects in real time to obtain the defect data of the vehicle surface paint, and transmit it to the classification analysis module; The classification and analysis module is used to mark the defect data transmitted by the dynamic monitoring module according to time, and classify and analyze the defect data within this cycle every other cycle, evaluate and obtain the severity of the defect data, and transmit the severity of the defect data to the staff's mobile phone terminal based on the severity of the defect data.

[0005] Preferably, the specific working steps of the imaging unit are as follows: Obtain the length L, width W and height H of the vehicle; Will , as the geometric center coordinates of the vehicle.

[0006] The vehicle coordinate system is established with the geometric center of the vehicle as the coordinate origin, the vehicle length direction as the X-axis, the vehicle width direction as the Y-axis, and the vehicle height direction as the Z-axis; The surface of the vehicle is divided into monitoring areas to obtain the number of monitoring areas and the coordinates of the center point of each monitoring area. A camera is installed in each monitoring area. The installation position coordinates of the corresponding camera are determined according to the coordinates of the center point of each monitoring area. The number of cameras is equal to the number of monitoring areas. Each camera is equipped with three light sources, including a high-angle light source, a low-angle light source and a strip light source. The coordinates of the corresponding three light sources are obtained according to the coordinates of the camera's installation position; The camera and the light source are installed according to the determined installation position coordinates of the camera and the corresponding coordinates of the three light sources, and the image data of the vehicle surface paint is acquired in real time and transmitted to the preprocessing module in real time.

[0007] Preferably, the specific steps of dividing the monitoring area on the surface of the vehicle are as follows: The acquisition area of ​​the camera is obtained in advance. The acquisition area is a square, and the length of the acquisition area is T. Divide the length of the vehicle by T, and round the value up to get the number of monitoring areas along the X-axis of the vehicle, P1. Divide the width of the vehicle by T, and round the value up to get the number of monitoring areas along the Y-axis of the vehicle, P2. Multiply P1 by P2 to get the total number of monitoring areas, P3, that is, the number of cameras installed. According to the formula , calculate and obtain the coordinates of the center point on the X-axis of the i-th monitoring area along the X-axis direction of the vehicle , where i is an integer ranging from 1 to P1; According to the formula , calculate and obtain the coordinates of the center point on the Y axis of the jth monitoring area along the vehicle's X axis direction , where j is an integer ranging from 1 to P2; Combined acquisition to obtain the center point coordinates of each monitoring area , set the safe distance d1 between the camera and the vehicle in advance, As the camera's installation position coordinates.

[0008] Preferably, the specific working steps of acquiring the coordinates of the three corresponding light sources according to the installation position coordinates of the camera are as follows: Get the camera's installation location coordinates , set the distance d2 between the light source and the camera and the horizontal rotation angle of the camera in advance ; For high-angle light sources, according to the formula , calculated and obtained; For low-angle light sources, according to the formula , calculated and obtained; For a strip light source, according to the formula , calculated and obtained.

[0009] Preferably, the specific working steps of the preprocessing module are as follows: First, the collected image data is grayed out to convert the RGB image into a grayscale image; then histogram equalization is performed, and then Gaussian filtering is used to remove noise, and finally normalization is performed to adjust the image pixel values ​​to a uniform range to obtain preprocessed image data, which is then transmitted to the dynamic monitoring module.

[0010] Preferably, the specific working steps of the dynamic monitoring module are as follows: Get the preprocessed image data. For each pixel in the preprocessed image data , calculate its local mean ; Then calculate the local contrast of each pixel in the preprocessed image data ; A threshold is set in advance. If the local contrast of a pixel If the absolute value of is greater than the threshold, the pixel is considered to be a defective point, and it is determined that there is a defect in the vehicle paint. An alarm signal is generated and transmitted to the staff's mobile phone terminal, and all defective points in the preprocessed image data are obtained, and adjacent defective points are merged to obtain the defective area, and the change data of the defective area is obtained in real time. The size and change data of the defective area are used as the defect data of the vehicle surface paint and transmitted to the classification analysis module.

[0011] Preferably, the local average and local contrast The specific way to obtain is: For the local mean , obtained by dividing the sum of the pixel values ​​of the image data after preprocessing by each camera by the number of pixels of the image data after preprocessing by each camera; For local contrast , through the pixel value of the point and its local average The difference calculation is used to obtain the local contrast .

[0012] Preferably, the specific working steps of the classification analysis module are: Receive defect data of the paint on the vehicle surface, calculate the area H and perimeter G of each defect area according to the defect data of the paint on the vehicle surface, divide the area H by the perimeter G, and obtain the severity value B of the defect; A threshold range T1 to T2 of the severity value B is set in advance, and T1 is less than T2. ​​If the severity value B is less than T1, the defect is marked as a scratch. If the severity value B is greater than or equal to T1 and less than T2, the defect is marked as a particle. If the severity value is greater than or equal to T2, the defect is marked as a pit, and the classification result of each defect is obtained; The classification result of each defect and the area H size of the defect are used as the severity of the defect data of the defect, and the severity of the defect data is transmitted to the mobile phone terminal of the staff.

[0013] Preferably, the specific calculation steps of the area H and the perimeter G of the defect area are: The area H of the defect area is obtained by counting the number of pixels in the defect area; For the perimeter G, first convert the image into a binary image, where the defect area is white with a pixel value of 255, the background is black with a pixel value of 0, and a 3×3 structure element is used to erode the binary image to remove the boundary pixels in the image; subtract the original binary image from the eroded image, and the result is the boundary point; Calculate the distances between adjacent boundary points and add these distances to get the boundary length, i.e. the perimeter G.

[0014] The present invention also proposes a dynamic defect monitoring and early warning method for rapid image recognition by machine vision, comprising the following steps: Step 1: Collect vehicle surface images by combining multi-angle light sources, and transmit the collected image data to the pre-processing stage in real time; Step 2: Preprocess the collected images, and then monitor the preprocessed images in real time to identify and locate defects on the paint surface. Monitor complex curved surfaces and edge areas from different angles to ensure full coverage and track defect change data in real time. Step 3: Mark, classify and analyze the monitored defect data to assess the severity of the defects.

[0015] Beneficial effects: The data acquisition module provides high-quality original image data, effectively reducing the interference of high reflective paint and covering complex curved surfaces and edge areas. The pre-processing module optimizes the image to further highlight the defect characteristics, improve the image quality of complex areas, and transmits the processed image data to the dynamic monitoring module; The dynamic monitoring module monitors and identifies paint surface defects in real time, generates defect data, location information and alarm signals (if any), and transmits relevant information to the classification and analysis module. The classification and analysis module marks, classifies and evaluates the severity of the defect data, and formulates corresponding quality control strategies, thereby achieving rapid and accurate monitoring and early warning of paint surface defects, providing a reliable solution for vehicle paint quality monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0017] Application scenarios: In actual use, the paint surface of a car usually has a high reflective property, especially the high-gloss paint surface after multiple layers of spraying. This property will cause the traditional machine vision system to have too many specular reflection components during the imaging process. The high-gloss reflective paint surface may cover up tiny scratches, particles or shrinkage holes during imaging, making defect monitoring difficult; Moreover, automobile bodies usually have complex curved surfaces and ridge structures. Defect monitoring in these areas is a weak link in existing systems. The ridges of the car body cannot form a flat effect, which will reduce the probability of the monitoring system detecting defects such as particles. For the edges and occluded areas of the car body, traditional vision systems may not be able to achieve full coverage, thereby missing potential defects.

[0018] like Figure 1As shown: A dynamic defect monitoring and early warning method and system for rapid machine vision image recognition, including a data acquisition module, a preprocessing module, a dynamic monitoring module and a classification analysis module: The data acquisition module includes an imaging unit and an image unit. The imaging unit is used to collect image data of the vehicle surface and transmit the image data to the image unit. The image unit is used to synchronously collect image data generated by the imaging unit and transmit it to the preprocessing module in real time. It should be noted that the imaging unit adopts polarized light imaging technology, eliminates mirror reflection through a polarizing filter, reduces interference from highly reflective surfaces, enhances image contrast, and makes tiny defects (such as scratches and particles) easier to identify. The imaging unit also combines multi-angle light sources and 3D deflection imaging technology to illuminate the paint surface from different directions, ensuring that defects in complex curved surfaces and edge areas can be clearly captured. The preprocessing module is used to receive the image data transmitted by the data acquisition module in real time, and preprocess the acquired image to further highlight the defect characteristics, obtain the preprocessed image data, and transmit it to the dynamic monitoring module; it should be noted that the problem of insufficient image quality caused by the complex texture and high reflective characteristics of the paint surface can be solved; The dynamic monitoring module is used to receive the image data preprocessed by the preprocessing module in real time, monitor it in real time, identify and locate the defects of the vehicle surface paint, and if it is determined that there are defects, generate an alarm signal and transmit it to the staff's mobile phone terminal, and track the defect change data in real time to obtain the defect data of the vehicle surface paint, and transmit it to the classification analysis module; it should be noted that the multi-view imaging technology of the imaging unit is used to monitor complex curved surfaces and edge areas from different angles to ensure comprehensive coverage, and track the defect change data in real time, especially in complex curved surfaces and edge areas, to ensure that defects will not be missed; The classification and analysis module is used to mark the defect data transmitted by the dynamic monitoring module according to time, and classify and analyze the defect data in this cycle every other cycle, evaluate and obtain the severity of the defect data, and transmit the severity of the defect data to the staff's mobile phone terminal according to the severity of the defect data. It should be noted that the detection rate of defects in complex surfaces and edge areas, especially defects that are difficult to monitor such as particles, is improved. According to the severity and location information of the defects, targeted strategies are formulated for subsequent quality control to ensure that defects in complex areas receive special attention.

[0019] As an optional embodiment: the specific working steps of the imaging unit are as follows: The length L, width W and height H of the vehicle are obtained; it should be noted that the vehicle is painted in a special paint room, and the floor of the paint room is horizontal by default. The length L, width W and height H of the vehicle are measured by the staff before painting, and the unit is cm; Will , as the geometric center coordinates of the vehicle. It should be noted that based on the symmetry and size of the vehicle, because the vehicle is symmetrical in all directions by default, the geometric center is located at the midpoint of the vehicle size; The vehicle coordinate system is established with the geometric center of the vehicle as the coordinate origin, the vehicle length direction as the X-axis, the vehicle width direction as the Y-axis, and the vehicle height direction as the Z-axis; The surface of the vehicle is divided into monitoring areas to obtain the number of monitoring areas and the coordinates of the center point of each monitoring area. A camera is installed in each monitoring area. The installation position coordinates of the corresponding camera are determined according to the coordinates of the center point of each monitoring area. The number of cameras is equal to the number of monitoring areas. Each camera is equipped with three light sources, including a high-angle light source, a low-angle light source and a strip light source. The coordinates of the corresponding three light sources are obtained according to the coordinates of the camera's installation position; The camera and the light source are installed according to the determined installation position coordinates of the camera and the corresponding coordinates of the three light sources, and the image data of the vehicle surface paint is acquired in real time and transmitted to the preprocessing module in real time.

[0020] As an optional embodiment: the specific steps of dividing the monitoring area of ​​the vehicle surface are as follows: The acquisition area of ​​the camera is obtained in advance. The acquisition area is a square, and the length of the acquisition area is T. It should be noted that in actual use, the acquisition area of ​​the camera is a circle. In order to maintain the accuracy of acquisition, the circle is used as the center point of the square, and the acquisition area is set to a square area smaller than the original acquisition area. Divide the length of the vehicle by T, and round the value up to get the number of monitoring areas along the X-axis of the vehicle, P1. Divide the width of the vehicle by T, and round the value up to get the number of monitoring areas along the Y-axis of the vehicle, P2. Multiply P1 by P2 to get the total number of monitoring areas, P3, that is, the number of cameras installed. According to the formula , calculate and obtain the coordinates of the center point on the X-axis of the i-th monitoring area along the X-axis direction of the vehicle , where i is an integer ranging from 1 to P1; According to the formula , calculate and obtain the coordinates of the center point on the Y axis of the jth monitoring area along the vehicle's X axis direction , where j is an integer ranging from 1 to P2; Combined acquisition to obtain the center point coordinates of each monitoring area , set the safe distance d1 between the camera and the vehicle in advance, It should be noted that the safe distance d1 between the camera and the vehicle is set by the camera manufacturer or staff, while the camera can capture the image of the vehicle surface without affecting the processing of the vehicle surface paint.

[0021] As an optional embodiment: the specific working steps of obtaining the coordinates of the three corresponding light sources according to the installation position coordinates of the camera are as follows: Get the camera's installation location coordinates , set the distance d2 between the light source and the camera and the horizontal rotation angle of the camera in advance It should be noted that the distance d2 between the light source and the camera is obtained by presetting the parameters of the camera, which is 10 cm in this embodiment, and the rotation angle of the camera in the horizontal direction is Obtained through the camera's installation position coordinates and the camera's corresponding monitoring area center coordinates; For high-angle light sources, according to the formula , calculated and obtained; For low-angle light sources, according to the formula , calculated and obtained; For a strip light source, according to the formula , calculated and obtained.

[0022] As an optional embodiment: the specific working steps of the preprocessing module are as follows: First, the collected image data is grayed out to convert the RGB image into a grayscale image; then histogram equalization is performed, and then Gaussian filtering is used to remove noise, and finally normalization is performed to adjust the image pixel values ​​to a uniform range to obtain preprocessed image data, which is then transmitted to the dynamic monitoring module.

[0023] As an optional embodiment: the specific working steps of the dynamic monitoring module are as follows: Get the preprocessed image data. For each pixel in the preprocessed image data , calculate its local mean ; Then calculate the local contrast of each pixel in the preprocessed image data ; A threshold is set in advance. If the local contrast of a pixel If the absolute value of is greater than the threshold, the pixel is considered to be a defective point, and it is determined that there is a defect in the vehicle paint. An alarm signal is generated and transmitted to the staff's mobile phone terminal, and all defective points in the preprocessed image data are obtained, and adjacent defective points are merged to obtain the defective area, and the change data of the defective area is obtained in real time. The size and change data of the defective area are used as the defect data of the vehicle surface paint and transmitted to the classification analysis module.

[0024] As an optional embodiment: the local average and local contrast The specific way to obtain is: For the local mean , obtained by dividing the sum of the pixel values ​​of the image data after preprocessing by each camera by the number of pixels of the image data after preprocessing by each camera; For local contrast , through the pixel value of the point and its local average The difference calculation is used to obtain the local contrast .

[0025] As an optional embodiment: the specific working steps of the classification analysis module are: Receive defect data of the paint on the vehicle surface, calculate the area H and perimeter G of each defect area according to the defect data of the paint on the vehicle surface, divide the area H by the perimeter G, and obtain the severity value B of the defect; A threshold range T1 to T2 of the severity value B is set in advance, and T1 is less than T2. ​​If the severity value B is less than T1, the defect is marked as a scratch. If the severity value B is greater than or equal to T1 and less than T2, the defect is marked as a particle. If the severity value is greater than or equal to T2, the defect is marked as a pit, and the classification result of each defect is obtained; The classification result of each defect and the size of the defect area H are used as the severity of the defect data of the defect, and the severity of the defect data is transmitted to the mobile phone terminal of the staff. It should be noted that according to the classification result of the defect, the staff can formulate corresponding quality control strategies: For scratches: If the scratch length exceeds a certain threshold, it is marked as a serious defect and needs to be repaired immediately. If the scratch is shorter, it is marked as a minor defect, recorded and repaired later.

[0026] For particles: If the particle area exceeds a certain threshold, it is marked as a serious defect and needs to be cleaned immediately. If the particle is small, it is marked as a minor defect, recorded and cleaned later.

[0027] For pits: If the pit depth exceeds a certain threshold, it is marked as a serious defect and needs to be repaired immediately. If the pit is shallow, it is marked as a minor defect, recorded and repaired later.

[0028] As an optional embodiment: the specific calculation steps of the area H and the perimeter G of the defect area are: The area H of the defect area is obtained by counting the number of pixels in the defect area; For the perimeter G, first convert the image into a binary image, where the defect area is white with a pixel value of 255, the background is black with a pixel value of 0, and a 3×3 structure element is used to erode the binary image to remove the boundary pixels in the image; subtract the original binary image from the eroded image, and the result is the boundary point; Calculate the distances between adjacent boundary points and add these distances to get the boundary length, i.e. the perimeter G.

[0029] The present invention also proposes a dynamic defect monitoring and early warning method for rapid image recognition by machine vision, comprising the following steps: Step 1: Collect vehicle surface images by combining multi-angle light sources, and transmit the collected image data to the pre-processing stage in real time; Step 2: Preprocess the collected images, and then monitor the preprocessed images in real time to identify and locate defects on the paint surface. Monitor complex curved surfaces and edge areas from different angles to ensure full coverage and track defect change data in real time. Step 3: Mark, classify and analyze the monitored defect data to assess the severity of the defects.

[0030] Working principle: The data acquisition module includes an imaging unit and an image unit. The imaging unit is used to collect image data of the vehicle surface and transmit the image data to the image unit. The image unit is used to synchronously collect image data generated by the imaging unit and transmit it to the preprocessing module in real time. It should be noted that the imaging unit adopts polarized light imaging technology, eliminates mirror reflection through a polarizing filter, reduces interference from highly reflective surfaces, enhances image contrast, and makes tiny defects (such as scratches and particles) easier to identify. The imaging unit also combines multi-angle light sources and 3D deflection imaging technology to illuminate the paint surface from different directions, ensuring that defects in complex curved surfaces and edge areas can be clearly captured. The preprocessing module is used to receive the image data transmitted by the data acquisition module in real time, and preprocess the acquired image, perform image enhancement, denoising and contrast adjustment on the image data, further highlight the defect characteristics, obtain the preprocessed image data, and transmit it to the dynamic monitoring module; it should be noted that the problem of insufficient image quality caused by the complex texture and high reflective characteristics of the paint surface can be solved; The dynamic monitoring module is used to receive the image data preprocessed by the preprocessing module in real time, monitor it in real time, identify and locate the defects of the vehicle surface paint, and if it is determined that there are defects, generate an alarm signal and transmit it to the staff's mobile phone terminal, and track the defect change data in real time to obtain the defect data of the vehicle surface paint, and transmit it to the classification analysis module; it should be noted that the multi-view imaging technology of the imaging unit is used to monitor complex curved surfaces and edge areas from different angles to ensure comprehensive coverage, and track the defect change data in real time, especially in complex curved surfaces and edge areas, to ensure that defects will not be missed; The classification and analysis module is used to mark the defect data transmitted by the dynamic monitoring module according to time, and classify and analyze the defect data in this cycle every other cycle, evaluate and obtain the severity of the defect data, and formulate corresponding strategies for subsequent quality control based on the severity of the defect data. It should be noted that the detection rate of defects in complex surfaces and edge areas, especially defects that are difficult to monitor such as particles, is improved. According to the severity and location information of the defects, targeted strategies are formulated for subsequent quality control to ensure that defects in complex areas receive special attention.

[0031] The above are only preferred implementations of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical staff in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of this template.

Claims

1. A dynamic defect monitoring and early warning system based on machine vision rapid image recognition, characterized in that: Including data acquisition module, preprocessing module, dynamic monitoring module and classification analysis module: The data acquisition module includes an imaging unit and an image unit, wherein the imaging unit is used to acquire image data of the vehicle surface and transmit the image data to the image unit, and the image unit is used to synchronously acquire image data generated by the imaging unit and transmit it to the preprocessing module in real time; The preprocessing module is used to receive the image data transmitted by the data acquisition module in real time, and preprocess the acquired image to further highlight the defect characteristics, obtain the preprocessed image data, and transmit it to the dynamic monitoring module; The dynamic monitoring module is used to receive the image data preprocessed by the preprocessing module in real time, monitor it in real time, identify and locate the defects of the vehicle surface paint, and if it is determined that there are defects, generate an alarm signal and transmit it to the mobile phone terminal of the staff, and track the change data of the defects in real time to obtain the defect data of the vehicle surface paint, and transmit it to the classification analysis module; The classification and analysis module is used to mark the defect data transmitted by the dynamic monitoring module according to time, and classify and analyze the defect data within this cycle every other cycle, evaluate and obtain the severity of the defect data, and transmit the severity of the defect data to the staff's mobile phone terminal based on the severity of the defect data.

2. According to claim 1, a dynamic defect monitoring and early warning system for rapid machine vision image recognition is characterized in that: The specific working steps of the imaging unit are as follows: Obtain the length L, width W and height H of the vehicle; Will , as the geometric center coordinates of the vehicle; The vehicle coordinate system is established with the geometric center of the vehicle as the coordinate origin, the vehicle length direction as the X-axis, the vehicle width direction as the Y-axis, and the vehicle height direction as the Z-axis; The surface of the vehicle is divided into monitoring areas to obtain the number of monitoring areas and the coordinates of the center point of each monitoring area. A camera is installed in each monitoring area. The installation position coordinates of the corresponding camera are determined according to the coordinates of the center point of each monitoring area. The number of cameras is equal to the number of monitoring areas. Each camera is equipped with three light sources, including a high-angle light source, a low-angle light source and a strip light source. The coordinates of the corresponding three light sources are obtained according to the coordinates of the camera's installation position; The camera and the light source are installed according to the determined installation position coordinates of the camera and the corresponding coordinates of the three light sources, and the image data of the vehicle surface paint is acquired in real time and transmitted to the preprocessing module in real time.

3. According to claim 2, a dynamic defect monitoring and early warning system for rapid machine vision image recognition is characterized in that: The specific steps of dividing the detection area of ​​the vehicle surface are as follows: The acquisition area of ​​the camera is obtained in advance. The acquisition area is a square, and the length of the acquisition area is T. Divide the length of the vehicle by T, and round the value up to get the number of monitoring areas along the X-axis of the vehicle, P1. Divide the width of the vehicle by T, and round the value up to get the number of monitoring areas along the Y-axis of the vehicle, P2. Multiply P1 by P2 to get the total number of monitoring areas, P3, that is, the number of cameras installed. According to the formula , calculate and obtain the coordinates of the center point on the X-axis of the i-th monitoring area along the X-axis direction of the vehicle , where i is an integer ranging from 1 to P1; According to the formula , calculate and obtain the coordinates of the center point on the Y axis of the jth monitoring area along the vehicle's X axis direction , where j is an integer ranging from 1 to P2; Combined acquisition to obtain the center point coordinates of each monitoring area , set the safe distance d1 between the camera and the vehicle in advance, As the camera's installation position coordinates.

4. According to claim 2, a dynamic defect monitoring and early warning system for rapid machine vision image recognition is characterized in that: The specific working steps of obtaining the coordinates of the three corresponding light sources according to the installation position coordinates of the camera are as follows: Get the camera's installation location coordinates , set the distance d2 between the light source and the camera and the horizontal rotation angle of the camera in advance ; For high-angle light sources, according to the formula , calculated and obtained; For low-angle light sources, according to the formula , calculated and obtained; For a strip light source, according to the formula , calculated and obtained.

5. The dynamic defect monitoring and early warning system of machine vision rapid image recognition according to claim 1 is characterized in that: The specific working steps of the preprocessing module are as follows: First, the collected image data is grayed out to convert the RGB image into a grayscale image; then histogram equalization is performed, and then Gaussian filtering is used to remove noise, and finally normalization is performed to adjust the image pixel values ​​to a uniform range to obtain preprocessed image data, which is then transmitted to the dynamic monitoring module.

6. The dynamic defect monitoring and early warning system of machine vision rapid image recognition according to claim 5 is characterized in that: The specific working steps of the dynamic monitoring module are as follows: Get the preprocessed image data. For each pixel in the preprocessed image data , calculate its local mean ; Then calculate the local contrast of each pixel in the preprocessed image data ; A threshold is set in advance. If the local contrast of a pixel If the absolute value of is greater than the threshold, the pixel is considered to be a defective point, and it is determined that there is a defect in the vehicle paint. An alarm signal is generated and transmitted to the staff's mobile phone terminal, and all defective points in the preprocessed image data are obtained, and adjacent defective points are merged to obtain the defective area, and the change data of the defective area is obtained in real time. The size and change data of the defective area are used as the defect data of the vehicle surface paint and transmitted to the classification analysis module.

7. The dynamic defect monitoring and early warning system of machine vision rapid image recognition according to claim 6 is characterized in that: The local mean and local contrast The specific way to obtain is: For the local mean , obtained by dividing the sum of the pixel values ​​of the image data after preprocessing by each camera by the number of pixels of the image data after preprocessing by each camera; For local contrast , through the pixel value of the point and its local average The difference calculation is used to obtain the local contrast .

8. The dynamic defect monitoring and early warning system of machine vision rapid image recognition according to claim 6 is characterized in that: The specific working steps of the classification analysis module are: Receive defect data of the paint on the vehicle surface, calculate the area H and perimeter G of each defect area according to the defect data of the paint on the vehicle surface, divide the area H by the perimeter G, and obtain the severity value B of the defect; A threshold range T1 to T2 of the severity value B is set in advance, and T1 is less than T2. ​​If the severity value B is less than T1, the defect is marked as a scratch. If the severity value B is greater than or equal to T1 and less than T2, the defect is marked as a particle. If the severity value is greater than or equal to T2, the defect is marked as a pit, and the classification result of each defect is obtained; The classification result of each defect and the area H size of the defect are used as the severity of the defect data of the defect, and the severity of the defect data is transmitted to the mobile phone terminal of the staff.

9. The dynamic defect monitoring and early warning system of machine vision rapid image recognition according to claim 8 is characterized in that: The specific calculation steps of the area H and perimeter G of the defect area are: The area H of the defect area is obtained by counting the number of pixels in the defect area; For the perimeter G, first convert the image into a binary image, where the defect area is white with a pixel value of 255, the background is black with a pixel value of 0, and a 3×3 structure element is used to erode the binary image to remove the boundary pixels in the image; subtract the original binary image from the eroded image, and the result is the boundary point; Calculate the distances between adjacent boundary points and add these distances to get the boundary length, i.e. the perimeter G.

10. A dynamic defect monitoring and early warning method based on rapid machine vision image recognition, characterized in that: A dynamic defect monitoring and early warning system for rapid machine vision image recognition applicable to any one of claims 1 to 9, comprising the following steps: Step 1: Collect vehicle surface images by combining multi-angle light sources, and transmit the collected image data to the pre-processing stage in real time; Step 2: Preprocess the collected images, and then monitor the preprocessed images in real time to identify and locate defects on the paint surface. Monitor complex curved surfaces and edge areas from different angles to ensure full coverage and track defect change data in real time. Step 3: Mark, classify and analyze the monitored defect data to assess the severity of the defects.

Images