Electric tricycle body spraying defect detection method and system based on machine vision

Abstract

The invention provides an electro-tricycle body spraying defect detection method and system based on machine vision, and relates to the technical field of defect detection, and the method comprises the steps: obtaining a machine vision perception module; performing multi-angle image acquisition through a panoramic lens; performing preprocessing and panoramic stitching on the multi-angle tricycle body image set; performing defect area identification and anchor frame marking on the panoramic image of the tricycle body; carrying out feature extraction quantization on the tricycle defect area anchor frame set by adopting a microscopic lens; and carrying out statistical evaluation analysis on the tricycle defect characteristic quantitative information according to the spraying defect index set to generate an electro-tricycle body spraying defect detection result. The technical problem that in the prior art, due to the fact that small defects are difficult to accurately position, the defect detection accuracy is poor can be solved, comprehensive detection of the spraying defects of the electro-tricycle body is achieved by integrating the panoramic lens and the microscopic lens, and the defect detection accuracy is improved.

Description

Technical Field

[0001] The present application relates to the technical field of defect detection, and in particular to a method and system for detecting body spray defects of an electric tricycle based on machine vision. Background Art

[0002] The spraying of the electric tricycle body is a key link to ensure its appearance and quality. The quality of spraying directly affects the appearance, durability and long-term performance of the vehicle. Defects in the spraying process (such as uneven coating, bubbles, scratches, paint peeling, color difference, etc.) not only affect the aesthetic appearance, but may also lead to reduced surface protection performance, affecting the corrosion resistance and service life of the vehicle. The existing methods for detecting defects in the spraying of electric tricycle bodies mainly include machine vision-based detection and multi-scale detection methods. However, due to the wide variety of spraying defects and their different forms, including scratches, bubbles, color difference, uneven coating, etc., the characteristics and manifestations of each defect are different. Existing methods are difficult to adapt to multiple types of defects at the same time, especially when dealing with multiple complex, small or variant defects. It may not be possible to accurately locate, resulting in missed detection or false detection.

[0003] In summary, the prior art has a technical problem that the accuracy of defect detection is poor due to the wide variety and complexity of defects, and it is difficult to accurately locate tiny, complex or hidden defects. Summary of the invention

[0004] The purpose of this application is to provide an electric tricycle body spray defect detection method and system based on machine vision, so as to solve the technical problem in the prior art that the defect types are numerous and complex, and it is difficult to accurately locate small, complex or hidden defects, resulting in poor accuracy of defect detection.

[0005] In view of the above problems, the present application provides a method and system for detecting paint defects on an electric tricycle body based on machine vision.

[0006] In the first aspect, the present application provides an electric tricycle body spray defect detection method based on machine vision, and the electric tricycle body spray defect detection method based on machine vision is implemented by an electric tricycle body spray defect detection system based on machine vision, wherein the electric tricycle body spray defect detection method based on machine vision includes: obtaining a machine vision perception module, and the machine vision perception module integrates a panoramic lens and a microscopic lens; performing multi-angle image acquisition on the target electric tricycle through the panoramic lens to obtain a multi-angle tricycle body image set; preprocessing and panoramic stitching the multi-angle tricycle body image set to obtain a panoramic image of the tricycle body; performing defect area identification and anchor frame marking on the panoramic image of the tricycle body to obtain a tricycle defect area anchor frame set; using the microscopic lens to perform feature extraction and quantification on the tricycle defect area anchor frame set to obtain tricycle defect feature quantification information; performing statistical evaluation and analysis on the tricycle defect feature quantification information according to the spray defect index set to generate an electric tricycle body spray defect detection result.

[0007] In the second aspect, the present application also provides an electric tricycle body spray defect detection system based on machine vision, which is used to execute the electric tricycle body spray defect detection method based on machine vision as described in the first aspect, wherein the electric tricycle body spray defect detection system based on machine vision includes: a lens construction unit, the lens construction unit is used to obtain a machine vision perception module, the machine vision perception module integrates a panoramic lens and a microscopic lens; an image acquisition unit, the image acquisition unit is used to perform multi-angle image acquisition of the target electric tricycle through the panoramic lens to obtain a multi-angle tricycle body image set; an image stitching unit, the image stitching unit The unit is used to preprocess and panoramically stitch the multi-angle tricycle body image set to obtain a panoramic image of the tricycle body; the defect recognition unit is used to identify the defect area and mark the anchor frame of the tricycle body panoramic image to obtain a tricycle defect area anchor frame set; the feature extraction unit is used to use the microscopic lens to perform feature extraction and quantification on the tricycle defect area anchor frame set to obtain tricycle defect feature quantification information; the statistical evaluation unit is used to perform statistical evaluation and analysis on the tricycle defect feature quantification information according to the spray defect index set to generate an electric tricycle body spray defect detection result.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages: By acquiring a machine vision perception module, the machine vision perception module integrates a panoramic lens and a microscopic lens; the panoramic lens is used to collect multi-angle images of the target electric tricycle to obtain a multi-angle tricycle body image set; the multi-angle tricycle body image set is preprocessed and panoramically stitched to obtain a panoramic image of the tricycle body; the defect area is identified and anchor frame marked on the panoramic image of the tricycle body to obtain a tricycle defect area anchor frame set; the microscopic lens is used to extract and quantify the features of the tricycle defect area anchor frame set to obtain the tricycle defect feature quantification information; the tricycle defect feature quantification information is statistically evaluated and analyzed according to the spray defect index set to generate the electric tricycle body spray defect detection result. In other words, by integrating different types of lenses, combining panoramic image stitching and microscopic feature extraction, comprehensive detection of electric tricycle body spray defects is achieved, and the accuracy of defect detection is improved.

[0009] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0011] Figure 1 This is a flow chart of the electric tricycle body spray defect detection method based on machine vision in this application.

[0012] Figure 2 This is a structural schematic diagram of the electric tricycle body spray defect detection system based on machine vision in this application.

[0013] Explanation of the reference numerals: lens construction unit 11 , image acquisition unit 12 , image stitching unit 13 , defect recognition unit 14 , feature extraction unit 15 , statistical evaluation unit 16 . DETAILED DESCRIPTION

[0014] This application provides a method and system for detecting defects in the spray coating of electric tricycle bodies based on machine vision, which solves the technical problem in the prior art that the defects are numerous and complex, and it is difficult to accurately locate small, complex or hidden defects, resulting in poor accuracy in defect detection. By integrating different types of lenses, combining panoramic image stitching and microscopic feature extraction, comprehensive detection of defects in the spray coating of electric tricycle bodies is achieved, and the accuracy of defect detection is improved.

[0015] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0016] For example, please refer to the attached Figure 1 The present application provides a method for detecting defects in the body painting of an electric tricycle based on machine vision, wherein the method for detecting defects in the body painting of an electric tricycle based on machine vision is applied to a system for detecting defects in the body painting of an electric tricycle based on machine vision, and the method for detecting defects in the body painting of an electric tricycle based on machine vision specifically comprises the following steps: Step 1: Obtain a machine vision perception module, wherein the machine vision perception module integrates a panoramic lens and a microscopic lens.

[0017] Specifically, in the detection of paint defects on the body of an electric tricycle, the machine vision perception module combines two types of lenses: panoramic lenses and micro lenses, so as to capture information on the body surface from different angles and scales. The machine vision perception module integrates multiple hardware devices and software algorithms to understand objects or scenes through image or video capture, processing and analysis, and usually includes cameras, sensors, image processing units (such as image acquisition cards), and software for image analysis.

[0018] Panoramic lenses are lenses that can capture images or videos over a wide range and at a wide angle. They are usually able to capture a wide range of scenes (such as multiple angles of a vehicle body) in a single shot, thus providing comprehensive perspective and scene information. Panoramic lenses are often used in situations where it is necessary to obtain images with a wide field of view, such as shooting the sides of a vehicle body or a panoramic image. The advantage of a panoramic lens is that it can provide a macro image covering the entire vehicle body, including multiple angles of the vehicle body, thereby avoiding the situation where defects are missed due to perspective issues.

[0019] A microscopic lens is a lens that can focus on and clearly capture very small objects or features. It has a high magnification and can reveal details that are difficult to detect with the naked eye. It is often used to detect subtle defects such as tiny scratches, bubbles or uneven spraying, etc. It facilitates the precise positioning and identification of tiny defects as a supplement to panoramic images.

[0020] During the inspection process, the panoramic lens is first used to collect multi-angle images of the vehicle body to obtain a large-scale image set of the vehicle body, ensuring that the full view of the vehicle body is obtained from multiple angles, such as front, back, left and right views, to ensure that all parts are included. After that, the defective areas in the panoramic image (such as areas where scratches, bubbles or uneven spraying may exist) are selected and these areas are magnified and photographed through the microscopic lens to capture more subtle defects. The machine vision perception module that integrates panoramic lenses and microscopic lenses can accurately capture tiny defects on the surface of the vehicle body while ensuring wide-angle image acquisition. It can not only capture a wide range of vehicle body information, but also accurately locate small defects, improving the comprehensiveness and accuracy of spray defect detection.

[0021] Step 2: Capture multi-angle images of the target electric tricycle through the panoramic lens to obtain a multi-angle tricycle body image set.

[0022] Specifically, a panoramic lens is used to capture multi-angle images of the target electric tricycle, and the target electric tricycle is photographed from different perspectives to ensure that all possible defect areas are covered, thereby obtaining a multi-angle tricycle body image set. A multi-angle tricycle body image set refers to a set of image data obtained by shooting at different angles, which includes views of various parts of the body surface. In spray defect detection, a single perspective may miss defects at some angles, especially on the curved surface, corners, or areas that are difficult to observe directly. By capturing images from multiple angles, all parts of the body can be fully covered, improving the accuracy and completeness of defect detection. A panoramic lens with a larger field of view can be selected, usually with a field of view of 90° to 180°, to cover a larger part of the body at one time and reduce the number of shots. In order to ensure that all parts of the body are covered, it is usually shot from multiple angles.

[0023] The surface of a car body usually contains multiple different areas, such as front and back, left and right, and top. Each area has different spraying defects, so images are collected from multiple angles (such as the front, side, back, oblique side, etc.). Assuming that the shooting angle of the car body has 4 different directions, images at different angles can be obtained by adjusting the position of the panoramic lens or rotating the car body. Start the panoramic lens and take images from each set angle. Each time the panoramic lens is shot, it captures a large range of images of the entire car body surface. Normally, the image resolution should be high enough (such as 4000×3000 pixels) to ensure that enough details are captured. Through the panoramic lens, images from multiple angles are collected to ensure that every part of the car body is fully covered, avoid missing spraying defects due to angle problems, and ensure the integrity of the inspection.

[0024] Step 3: Preprocess and panoramically stitch the multi-angle tricycle body image set to obtain a panoramic image of the tricycle body.

[0025] Specifically, the multi-angle tricycle body image set is preprocessed, including noise removal, brightness adjustment, contrast enhancement, edge detection, etc., to remove irrelevant information and highlight useful features in the image, thereby improving the accuracy and usability of the image. First, a noise recognition algorithm is used to identify the type of noise in the image, such as Gaussian noise (which conforms to the normal distribution and is usually caused by uneven sensor brightness, circuit component noise or excessive temperature, and is manifested as random disturbances in grayscale values), salt and pepper noise (manifested as random white or black points in the image), etc. According to the determined noise characteristic information, a suitable filtering algorithm is selected, such as median filtering for removing salt and pepper noise and Gaussian filtering for removing Gaussian noise. The filter parameters, such as window size and filter strength, are adaptively adjusted. The determined noise filtering adaptive algorithm is applied to filter the multi-angle tricycle body image set to obtain a preprocessed standard image set.

[0026] Use feature point detection and matching algorithms (such as SIFT, SURF, or ORB) to detect key points in the image, which can be points with high contrast and stability on the car body, such as windows, lights, wheels, etc. SIFT (Scale Invariant Feature Transform) detects key points in different scale spaces and generates rotation-invariant descriptors for each key point. It is often used in scenarios that require high-precision matching, but the computational complexity is high. SURF (Speeded Up Robust Features) is an efficient and improved version of SIFT. It uses box filters to accelerate the detection of feature points and the generation of descriptors. The calculation speed is faster than SIFT, but there is still a certain computational complexity. ORB combines the FAST corner detector and the BRIEF descriptor. It generates feature descriptors by detecting corner points in the image and calculating their directions. The descriptor is defined using a binary string, which has a dimension of 32, takes up less storage space, and has a fast calculation speed. For example, use the SIFT algorithm to extract key points in each image. Assuming that the number of feature points in two car body images is 500 and 600 respectively, use the feature matching algorithm to find the corresponding key points between these images.

[0027] The transformation relationship between images is calculated by matching feature points, and false matches are eliminated to obtain accurate transformation relationships. The images are rotated, translated, scaled, and other transformations are performed to correctly connect them. The transformed images are fused and spliced ​​to generate the final panoramic image. The goal of image fusion is to seamlessly connect images and avoid gaps or repeated areas in the splicing. After image preprocessing and panoramic stitching, the final panoramic image of the tricycle body can present a complete view of the entire body, combining information shot from multiple angles, avoiding blind spots caused by any one perspective, and providing a comprehensive display of the body surface.

[0028] Step 4: performing defect area identification and anchor frame marking on the panoramic image of the tricycle body to obtain a tricycle defect area anchor frame set.

[0029] Specifically, defective area identification of the panoramic image of the tricycle body is usually performed based on a standard spray-painted image. The standard spray-painted image is an image of the electric tricycle body without any spray-painted defects, taken under ideal conditions using a high-resolution camera under good lighting conditions. The features of the standard spray-painted image and the panoramic image of the tricycle body are extracted, and the difference between the two feature sets is compared through a loss function to obtain a spray-painted feature loss data set to find areas where defects may exist.

[0030] Feature extraction is achieved by inputting the standard painting image and the panoramic image of the tricycle body into the two sub-networks of the painting feature twin network. Each sub-network independently learns the feature representation of the input image. The painting feature twin network identifies and quantifies the differences between the images, i.e., painting defects, by comparing the feature vectors output by the two sub-networks. The painting feature twin network is a special deep learning network architecture that includes two symmetrical sub-networks for comparing the differences between the two input images. It is usually composed of multiple convolutional layers, each followed by a pooling layer and an activation function, which work together to extract features from the input image. The painting feature twin network outputs two feature sets - the standard painting feature set and the global painting feature set, which identify and quantify the differences between the standard painting image and the actual panoramic image of the vehicle body.

[0031] The difference between the standard spraying image and the actual body panoramic image is a potential image segment that may contain defects. For the identified differences, an anchor frame is generated, and its position and size are finely adjusted to accurately cover the defect area. An anchor frame is a rectangular frame used to mark the region of interest (defect area) in the image. The process of anchor frame marking is to generate rectangular marks in the image according to the identified defect areas to clarify the position and size of these areas. Each anchor frame is assigned a defect category label (such as uneven spraying, bubbles, scratches, etc.) and a confidence score. The panoramic image of the tricycle body is marked with regional anchor frames to obtain all defect area anchor frames. The position information, category label and confidence of each anchor frame are integrated into the tricycle defect area anchor frame set, including all areas in the tricycle panoramic image that may have defects. Each anchor frame includes position (coordinates) and size information, as well as possible defect categories. Through anchor frame marking, defect areas can be accurately marked in the panoramic image, which improves the accuracy of defect detection.

[0032] Step 5: Use the microscopic lens to perform feature extraction and quantification on the anchor frame set of the tricycle defect area to obtain quantitative information of the tricycle defect features.

[0033] Specifically, a microscopic lens is used to collect fine images of the calibrated defect area anchor frame set, that is, a lens with high resolution that can capture object details is used to detect tiny defects or surface textures. The collected high-resolution images are preprocessed, including denoising, contrast enhancement, and improving texture recognizability. The defect assessment network is trained based on feature evaluation indicators such as morphology, area, and distribution. Defect assessment is performed on the collected high-resolution images to identify morphological features, area features, distribution features, etc. The shape of the defect (such as the straight line of the crack, the circularity of the bubble, etc.) is extracted through contour detection or shape recognition algorithms; the area of ​​the defect area is calculated through regional segmentation algorithms (such as threshold segmentation or semantic segmentation based on deep learning); the position distribution of defects on the entire vehicle body is determined (such as whether they are concentrated in a certain area). The extracted features are converted into quantitative information and expressed through numerical indicators.

[0034] The extracted features such as shape, area, and distribution are comprehensively evaluated to obtain quantitative information of each defect area. The influence of each feature is integrated through algorithms (such as weighted average) to generate a final defect evaluation value or quality score. The quantitative information of all defects is summarized to generate the overall defect feature quantitative information of the tricycle body. By using a microscopic lens for feature extraction and quantification, high-precision identification and quantification of the tricycle defect area is achieved, which not only improves the accuracy of defect detection, but also provides detailed information on defect features.

[0035] Step six: Perform statistical evaluation and analysis on the quantitative information of the defect characteristics of the tricycle according to the spray defect index set to generate the electric tricycle body spray defect detection results.

[0036] Specifically, based on the feature extraction of the spray paint defects of the tricycle body, it is quantified into a specific data set, including the shape, area, distribution of each defect, etc., and the defect shape feature set, defect area feature set and defect distribution feature set are sorted out respectively. For example, the defect shape feature set includes data on indicators such as roundness and edge roughness; the defect area feature set includes data on the area size of different defects; and the defect distribution feature set includes data on the location of defects on the body.

[0037] The spray defect index set is a set of key performance indicators used to evaluate the spray quality of the electric tricycle body, including the type, area, shape, distribution, quantity, and impact of defects. Different defect types are extracted from the spray defect index set to form a spray defect type set. According to the spray type, the defect morphology feature set, defect area feature set, and defect distribution feature set are clustered, and similar types are classified into the same category. The feature set corresponding to each type after clustering is statistically evaluated and analyzed, that is, the defect morphology feature set, defect area feature set, and defect distribution feature set corresponding to each defect type are statistically analyzed to obtain comprehensive features.

[0038] According to the influence of defect morphology index, defect area index and defect distribution index, a weight is assigned to each index, and the score of the cluster spray defect comprehensive feature set corresponding to each type is calculated. Then, according to the influence of each defect type on the vehicle body, a corresponding weight coefficient is assigned to each defect type, and it is weighted with the score corresponding to each defect type to obtain the total detection result, that is, the electric tricycle body spray defect detection result. The electric tricycle body spray defect detection result is a test report obtained after statistical analysis, which contains detailed information about the body spray defects, such as the type, location, number, area size, and impact of the defects. According to the results of the evaluation and analysis, a detailed defect detection report is generated, listing all defect types and their numbers, the locations where the defects are concentrated, the severity of the defects, etc., and finally generating a priority repair sequence to ensure that the most serious and most influential defects can be repaired first. By statistically analyzing the various characteristics of spray defects, the impact of defects can be accurately evaluated, more serious defect types can be identified, and missed detection can be avoided.

[0039] Further, step three of this application includes: The method comprises the following steps: performing noise identification and classification on each body image in the multi-angle tricycle body image set in turn to obtain noise characteristic information of the image data; determining a noise filtering adaptive algorithm according to the noise characteristic information of the image data, and performing filtering preprocessing on the multi-angle tricycle body image set based on the noise filtering adaptive algorithm to obtain a standard multi-angle body image set; performing feature point detection on each body image in the standard multi-angle body image set to obtain a multi-angle body feature point set; performing matching and panoramic stitching on the standard multi-angle body image set based on the multi-angle body feature point set to obtain a panoramic image of the tricycle body.

[0040] Specifically, noise recognition is performed once for each image in the multi-angle tricycle body image set to identify and distinguish the noise and real image content in the image, and obtain the image data noise characteristic information, that is, the specific manifestation of noise in the image, such as the type of noise (such as Gaussian noise, salt and pepper noise, etc.), intensity, distribution, etc. Noise is usually irrelevant information introduced by factors such as camera equipment and environmental interference, which may be random pixel changes or irregular point interference. Noise recognition classification distinguishes noise from valid information by analyzing the texture, color distribution and other features of the image. Common noise types include Gaussian noise (electronic noise generated by the image sensor) and salt and pepper noise (errors generated during image transmission or storage). Through statistical analysis methods, the mean and standard deviation of pixels in the image are calculated to identify noise areas. If the pixel fluctuations in certain areas exceed the normal range, they can be marked as noise areas.

[0041] According to the noise characteristic information of the identified image data, the noise filtering adaptive algorithm is determined, that is, the filter parameters are automatically selected and adjusted according to the characteristics of the noise in the image to remove the noise, such as the mean filter, median filter, Gaussian filter, Kalman filter and other noise filtering methods. The adaptive algorithm will dynamically adjust the filtering strategy according to the changes in the noise type and image content to maximize the retention of image details. For example, if the noise characteristic of the image is Gaussian noise when processing a multi-angle vehicle body image set, the standard deviation of the Gaussian filter is adjusted according to the intensity and distribution of the noise to optimize the clarity of the image.

[0042] The multi-angle tricycle body image set is filtered and preprocessed respectively through the noise filtering adaptive algorithm, and the filtering parameters are dynamically adjusted according to the noise intensity of the local area of ​​the image, so as to retain the details of the image while removing the noise, and obtain a standard multi-angle body image set, which is usually a clean, high-quality image set with noise removed or reduced.

[0043] Perform feature point detection on each image in the standard multi-angle vehicle body image set with noise removed, extract feature points with high recognition and robustness, and obtain a multi-angle vehicle body feature point set, which is usually points in the image that are easy to identify and insensitive to position changes, such as corners, edges or areas with significant texture changes, and are invariant in different image viewing angles. For example, use SIFT (Scale Invariant Feature Transform) or detect feature points, including more significant features such as vehicle body edges, windows, and lights.

[0044] According to the obtained multi-angle vehicle body feature point set, the standard multi-angle vehicle body image set is panoramically stitched. By matching the feature points, the transformation matrix between each image is calculated, so that the images from multiple angles are seamlessly stitched into a complete vehicle body image. Panoramic stitching refers to the use of algorithms to synthesize images from different perspectives or angles into a complete image in order to show a complete scene. In other words, panoramic stitching integrates vehicle body images from different angles into a complete vehicle body view. For example, assuming that vehicle body images are taken from four angles, each image contains 500 feature points after feature point detection. By matching these feature points, the transformation relationship between each pair of images (such as rotation matrix, translation matrix, etc.) is calculated. These matrices are used for image stitching, so that the images from each perspective are seamlessly merged, and finally a complete panoramic image of the vehicle body is obtained. The image quality is significantly improved through noise recognition and filtering. Noise filtering removes interference caused by the camera or environmental factors and reduces errors. Feature point detection extracts recognizable points from the vehicle body image, has good robustness, and can stably match images at different angles. Accurate feature point matching and seamless stitching of vehicle body images taken from multiple angles generate a complete panoramic image of the vehicle body, improving the accuracy and comprehensiveness of defect detection and ensuring that the detection results cover the entire vehicle body.

[0045] Further, step 4 of this application includes: The standard spraying image information of the target electric tricycle is acquired, and a spraying feature twin network is built based on the standard spraying image information and the panoramic image of the tricycle body, wherein the spraying feature twin network includes a standard twin sub-network and a defect twin sub-network; based on the spraying feature twin network, feature extraction is performed on the standard spraying image and the panoramic image of the tricycle body to obtain a standard spraying feature set and a global spraying feature set; a feature loss function is introduced to perform loss analysis on the standard spraying feature set and the global spraying feature set to obtain a spraying feature loss data set; based on the spraying feature loss data set, the panoramic image of the tricycle body is marked with a regional anchor frame to obtain the tricycle defect regional anchor frame set.

[0046] Specifically, a standard spray-painted image of the target electric tricycle is collected through a high-resolution camera (such as a digital camera or an industrial camera) as a defect-free reference image, and appropriate lighting and shooting conditions are used to ensure image quality. The standard spray-painted image refers to an ideal or standard image of the surface spraying of the electric tricycle body, which represents the normal, defect-free spraying effect during the spraying process. A spray-painted feature twin network is built based on the standard spray-painted image information and the panoramic image of the tricycle body. The spray-painted feature twin network contains two sub-networks, namely the standard twin sub-network and the defect twin sub-network, which receive different inputs and compare them in some way.

[0047] The standard twin network is used to process standard spraying images and extract feature information of normal spraying, while the defect twin network is used to extract spraying features in the actual panoramic image of the vehicle body, including spraying defects such as scratches, bubbles, uneven spraying, etc. Comparison of the two can help identify differences or defects in the spraying process. The two subnetworks of the spraying feature twin network (standard twin network and defect twin network) share the same network structure and weights. By inputting standard images and vehicle body images, feature information is extracted and compared.

[0048] The standard spraying image and the panoramic image of the tricycle body are extracted through the spraying feature twin network to obtain the standard spraying feature set and the global spraying feature set. The standard spraying feature set and the global spraying feature set contain the same number of feature vectors. The standard spraying feature set represents various properties of normal spraying, such as spraying uniformity, color consistency, surface flatness, etc. The global spraying feature set represents all the information of the actual spraying on the body surface, including defective areas, unevenly sprayed parts, etc. For example, for a standard spraying image, the standard twin network may extract smooth and uniform surface texture features; while the defect twin network may extract features of defective areas on the body surface, such as local uneven spraying or bubbles.

[0049] The feature loss function is used to analyze the difference between the standard spray feature set and the global spray feature set, that is, to analyze the difference between the standard spray feature and the body image feature. The feature loss function is used to measure the size of the feature difference between the two images (the standard spray image and the actual body image). The loss function is used to optimize the weights in the twin network so that the standard spray feature set and the feature set of the actual body image can be closer, thereby helping to better identify defects. The loss function calculates the difference between the two feature sets and generates a spray feature loss data set. The larger the loss, the larger the gap between the two, which means that there may be defects in the body image. If there are obvious differences in color and texture between the standard spray feature set and the global spray feature set, the loss function will output a larger loss value, indicating that there are defects in the body spray. The feature loss data set calculated by the feature loss function records the difference between the standard spray image features and the actual body image features, and identifies defects that occur during the spraying process, such as uneven coating or bubbles.

[0050] The regional anchor frame marking of the panoramic image of the tricycle body is performed according to the spray feature loss dataset. By analyzing the loss data, the area where defects may exist is determined, and the anchor frame is automatically generated in the image to mark the specific location of the defective area. Regional anchor frame marking is to divide the image into multiple small areas and identify potential defects in these areas. These areas are called anchor frames, which indicate where defects may occur. By marking anchor frames in the panoramic image of the body, the defective area can be further analyzed and located. Finally, the defective area anchor frame set of the electric tricycle is obtained, marking all areas where spray defects may exist. The defective area anchor frame set of the tricycle is the area where defects may exist marked in the panoramic image after feature loss analysis. Each anchor frame represents an area of ​​interest in the image, containing defects such as uneven spraying, bubbles or scratches. By using the spray feature twin network to effectively extract the features of the standard spray image and the actual body panoramic image, and analyzing the difference between the two through the feature loss function, the area that may contain spray defects can be accurately identified and marked.

[0051] Further, step five of this application includes: Based on the anchor frame set of the defect area of ​​the tricycle, a defect attention area set is determined; the defect image of the defect attention area set is collected using the microscopic lens to obtain a microscopic image set of the defect area; a spray defect feature assessment target is obtained, and a spray defect assessment network is constructed based on the spray defect feature assessment target training; based on the spray defect assessment network, a defect quantitative assessment is performed on the microscopic image set of the defect area to obtain the quantitative information of the defect features of the tricycle.

[0052] Specifically, based on the anchor frame set of the tricycle defect area, the attention mechanism is used to weight the confidence of the frame, defect category and other features to filter out high-priority areas. An attention score is assigned to each anchor frame, and the areas with higher scores are placed at the front, that is, the areas that need more attention, forming a defect attention area set. A microscopic lens is used to collect defect images of the defect attention area set, and high-resolution images are collected for each defect attention area. The light source is adjusted to eliminate light and shadow interference to obtain a microscopic image set of the defect area. The defect area microscopic image set is a collection of images collected from the defect attention area set using a microscopic lens, which shows the details of the defect area.

[0053] The spray defect assessment objectives were clarified, and three submodules were designed, including a morphology assessment module, an area assessment module, and a distribution assessment module. The collected microscopic images were annotated in detail using annotation tools, including shape (such as ellipse, crack), area (unit: mm²), and position (relative coordinates). The three assessment modules were integrated into a unified network using a multi-task learning framework to obtain a spray defect assessment network. The microscopic image set of the defect area was input into the spray defect assessment network, and the characteristics of the defects (such as shape, size, position, etc.) were accurately numerically analyzed to obtain quantitative information on the defect characteristics of the tricycle, and the defect shape category, area, and distribution hotspot map were output. Based on the microscopic images of the defect area, the network can accurately extract features such as morphology, area, and distribution, narrow the detection range through the attention mechanism, and combine the microscopic lens acquisition and evaluation network to achieve accurate identification and quantification of spray defects, which not only improves the accuracy of defect detection, but also provides detailed information on defect characteristics.

[0054] Furthermore, the present application also includes the following steps: Acquire a set of spray defect images of the target electric tricycle; decompose the defect feature assessment target to obtain defect morphology indicators, defect area indicators and defect distribution indicators; perform defect annotation training on the spray defect image set based on the defect morphology indicators, defect area indicators and defect distribution indicators to generate a defect morphology recognition network, a defect area recognition network and a defect distribution recognition network; perform parallel weighted fusion of the defect morphology recognition network, the defect area recognition network and the defect distribution recognition network to construct the spray defect assessment network.

[0055] Specifically, use a high-resolution camera or a microscopic lens to capture spray images of different parts of the vehicle body to ensure that the defect features are clearly visible. A set of spray defect images of the target electric tricycle containing different types of spray defects is obtained, which records various defects existing in the spraying process of the electric tricycle body, including irregular shape, abnormal area, uneven distribution, etc. The defect feature evaluation target is broken down into multiple indicators, such as morphology (shape and type of defects), area (size of defects), and distribution (spatial distribution of defects on the vehicle body). The defect morphology index refers to the appearance characteristics of the defect, such as shape category (circular, elliptical, strip, etc.); the defect area index is used to measure the size of the defect area in mm²; the defect distribution index is used to count the distribution of defects in various areas of the vehicle body surface, such as concentrated on the edge, local, etc. Specific evaluation criteria are formulated for each indicator. For example, an area deviation of more than 5mm² is considered a serious defect.

[0056] The spray defect image set is annotated, and detailed label information (such as morphological category, area range, and location) is provided for the defect area of ​​each image through defect morphological indicators, defect area indicators, and defect distribution indicators. The annotated data sets are used to train the defect morphological recognition network, defect area recognition network, and defect distribution recognition network, which are used to identify defect shapes, quantify defect areas, and analyze the spatial distribution of defects, respectively. The defect morphological recognition network is constructed using the convolutional neural network architecture, and its classification performance is used to identify defect morphology. Add a fully connected layer with the number of output nodes equal to the defect category (such as scratches, bubbles, and spots), input the preprocessed defect image, and output the defect category. Iteratively train the defect morphological recognition network, use the cross entropy loss function to optimize the classification accuracy, adjust the model parameters until the accuracy reaches the preset accuracy, stop training, and obtain the defect morphological recognition network.

[0057] The target detection model is used to detect and calculate the defect area, and the defect boundary and area value are identified according to the defect image. The bounding box and area information of each defect are annotated to build a training data set. The IoU (Intersection over Union) loss function is used as the evaluation index to optimize the bounding box prediction accuracy. The defect area recognition network is obtained by iterative training through the defect area index and the spray defect image set. The segmentation model is used to build a defect distribution recognition network, and the distribution characteristics of defects on the panoramic image are identified according to the panoramic defect image. The segmentation model is used to segment the defects at the pixel level, add the Softmax layer, and output the probability that each pixel belongs to the defect area. The pixels in the defect area are annotated (that is, whether each pixel belongs to a defect), and the model is optimized using the pixel-level cross entropy as the loss function. The network parameters are optimized using the pixel-level segmentation cross entropy to obtain the defect distribution recognition network. The defect morphology recognition network, defect area recognition network, and defect distribution recognition network generate targeted models by training their respective data sets. Each model focuses on one evaluation target to ensure the accuracy and professionalism of the recognition results.

[0058] The output results of the defect morphology recognition network, defect area recognition network, and defect distribution recognition network are connected in parallel to extract the intermediate features of each network. Through the weighted fusion mechanism, the prediction results of the three networks are combined to generate a comprehensive evaluation result. The weight of each network is optimized based on the performance of the validation set to construct the final spray defect evaluation network. Through the weighted fusion of three independent networks, the comprehensive evaluation capability of complex defects is improved, and the false detection rate and missed detection rate are significantly reduced.

[0059] Furthermore, step six of this application includes: According to the quantitative information of the defect characteristics of the tricycle, the defect morphology feature set, the defect area feature set and the defect distribution feature set are determined; the defect types are extracted from the spray defect index set to obtain the spray defect type set; the defect morphology feature set, the defect area feature set and the defect distribution feature set are clustered according to the spray defect type set to obtain the spray defect clustering feature set; the spray defect clustering feature set is statistically evaluated and analyzed to obtain the electric tricycle body spray defect detection result.

[0060] Specifically, according to the quantitative information of the defect characteristics of the tricycle, the morphology, area and distribution characteristics of the defects are determined respectively, and the defect morphology feature set, defect area feature set and defect distribution feature set are determined. The defect morphology feature set refers to the feature set about the appearance shape of the defect extracted from the image, usually including the edge contour, size, symmetry, etc. of the defect, which helps to distinguish different types of defects; the defect area feature set refers to the feature set describing the area of ​​the defect region, usually based on the size of the area, including the actual coverage area of ​​the defect, the proportion of the body surface occupied, etc.; the defect distribution feature set refers to the distribution characteristics of the defects on the body surface, including the frequency of occurrence of defects, location distribution, regional concentration, etc.

[0061] The spray defect index set is a parameter or standard used to evaluate spray defects. Different defect types are extracted from the spray defect index set to obtain a spray defect type set, including scratches, bubbles, cracks, uneven coating, etc. According to the spray defect type set, the defect morphology feature set, defect area feature set, and defect distribution feature set are clustered. Clustering analysis is performed on spray defects using clustering algorithms (such as K-means clustering and DBSCAN). Defects with similar features are classified into the same category to form a spray defect clustering feature set. Feature clustering is the process of classifying defects with similar features into the same category. Through the clustering algorithm, similar defect types in the defect morphology feature set, defect area feature set, and defect distribution feature set are classified into one category. The K-means clustering algorithm usually has a pre-specified number of categories (such as 4), and the defects are divided into 4 categories according to their characteristics. The center of each cluster is optimized through an iterative process, and each defect is ensured to be assigned to the cluster with the most similar characteristics. DBSCAN (density clustering algorithm) is different from K-means. It does not require the number of clusters to be specified in advance. Instead, it automatically discovers defect clusters based on density distribution and identifies defect clusters in dense areas.

[0062] The spray defect cluster feature set is analyzed in detail by statistical methods, including the calculation of mean, variance, maximum, minimum and other statistics, in order to evaluate the severity, distribution and overall quality of the defects. According to the clustered spray defect feature set, the relevant features of each cluster are extracted, including the morphological features (such as shape, contour), area (such as the size of the defect area) and distribution (such as the distribution density and regional distribution of the defect). The mean and standard deviation of each cluster feature are calculated to measure the distribution of defects. According to the influence weights of the defect morphological index, defect area index and defect distribution index, the relevant features of each extracted cluster are weighted and evaluated to obtain the influence score of each clustered spray defect feature set. Then, according to the degree of influence of each defect type, an influence score is assigned to each defect type, and a corresponding weight is assigned to each defect type.

[0063] According to the cluster analysis results, the defect types are prioritized. Generally, larger, denser, and more abnormal defects are considered as priority for repair. The statistical data and severity scores of all defect types are weighted and combined to obtain the final spray defect detection results, which are a comprehensive score or report that lists the number, location, severity, etc. of each defect type. By statistically analyzing the cluster characteristics, the severity, area, and distribution of each defect type can be more accurately assessed, and defects can be repaired in a targeted manner, especially those that have a greater impact on the quality of the vehicle body, thereby improving production efficiency and product quality.

[0064] Furthermore, the present application also includes the following steps: Perform an influence degree analysis on each defect type in the spray defect type set to obtain a spray defect type influence coefficient set; perform feature information statistics on the spray defect cluster feature set to obtain a cluster spray defect comprehensive feature set; perform influence weight allocation on the defect morphology index, defect area index and defect distribution index to obtain a defect index influence factor set; use the defect index influence factor set to perform weighted integral evaluation on the cluster spray defect comprehensive feature set to obtain a cluster spray defect influence degree set; perform defect weighted analysis on the cluster spray defect influence degree set based on the spray defect type influence coefficient set to obtain the electric tricycle body spray defect detection result.

[0065] Specifically, the impact degree of each defect type in the spray defect type set is analyzed, and an impact coefficient is assigned to each defect type to obtain the spray defect type impact coefficient set. The spray defect type impact coefficient is evaluated based on the different impacts of the defect type on the car body, which is usually determined by experience and actual application results. The spray defect type impact coefficient set represents the degree of impact of different spray defect types on the overall quality. For example, bubbles may have a greater impact on the spray appearance, while scratches may have a greater impact on the structure. The impact coefficient will be assigned according to the importance and severity of the defect type.

[0066] The characteristic information of the spray defect clustering feature set is statistically analyzed, and the defect morphology feature set, defect area feature set, and defect distribution feature set corresponding to each defect type are extracted. The statistical information such as the mean and variance of each defect type is calculated to obtain a comprehensive feature set. According to the different characteristics of spray defects, an influencing weight factor is assigned to each defect indicator (morphology, area, distribution). Usually, a survey is conducted based on experimental data or industry standards. By comparing the degree of influence of different defect types on the vehicle body quality, the weight of each indicator is adjusted. The defect indicator influencing factor set is a set of factors obtained after weighting the defect morphology indicator, defect area indicator, and defect distribution indicator.

[0067] The comprehensive feature set of clustered spraying defects is weighted according to the defect index influencing factor set. An overall evaluation score is obtained by weighted calculation of the defect morphology index, defect area index and defect distribution index of each defect type after clustering and their corresponding weights. The clustered spraying defect influence degree set is obtained, which represents the influence degree of each defect type on the vehicle body quality.

[0068] According to the spray defect type influence coefficient set, that is, the influence degree of each defect type, the clustered spray defect influence degree set is subjected to defect weighted analysis, that is, the evaluation score of each defect type is weighted according to its influence degree, and the influence of different defect types is comprehensively considered to obtain the final electric tricycle body spray defect detection result. By conducting influence degree analysis and feature information statistics, combined with weighted integral evaluation and defect weighted analysis, a comprehensive evaluation and accurate detection of electric tricycle body spray defects can be achieved, which improves the accuracy and comprehensiveness of defect detection.

[0069] In summary, the electric tricycle body spray defect detection method based on machine vision provided in this application has the following technical effects: By acquiring a machine vision perception module, the machine vision perception module integrates a panoramic lens and a microscopic lens; the panoramic lens is used to collect multi-angle images of the target electric tricycle to obtain a multi-angle tricycle body image set; the multi-angle tricycle body image set is preprocessed and panoramically stitched to obtain a panoramic image of the tricycle body; the defect area is identified and anchor frame marked on the panoramic image of the tricycle body to obtain a tricycle defect area anchor frame set; the microscopic lens is used to extract and quantify the features of the tricycle defect area anchor frame set to obtain the tricycle defect feature quantification information; the tricycle defect feature quantification information is statistically evaluated and analyzed according to the spray defect index set to generate the electric tricycle body spray defect detection result. In other words, by integrating different types of lenses, combining panoramic image stitching and microscopic feature extraction, comprehensive detection of electric tricycle body spray defects is achieved, and the accuracy of defect detection is improved.

[0070] Embodiment 2, based on the same inventive concept as the electric tricycle body spraying defect detection method based on machine vision in the aforementioned embodiment 1, the present application also provides an electric tricycle body spraying defect detection system based on machine vision, please refer to the attached Figure 2 , the electric tricycle body spraying defect detection system based on machine vision includes: A lens construction unit 11, the lens construction unit 11 is used to obtain a machine vision perception module, and the machine vision perception module integrates a panoramic lens and a microscopic lens; an image acquisition unit 12, the image acquisition unit 12 is used to perform multi-angle image acquisition on the target electric tricycle through the panoramic lens to obtain a multi-angle tricycle body image set; an image stitching unit 13, the image stitching unit 13 is used to pre-process and panoramic stitch the multi-angle tricycle body image set to obtain a panoramic image of the tricycle body; a defect recognition unit 14, the defect recognition unit 14 is used to identify defect areas and anchor frame marking on the panoramic image of the tricycle body to obtain a tricycle defect area anchor frame set; a feature extraction unit 15, the feature extraction unit 15 is used to use the microscopic lens to perform feature extraction and quantification on the tricycle defect area anchor frame set to obtain tricycle defect feature quantification information; a statistical evaluation unit 16, the statistical evaluation unit 16 is used to perform statistical evaluation and analysis on the tricycle defect feature quantification information according to the spray defect index set to generate an electric tricycle body spray defect detection result.

[0071] Furthermore, the image stitching unit 13 in the electric tricycle body spraying defect detection system based on machine vision is also used for: The method comprises the following steps: performing noise identification and classification on each body image in the multi-angle tricycle body image set in turn to obtain noise characteristic information of the image data; determining a noise filtering adaptive algorithm according to the noise characteristic information of the image data, and performing filtering preprocessing on the multi-angle tricycle body image set based on the noise filtering adaptive algorithm to obtain a standard multi-angle body image set; performing feature point detection on each body image in the standard multi-angle body image set to obtain a multi-angle body feature point set; performing matching and panoramic stitching on the standard multi-angle body image set based on the multi-angle body feature point set to obtain a panoramic image of the tricycle body.

[0072] Furthermore, the defect recognition unit 14 in the electric tricycle body spraying defect detection system based on machine vision is also used for: The standard spraying image information of the target electric tricycle is acquired, and a spraying feature twin network is built based on the standard spraying image information and the panoramic image of the tricycle body, wherein the spraying feature twin network includes a standard twin sub-network and a defect twin sub-network; based on the spraying feature twin network, feature extraction is performed on the standard spraying image and the panoramic image of the tricycle body to obtain a standard spraying feature set and a global spraying feature set; a feature loss function is introduced to perform loss analysis on the standard spraying feature set and the global spraying feature set to obtain a spraying feature loss data set; based on the spraying feature loss data set, the panoramic image of the tricycle body is marked with a regional anchor frame to obtain the tricycle defect regional anchor frame set.

[0073] Furthermore, the feature extraction unit 15 in the electric tricycle body spraying defect detection system based on machine vision is also used for: Based on the anchor frame set of the defect area of ​​the tricycle, a defect attention area set is determined; the defect image of the defect attention area set is collected using the microscopic lens to obtain a microscopic image set of the defect area; a spray defect feature assessment target is obtained, and a spray defect assessment network is constructed based on the spray defect feature assessment target training; based on the spray defect assessment network, a defect quantitative assessment is performed on the microscopic image set of the defect area to obtain the quantitative information of the defect features of the tricycle.

[0074] Furthermore, the feature extraction unit 15 in the electric tricycle body spraying defect detection system based on machine vision is also used for: Acquire a set of spray defect images of the target electric tricycle; decompose the defect feature assessment target to obtain defect morphology indicators, defect area indicators and defect distribution indicators; perform defect annotation training on the spray defect image set based on the defect morphology indicators, defect area indicators and defect distribution indicators to generate a defect morphology recognition network, a defect area recognition network and a defect distribution recognition network; perform parallel weighted fusion of the defect morphology recognition network, the defect area recognition network and the defect distribution recognition network to construct the spray defect assessment network.

[0075] Furthermore, the statistical evaluation unit 16 in the electric tricycle body spraying defect detection system based on machine vision is also used for: According to the quantitative information of the defect characteristics of the tricycle, the defect morphology feature set, the defect area feature set and the defect distribution feature set are determined; the defect types are extracted from the spray defect index set to obtain the spray defect type set; the defect morphology feature set, the defect area feature set and the defect distribution feature set are clustered according to the spray defect type set to obtain the spray defect clustering feature set; the spray defect clustering feature set is statistically evaluated and analyzed to obtain the electric tricycle body spray defect detection result.

[0076] Furthermore, the statistical evaluation unit 16 in the electric tricycle body spraying defect detection system based on machine vision is also used for: Perform an influence degree analysis on each defect type in the spray defect type set to obtain a spray defect type influence coefficient set; perform feature information statistics on the spray defect cluster feature set to obtain a cluster spray defect comprehensive feature set; perform influence weight allocation on the defect morphology index, defect area index and defect distribution index to obtain a defect index influence factor set; use the defect index influence factor set to perform weighted integral evaluation on the cluster spray defect comprehensive feature set to obtain a cluster spray defect influence degree set; perform defect weighted analysis on the cluster spray defect influence degree set based on the spray defect type influence coefficient set to obtain the electric tricycle body spray defect detection result.

[0077] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. Figure 1The electric tricycle body spraying defect detection method based on machine vision and the specific examples in the first embodiment are also applicable to the electric tricycle body spraying defect detection system based on machine vision in this embodiment. Through the above detailed description of the electric tricycle body spraying defect detection method based on machine vision, those skilled in the art can clearly know the electric tricycle body spraying defect detection system based on machine vision in this embodiment, so for the sake of brevity of the specification, it will not be described in detail here. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0078] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0079] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.

Claims

1. A method for detecting defects in spray coating of an electric tricycle body based on machine vision, characterized in that: include: Acquire a machine vision perception module, wherein the machine vision perception module integrates a panoramic lens and a microscopic lens; Capturing multi-angle images of the target electric tricycle through the panoramic lens to obtain a multi-angle tricycle body image set; Preprocessing and panoramic stitching the multi-angle tricycle body image set to obtain a panoramic image of the tricycle body; Perform defect area recognition and anchor frame marking on the panoramic image of the tricycle body to obtain an anchor frame set of the tricycle defect area; Using the microscopic lens to perform feature extraction and quantification on the anchor frame set of the tricycle defect area, to obtain quantified feature information of the tricycle defect; According to the spray defect index set, the quantitative information of the defect characteristics of the tricycle is statistically evaluated and analyzed to generate the spray defect detection results of the electric tricycle body.

2. The method for detecting defects in spray coating of an electric tricycle body based on machine vision according to claim 1, characterized in that: The method of obtaining a panoramic image of a tricycle body comprises: Performing noise recognition and classification on each body image in the multi-angle tricycle body image set in turn to obtain image data noise characteristic information; Determine a noise filter adaptive algorithm according to the noise characteristic information of the image data, and perform filter preprocessing on the multi-angle tricycle body image set based on the noise filter adaptive algorithm to obtain a standard multi-angle body image set; Performing feature point detection on each vehicle body image in the standard multi-angle vehicle body image set to obtain a multi-angle vehicle body feature point set; Based on the multi-angle vehicle body feature point set, the standard multi-angle vehicle body image set is matched and panoramically stitched to obtain the panoramic image of the tricycle body.

3. The method for detecting body spraying defects of an electric tricycle based on machine vision as claimed in claim 1, characterized in that: The obtaining of the anchor frame set of the tricycle defect area comprises: Acquire standard spraying image information of the target electric tricycle, and build a spraying feature twin network based on the standard spraying image information and the panoramic image of the tricycle body, wherein the spraying feature twin network includes a standard twin sub-network and a defect twin sub-network; Based on the spray feature twin network, feature extraction is performed on the standard spray image and the panoramic image of the tricycle body to obtain a standard spray feature set and a global spray feature set; Introducing a feature loss function to perform loss analysis on the standard spray feature set and the global spray feature set to obtain a spray feature loss data set; Based on the spray feature loss data set, the panoramic image of the tricycle body is marked with regional anchor frames to obtain the tricycle defect region anchor frame set.

4. The method for detecting defects in spray coating of an electric tricycle body based on machine vision as claimed in claim 1, characterized in that: The method of obtaining the quantitative information of the defect characteristics of the tricycle includes: Determine a defect attention region set based on the tricycle defect region anchor frame set; Using the microscopic lens to collect defect images of the defect attention area set to obtain a defect area microscopic image set; Obtaining a spraying defect feature evaluation target, and training and constructing a spraying defect evaluation network based on the spraying defect feature evaluation target; Based on the spray defect assessment network, defect quantitative assessment is performed on the microscopic image set of the defect area to obtain quantitative information on the defect characteristics of the tricycle.

5. The method for detecting body spraying defects of an electric tricycle based on machine vision as claimed in claim 4, characterized in that: The construction of a spraying defect assessment network based on the spraying defect feature assessment target training comprises: Acquiring a set of spraying defect images of the target electric tricycle; Decomposing the defect characteristic evaluation target into indicators to obtain defect morphology indicators, defect area indicators and defect distribution indicators; Based on the defect morphology index, defect area index and defect distribution index, defect annotation training is performed on the spray defect image set to generate a defect morphology recognition network, a defect area recognition network and a defect distribution recognition network; The defect morphology recognition network, the defect area recognition network and the defect distribution recognition network are weightedly fused in parallel to construct the spraying defect assessment network.

6. The method for detecting body spraying defects of an electric tricycle based on machine vision as claimed in claim 5, characterized in that: The generating of the electric tricycle body spraying defect detection result comprises: Determine a defect morphology feature set, a defect area feature set, and a defect distribution feature set according to the quantified defect feature information of the tricycle; Extracting defect types from the spraying defect indicator set to obtain a spraying defect type set; Performing feature clustering on the defect morphology feature set, the defect area feature set, and the defect distribution feature set according to the spraying defect type set to obtain a spraying defect clustering feature set; Statistical evaluation and analysis are performed on the spray defect clustering feature sets to obtain the electric tricycle body spray defect detection results.

7. The method for detecting defects in spraying of an electric tricycle body based on machine vision as claimed in claim 6, characterized in that: The method of obtaining the electric tricycle body spraying defect detection result comprises: Analyzing the influence degree of each defect type in the spraying defect type set to obtain a spraying defect type influence coefficient set; Performing feature information statistics on the spray defect cluster feature sets respectively to obtain a cluster spray defect comprehensive feature set; Assigning influence weights to the defect morphology index, defect area index, and defect distribution index to obtain a defect index influence factor set; Using the defect index influencing factor set to perform weighted integral evaluation on the clustered spray defect comprehensive feature set, to obtain a clustered spray defect influence degree set; Based on the spray defect type influence coefficient set, a defect weighted analysis is performed on the clustered spray defect influence degree set to obtain the electric tricycle body spray defect detection result.

8. The electric tricycle body spraying defect detection system based on machine vision is characterized by: The steps for implementing the method for detecting defects in spraying of an electric tricycle body based on machine vision according to any one of claims 1 to 7, the electric tricycle body spraying defect detection system based on machine vision comprises: A lens construction unit, the lens construction unit is used to obtain a machine vision perception module, the machine vision perception module integrates a panoramic lens and a microscopic lens; An image acquisition unit, the image acquisition unit is used to acquire multi-angle images of the target electric tricycle through the panoramic lens to obtain a multi-angle tricycle body image set; An image stitching unit, the image stitching unit is used to pre-process and stitch the multi-angle tricycle body image set to obtain a panoramic image of the tricycle body; A defect recognition unit, the defect recognition unit is used to identify defect areas and mark anchor frames on the panoramic image of the tricycle body to obtain an anchor frame set of defect areas of the tricycle; A feature extraction unit, the feature extraction unit is used to use the microscopic lens to perform feature extraction and quantification on the anchor frame set of the tricycle defect area to obtain quantified feature information of the tricycle defect; A statistical evaluation unit is used to perform statistical evaluation and analysis on the quantified information of the defect characteristics of the tricycle according to the spray defect index set, and generate a spray defect detection result for the electric tricycle body.

Images