A high-speed guardrail board plastic spraying defect detection method and system

Abstract

The application relates to the technical field of high-speed guardrail plate plastic spraying defect detection, in particular to a high-speed guardrail plate plastic spraying defect detection method and system, wherein the high-speed guardrail plate plastic spraying defect detection method comprises image preprocessing, data set making, model training, image detection, image splicing and post-processing, different defects with the same characteristics can be distinguished, normal images and defect images are obtained, the defect images are further detected, a screening threshold is set, the defect images are further distinguished into necessary treatment type images and unnecessary treatment type images through calculation of the number, area and spacing of the defects, and a high-speed guardrail plate plastic spraying defect detection system is built according to the high-speed guardrail plate plastic spraying defect detection method. The application has the effects of quantitatively evaluating the severity of defects and improving detection efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of high-speed guardrail plate plastic spraying defect detection, in particular to a high-speed guardrail plate plastic spraying defect detection method and system. BACKGROUND

[0002] Guardrail plate plastic spraying is a surface treatment technology, mainly used to improve the corrosion resistance, appearance and service life of the guardrail plate; defects are inevitably produced in the process of guardrail plate plastic spraying, and the plastic spraying defects may cause the quality, performance and appearance of the guardrail plate to not meet the requirements.

[0003] The plastic spraying defects usually include bubbles, particles, sagging, pits, scratches, scratches, damage and deformation, etc. Among them, some will affect the corrosion resistance of the guardrail plate, such as bubbles and scratches; and some defects only affect the appearance of the guardrail plate and do not damage the integrity of the plastic spraying coating, such as sagging; in actual operation, some defects may be accepted or tolerated due to various reasons (such as cost, time, process difficulty, etc.), in which case only the defects that must be treated need to be detected and distinguished, but defects with different degrees of influence may have similar characteristics and need to be distinguished.

[0004] In the prior art, the method for detecting high-speed guardrail plate plastic spraying defects usually detects all types of defects, however, the existing image processing method can only identify the existence of defects, and cannot quantitatively evaluate the severity of the defects, so manual judgment is needed to determine whether repair is needed, resulting in low detection efficiency. SUMMARY

[0005] In order to improve the detection efficiency, the present application provides a high-speed guardrail plate plastic spraying defect detection method.

[0006] In the first aspect, the present application provides a high-speed guardrail plate plastic spraying defect detection method, which adopts the following technical scheme:

[0007] A high-speed guardrail plate plastic spraying defect detection method, comprising the following steps:

[0008] Image preprocessing: obtaining historical images and performing image preprocessing, the historical images containing different types of defects with the same characteristics;

[0009] Data set making: data labeling and making training set and validation set for different types of defects with the same characteristics in the historical images;

[0010] Model training: training the classification model using the training set, and evaluating the classification accuracy of the model using the validation set to obtain the trained model;

[0011] Image detection: acquire an image to be detected, input the image to be detected into the trained model for detection, and obtain an identification result image with defect identification marks;

[0012] Image stitching: stitch multiple identification result images combined into the same guardrail plate to synthesize a complete guardrail plate image, each guardrail plate image containing one guardrail plate; according to the identification result, the guardrail plate image is re-divided into normal images and defect images;

[0013] Post-processing: including quantity statistics step, judgment step, repair step and qualified evaluation step;

[0014] Quantity statistics: counting the number of defects in the defect images in the unit area within the range, and setting a threshold value of the number of defects allowed to appear in the unit area ; ;

[0015] The calculation model of the number of defects is as follows:

[0016] = ,

[0017] Wherein, represents the total number of defects in the unit area , i = 1, 2, 3…; represents the number of defects of the i-th type in the unit area ; ;

[0018] Judgment: when the number of defects is less than or equal to the set number threshold , the defect image is divided into a non-essential processing type image, and the qualified evaluation step is executed; when the number of defects is greater than the set number threshold , the defect image is divided into a necessary processing type image, and the repair step is executed; Qualified evaluation: the normal image and the non-essential processing type image are qualified images and do not need to be processed;

[0019] Repair: the necessary processing type image is unqualified image and needs to be repaired.

[0020]

[0021] ​​​​​By adopting the above technical solution, this application first acquires historical images, obtains training and validation sets through image preprocessing and data annotation, then trains the classification model using the training set and evaluates the model's classification accuracy using the validation set to obtain a trained model; inputs the image to be detected into the trained model to obtain the recognition result image, stitches the recognition result images together, combines multiple recognition result images located on the same guardrail into one, and distinguishes between normal and defective images for subsequent processing; finally, post-processing is performed to count the number of defects in the defective images and set a threshold for the number. The defective images are further divided into images that do not require processing and images that require processing, and then qualified and repaired. It can not only identify the defect type, but also determine whether repair is needed by counting the number of defects. This realizes the quantification of the severity of defects, reduces manual identification, effectively improves detection efficiency, and reduces the uncertainty of manual identification.

[0022] Optionally, a reclassification step may be included between the judgment step and the conformity assessment step;

[0023] Reclassification includes the steps for calculating the impact health value and the re-judgment step;

[0024] Impact Health Value Calculation: Different weights are assigned to different types of defects based on their degree of impact, and the impact per unit area is calculated. The impact of the number of defects on health value And set a health value threshold. ;

[0025] Impact on health value The calculation model is as follows:

[0026] ,

[0027] in, Indicates the first The health value of the influence per unit area , It is the first The number of class defects, , These are the corresponding weights, and the sum of the weights is... ;

[0028] Further assessment: When the impact on health value Less than or equal to the health value threshold When necessary, images that do not require unnecessary processing are stored as images that do not require unnecessary processing, and a conformity assessment step is performed; when the impact on health value... Greater than the health threshold If necessary, the image that was not processed is saved as an image that requires processing, and the repair steps are performed.

[0029] By adopting the above technical solution, the impact of defects on the quality of guardrails is quantified based on defect type, tolerability, and degree of influence. Different weights are assigned to the number of different types of defects, with high weights given to defects with high impact and low weights given to defects with low impact. This can influence the setting of the health value, increase the proportion of high-impact defects in the health value, and enhance the data value and credibility of the health value. By assessing the severity of defects through the health value, the quality of defect assessment can be effectively improved, and the determination of whether defect images need to be processed can be more accurate.

[0030] Optionally, a defect area statistics step and a third judgment step are set between the re-judgment step and the conformity assessment step;

[0031] Defect area statistics: Defect area of ​​images that do not require necessary processing. Perform calculations and set a threshold for the defect area. ;

[0032] Defect area The calculation model is as follows:

[0033] ,

[0034] in, Indicates the first unit area The total area of ​​defects within. ; Represents unit area The area of ​​the i-th type of defect, ;

[0035] Third judgment: When the defect area Less than or equal to the set defect area limit threshold When necessary, images that do not require unnecessary processing are stored as images that do not require unnecessary processing, and the conformity assessment steps are performed; when the defect area The defect area exceeds the set threshold. At that time, images that do not require processing are classified as images that require processing, and repair steps are performed.

[0036] Defect area limit threshold The calculation model is as follows:

[0037] ,

[0038] Where P is a proportionality coefficient. The range of values ​​is , Indicates unit area.

[0039] By adopting the above technical solution, the defect area of ​​non-essential processing images that meet the defect quantity requirements is calculated. The calculation, when the defect area Not exceeding the set area threshold When necessary, images that do not require unnecessary processing will continue to be stored as images that do not require unnecessary processing; when the defect area Exceeding the set area threshold At that time, images that do not require processing are classified as images that require processing; by introducing defect area, the number of defects can be reduced to meet requirements, but product quality problems caused by excessive defect coverage area can be addressed by setting an area threshold. It can reduce the unit area The defect coverage area is limited to an acceptable range; by comprehensively assessing the severity of defects through the number and area of ​​defects, the limitations of a single assessment standard can be reduced, and the quality of product testing can be improved.

[0040] Optionally, a defect merging and discrimination step is set between the third judgment step and the conformity assessment step;

[0041] Defect merging and discrimination: including defect distance statistics, distance judgment, and defect adjustment steps;

[0042] Defect statistics: the distance between defects in images that do not require unnecessary processing. Perform calculations and set spacing thresholds. ;

[0043] Distance judgment: when the defect spacing Perform calculations and set spacing thresholds. ;

[0044] Distance judgment: when the defect spacing Less than the set spacing threshold When the defect spacing is... Greater than or equal to the set spacing threshold At the same time, maintain the defect area The parameters remain unchanged, and the conformity assessment procedure is performed; the spacing threshold is also unchanged. It is a constant representing the minimum distance between defects;

[0045] Defect adjustment: Adjustment of defect area Adjust the coefficients and update the defect area. The calculation model is as follows:

[0046] ,

[0047] in, Indicates the first unit area The total area of ​​defects within. ; Represents unit area The area of ​​the i-th type of defect, , n represents the defect spacing Less than the set spacing threshold The number of times.

[0048] By adopting the above technical solution, when the defect spacing Less than the set spacing threshold At the same time, the adhesion of the coating between defects will also decrease, making it easier for the defect area to expand. Increased, and the defect spacing Less than the set spacing threshold The more times it is done, the larger the defect area becomes. The larger the defect, the smaller the defect area can be through defect adjustment. With defect spacing Less than the set spacing threshold The number of times the defect is applied increases, resulting in a larger defect area. The results are more accurate, providing stronger data support for the further classification of images that do not require unnecessary processing, and effectively improving the detection quality and efficiency of the product.

[0049] Optionally, different quantity thresholds can be set for the outer and inner surfaces of the guardrail. Health threshold Defect area limit threshold and spacing threshold Wherein, the outer surface is the main surface that can be directly seen during normal use, the inner surface is the surface that cannot be observed during normal use, and the inner surface has a higher tolerance for defects than the outer surface.

[0050] By adopting the above technical solution, different defect detection standards can be set for the different tolerance levels of defects on different surfaces of the guardrail. The outer surface is the main surface and is easily in direct contact with vehicles, flying sand and gravel, wind and rain, etc., so the highest detection standard can be set. The inner surface is the surface that is not in contact during normal use, so the detection standard can be relatively lower than that of the outer surface. This helps to reduce the rework rate of the guardrail, save costs, and improve work efficiency.

[0051] Optionally, between the distance determination step and the conformity assessment step, an edge detection step and a fourth determination step are also provided;

[0052] Edge detection step: Perform edge detection on defects in non-essential processing images to determine whether the defects are located at the edge of the guardrail.

[0053] Fourth judgment step: When the defect is not located at the edge of the guardrail, the non-essential processing image is stored as a non-essential processing image and the pass / fail assessment step is performed; when the defect is located at the edge of the guardrail, the non-essential processing image is directly classified as an essential processing image and the repair step is performed.

[0054] By adopting the above technical solution, guardrails with defects at the edges can be directly classified as images requiring rework and repair; because defects at the edges are prone to friction and breakage, leading to detachment and affecting the quality of the guardrails.

[0055] Secondly, the powder coating defect detection system for highway guardrails provided in this application adopts the following technical solution:

[0056] A system for detecting defects in powder coating on highway guardrails includes:

[0057] Image preprocessing module: used to retrieve historical images from the database and perform image preprocessing; the image preprocessing includes grayscale conversion, edge detection, and cropping of the historical images;

[0058] Dataset creation module: The input end is electrically connected to the output end of the image preprocessing module, and is used to annotate different types of defects with the same characteristics in historical images and create training and validation sets;

[0059] Model training module: The input end is electrically connected to the output end of the dataset creation module, and is used to train the classification model using the training set and evaluate the classification accuracy of the model using the validation set to obtain the trained model;

[0060] Image detection module: includes a pre-trained model; used to retrieve the image to be detected from the file system, input the image to be detected into the pre-trained model for detection, and obtain the recognition result image with defect recognition markers;

[0061] Image stitching module: The input end is electrically connected to the output end of the image detection module, and is used to stitch together multiple recognition result images that are combined to form the same guardrail panel to synthesize a complete guardrail panel image. Each guardrail panel image contains one guardrail panel. According to the recognition results, the guardrail panel images are reclassified into normal images and defective images.

[0062] Post-processing module: The input end is electrically connected to the output end of the image stitching module. It is used to perform quantitative analysis on the defects in the defect image. By counting the number of defects per unit area and setting a defect number threshold, it judges whether the number of defects per unit area meets the requirements and further divides the defect image into images that do not require processing and images that require processing.

[0063] In summary, this application includes at least one of the following beneficial technical effects:

[0064] 1. This application can not only identify the type of defect, but also determine whether repair is needed by counting the number of defects, thereby quantifying the severity of defects, reducing manual identification, effectively improving detection efficiency, and reducing the uncertainty of manual identification.

[0065] 2. This application quantifies the impact of defects on guardrail quality based on defect type, tolerability, and degree of influence. Different weights are assigned to the number of different types of defects, with higher weights for defects with greater influence and lower weights for defects with less influence. This can influence the setting of the health value, increase the proportion of defects with greater influence in the health value, and enhance the data value and credibility of the health value. By assessing the severity of defects through the health value, the quality of defect assessment can be effectively improved, and the determination of whether defect images need to be processed can be more accurate.

[0066] 3. This application calculates the defect area of ​​non-essential images that meet the defect quantity requirements. When the defect area does not exceed the set area threshold, the non-essential images are stored as non-essential images. When the defect area exceeds the set area threshold, the non-essential images are classified as essential images. By introducing the defect area, product quality problems caused by excessively large defect coverage areas despite meeting the defect quantity requirements can be reduced. By setting an area threshold, the defect coverage area per unit area can be limited to an acceptable range. By comprehensively evaluating the severity of defects through the combination of defect quantity and defect area, the limitations of a single evaluation standard can be reduced, and the quality of product inspection can be improved.

[0067] 4. This application, through defect adjustment, can increase the defect area as the number of times the defect spacing is less than the set spacing threshold increases, making the obtained defect area result more accurate, providing stronger data support for further segmentation of images that do not require unnecessary processing, and effectively improving the detection quality and efficiency of the product. Attached Figure Description

[0068] Figure 1 This is a flowchart of Embodiment 1 of this application;

[0069] Figure 2 This is a flowchart of Embodiment 2 of this application;

[0070] Figure 3 This is a flowchart of Embodiment 3 of this application;

[0071] Figure 4 This is a flowchart of Embodiment 4 of this application. Detailed Implementation

[0072] The following combination Figures 1 to 4 This application will be described in further detail.

[0073] This embodiment discloses a method for detecting defects in the powder coating of highway guardrails.

[0074] Example 1: Refer to Figure 1 A method for detecting defects in powder coating on highway guardrails, comprising the following steps:

[0075] Image preprocessing: Historical images are acquired and preprocessed. These images contain different types of defects with similar characteristics. Obvious defects such as missed sprays and scratches can be directly identified as requiring processing. However, bulging defects are difficult to distinguish visually due to their similar shapes, and there are many types of bulging defects with varying degrees of impact. Therefore, this embodiment will differentiate bulging defects, which typically include bubbles, particles, and runs. Bubbles are cavities of varying sizes, either round or elliptical. Particles are granular substances embedded in the coating layer, mostly appearing as irregular dots. Runs are marks formed by the coating flowing downwards due to gravity during the drying process, creating curtain-like or tear-like marks of varying widths and lengths. Based on the characteristics of these defects, bubbles and particles easily lead to coating peeling and cracking, severely affecting the appearance quality and performance of the coating. Runs affect the uniformity and aesthetics of the coating.

[0076] Specifically, image preprocessing includes grayscale conversion, edge detection, and cropping of historical images. Grayscale conversion uses a grayscale algorithm to convert a color image into a grayscale image. In this embodiment, a weighted average method is used for grayscale conversion. Grayscale images can highlight the structure and shape information of the image, simplify image information, and reduce the complexity of image processing. Edge detection is used to extract the contours of defects, providing a basis for subsequent image segmentation algorithms and edge discrimination steps. Cropping is used to remove unnecessary parts from historical images and crop them to a specific size to meet the input requirements of the classification model and unify the image size in the dataset. During the cropping process, an image cropping algorithm is used to crop the images and the cropped images are numbered to facilitate identification of the initial position of the cropped images, correspondence of defect positions, and subsequent image stitching algorithms to stitch and restore the images.

[0077] Dataset creation: Data annotation is performed on different types of defects with the same characteristics in historical images, and training and validation sets are created. Data annotation can be done manually or using semi-automated annotation tools. Algorithms are used to perform preliminary annotation on historical images, followed by manual correction. During the data annotation process, different attribute values ​​need to be assigned to different types of defects.

[0078] Compared to flawless images, the number of defective images is very small. To address this issue, upsampling and image augmentation techniques are used to expand the dataset for defective images, increasing the diversity and quantity of the data and improving training performance.

[0079] Model training: The classification model is trained using the training set and evaluated using the validation set to obtain a trained model. Based on the data characteristics of small samples in multi-class classification in this embodiment, the classification model is either ResNet or U-Net, and the classification model is trained and validated using the dataset.

[0080] Specifically, the classification accuracy of the model is evaluated using basic evaluation metrics, including precision, recall, and F1 score. Precision measures the proportion of instances that the model predicts as positive but are actually positive. Recall measures the model's ability to identify positive instances. The F1 score is the harmonic mean of precision and recall, used to comprehensively measure the model's performance.

[0081] Image detection: The image to be detected is acquired by a high-resolution camera and preprocessed to create a detection dataset. The preprocessed image to be detected is then input into a trained model for detection to obtain a recognition result image with defect identification markers. The preprocessing process for the image to be detected is the same as that for historical images.

[0082] Image stitching: Based on the image numbering during the image cropping process, multiple recognition result images that are combined to form the same guardrail are stitched together using an image stitching algorithm to synthesize a complete guardrail image. Each guardrail image contains a complete guardrail, which facilitates the statistical analysis of defects located on the same guardrail and lays the groundwork for subsequent work. Based on whether there are recognition defects in the stitched image, the guardrail images are distinguished into normal images and defective images.

[0083] Specifically, the images are stitched together according to the shooting location, sequence, and cropping number. After stitching, the location of defects is determined through visual inspection. Images without defects are classified as normal images, and images containing defects are classified as defective images, so as to conduct further quantitative analysis on the defective images.

[0084] Post-processing includes quantity statistics, judgment, repair, and conformity assessment.

[0085] Quantitative statistics: per unit area of ​​defective images The number of defects within the specified range is counted, and the unit area is... The selection is made by gradually moving the displacement from the upper left corner of the defect image, with a unit area... The displacement is determined based on the minimum acceptable defect area, and a unit area is set. Threshold for the number of defects allowed within ;

[0086] Number of defects The calculation model is as follows:

[0087] ,

[0088] in, Indicates the first unit area The total number of defects within. ; Represents unit area The number of type i defects, ;

[0089] In this embodiment, the first type of defect is defined as bubbles, the second type of defect as particles, and the third type of defect as runs; according to the detection requirements, the unit area is... Set as threshold Set to 3, that is, the th unit area unit area Total number of defects unit area The maximum number of defects allowed is 3.

[0090] Judgment: When the number of defects Less than or equal to the set threshold quantity When the number of defects is high, the defect images are classified into non-essential processing images, and a conformity assessment procedure is performed; Greater than the set threshold number When necessary, defective images are classified into images requiring necessary processing, and repair steps are performed.

[0091] Conformity assessment: Images of guardrails without defects are classified as normal images, and images with defects whose severity is within an acceptable range are classified as images that do not require processing. Both normal images and images that do not require processing are qualified images and do not need to be processed. This can improve the quality and efficiency of defect detection.

[0092] Repair: Images with defects whose severity exceeds the acceptable range are classified as images requiring necessary processing. Images in the necessary processing category are unacceptable images and need to be reworked and repaired.

[0093] The implementation principle of Embodiment 1 of this application is as follows: First, historical images are acquired. Training and validation sets are obtained through image preprocessing and data annotation. Then, the classification model is trained using the training set and evaluated using the validation set to obtain a trained model. The image to be detected is input into the trained model to obtain the recognition result image. The recognition result images are then stitched together, combining multiple recognition result images located on the same guardrail panel into one image, and classifying them into normal and defective images. Finally, post-processing is performed to count the number of defects in the defective images and set a threshold for the number of defects. The defective images are further divided into images requiring no processing and images requiring processing, and then assessed for compliance and repaired. This not only identifies the defect type but also assesses the defect per unit area. The number of defects within the scope is counted and judged, and the severity of defects is quantitatively assessed, which effectively improves the detection efficiency. Example

[0094] Reference Figure 2 The difference between this embodiment and embodiment 1 is that a reclassification step is set between the judgment step and the conformity assessment step; the reclassification step includes an impact health value calculation step and a rejudgment step.

[0095] Impact Health Value Calculation: Different weights are assigned to different types of defects based on their degree of impact, and the impact per unit area is calculated. The impact of the number of defects on health value And set a health value threshold. ;

[0096] Impact on health value The calculation model is as follows:

[0097] ,

[0098] in, Indicates the first The health value of the influence per unit area ; It is the first The number of class defects, , These are the corresponding weights, and the sum of the weights is... .

[0099] In this embodiment, based on product characteristics and industry experience, bubbles have hollow interiors and thin coatings, making them prone to breakage; particles contain impurities and are easily detached under external force; while sagging has a solid internal structure and is generally not prone to breakage. Based on the characteristics of the above defects, the degree of influence of the defect is set as that of bubbles. Particles Flowing, that is ,and Set health value threshold , and quantity threshold equal.

[0100] Further assessment: When the impact on health value Less than or equal to the health value threshold When necessary, images that do not require unnecessary processing are stored as images that do not require unnecessary processing, and the conformity assessment steps are performed; when the impact on health value... Greater than the health threshold If necessary, images that do not require processing are saved as images that require processing, and repair steps are performed.

[0101] The implementation principle of Embodiment 2 of this application is as follows: the impact degree of different types of defects is classified into levels, and different weights are set for each level. Defects with a high impact degree are given high weights, and defects with a low impact degree are given low weights, thereby increasing the impact degree of defects with a high impact degree in the impact health value. The proportion of [something] increases the impact of health value. The value and credibility of the data.

[0102] Example 3:

[0103] Reference Figure 3 The difference between this embodiment and embodiment 2 is that a defect area statistics step and a third judgment step are set between the re-judgment step and the conformity assessment step.

[0104] Among them, the defect area statistics are the defect area of ​​images that do not require unnecessary processing. Perform calculations and set a threshold for the defect area. ;

[0105] Defect area The calculation model is as follows:

[0106] ,

[0107] in, Indicates the first unit area The total area of ​​defects within. ; Represents unit area Inner The area of ​​the defect type, In this embodiment, the first unit area Total area of ​​defects within .

[0108] Third judgment: When the defect area Less than or equal to the set defect area limit threshold When necessary, images requiring unnecessary processing are stored as such, and the conformity assessment steps are performed; when the defect area... The defect area exceeds the set threshold. At that time, images that do not require processing are classified as images that require processing, and repair steps are performed.

[0109] Defect area limit threshold The calculation model is as follows:

[0110] ,

[0111] in, It is a proportionality coefficient. The range of values ​​is , This represents the unit area; in this embodiment, the detection standard is based on the allowable number of defects per unit area during the powder coating inspection process. Within the range, There can be no more than 3 bulges, and P is set to a value of 1. That is, per unit area Defect area within The proportion does not exceed .

[0112] Furthermore, a defect merging and discrimination step is set up between the third judgment step and the conformity assessment step; the defect merging and discrimination step is applicable to defects of the easy-to-fall-off type, and the defect merging and discrimination includes a defect distance statistics step, a distance judgment step, and a defect adjustment step.

[0113] Defect statistics: By analyzing the distance between defects in images that do not require unnecessary processing. Perform calculations and set spacing thresholds. It is used to limit the distance between defects.

[0114] Specifically, the minimum distance between two defect edges is set as the defect spacing. and the defect spacing Spacing threshold Compare them.

[0115] Distance judgment: when the defect spacing Less than the set spacing threshold When the defect spacing is... Greater than or equal to the set spacing threshold At the same time, maintain the defect area The process remains unchanged, and the conformity assessment procedure is performed; due to the acceptable defect spacing. It is a fixed value that does not change with other variables, so the spacing threshold is a constant, representing the minimum acceptable spacing between defects.

[0116] Defect adjustment: Adjustment of defect area Adjust the coefficients and update the defect area. The calculation model is as follows:

[0117] ,

[0118] in, Indicates the first unit area The total area of ​​defects within. ; Represents unit area Inner The area of ​​the defect type, , , Indicates the defect spacing Less than the set spacing threshold The number of times.

[0119] In this embodiment, the first unit area Total area of ​​defects within That is, with the defect spacing Less than the set spacing threshold With the increase in the number of times, the defect area The faster the growth rate.

[0120] The implementation principle of Embodiment 3 of this application is as follows: The defect image that meets the defect quantity requirements in Embodiment 2 is processed by measuring the defect area. The calculation divides the defect images into non-necessary processing images and necessary processing images to filter out individual defects with large areas; and based on the defect spacing... For defects that are prone to detachment, adjust the defect area and update the defect size. The calculation includes the area between two closely spaced defects, resulting in the obtained defect area. More precise.

[0121] In other embodiments, different inspection standards are set for different surfaces of the guardrail. These different surfaces typically include outer and inner surfaces. The outer surface is the main surface, which is easily in direct contact with vehicles, flying sand and gravel, wind and rain, etc., and therefore has a high inspection standard. The inner surface is not touched during normal use, and therefore has a lower inspection standard compared to the outer surface. Different quantity thresholds can be set according to specific circumstances. Health threshold Defect area limit threshold and spacing threshold This reduces the repair rate of guardrail panels, saves costs, and improves work efficiency.

[0122] In other implementations, edge detection is performed on images that do not require processing. Defects located on the edge of the guardrail are directly classified as images that require processing, regardless of their degree of impact, and are then reworked and repaired. Example

[0123] Reference Figure 4 This embodiment discloses a defect detection system for powder coating on highway guardrails, the detection system comprising:

[0124] Image preprocessing module: used to retrieve historical images from the database and perform image preprocessing, including grayscale conversion, edge detection, and cropping of historical images.

[0125] Dataset creation module: The input end is electrically connected to the output end of the image preprocessing module. It is used to annotate different types of defects with the same characteristics in historical images and create training and validation sets. Different attribute values ​​are assigned to different types of defects. The dataset can also be expanded through upsampling and image enhancement techniques.

[0126] Model training module: The input end is electrically connected to the output end of the dataset creation module. It is used to train the classification model using the training set and evaluate the classification accuracy of the model using the validation set to obtain the trained model. The classification accuracy of the model is evaluated using basic evaluation metrics to comprehensively measure the performance of the model.

[0127] Image detection module: The input end is electrically connected to the output end of the model training module, including the trained model; it is used to obtain the image to be detected from the file system. The image to be detected can be obtained by a high-resolution camera. The image to be detected is preprocessed to form a detection dataset. The preprocessing of the detection dataset is consistent with the standard of the training dataset. The preprocessed image to be detected is input into the trained model for detection to obtain the recognition result image with defect recognition labels.

[0128] Image stitching module: The input end is electrically connected to the output end of the image detection module. It is used to stitch together multiple recognition result images that are combined to form the same guardrail, and synthesize a complete guardrail image. Each guardrail image contains one guardrail. Based on whether there are recognition defects in the stitched image, the guardrail image is re-distinguished into normal images and defective images.

[0129] Post-processing module: The input end is electrically connected to the output end of the image stitching module. It is used to quantitatively analyze the defects in the defect image. By counting the number of defects per unit area and setting a defect number threshold, it judges whether the number of defects per unit area meets the requirements. When the number of defects is less than or equal to the set threshold, the defect image is classified as an image that does not require processing and is rated as a qualified image that does not require processing. When the number of defects is greater than the set threshold, the defect image is classified as an image that requires processing and the repair steps are executed for rework repair.

[0130] Furthermore, by setting an impact health value, different weights are assigned to different types of defects, and the impact health value of the number of defects per unit area is calculated, with a health value threshold set. When the impact health value is less than or equal to the health value threshold, images that do not require necessary processing are stored as images that do not require necessary processing and are rated as qualified images without processing. When the impact health value is greater than the health value threshold, images that do not require necessary processing are stored again as images that require necessary processing, and a repair step is performed for rework repair.

[0131] Furthermore, by statistically analyzing the defect area per unit area and setting a defect area limit threshold, when the defect area is less than or equal to the set defect area limit threshold, images that do not require necessary processing are stored as images that do not require necessary processing and are rated as qualified images without processing; when the defect area is greater than the set defect area limit threshold, images that do not require necessary processing are classified as images that require necessary processing and repair steps are performed for rework repair.

[0132] Furthermore, for defects that are prone to detachment, the defect spacing between defects in images that do not require processing is calculated, and a spacing threshold is set. When the defect spacing is greater than or equal to the set spacing threshold, the defect area remains unchanged and the image is rated as qualified without processing. When the defect spacing is less than the set spacing threshold, a defect adjustment step is performed, and the adjusted area is judged as the defect area.

[0133] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for detecting defects in powder coating on highway guardrails, characterized in that: Includes the following steps: Image preprocessing: Acquire historical images and perform image preprocessing, wherein the historical images contain different types of defects with the same characteristics; Dataset creation: Data annotation is performed on different types of defects with the same characteristics in the historical images, and training and validation sets are created; Model training: The classification model is trained using the training set, and the classification accuracy of the model is evaluated using the validation set to obtain the trained model; Image detection: Acquire the image to be detected, input the image to be detected into the trained model for detection, and obtain the recognition result image with defect recognition labels; Image stitching: Multiple images of the recognition results that are combined to form the same guardrail panel are stitched together to synthesize a complete guardrail panel image, with each guardrail panel image containing one guardrail panel; based on the recognition results, the guardrail panel images are reclassified into normal images and defective images; Post-processing includes quantity statistics, judgment, repair, and conformity assessment steps. Quantitative statistics: per unit area The defect images within the specified range are used to count the number of defects, and a unit area is set. Threshold for the number of defects allowed within ; Number of defects The calculation model is as follows: = ， in, Indicates the first unit area The total number of defects within. ; Represents unit area The number of type i defects, ; Judgment: When the number of defects Less than or equal to the set threshold quantity When the defective images are classified into non-essential processing categories, a conformity assessment step is performed; when the number of defects... Greater than the set threshold number When necessary, the defective images are classified into images requiring necessary processing, and repair steps are performed. Acceptance assessment: The normal images and the images that do not require necessary processing are considered acceptable images and do not require further processing. Repair: The images described in the necessary processing category are unqualified and need to be reworked and repaired.

2. The method for detecting defects in powder coating of highway guardrails according to claim 1, characterized in that: A reclassification step is also included between the judgment step and the conformity assessment step; Reclassification includes the steps for calculating the impact health value and the re-judgment step; Impact Health Value Calculation: Different weights are assigned to different types of defects based on their degree of impact, and the impact per unit area is calculated. The impact of the number of defects on the health value W, and setting a health value threshold. ; Impact on health value The calculation model is as follows: ， in, Indicates the first The health value of the influence per unit area ; It is the first The number of class defects, , These are the corresponding weights, and the sum of the weights is... ; Further judgment: When the impact on health value is... Less than or equal to the health value threshold When necessary, images that do not require unnecessary processing are stored as images that do not require unnecessary processing, and a conformity assessment step is performed; when the impact on health value... Greater than the health threshold If necessary, the image that was not processed is saved as an image that requires processing, and the repair steps are performed.

3. The method for detecting defects in powder coating of highway guardrails according to claim 2, characterized in that: Between the re-judgment step and the conformity assessment step, there is also a defect area statistics step and a third judgment step; Defect area statistics: Defect area of ​​images that do not require necessary processing. Perform calculations and set a threshold for the defect area. ; Defect area The calculation model is as follows: ， in, Indicates the first unit area The total area of ​​defects within. ; Represents unit area The area of ​​the i-th type of defect, ; Third judgment: When the defect area Less than or equal to the set defect area limit threshold When necessary, images that do not require unnecessary processing are stored as images that do not require unnecessary processing, and the conformity assessment steps are performed; when the defect area The defect area exceeds the set threshold. At that time, images that do not require processing are classified as images that require processing, and repair steps are performed. Defect area limit threshold The calculation model is as follows: ， in, It is a proportionality coefficient. The range of values ​​is , Indicates unit area.

4. The method for detecting defects in powder coating of highway guardrails according to claim 3, characterized in that: Between the third judgment step and the conformity assessment step, there is also a defect merging judgment step; Defect merging and discrimination: including defect distance statistics, distance judgment, and defect adjustment steps; Defect statistics: the distance between defects in images that do not require unnecessary processing. Perform calculations and set spacing thresholds. ; Distance judgment: when the defect spacing Less than the set spacing threshold At that time, perform the defect adjustment steps; When the defect spacing Greater than or equal to the set spacing threshold At the same time, maintain the defect area The process remains unchanged, and the conformity assessment procedures are performed; among which... Spacing threshold It is a constant representing the minimum distance between defects; Defect adjustment: Adjustment of defect area Adjust the coefficients and update the defect area. The calculation model is as follows: ， in, Indicates the first unit area The total area of ​​defects within. , Represents unit area The area of ​​the i-th type of defect, , Indicates the defect spacing Less than the set spacing threshold The number of times.

5. The method for detecting defects in powder coating of highway guardrails according to any one of claims 1-4, characterized in that: Different quantity thresholds are set for the outer and inner surfaces of the guardrail. Health threshold Defect area limit threshold and spacing threshold Wherein, the outer surface is the main surface that can be directly seen during normal use, the inner surface is the surface that cannot be observed during normal use, and the inner surface has a higher tolerance for defects than the outer surface.

6. The method for detecting defects in powder coating of highway guardrails according to claim 4, characterized in that: Between the distance determination step and the conformity assessment step, there are also an edge discrimination step and a fourth determination step; Edge detection step: Perform edge detection on defects in non-essential processing images to determine whether the defects are located at the edge of the guardrail. Fourth judgment step: When the defect is not located at the edge of the guardrail, the non-essential processing image is stored as a non-essential processing image and the pass / fail assessment step is performed; when the defect is located at the edge of the guardrail, the non-essential processing image is directly classified as an essential processing image and the repair step is performed.

7. A system for detecting defects in powder coating on highway guardrails, wherein the system is applicable to the detection method described in any one of claims 1-6, characterized in that, include: Image preprocessing module: used to acquire historical images and perform image preprocessing; Dataset creation module: used to annotate different types of defects with the same features in historical images and create training and validation sets; Model training module: Used to train the classification model using the training set and evaluate the classification accuracy of the model using the validation set to obtain the trained model; Image detection module: used to acquire the image to be detected, input the image to be detected into the trained model for detection, and obtain the recognition result image; Image stitching module: This module stitches together multiple recognition result images that combine to form the same guardrail panel, creating a complete guardrail panel image. Each guardrail panel image contains one guardrail panel. Based on the recognition results, the guardrail panel images are distinguished into normal images and defective images. Post-processing module: used to perform quantitative analysis on defects in the defect image, by counting the number of defects per unit area and setting a defect number threshold, to determine whether the number of defects per unit area meets the requirements, and further divide the defect image into images that do not require processing and images that require processing.

Images