Search

CN-122009642-A - Wire harness test labeling system

CN122009642ACN 122009642 ACN122009642 ACN 122009642ACN-122009642-A

Abstract

The invention discloses a wire harness test labeling system, which relates to the technical field of image processing, and comprises the steps of obtaining an initial image to be labeled, clustering the initial image to obtain a clustered image, carrying out enhancement operation on the clustered image to output an enhanced image, enhancing the small target defect feature expression capability through the clustering and enhancement operation on the initial image, inputting the enhanced image into a defect detection model to output a detection result, realizing accurate classification labeling by means of the defect detection model, combining an alarm device to quickly feed back the detection result, controlling the alarm device to carry out alarm prompt according to the detection result, effectively improving the accuracy of label defect detection, reducing the condition of missed detection or misjudgment, and effectively guaranteeing the quality control effect of wire harness labeling.

Inventors

  • HU GUANGHUA

Assignees

  • 安徽隆华瑞达科技有限公司

Dates

Publication Date
20260512
Application Date
20260227

Claims (10)

  1. 1. The utility model provides a pencil test subsides mark system, its characterized in that, the system includes cluster reinforcing module, detection module, alarm module, wherein: The clustering enhancement module is used for performing clustering operation on pixels in the initial image to obtain a clustered image, and performing enhancement operation on a target feature region to output an enhanced image, wherein the target feature region is a defect feature region in the clustered image; The detection module is used for inputting the enhanced image into a defect detection model and outputting a detection result, wherein the defect detection model is a target model obtained based on YOLOv improvement, and the improvement comprises adding 1 PDC module between a 16 th layer and a 17 th layer in a YOLOv model neck network, wherein the PDC module is used for enhancing texture characteristics of label folds; And the alarm module is used for controlling alarm equipment to carry out alarm prompt according to the detection result.
  2. 2. The wire harness test labeling system of claim 1, further comprising a preprocessing module comprising a scaling module and a conversion module, wherein: The scaling module is used for obtaining an original image to be labeled, and calling an image scaling function for the original image to adjust the original image to a standard size to obtain a scaled image, wherein the original image is an RGB image, and the labeling size is 256 multiplied by 256; the conversion module is used for converting the scaled image into a gray image through a gray conversion function, and denoising the gray image to obtain an initial image.
  3. 3. The wire harness test labeling system of claim 1, wherein the cluster enhancement module comprises a first computing module, a query module, a selection module, a processing module, wherein: the first calculation module is used for respectively testing clustering centers with different values for pixels in the initial image, calculating intra-cluster variance corresponding to each clustering center and drawing to obtain an intra-cluster curve; The query module is used for querying points with the largest slope in the curve in the cluster to define inflection points, and determining the number N of clusters according to the inflection points; The selecting module is used for randomly selecting N non-repeated values from the pixel intensity values of the initial image to serve as initial clustering centers; And the processing module is used for processing the initial image according to the initial clustering center to obtain a clustered image.
  4. 4. The wire harness test labeling system of claim 3, wherein the processing module comprises an iteration module and an addition module, wherein: The iteration module is used for executing the first step to the third step, and specifically comprises the following steps: calculating the distance between the intensity value of each pixel in the initial image and N initial clustering centers, and distributing target pixels to the clustering labels with the shortest distance, wherein the target pixels are any one of the pixels of the initial image; Counting all pixels contained in each cluster label, calculating the average value of intensity values of all pixels, taking the average value as a new cluster center, and replacing the cluster center generated in the previous round; Step three, repeating the step two and the step three, calculating the total variation of the intensity values of N new clustering centers after each replacement, if the total variation is not smaller than a threshold value or the iteration number does not reach the preset iteration number, continuing iteration, otherwise, stopping iteration, and finishing convergence to obtain a clustering label corresponding to the target pixel; the adding module is used for traversing each pixel of the initial image, and adding the clustering label to the corresponding pixel in the initial image to obtain a clustered image.
  5. 5. The wire harness test labeling system of claim 4, wherein the cluster enhancement module further comprises a boundary generation module, a weighting module, a second calculation module, a dynamic adjustment module, wherein: The boundary generation module is used for extracting boundaries among different clustering areas in the clustered images by adopting a preset algorithm to generate boundary images; The weighting module is used for carrying out weighted superposition on the boundary image and the clustered image to obtain a boundary enhanced image; the second calculation module is used for performing morphological operation on the boundary enhanced image to obtain a purified image and calculating a gray level histogram of the purified image; and the dynamic adjustment module is used for dynamically adjusting gray level distribution according to the gray level histogram, and carrying out local contrast enhancement on each connected region defined by the clustering label in the purified image to obtain an enhanced image.
  6. 6. The wire harness test labeling system of claim 1, wherein the defect detection model is developed in accordance with YOLOv a improvement comprising: Replacing a layer 3, a layer 5, a layer 7 and a layer 9C 3K2 module in the YOLOv model backbone network with a C3K2_RCV module, and replacing a layer 10 SPPF module with a SZASPPF module, wherein the C3K2_RCV module is used for accurately extracting small target features on a label on the premise of light weight, and SZASPPF is used for strengthening the extraction of small target key features and inhibiting background clutter interference; the layer 14, layer 17C 3K2 modules in the YOLOv model neck network are replaced with c3k2_rcv modules.
  7. 7. The wire harness test labeling system of claim 6, wherein the c3k2_rcv module operates on the following principle: the method comprises the steps of obtaining an input original image as a 1 st original feature image, inputting the 1 st original feature image into a CBS module to obtain a2 nd original feature image, and splitting the 2 nd original feature image into a 1 st channel feature image and a2 nd channel feature image according to channel number halves; Inputting the 1 st channel characteristic diagram into a1 st RepViTBlock th module to obtain a 3 rd original characteristic diagram, and inputting the 3 rd original characteristic diagram into a 2 nd RepViTBlock th module to obtain a 4 th original characteristic diagram; splicing the 1 st channel feature map, the 2 nd channel feature map, the 3 rd original feature map and the 4 th original feature map to obtain a 5 th original feature map; and inputting the 5 th original feature map into a Conv module for convolution to output the 6 th original feature map, and taking the 6 th original feature map as the input of the next module.
  8. 8. The wire harness test labeling system of claim 6, wherein the SZASPPF module operates on the following principle: The method comprises the steps of obtaining a 6 th original feature map and serving as a 1 st target feature map, inputting the 1 st target feature map into a Conv module to obtain a 2 nd target feature map, carrying out maximum pooling on the 2 nd target feature map to obtain a 3 rd target feature map, and carrying out maximum pooling on the 3 rd target feature map to obtain a 4 th target feature map; Inputting the 2 nd target feature map into the channel attention output 5 th target feature map; splicing the 2 nd target feature map, the 3 rd target feature map, the 4 th target feature map and the 5 th target feature map to obtain a 6 th target feature map; And sequentially inputting the 6 th target feature map into a spatial attention and Conv module to output the 7 th target feature map, and taking the 7 th target feature map as the input of the next module.
  9. 9. The wire harness test labeling system of claim 1, wherein the PDC module operates according to the following principle: The method comprises the steps of obtaining an input feature map, sequentially inputting the input feature map into two Conv modules for convolution to obtain a1 st convolution feature map, and inputting the 1 st convolution feature map into PWConv modules for convolution to obtain a2 nd convolution feature map; sequentially inputting the 1 st convolution feature map into two DWConv modules for deep convolution to output a 3 rd convolution feature map; fusing the 2 nd convolution feature map and the 3 rd convolution feature map to obtain a 4 th convolution feature map; And sequentially performing average pooling and ReLU function activation on the 4 th convolution feature map to obtain a 5 th convolution feature map, and taking the 5 th convolution feature map as the input of the next module.
  10. 10. The wire harness test labeling system of claim 1, wherein the alarm module comprises a statistics module, wherein: the statistics module is used for counting the number of defect targets marked as unqualified in the enhanced image according to the detection result; If the defect target number is greater than a threshold value, generating a first control signal, wherein the first control signal is used for controlling alarm equipment to execute a first alarm mode; generating a second control signal if the defect target number is not greater than a threshold value, wherein the second control signal is used for controlling the alarm equipment to execute a second alarm mode; and stopping the operation after the operation is continued in the first alarm mode until the operation is manually confirmed or the preset time is reached, and automatically stopping the operation after the operation is continued in the second alarm mode for the preset short time.

Description

Wire harness test labeling system Technical Field The invention belongs to the technical field of image processing, and particularly relates to a wire harness test labeling system. Background The wire harness labeling is a wire harness consisting of a plurality of wires, cables and accessories, and is marked with information by means of labeling, code spraying and the like, wherein the label generally comprises the contents of wire harness specification, application, number, production date, belonging equipment and the like, has the core effects of facilitating production traceability, quick identification during installation and maintenance, being widely applied to the fields of automobiles, electronics and the like, greatly influencing production efficiency due to mismatching and reworking of installation and construction caused by incorrect or defective printing of the label, increasing management and control cost, prolonging maintenance time and improving fault risk due to information deletion or misread during the later maintenance. The prior publication number is CN117474924A, which discloses a label defect detection method based on machine vision, comprising the steps of acquiring a label image of a label to be detected, preprocessing the label image, respectively extracting contours of the target label image and a target template image which is acquired in advance, carrying out preliminary matching on the contours of the label to be detected and the contours of all templates in a template label contour set, classifying the contours of the label to be detected in the contour set of the label to be detected, carrying out refined matching on the contours of the label to be refined and the contour set to be matched corresponding to the contours of the label to be refined, and generating label defect information corresponding to the label to be detected. According to the method, the problem that the defect detection effect is low under the condition of small local part is solved by carrying out contour extraction on the label image and carrying out refinement treatment on the extracted label contour, but in some industrial environments, background illumination can reduce the extraction of defect features in the label image, and some small target defect features (such as folds, breakage, font dislocation and font blurring) of the label cannot be solved only by contour treatment, so that the problem of missed judgment or misjudgment on defect identification can be caused, and the efficiency of labeling the wire harness is affected. Disclosure of Invention The invention aims to solve the problem that the extraction of the wrinkle defect characteristics in the label image is greatly influenced by the background illumination in the industrial environment, so that the defect identification is missed or misjudged, and provides a wire harness test labeling system. The invention provides a wire harness test labeling system, which comprises a clustering segmentation module, a detection module and an alarm module, wherein: The clustering enhancement module is used for performing clustering operation on pixels in the initial image to obtain a clustered image, and performing enhancement operation on a target feature region to output an enhanced image, wherein the target feature region is a defect feature region in the clustered image; The detection module is used for inputting the enhanced image into a defect detection model and outputting a detection result, wherein the defect detection model is a target model obtained based on YOLOv improvement, and the improvement comprises adding 1 PDC module between a 16 th layer and a 17 th layer in a YOLOv model neck network, wherein the PDC module is used for enhancing texture characteristics of label folds; And the alarm module is used for controlling alarm equipment to carry out alarm prompt according to the detection result. Optionally, the system further comprises a preprocessing module, the preprocessing module comprising a scaling module and a conversion module, wherein: The scaling module is used for obtaining an original image to be labeled, and calling an image scaling function for the original image to adjust the original image to a standard size to obtain a scaled image, wherein the original image is an RGB image, and the labeling size is 256 multiplied by 256; the conversion module is used for converting the scaled image into a gray image through a gray conversion function, and denoising the gray image to obtain an initial image. Optionally, the cluster enhancement module includes a first calculation module, a query module, a selection module, and a processing module, where: the first calculation module is used for respectively testing clustering centers with different values for pixels in the initial image, calculating intra-cluster variance corresponding to each clustering center and drawing to obtain an intra-cluster curve; The query module is used for querying