CN-121999282-A - Multi-view-based appearance defect detection method, device, equipment and storage medium

CN121999282ACN 121999282 ACN121999282 ACN 121999282ACN-121999282-A

Abstract

The application relates to a multi-view-based appearance defect detection method, a device, equipment and a storage medium, which are applied to a control module of a multi-view detection system, wherein the multi-view detection system further comprises a detection platform, multi-light source illumination equipment and an image acquisition unit, wherein an object to be detected is placed on the detection platform; and carrying out defect detection on the a pieces of second images through a preset multi-view visual defect detection model to obtain a target visual defect detection result of the object to be detected. By adopting the embodiment of the application, the accuracy of appearance defect detection can be improved.

Inventors

XU KAIYUAN
CHEN PENGGUANG
FENG JIAPEI
LIU SHU

Assignees

深圳思谋信息科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260115

Claims (10)

1. The multi-view visual angle-based appearance defect detection method is characterized by being applied to a control module of a multi-view angle detection system, wherein the multi-view angle detection system further comprises a detection platform, multi-light source illumination equipment and an image acquisition unit, an object to be detected is placed on the detection platform, and the method comprises the following steps: Sequentially illuminating the object to be detected by adopting different light sources through the multi-light source illuminating equipment, and acquiring images of the object to be detected at different visual angles through the image acquisition unit in each illuminating process to obtain a first images, wherein a is an integer larger than 1; preprocessing the a first images to obtain a Zhang Dier images; and performing defect detection on the a second images through a preset multi-view visual defect detection model to obtain a target visual defect detection result of the object to be detected.
2. The method of claim 1, wherein the multi-view appearance defect detection model comprises a channel merge weighting layer, a convolution layer, a cross-resolution fusion layer, a feature aggregation layer, a global average pooling layer, a full connection layer, a Softmax layer; Performing defect detection on the a second images through a preset multi-view appearance defect detection model to obtain a target appearance defect detection result of the object to be detected, wherein the method comprises the following steps: Tensor processing is carried out on the a second images through the channel merging weighting layer, so that a first tensor is obtained; step-by-step feature extraction is carried out on the first tensor through the convolution layer to obtain b feature graphs, wherein the resolution ratios of the b feature graphs are different; the b feature images are fused through the cross-resolution fusion layer, so that fusion feature images are obtained; the first tensor and the fusion feature map are polymerized through the feature polymerization layer, so that target polymerization features are obtained; carrying out global average pooling on the target aggregation features through the global average pooling layer to obtain target vectors; Processing the target vector through a full connection layer to obtain classification probability distribution corresponding to the preset defect class number; normalizing the classification probability distribution through the Softmax layer to obtain a target probability distribution; determining target defect type label information corresponding to the target probability distribution; And determining a target appearance defect detection result of the object to be detected according to the target defect type label information and the a second images.
3. The method according to claim 2, wherein the tensor processing the a-pieces of the second image by the channel merge weighting layer to obtain a first tensor includes: mapping each image in the a-pieces of second images into tensors to obtain a-pieces of tensors; splicing the a tensors to obtain spliced tensors; Determining a target channel number corresponding to the splice tensor; determining global scalar weights with the same number as the target channel number to obtain c global scalar weights, wherein c is the target channel number; and adjusting the spliced tensor according to the c global scalar weights to obtain a first tensor.
4. A method according to claim 2 or 3, wherein the fusing the b feature maps by the cross-resolution fusion layer to obtain a fused feature map includes: aligning the resolutions of the b feature images to obtain b first feature images; And aligning and fusing the channels of the b first feature images to obtain a fused feature image.
5. A method according to claim 2 or 3, wherein the aggregating the first tensor and the fused feature map by the feature aggregation layer to obtain a target aggregate feature comprises: determining a second feature map corresponding to the first tensor; performing resolution adjustment on the second feature map and the fusion feature map to obtain a third feature map and a fourth feature map, wherein the resolution of the third feature map is the same as that of the fourth feature map; Splicing the third characteristic diagram and the fourth characteristic diagram according to the channel dimension to obtain a spliced characteristic diagram; and performing dimension reduction and smoothing treatment on the spliced feature map to obtain the target aggregation feature.
6. A method according to claim 2 or 3, wherein said determining a target appearance defect detection result of the object to be detected based on the target defect class label information and the a-sheet second image comprises: Determining a target mask corresponding to the target defect type label information; labeling the a second images according to the target mask to obtain a labeled images; And determining a target appearance defect detection result of the object to be detected according to the a Zhang Biaozhu image.
7. A method according to claim 2 or 3, wherein said determining target defect class label information corresponding to said target probability distribution comprises: determining a reference output tensor corresponding to the target probability distribution; Determining a reference defect label graph corresponding to the reference output tensor; and determining target defect type label information according to the reference defect label graph.
8. The utility model provides an appearance defect detection device based on multi-view, its characterized in that is applied to the control module of multi-view detecting system, multi-view detecting system still includes testing platform, many light source lighting apparatus, image acquisition unit, the test platform has placed the object of waiting to detect, and the device includes: The acquisition module is used for sequentially illuminating the object to be detected by adopting different light sources through the multi-light source illumination equipment, and acquiring images of the object to be detected at different visual angles through the image acquisition unit in each illumination process to obtain a first images, wherein a is an integer larger than 1; the preprocessing module is used for preprocessing the a first images to obtain a Zhang Dier images; and the defect detection module is used for carrying out defect detection on the a second images through a preset multi-view appearance defect detection model to obtain a target appearance defect detection result of the object to be detected.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 7 when the computer program is executed.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 7.

Description

Multi-view-based appearance defect detection method, device, equipment and storage medium Technical Field The present application relates to the field of image detection technologies, and in particular, to a method, an apparatus, a device, and a storage medium for detecting an appearance defect based on multiple views. Background With the rapid development of electronic manufacturing industry and automatic detection technology, product appearance defect detection has become a key link in production quality control. At present, a single-light-source imaging scheme is adopted in the appearance defect detection method based on deep learning, automatic defect identification is achieved through a convolutional neural network, and the problems that illumination coverage is insufficient (different defect characteristics of a complex surface cannot be displayed), a detection result is unstable (accuracy is reduced due to angle and illumination change) and the like still exist, so that the accuracy of appearance defect detection is low. Therefore, how to improve the accuracy of appearance defect detection is a problem to be solved. Disclosure of Invention In view of the foregoing, it is desirable to provide a method, an apparatus, a device, and a storage medium for detecting an appearance defect based on multiple angles, which can improve the accuracy of appearance defect detection. In a first aspect, the application provides a multi-view-based appearance defect detection method, which is applied to a control module of a multi-view detection system, wherein the multi-view detection system further comprises a detection platform, a multi-light source lighting device and an image acquisition unit, an object to be detected is placed on the detection platform, and the method comprises the following steps: Sequentially illuminating an object to be detected by adopting different light sources through a multi-light source illuminating device, and acquiring images of the object to be detected at different visual angles through an image acquisition unit in each illuminating process to obtain a first images, wherein a is an integer larger than 1; preprocessing the a first images to obtain a Zhang Dier images; and performing defect detection on the a second images through a preset multi-view visual defect detection model to obtain a target visual defect detection result of the object to be detected. In a second aspect, the application provides an appearance defect detection device based on multiple views, which is applied to a control module of a multiple view detection system, wherein the multiple view detection system further comprises a detection platform, multiple light source illumination equipment and an image acquisition unit, an object to be detected is placed on the detection platform, and the device comprises: The system comprises an acquisition module, a first image acquisition module, a second image acquisition module and a display module, wherein the acquisition module is used for sequentially illuminating an object to be detected by adopting different light sources through a multi-light source illumination device, and acquiring images of the object to be detected under different visual angles through an image acquisition unit in each illumination process to obtain a first images; The preprocessing module is used for preprocessing the a first images to obtain a Zhang Dier images; And the defect detection module is used for carrying out defect detection on the a second images through a preset multi-view appearance defect detection model to obtain a target appearance defect detection result of the object to be detected. In a third aspect, the application provides a computer device comprising a memory storing a computer program and a processor implementing the steps of the method described above when the processor executes the computer program. In a fourth aspect, the present application provides a computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of the method described above. In a fifth aspect, the application provides a computer program product comprising a computer program which, when executed by a processor, implements the steps of the method described above. According to the multi-view-angle-based appearance defect detection method, different spectrum or morphological characteristics (such as coaxial light-induced scratches and strip-shaped light-induced stains) of defects are highlighted through different light sources, then images of different view angles of an object to be detected are collected under each light source condition to obtain a first image, so that the problem that the defect characteristics are easy to be blocked or weakened under a single view angle or light source can be solved, a more comprehensive image data basis is provided for detection, then the multi-view-angle appearance defect detection model is specially ad