A defect detection method, device, medium and equipment based on semantic segmentation model

Abstract

The invention discloses a defect detection method based on a semantic segmentation model, comprising: acquiring image data of a workpiece to be detected; inputting the image data into a trained semantic segmentation model to obtain a defect prediction Label graph of the workpiece; the semantic segmentation model includes The feature extraction network and feature prediction network connected in sequence; the feature extraction network is used to predict the location of the defect area, which includes a feature extraction layer, a feature compression layer, a feature flattening layer, and a feature classification layer; the feature prediction network is used to classify the feature Extract each pixel in the defect area determined by the network to classify, and output the defect prediction Label map of the workpiece after locating and segmenting the defect area; the defect detection of the workpiece based on the defect prediction Label map can effectively reduce the calculation amount of the model, and is fast and high-precision Realize the visual detection of workpiece defects.

Description

technical field [0001] The invention belongs to the field of computer vision and deep learning, and more specifically relates to a workpiece defect detection method, device, storage medium and computer equipment based on a semantic segmentation model. Background technique [0002] With the rapid development of Industry 4.0, big data analysis and artificial intelligence have been applied in the field of industrial visual inspection. Compared with the image extraction features that cannot be quantified by traditional visual technology, the deep learning neural network can effectively optimize the processing, and to a certain extent can improve Accuracy and stability of visual detection of workpiece defects. [0003] Through the 2D camera to effectively collect sample images of workpiece defects, the deep learning semantic segmentation model can predict the defect area of ​​the workpiece with high precision, which is convenient for subsequent defect quantification processing. ...

AI Technical Summary

Problems solved by technology

[0005] (1) The 2D camera cooperates with the backlight to calculate the area of ​​the shadow area, but its mechanical structure and facilities are relatively complex, and the 3D defects of the workpiece obtained under this method are mostly in the form of shadows, the relationship between the actual defect area and the projected area is unknown, and the defect quantification accuracy is low ;

[0006] (2) 3D point cloud defect area segmentation technology, but this method cannot balance the amount of data calculation and defect accuracy, that is, if the point cloud is relatively dense, the defect area is relatively clear, but the corresponding data calculation amount is large, and it relies too much on high-dimensional Data processing; if the point cloud is down-sampled, the defect detection rate can be guaranteed to a certain extent, but the 3D point cloud data is relatively sparse, which will reduce the defect detection accuracy

Images