Detection device in industrial quality inspection

Abstract

The invention provides a detection device in industrial quality inspection, comprising: a classification residual network, including: a series of first to fourth stage classification residual networks, wherein, in the second to fourth stage classification residual networks, insert Deformable convolutional layer; feature extraction network, used to extract the feature map of the picture whose classification result is defect; region extraction network, used to generate at least one candidate region according to the feature map; candidate region processing network, used to normalize the candidate region The fixed-size candidate region is optimized to obtain a fixed-size feature map; the circular convolutional neural network is used to output the category to which the candidate region belongs and the position of the candidate region in the image according to the fixed-size feature map. The present invention inserts a deformable convolution layer into the classification residual network, which can enhance the detection effect on the deformation, rotation, scaling, etc. of the detected small objects, and further input the defective pictures to the target detection network for processing, so that The extracted features are more accurate.

Description

technical field [0001] The invention relates to the technical field of industrial quality inspection, in particular to a detection device in industrial quality inspection. Background technique [0002] In the field of industrial quality inspection, various process defects (cracks, scabs, curling, cavities, wear, etc.) in the industrial manufacturing process have a great impact on the quality, appearance level and performance parameters of industrial products influences. In the context of "Intelligent Manufacturing Power", the requirements for defect inspection are getting higher and higher, and the requirements of customers are also getting higher and higher. Improving the accuracy and performance of inspection has become the top priority, which means that customer satisfaction is improved and business significance very big. [0003] Due to the complex optical environment of the defect detection site and the large differences in the scale and shape of the detected defects,...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the above technical problems, the present invention proposes a detection device in industrial quality inspection. The present invention inserts a deformable convolutional layer into the classification residual network, which can enhance the detection of small objects, such as deformation, rotation, scaling, etc. The detection effect can better adapt to the influence of the scale, posture, and viewing angle of the detection target, and the defective pictures are further input to the target detection network for processing, so that the extracted features have higher accuracy and better meet the needs of the site and customers. The requirements of the industrial field have very good generalization ability and robustness

Images