Outer package defect detection method

Abstract

The invention provides an outer package defect detection method. The method comprises the steps of obtaining an outer package picture as a data set; performing data preprocessing on the data set; setting the size of a defect target candidate box through a clustering algorithm; inputting the preprocessed data set into a Cascade R-CNN model, and performing region-of-interest extraction on the data set; normalizing the size of the feature map of the region of interest; in the Cascade R-CNN model, adopting a pre-trained ResNet-50 with a deformable convolution v2 as a backbone network, carrying outconvolution, and extracting features; and S44, framing out defects and marking defect categories. The outer package defect detection method is simple in calculation process and capable of rapidly positioning and recognizing defects, the FPN feature extraction method is adopted, features with different resolutions are combined, the combined features are used for enhancing multi-level features in FPN, and recognition is more accurate.

Description

technical field [0001] The invention relates to the field of computers, in particular to an outer packaging defect detection method. Background technique [0002] Defect detection is in great demand in industrial applications such as textiles, glass products, steel, road traffic, chips, etc. Traditionally, manual and machine vision-based detection methods are generally used. However, manual detection is inefficient and slow; traditional machine vision detection needs to rely on some artificially designed features and has poor robustness. In recent years, with the improvement of computer computing power and the rapid development of deep neural networks, the use of deep learning methods, especially the use of target detection methods for defect detection has gradually become one of the research hotspots, and the detection accuracy of current detection methods still needs to be improved. . SUMMARY OF THE INVENTION [0003] In order to overcome the above-mentioned defects in...

AI Technical Summary

Problems solved by technology

However, manual detection is inefficient and slow; while traditional machine vision detection needs to rely on some artificially designed features and has poor robustness

In recent years, with the improvement of computer computing power and the rapid development of deep neural networks, the use of deep learning methods, especially the use of target detection methods for defect detection has gradually become one of the research hotspots, but the detection accuracy of current detection methods still needs to be improved.

Images