Ream paper defect segmentation and classification detection method

Abstract

The invention relates to the technical field of image processing and order paper defect detection, in particular to a ream paper defect segmentation and classification detection method, which comprises the steps of S1, training a segmentation and classification network module; training a segmentation network to learn to obtain representation features of the ream paper defects, so that the networkcan segment token defect areas from the input ream paper images, wherein the input of the classification network module is an H * W * C feature tensor, the feature tensor comes from ROI features of acorresponding network layer, and the output of the classification network module is an order paper defect category; s2, inputting the ream paper image into a segmentation network module, and obtainingabstract ream paper defect features through an Encoder-Decoder neural network; s3, obtaining a defect external rectangle and a corresponding proper network layer number by using an ROI calculation module; s4, extracting a feature tensor of the size of H * W * C of the defect area by using an ROI information extraction module; and S5, enabling the classification network module to perform classification judgment on the ream paper defects, and output ream paper defect categories through Softmax to obtain a ream paper defect detection result.

Description

technical field [0001] The invention relates to the technical fields of image processing and ream defect detection, in particular to a detection method for segmenting and classifying ream defects. Background technique [0002] The ream paper refers to a certain number of identical paper sheets in the paper industry. Generally, 500 identical paper sheets are used as a ream paper, that is, a ream of paper. For example, on office paper, a pack of ordinary A4 printing paper is generally 500 pages. Ream paper needs to go through quality inspection and screening on the automatic production line of papermaking enterprises. Defective ream paper products are removed from the production line queue and qualified normal ream paper enters the next packaging process. [0003] The level of mechanization and automatic information production of large paper-making enterprises is increasing day by day. However, due to the influence of a series of factors such as incoming materials, knives, as...

AI Technical Summary

Problems solved by technology

Due to the large span of defects in the detection process of ream paper defects, large defects can occupy the entire ream image, and small defects are only a few pixels in size. Proportion of candidate boxes to cover all cases

Images