Paper defect detection method and device, detection equipment and storage medium

Abstract

The invention relates to the technical field of deep learning, and discloses a method for paper defect detection, which comprises the following steps: collecting a depth image of detected paper; performing visualization processing on the depth image; inputting the visualized depth image into the tested segmentation network model to obtain depth point cloud data of a foreground region in the depth image; performing normalization processing on the depth point cloud data of the foreground region; and inputting the normalized depth point cloud data of the foreground region into the tested classification network model for defect analysis. According to the scheme, the depth information of the image is reserved while the background region is removed by using the segmentation network model. Compared with a mode of extracting features through dimension reduction in the prior art, depth information in the three-dimensional image is reserved, so that the accuracy of paper defect detection is improved. The invention further discloses a device for paper defect detection, detection equipment and a storage medium.

Description

technical field [0001] The present application relates to the technical field of deep learning, and in particular, to a method, an apparatus, a detection device and a storage medium for paper defect detection. Background technique [0002] At present, most of the traditional paper defect detection is based on visible light defect detection, while deep defects belong to non-visible light defect detection, so they are not visible on the image, and traditional algorithms cannot perform defect detection. In the prior art, most of the paper detection technologies for deep defects are performed manually, that is, to observe whether the paper has defects with the naked eye. However, manual detection has problems such as high labor cost, strong detection subjectivity, and long detection time. [0003] The related art discloses a paper defect detection method based on machine vision, which includes: acquiring a paper image; placing a defect model to detect the paper, and obtaining t...

AI Technical Summary

Problems solved by technology

[0004] The problem with related technologies is that after obtaining the paper image, it is directly put into the defect model, and the image processing method used in the model is to extract the main features of the image through dimensionality reduction, thus ignoring the depth in the image to the defect detection. Influence, reducing the accuracy of paper defect detection, especially when the detection accuracy is at the micron level, the depth information in the image is ignored, and the accuracy of paper defect detection cannot be guaranteed

Images