Surface Defect Detection of Hot Rolled Steel Strip Based on Image Compression

Abstract

In order to solve the problem of high error rate of manual methods for the detection of surface defects of hot rolled steel strips, two methods for detecting surface defects of hot-rolled steel strips based on image compression is proposed, namely reconstructed image based detection method and compressed representation based detection method. The deep learning YOLO-tiny network is used as the backbone framework for detection, and the convolutional self-encoder CAE is used as the compression preprocessing framework to realize the detection of surface defects of hot rolled steel strip. The NEU hot rolled steel strip surface defect data set is used for model training, and the Mean average precision (mAP) is used to evaluate the test results. The experimental results show that, compared with the original YOLO-tiny, both methods obtain higher mAP evaluation indexes.