Design of Steel Surface Defect Detection System Based on deep Learning

Abstract

A steel surface defect detection system based on deep learning is proposed to solve a series of problems in the current steel surface defect detection (mainly including relying on manual labor, difficult detection, high miss-detection rate, error-prone, etc.). After the system independently develops the defect feature recognition algorithm and obtains the steel surface image, it then recognizes and extracts the steel surface defect features, performs a series of mathematical morphological operations, performs feature library learning based on the defect features, and combines design development Based on the deep learning defect recognition and detection algorithm, and use the U-net network to quickly identify and detect defects. The system can completely identify and detect steel surface defects, and the detection speed is faster than manual, which meets the requirements of automatic steel surface quality inspection production lines. The system has realized certain academic research value and practical application value.