Abstract:Forged steel surface cracks are the main defects in reducing the pass rate of finished products during the processing of special steel. The common method is manual detection, that is, the human eye observes the surface of the forged steel to find cracks under the UV lamp in the dark room. The manual detection method has high detection accuracy, but high labor intensity and long-term absorption of ultraviolet radiation by the human body. To this end, researchers have developed machine vision methods. Compared to manual methods, the difficult problems of machine vision include rapid detection of narrow cracks and removal of complex background interference. Aiming at this problem, this paper adopts high-resolution camera component, straight line segment detection algorithm based on Hough transform and blade background elimination technology to realize the narrow crack detection of forged steel surface based on machine vision, wherein the detection width is 124mm×148mm, and the detection time is no more than 3s.