摘要：This paper deals with the automated online-inspection of trimmed edges of heavy steel plates. These edges sometimes exhibit defective fractured areas and burr. A novel algorithm is presented, which segments the images recorded by a CCD camera into burnished area, defective fractured area, burr, and area which exhibits a high quality. The segmented images will serve as a basis for quality control, generating data for machine learning, and informing the plant operator in case of insufficient quality. The segmentation is performed via thresholding, morphologic operations, estimation methods as well as statistical methods. The underlying framework is presented and example images demonstrate the capability of the algorithm to segment trimmed edges with various properties.
