Automated detection of surface defects on sphere parts using laser and CDD measurements

Automated detection of surface defects on sphere parts is a requirement to assure high quality of parts. Unlike most common techniques, which are frequently based on manual estimations that lead to significant time and financial constraints, this paper presents a novel automated detection system of surface defects on sphere parts, by using laser and CCD measurements. The proposed system focuses on the techniques of measuring and identifying defects. The work uses the surface defect detection (SDD) system developed with improved accuracy, flexibility and advanced measuring efficiency advantages. To perform initial locating of surface defects, the possible defects are extracted using empirical mode decomposition (EMD)-based extracting algorithm developed in this paper. Then, the system applies Canny edge algorithm to narrow down the search space of these defects, and backpropagation (BP) neural network to classify five defects with complex shapes, specifically, bulge, dent, abrasion, oxidation, and exfoliation defects. Once these defects are classified, 3D surface morphology of the extracted bulge and dent defects can be obtained and evaluated. Experimental results on automated defect detection of a variety of the sphere parts are reported to show the effectiveness of our system.