Abstract:In order to detect the surface flaws of the spherical powder plate quickly and accurately, the paper analyzes the key problems, such as the diversity of flaws and the optical inconsistency for different types, and the corresponding solutions are proposed. A correction template is created based on a template image which is collected from the flawless surface to realize the optical homogenization of the detected surface. The binary threshold is extracted automatically according to the template image of different type plate to improve the robustness of the software execution. A custom algorithm, which can effectively improve the precision and reliability of detecting the scratch type flaw, is used to compensate the loss of the fine line flaws in screening the flaws. In addition, screening model is established to remove the noise coming from the spherical powder according to the accuracy requirements. A lots of experiments show that the system can accurately and efficiently detect flaws on the spherical powder plate surface and achieve the identification, the location and the area measurement of the surface flaws in the production line, and meet the target accuracy. The system can be well qualified for online detection the flaws in surface of the spherical powder plate. Experimental results show that the detection system is very efficient and accurate, which greatly improves the automation of production line.
