Development Of USB Plugs Surface Defects Detection Algorithm Based On Embedded GPU

Currently,the common detecting method of USB plugs surface defects is manual inspection,and their aim is to exclude the USB plugs which have defects.Manual inspection which increased the human cost is easy to cause miss detection or false detection because of the eye fatigue.Therefore,this research develops a USB plugs surface defects detection system based on embedded GPU,which can accurately detect defects like dot type,spot type and planar type.The system has high value in engineering application.Based on machine vision,the system captures images of USB plugs in lines,extracts USB plugs surface region.And it uses saliency detection on USB plugs surface to highlight the spot type,planar type defects which are difficult to detect.The USB plugs surface coating easily leads to asymmetric illumination and is thus unsuited to be considered from the whole surface.So the superpixel segmentation algorithms can be used to solve the asymmetric illumination problems.Using the superpixel segmentation,USB plugs surface can be divide into multiple superpixel regions.Then,these regions will be detected separately.In order to improve the performance of detection algorithm,this research implement the algorithm by means of parallel computation.Analyzing the runtime of algorithm's modules and taking advantage of the embedded GPU,this research implement the CUDA-based optimize algorithm.It was shown in system tests that,the developed defects detection system meets the functional requirements and performance requirements of the USB plugs surface defects detection.The detection algorithm takes 61ms or less,and the accurate rate is higher than 95%.