Research Of Mobile Phone USB Flash Disk Defects Detection System Based On Computer Vision Technology

With the rapid development of computer technology, computer vision technology has been widly applicated in medical diagnosis, intelligent robot, industrial areas such as automatic detection and control.In the manufacturing of. product appearance quality inspection, measurement, quantitative aspects of the developed especially rapidly, the industry has made great economic and social benefits.In the production of electronic products,computer vision technology has always been a very hot research topic. Using computer vision system for defect detection of mobile phone USB driver, not only avoid the randomness of manual detection problems, poor reproducibility; but aslo solve quality deviation and solder joint analysis components manual inspection which can not be quantitatived, it has greatly optimized and improved the manufacturing process. In this paper, the size of the mobile USB driver and the surface appearance of the solder joint inspection has been researched, a set of defect detection system based on computer vision has been designed.In this paper, firstly, the present development situation of computer vision detection technology is summarized. The hardware of system has been analyzed and selected, and the hardware platform is build. The image is grabbed and processing by using image processing procedures including preprocessing, edge detection, Hough transform, parameter measurement and other image processing steps of calculation, to calculate the length and the deflection of the mobile phone USB flash disk samples. In the image pre-processing, gray processing, image sharpening are used to lay a foundation for the following edge detection. The applicability of several common edge detection operators are compared, and finally select the Canny operator to extract the edge of the external image. Using Hough transform to detect the edge of the straight line extraction, and calculating the size and the angle.After using median filter to eliminate the noise. A series of image processing methods including pre-treatment, tilt correction, threshold segmentation, cancroids extraction are adopted to locate solder joints. Vertical projection histogram and connected region area are adopted to recognize solder bridging(short circuit) and incomplete welding(broken circuit). Lastly, the calibration method of system is introduced and the detection rate of size and angle detection accuracy and the solder joint defect is verified by experiments. Experiments show that the recognition method can accurately locating the solder joint defects, and has good recognition effect. It has certain practical value.