SCRATCH DETECTION SYSTEM OF METAL SHEETS BASED ON FPGA IMAGE PROCESSING

• Main Authors: Rui-Tang Huang, 黃睿堂
• Other Authors: Lu, Ruei-Chang
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/9v59rc

碩士 === 國立宜蘭大學 === 電子工程學系碩士班 === 107 === The history of human industrial development, from the beginning of the industrial revolution's steam to replace human and animal power, then to the power-driven modern industry and then to the future of the unmanned factories, is basically a process of continuously improving the degree of automation. In the automation industry, automated product testing is also an important part of it. The high difficulty of automated product testing lies in the need to address the many possible defects in production. Therefore, the algorithm for automated detection is more complicated, resulting in higher hardware requirements for the device and higher cost. This paper proposes a scratch detection system based on field programmable gate array (FPGA) for metal sheets. This system first photographing the image of the metal sheet under different light source intensity and incident angle conditions. Then the image is gray-scaled and is performed by Sobel edge detection image processing. The processed image is then partitioned to define effective and ineffective areas. Base on the light conditions of the picture, the system will assign various filter and binarization thresholds to differentiate noise and shading. Whether there are abnormal lines or spots in the effective area is used to achieve the purpose of automatic detection. In the back of the paper is the result of experimental tests on several types of scratches, as well as statistics on experimental data such as correctness and processing time. We present a low-cost and fast processing detection system based on its pure hardware architecture and simple algorithms.