| Summary: | 碩士 === 國立臺灣科技大學 === 電機工程系 === 89 === Computer vision has been widely used in on-line inspection of electronic components. In this paper, we present a printed circuit board (PCB) inspection system based on computer vision using several methods, which provides us with an efficient solution for inspecting all kinds of mistakes of component. The PCB inspection system can be divided into two phases: (1) image segmentation and (2) image processing. In image segmentation, we locate the position of PCB by using sobel operator in intensity of image. Then we process our color image with bi-level quantization by using back-propagation neural network and fuzzy c-means clustering method in combination with the algorithm of compared with cluster centers. In part of image processing, we use run-length smoothing in noise cancellation and use run-length labeling to fix the position of component; furthermore we extract character of those components by left and right fallen-point feature extraction. Finally, we simulate the best parameters for all methods, and build the accurate, repeating, and flexible system to inspect components in PCB.
|
|---|
