| Summary: | 碩士 === 元智大學 === 工業工程研究所 === 88 === Various automated visual inspection systems for printed circuit boards (PCBs) have been developed in the past years. However, most of the visual inspection techniques use only gray-level information of PCB images and focus mainly on line-etched defects. In this study, we employ color machine vision to inspect defects on electroplated surfaces of PCBs and in particular, edge connectors.
The electroplated surfaces of edge connectors can be considered as a homogeneous texture. Traditional texture analysis techniques such as co-occurrence matrix methods in the spatial domain, and Fourier-based features in the spectral domain are too computationally expensive to develop an efficient inspection system. In this study, we develop two entropy measures to evaluate the homogeneity of edge connector surfaces. One entropy measure uses two color features to detect color anomalies such as oxygenation, and the other uses edge angles to detect structural defects such as scratches on electroplated surfaces. Experimental results have shown that the purposed method is reliable in detection and efficient in computation. It takes only 1 second to detect 7 edge-connector pins in one image.
|
|---|
