| Summary: | 碩士 === 明新科技大學 === 工程管理研究所 === 94 === Taiwan is a well-known country of manufacturing affluent electronic products such as main board, digital camera, mobile phone, and SOC chip. During the manufacturing process, it is inevitable to detect the defects and classify those defects into the correct categories.
Recently, researches related to the applications of computer vision have been developed. Many researchers have tried their best to explore every possibility to their applications from the face recognition, object tracing, and the automated defect detection. Factory owners have also invested in the research work which can take over the human labor. In this research, we devoted ourselves to defect detection and classification in Copper Clad Laminate (CCL) that is one of the most important raw materials in printed circuit board.
This research was to conduct the CCL defects detection, which are obtained by the wavelet transformation. After transformation the defects, we used blob analysis which extracted the eleven geometric features. These eleven features were feed into the SVM (Support Vector Machines). The multi-class approach one–against-all method and one–against–one method were then used for classification. Finally, back-propagation neural network is used to compare the results with those of SVM. From the test result, it was concluded that SVM in one-against-all approach had almost the same classification accuracies with those of back-propagation neural network.
|
|---|
