| Summary: | 碩士 === 朝陽科技大學 === 資訊工程系 === 107 === At present, computer vision technology is quite mature and widely used in the manufacturing industry, also known as Automated Optical Inspection (AOI). This paper focuses on the design of AOI methods for LED manufacturing. In the early stages of LED defect detection, artificial vision methods are used. However, due to the continuous improvement of production speed, the artificial visual inspection speed cannot meet the detection requirements, and the manual detection is also prone to fatigue problems. Therefore, this study designed the AOI method to detect LED defects. Defects are classified into lack of glue, overflow of glue and lack of luminescent crystals. This thesis mainly uses the industrial computer vision library (Euresys Open eVision) detection. The main method consists of three stages. The first stage is LED positioning, and the object matching method is realized by using template matching. The second stage is image preprocessing, which reduces the noise in the image by a Gaussian filter. The third stage is defect detection, which uses edge detection to detect. Currently, 627 images were collected for three defect categories, including 225 defect images, 258 glue defect images, and 144 missing crystal defect images. The recognition accuracy of the lack of glue defects is 98.67%, the recognition accuracy of the glue defects is 96.52%, and the accuracy of the recognition of the luminescent crystal defects is 91.04%, which achieves the purpose of accurately identifying the LED defects.
|
|---|
