|
For the current semiconductor testing industry, Testing, is the last line of defense before shipment. Although, there are still operators utilizing visual inspection as the final inspection before shipment, the visual inspection is not only time-consuming, but also inefficient. Hence, the automated testing is currently the optimal way to control quality regulation of products. Although in the industry, there are many advanced visual inspection systems that can apply to various flaw detection machines, they all require additional high costs. This study will apply Image Processing Technology and the theory of neural network as the last protection before shipment. In order to avoid exceptional situations in production, it retrieve and identify the product model by Image Processing Technology and Optical Character Recognition (OCR) technology with the existing machine, charge-coupled Device’s (CCD’s) imaging function for flaw detection. In this process, many technologies are applied, such as Image Processing Technology, Color Space Conversion, Image Geometric Transformation, Image Binarization and Morphological Processing. In order to improve the character recognition rate of the system, this study introduces the theory of Back Propagation Neural Network, and establishes a sample database for character training, which can effectively improve the accuracy of Optical Character Recognition (OCR) technology. This system has great benefits in testing equipment, specifically the identification efficiency of the product models and Engineer maintenance. It increased character recognition performance by 50.77%. It not only improves the automation capability of the production machine, but also reduces the time and the cost of labor.
|