Application of Machine Vision Technology in Defect Detection of High Performance Phase Noise Measurement Chips

Abstract

Abstract Defect detection plays a crucial role in chip quality control, and the chip field has widely researched and applied machine vision-based surface defect detection methods due to their high efficiency, accuracy, and real-time performance. In this paper, we utilize imaging equipment to collect images of high-performance phase noise measurement chips, and we use a mean filtering algorithm and a Sobel operator to preprocess the collected chip images. Then, the PCA method is applied to downscale the extracted chip shape and texture features, and the improved support vector machine algorithm using a genetic algorithm is used to classify and recognize chip defect features. The test results show that the error rate of the defect detection method for high-performance phase noise measurement of chip surface defects is only 1.82% at the highest, and the average error of the measurement of the chip pin width and pitch is much lower than the actual production of the specified error rate. Meanwhile, the design requirement of 3 pcs/s detection efficiency in the actual production of high-performance phase noise measurement chips is satisfied by the theoretical maximum defect detection efficiency of the method. The chip defect detection method presented in this paper has both practical application value and theoretical research significance.