Wafer Dies Counting Based On Computer Vision

Wafer die counting based on machine vision can be used to improve the efficiency of semiconductor manufacturing. Many domestic and international IC manufacturing enterprises still count off-line wafer dies manually, this approach is vulnerable to a variety of human factors, the efficiency and accuracy are relatively low. The manual work needs urgent improvement to accommodate the urgent need for industrial upgrading.The previous published papers related with wafer are mostly related to the defect detection of wafer dies using microscopy techniques. Their testing equipments are complex and expensive. Some papers are focusing on the theoretical die count per waferwith the aim of manufacturing more dies on a wafer. To the best of our knowledge, no paper related to the visual wafer dies counting has been published.This paper proposes the wafer dies counting method which based on machine vision. First Robert operator for edge detection was used, the method obtain the outline of the wafer, fitting round use RANSAC algorithm, determine the wafer position in the image, implement RANSAC algorithm in view of the wafer area for more straight line detection. Then using the wafer’s layout characteristics, on the basis of straight line bearing the detected straight line can divide into two categories, eliminate the wrong line and calculate the distance between two adjacent lines. Using the extraction of the geometrical characteristics for the completion of leak detection line in order to complete the wafer internal structure layout information. Finally based on the mathematical morphology method, the connected domain can be found, filter pieces are connected to the side of wafer circle, segment the complete chip and counting. Through experimental checking, this paper proposed method of the average count time is6.4s, the average counting accuracy reached98.4%. It indicates that using this method for count is faster than artificial count, and the accuracy is higher, can help to improve the test automation of production line.