Visual wafer dies counting using geometrical characteristics

Abstract

This study presents a computer vision-based wafer dies counting algorithm. It utilises the single model and multi-model RANSAC (Random Sample Consensus) algorithms to detect the circular contour of a wafer and its dicing lanes, respectively. Statistical analysis is performed to extract the characteristics of layout so that missed dicing lanes can be supplemented. The number of integral dies is counted after segmenting individual dies based on the complete layout information. The experimental results show that the proposed algorithm has high counting accuracy and good computational efficiency.