Detection and Recognition of Mixed-Type Defect Patterns in Wafer Bin Maps via Tensor Voting
Rui Wang, Nan Chen
Abstract
Spatial defect patterns on semiconductor wafer bin maps can provide valuable information on the root causes of process abnormalities. Thus, the identification of these patterns is important for quality management and yield improvement. To date, most research has focused on the case of a single defect pattern. The recognition of mixed-type patterns is challenging as the patterns need to be separated into clusters and each cluster classified as a predefined type of defect pattern. In this study, we propose a novel method for mixed-type defect pattern detection and recognition in wafer bin maps. We separate mixed-type defect patterns into clusters using tensor voting. Marching algorithms are then developed to extract region and curve patterns based on the structural saliency information of the voting process. Our method is inherently robust against noise and sufficiently flexible to deal with complex defect patterns. The results obtained using real and simulated data demonstrate the effectiveness of the proposed method in the detection and recognition of both single and mixed-type defect patterns.