Automatic Recognition for Losing of Train Bogie Center Plate Screw Based on Multiple-Fuzzy Relation Tree

摘要

Automatic recognition of losing of bogie center plate screw (BCPS) on digit images is a hard problem that due to the vast differences and impreciseness associated with the poor sampling environment and moving photographing. This paper presents a multiple-fuzzy relation tree to solve this problem. Single sample is preprocessed by the method such as finite-radius Hough transformation, constraint-rectangle calculator and the gray average of local domain to obtain elements construct the fuzzy sets. Multiple-fuzzy relation tree was designed based on the rules based on fuzzy sets. Demonstrated by application, the multiple-fuzzy relation tree has an advantage with the recognition of complex pattern, strong correlation and low contrast images. The recognition accuracy is achievable up to 90%.