Complex Image Recognition Method of Train Wheels Based on Contour Feature

摘要

At present, most of the domestic train wheel detection is still manual. Existing image detection methods do not fully consider the characteristics of wheel profiles, as the algorithm is very complicated, they do not apply to fast dynamic monitoring. The paper proposes a dynamic image recognition method of train wheels based on contour feature. Firstly, wheel contour curve is extracted using RCD circular arc detection algorithm with constraint condition. Then, after analyzing the wheel contour curves, vector contour is constructed to describe them. At last, images of wheels are rapid recognized, and the recognition is based on identification function. Experimental results show that this method can achieve 95% accuracy when the similarity of complete wheel is set to 0.98 or more. This method can adapt to railway environment and has a certain application value.