Research Of The Detection Algorithm On Railway Fastener Defects Based On Image

Railway fastener is the important part of rail line. The normal state of the fastener is the guarantee of the rail transportation security. The missing or broken fastener may be the tremendous threaten to the security, or even cause the serious traffic accidents. The detection of the railway fastener state in our country those days is still in the manner of the manual detection. But this manner of detection is inefficiency and lack of objectivity, because it relies on the worker’s technical proficiency. With the development of railway transportation in our country to high-speed and heavy-load, the railway defects including fastener defects grow in number and the manual detection can’t meet the needs of modern railway transportation safety. Therefore, the automatic detection of railway fastener state is the problem that the railway development has to face. Nowadays, the automatic detection of fastener state by using image processing technology is the most widely used method in the world. Fastener defects detection algorithm based on image procesing technology can improve the detection performance, and it can also reduce the detection time to guarantee the higher frequency of line detection, which has the advantages of economic, efficient, high degree of automation, adaptability, etc.The paper mainly studies fastener defects detection algorithm based on image procesing technology. The main work is as follows:For the lack of existing fastener positioning algorithm, adaptive "cross" method is designed to locate the fastener area, which means using the node of lower edge of rail and the sleeper center line to locate the fastener. This algorithm can get over the disavantages of existing positioning algorithm. This algorithm with strong robustness and stability can’t be affected by sunlight or curved line.This paper introduces the concept and theory of template matching algorithm. The standard fastener template is designed according to the specific fastener shape in our country, and the accurate extraction of fastener target is realized by using the template matching technique. The preliminary classfication of fastener state based on the template matching coefficient can greatly reduce the difficulty and the recognition time of the subsequent fastener state classification.The reprentation of the Haar-like rectangle features and the calculation process of feature values are expounded in this paper. And the integral image is introduced to accelerate the calculation of rectangle feature. And it can also significantly reduce the computation of many rectangle features and improve the classifier training and the recognition speed. The rectangle feature represented by single rectangle is introduced because of the specific target of fastener. Fastener state recognition algorithm including the strong classifier based on AdaBoost, the weak classifier based on the single rectangular feture and the weak classifier based on template matching coefficient is designed to achieve the classification of normal fastener, missing fastener and broken fastener.The training sample library and the test sample library for the fastener state recognition is established. The fastener positioning algorithm is tested on Matlab platform. Realized the classifier in program and the classifier has also been trained and tested. The experimental result shows that the fastener state recognition algorithm based on image processing technology can meet the requirement of practical applications.