Research On Detection Method For Railway Fastener Defects Based On Machine Vision

In recent years, duo to the rapid development of the high-speed and heavy train, higher requirements are put forward by the railway line maintenance work. The traditional manual maintenance has been far from satisfying the needs of the modern railway safety. The field investigation data shows that once there are consecutively three or more rail fasteners defects, it may cause the train derailment accidents. However, due to the lack of technology, the current railway fastener defects detection mode mostly still relys on manual visual inspections along with the line, and the method can not meet the requirement of periodic maintenance. In addition, the safety of the workers can not get effective guarantee. Therefore, the method has been unable to guarantee for the high-speed train to provide a safe and reliable environment. In this environment, to develop a reliable and general method of automatic of rail fastener defect detection is particularly important.Aiming at the shortcomings of the existing railway fasteners defect detection method, based on machine vision and image processing, the fastener defects automatic detection and classification are preliminary realized and the rail fastener defect detection system software interface is designed. First, the acquired frame image needs to gray processing, in order to reduce the amount of information of the image. In addition, the image denoising and enhancement algorithm are used to improve image clarity. For the problem of poor fasteners positioning, Line Segment Detector method which combines with the edges feature of the image area to locate the rail edge position is used to accurately acquire the fastener pillow shoulder position. Based on the relationship between the structure information of pillow shoulder position and fasteners area, the accurate fasteners positioning is realized. Finally, fusion with layered features of Local Binary Pattern and Histogram of Oriented Gradient algorithm is used to extract features of various types of fasteners, and made it as a basis for railway fasteners defects classification. Offline training of Support Vector Machine is completed through one-to-one classification method, and online railway fasteners defects classification is executed by "referendum rules". By selecting a large number of positive and negative samples simulations, this method is effective to provide the accuracy of recognition and classification system.The software interface platform surveillance system for railway fastener defects is designed, and the main functions include six functional modules. Such as login, video capturing, video storage, positioning and defects detection and alarm management. When the defective fasteners appear in monitor window, it can be automatically identified and classified. The classification results are sent via a wireless transmission system to the work department. The results will be as important evidence of maintenance.