Research On The Detection Algorithm Of Railway Fastener Based On Computer Vision

Rail fastener as an important part of the orbit is mainly used for linking the rail with the sleeper, fixing the correct position of the rails and preventing the rail displacement and tilting. In the actual railway line, the rupture and loss of rail fastener will be happened which directly affect the safe operation of the train. So it’s an important aspect in the process for daily operation and maintenance of fastener state detection. At present, on the international, railway fasteners detection system based on computer vision technology as a very good solution has been widely applied because of simple implementation and high efficiency.This article mainly aimed at the rail fastener detection system based on computer vision technology in railway fasteners detection algorithm, the main contents are as follows:Aiming at the shortcomings of the existing railway fastener localization algorithm, this paper puts forward a method of two-step positioning fasteners. First confirmed the position of the rail and sleeper in the original railway image, and according to the location of the relationship between them and fasteners coarse located fasteners area. Then on the basis of the coarse positioning image and the block shoulder information of nearby top and bottom edge of the fasteners, achieved precise localization of fasteners through template operations on gradient image outstanding block shoulder characteristics and suppress the noise and the effect of implement precise localization of fasteners. And using localization algorithm test sample library collectted on the actual line verified the effectiveness of the algorithm in this paper.This algorithm can effectively solve the problem that the positioning area of existing fastener localization algorithm is too large. It is subject to the influence of illumination and circuit bending mildly and has a strong robustness.For single feature easily affected by noise and illumination variation, the recognition rate is low, the fusion of PHOG characteristics which focus on the reflection of the contour of image multi-scale structural information and LBP features which focus on local texture features in the image is proposed in this paper, and combining support vector machine (SVM) method based on the fusion feature realized the state recognition of railway fasteners in images. And using identification algorithm test sample library colleetted on the actual line verified the effectiveness of the algorithm in this paper. Compared with the single feature, the fusion feature can obtain better recognition effect.