Research On Automatic Railway Fastener Defect Detection System Based On Image Processing

At present,railway transport plays an important role in the development of society and it also an indispensable element in the rapid development of the economic.Railway transport is not only the pillar of national economic,but also has the characteristics of safety and comfort,economic benefit,energy conservation and emission reduction,high speed and efficiency.These characteristics result that people increasingly rely on railway transport and that its position in Chinese transportation system is important.Its role and influence on economic and social development should not be underestimated.Therefore,very important to ensure the safety of railway traffic.In our country,due to the restriction of some factors such as the development of science and technology,we rely on the manual inspection for a long time.However,the manual inspection is inefficient and wastes a great deal of manpower and time.People's judging standards are different.As people's increasing demand of travelling with the railway,we need an automated inspection system instead of ordinary manual inspection to increase the accuracy of inspection.Currently,we mainly detect the exceptions of the fasteners which are connected the railroad track with the sleepers in actual application.There are mainly two types of fasteners: the type-W and the type-6 Fasteners.This paper primarily study following two parts: the position and the defect detection of fastener,especially the discussion of defect detection method.The part of positioning.Using the method of projection to locate the position of rail and sleeper,and then positioning the fastener through the position of the two periodically.(2)The part of detection.In this paper,we use the traditional HOG feature extraction method combined with SVM classifier and find the most suitable feature classification size by the principle of image bilinear interpolation scaling.In this paper,we primarily introduce the second part and give detailed description of HOG features,SVM and convolution neural network.We perform a lot of experiments on these methods.The result shows that both methods can detect the defect of fasteners in different degree.The two methods improve detection efficiency for railway fasteners and ensure the safety and stability of the train driving,which has a certain engineering significance.