Research On Trouble Detection Method Of Emu's Running Gear Based On Image Processing

At present,Chinese economy is at a rapid growth stage,and the railway is entering a new stage of leapfrog development.As a new means of transportation,high-speed trains began to run in large numbers,and the safety of its running state has been closely watched.The structure of the emu is complex,and the small parts are numerous,the traditional detection method is affected by the technical level of the detection personnel and the fatigue degree of the human body.It is inevitable that there will be negligence and careless mistakes in the detection process,which has a great limitation.Once the train breaks down during operation,it will cause great loss of manpower,material resources and even personal safety.So researching and developing trouble of moving EMU detection system(TEDS),realizing automatic positioning and trouble detection to enhance the accuracy of the emu repair work,security and efficiency has important practical significance.Based on this,this article according to the emu's running gear structure characteristics and features of trouble features,combining with the feature extraction and target recognition,studied the emu's running gear of typical trouble location and trouble detection algorithms,including the axle box bolts missing,traction rod crack and lateral deformation of antiroll torsion bar three categories.This article researches the emu's typical trouble of running gear with image detection technology inland and abroad research status quo,mastered the trouble of moving EMU detection system using inland and abroad,working principle and structure,etc.On the basis of the first going to the emu's running gear of image preprocessing image,including image enhancement,image segmentation,edge extraction and uneven illumination correction,established on the basis of the target image and the target image of positive,negative samples library,lay a foundation for the following feature extraction and trouble identification.This article uses the HOG feature extraction algorithm for feature extraction to axle box,next using the extracted feature training SVM classifier.The use of the training model of the axle box in the original image to identify and intercept,implements the emu's direction of axle box positioning.Then according to the loss after the formation of the axle box bolt and the surrounding gray contrast the characteristics of the large dark round hole,put forward the corresponding trouble recognition algorithm to solve the emu line of axle box bolt missing detection problem.This article uses the LBP feature extraction algorithm for feature extraction.The traction rod after extraction of the feature to the Opencv use trained classifier Adaboost algorithm,using the classifier model is used to identify the traction rod location in the image.Then,according to the geometrical characteristics of the crack,the method of thresholding is used to perform the binarization of the traction pull rod after the Angle correction,and the pull rod and crack are segmented to realize the crack detection of the traction rod.This article is based on template matching SURF feature recognition algorithm to achieve the goal of antiroll torsion bar.Then according to the lateral deformation of antiroll torsion bar trouble location is not fixed,geometric shapes and characteristics is not obvious characteristics such as large change,proposing the corresponding trouble recognition algorithm and improving the robustness of trouble detection.The research of this article provides a theoretical and algorithmic basis for the automatic identification of the typical troubles in the running gear of the emu,which is accurate,reliable and effective,and it has a good application prospect.