The Study On Detection Of Pantograph And Location Of Slidebased On Image Processing

With the continuous increasement of train speed,the running safety of high-speed trains is facing far bigger challenges.The electrical energy required for high-speed train is obtained from the contact line through the slide installed on the roof of the train.The state of pantograph directly affects the operation of the electric locomotive.At present,a large number of pantograph and slide state data are obtained through power supply safety inspection and monitoring system.It has become a challenging and important problem for how to effectively use the monitoring data obtained by long-term tracking experiments to grasp the pantograph service state.Based on these monitoring data,pantograph and slide plate are taken as the research objects and machine learning theory and image processing technology are applied to study the method of pantograph and the location method of slide.The main contents are summarized as follows:(1)Aiming at the problem of overexposure or overshadow in the planform of high-speed train,the MSRCR algorithm is used to enhance the image in vertical view of the train.The algorithm solves the problem that the brightness of the image is too high or too low and the image quality is not good.(2)The method of decision fusion based on HOG and LBP feature is used to detect the pantograph area in vertical view of the train.First,the Sobel operator filtering and the linear Hough transform are used to detect the pantograph area roughly before the exact detection.On the basis of rough detection,the LBP features and HOG features are extracted from the pantograph area,and SVM is used to train and classify the extracted features.Finally,the decision fusion is carried out based on the classified information.By this way,the final detection results are obtained.The accuracy of this method for detecting two data sets can reach 98.7% and 97.2%,respectively.Thus,this method can better achieve the exact detection of the pantograph area and reach a higher accuracy.(3)Aiming at the problem that the slide tilt in the image is not conducive to subsequent positioning,the method which combines linear Hough transform with perspective transform is proposed.The method corrects the extracted pantograph image and solves the problem of the slide tilt in the image.(4)The location of the slide in the image of the pantograph is completed.In the case of fewer samples,the data of slide is enhanced by enlarging the number of samples.The network of LeNet-5 is improved and the final CNN network structure is designed by contrast experiments of different parameters.The final CNN network has high accuracy to the location of the slide,and the validity of the network is verified.