Research On Location And Recognition Of Contact Signs

With the rapid and efficient development of the city,the high-speed and safe track environment is more and more important for the transportation of people and goods,and the corresponding requirements for the track detection are higher and higher.Real-time and accurate positioning is a basic condition for track detection and also a prerequisite for accurate detection of track infrastructure.At present,there are many kinds of track positioning technologies,but there are some problems more or less,such as the laying of the trackside transponder in the urban rail line and the later maintenance needs a lot of manpower and material resources;GPS may lose signals when entering the tunnel;GYK(railway vehicle operation control equipment)has a range accumulation error problem,ranging from tens of meters to hundreds of meters.It is of great significance to study a new orbital positioning method.Considering the existing subway track environment,this paper takes the video information of a section of Shanghai metro line 9 as the research object.After analyzing the advantages of digital image processing,this paper proposes the positioning method of catenary pillar based on visual significance detection to solve the positioning detection of catenary pillar.The problem of sign number recognition is solved by using the neural network classifier,and the feasibility and accuracy of the method are verified by experiments.Main research contents of this paper:(1)In view of the problem of data redundancy and complex background in the catenary pillar image along the track in the video acquisition format,the pre-processing method of catenary pillar video was studied.Since there are a lot of useless background images between the catenary pillar image sequences,it is necessary to separate the frames of the continuous catenary column video images.The algorithm in this paper calculates and statistics the feature of the Histogram of Oriented Gradient(HOG)of the image,obtains the feature information of positive sample and negative sample,and then USES theSupport vector Machine(SVM)classifier to obtain the image key frame containing the pillar to achieve the goal of fast image classification.(2)In order to improve the efficiency and accuracy of positioning and detection of catenary pillars,simple linear iterative clustering(SLIC)superpixel segmentation and markov chain significance detection algorithm are used to segment the foreground of catenary columns.Super pixel algorithm for the catechin column image segmentation,the algorithm will be divided into several super pixels,remove the adjacent pixels similar color and texture redundancy,super pixel blocks replace pixels as the most basic image processing unit,greatly reduce the post-order operation time,improve the processing efficiency.Using absorbing markov chain based on background prior to significant detection to extract the catenary post area,is firstly established after super division of pixels on the figure of graph model,using the strong background prior to obtain the absorption of absorbing markov chain nodes,by absorbing markov chain model to calculate the super nodes pixels to absorb the expectations of the time as a significant value,eventually get significant figure,improves the detection accuracy and absorption efficiency catenary post area.(3)Finally,the multi-layer Perceptron(MLP)neural network classifier was built and the corresponding parameters were set.After a large number of training samples of signboard images were input,the convergence accuracy was achieved after several iterations,and the classifier model was finally obtained.The test set is identified by random sampling,and the corresponding number information of the sign can be identified by inputting the sign area.Finally,according to the actual detection needs,on the basis of the research content of this paper,the software of the overhead contact network column positioning and sign recognition system under the Windows system is developed,which is used for the actual line detection,and the track positioning based on the sign plate is finally completed.