Font Size: a A A

Research On Train Positioning Method Based On Landmark Recognition

Posted on:2019-06-05Degree:MasterType:Thesis
Country:ChinaCandidate:Y X WangFull Text:PDF
GTID:2392330611493298Subject:Optical Engineering
Abstract/Summary:PDF Full Text Request
The rail train is important for the transportation industry.The autonomy,accuracy and real-time of the train positioning system are significant for the safe and effective operation of the train.The common methods of train positioning are GPS,odometers,balise and so on.However,they have problems of signal loss or error accumulation.The landmark recognition method avoids processing multi-frame pictures and uses landmark information to complete train positioning.In order to realize the high-autonomous,high-precision and low-cost train positioning method,the paper focuses on the train positioning method for landmark recognition.The main research contents are as the following parts:This paper introduces the basic principles commonly used in visual positioning,including the imaging principle of photogrammetry and the conversion relationship between different coordinate systems.Based on the conversion relationship,the basic principle of calculating train position information using kilometer sign latitude and longitude information is derived.Since the positioning method needs to use the corner pixel coordinates of the kilometer sign,two methods to extract the corner pixel coordinates are proposed and compare the positioning errors through experiments.The design uses digital image processing combined with landmark features to obtain candidate regions.After that,they are classified and identified by SVM to obtain the kilometer sign.In order to make the algorithm stable,adaptive threshold combined with whole threshold is used to complete the interception of the kilometer sign.Then the template matching and BP neural network are used to identify the number on the kilometer sign.Two different methods are designed to realize the train positioning,which utilizes the actual width constraint of the kilometer and the one-dimensional simplified imaging projection model based on photogrammetry.The principle for two positioning methods and the main source of positioning errors are analyzed.The static experiments were designed to verify the feasibility of the two methods and the positioning error is within the theoretical error range.Finally,according to the whole train positioning process,dynamic experiment is designed to verify the feasibility of the method under dynamic conditions.Firstly,the latitude and longitude database of kilometer signs is established.Then the experimental equipment is built to carry out the vehicle dynamic experiment verification.The experimental results show that the error of two methods are 3m and 2m,respectively,which can meet the demand accuracy.
Keywords/Search Tags:Landmark identification, Train positioning, Computer vision, Corner pixel coordinates
PDF Full Text Request
Related items