Research Of Train Satellite Positioning Preprocessing Method Based On Computer Vision

With the rapid development of the GNSS (Global Navigation Satellite System) system in the field of intelligent transportation system, GNSS technology has been widely used in the train positioning. GNSS technology has great advantages for train positioning, but because of the poor environment along the railway, GNSS satellite signal is easily blocked, and reflected or refracted, which makes multipath effect and affects the positioning accuracy. So, it has theoretical value and practical significance to detect obstacles and to reconstruct them to exclude the satellites which may produce multipath effect to improve the positioning accuracy.A preprocessing method is proposed in this thesis to aid the satellite selection for train GNSS positioning system, which is based on computer vision technology to get the shadow condition along the track. Using cameras to form groups of binocular vision systems and record the scenes along the railway, the angle database can be got by processing the image record to detect obstacles and get the shadowing angle and azimuth information around the antenna depending on stereo vision. The database can be applied for satellite selection, preventing the influence of multipath effect.The main research contents include:1. Camera calibration: accomplished camera calibration and the conversion matrix between the left and right camera coordinates based on 2D chessboard calibration method;2. Image preprocessing: a more suitable preproccess method for the trackside scenes is proposed, based on gray scale transform, can improve the contrast and enhance the image details;3. Feature extraction of the obstacles: accomplished feature extraction of the obstacles, extracted corner and edge information, modified the Harris operator;4. Feature match: accomplished feature match, proposed a method for rough match via the combination of monogenic signal and regional match method; after that epipolar geometry is applied for accurate match.5. 3D reconstruction and angle information solution: reconstruction is finished and angle information is got, which layed the foundation for satellite selection.Algorithms are detailedly deduced in this thesis, and through the simulationexperiment, good results are manifested.