The Research Of Vehicle Trajectory Algorithm Based On SLAM

With the continuous development of railway and urban rail system, track irregularity phenomenon is concerned by more and more people, through the detection of track irregularities can determine the safety of the train. Traditional track detection technology in super long wave detection research also can’t fully meet the requirements. In this thesis, proposed a new detection method based on simultaneous localization and mapping(SLAM) algorithm, put the emphasis of the detection on the train, by detecting the vehicle trajectory indirectly reflect the track irregularities.The coordinate conversion algorithm was further studied under continuous motion state of vehicles. In this thesis, a relationship between a dynamic coordinate system and the static reference point was built to realize continuous motion measurement, which was based on the principle of SLAM and the coordinate transformation method of binocular vision system. At the same time, explored the matching method of geometric features, on the basis of classical SIFT algorithm, through the rotation and translation relationship between two images by binocular stereo vision system, proposed four images matching algorithm based on binocular stereo vision system, and introducd RANSAC algorithm to purify matching points. Designed experiments to verify algorithm by using matlab and got the image finally which matching accuracy is higher. Error evaluation method was introduced for errors in the coordinate transformation, mainly including coordinate transformation model, evaluation model and evaluation method of error coordinate system, divided feature points into conversion point set and test point set, and analyzed the impact on conversion accuracy when the number of point and the spatial distribution of point is different through the coordinates of the experiment.Finally, designed two experiments to verify this algorithm and get the simulation of the vehicle trajectory. Among them, the experiment based on the motion control platform mainly verified the coordinate transformation algorithm, control the platform continuous motion to collect data and convert coordinates through the designed line. Three axis coordinate transformation maximum error of the experiment is 0.74 mm, and the triaxial absolute displacement error of the curve is less than 1.2mm. On the simulate track model constructed in the laboratory, measured the multistage interval by the simulation railcar which is controlled by train control software, and designed image acquisition system based on MFC. X,Y, Z axis data of trajectories were analyzed quantitatively, three axis error of the experimental results is less than 3mm compared with the actual line. The experimental results show that the vehicle trajectory algorithm mentioned in this thesis is feasible and the error is small, completely conform to the requirements of the test.