The Vehicle Tracking Research And Realization Based On The AIMM And Multi-Feature Integration

The moving vehicle detection and tracking system based on video is one of the important research contents of the intelligent transportation system. It has incorporated the research achievements in image processing, artificial intelligence, automatic control and other related areas, and been applied widely in the transport and management fields. Therefore, the researches on the key technology of intelligent transportation system based on the video and improving the performance of the intelligent transportation system have the important theoretical significance and broad application value.Aiming at the main problems existed in the complex scene of vehicle detection and tracking, we make the researches, and put forward the solutions in order to deal with the problems, such as removeing interferences, tracking accurately and etc. This paper's main research works are as follows:In the image-preprocessing respect, because of noises existed in the collected video, we use image-processing technologies as the pretreatment to the video, e.g. the image gray-scale equalization and median filtering, in order to make the image quality improved and make ready for the following vehicle detection and tracking.In the aspects of detecting moving vehicle, firstly, we will make discussion and simulation test to background difference method and frame difference method. We find that background difference method is very sensitive to the light change, dynamic scene change and other outside interferences; frame difference method cannot extract all of the feature pixels well enough.This paper adopts the self-adaptive background difference method and combines image processing techniques. Without increasing the amount of calculation, they can adapt to a certain degree of scene changes and can extract the comparatively complete feature pixels, and then extracted complete movement target vehicle,making preparations for the following work of vehicles feature extraction.In the aspects of tracking moving vehicle, firstly, the self-adaptive interactive multiple model kalman filter are used to establish uniform and constant acceleration model, in order to predict the possible position that tracked vehicles may arise in the next frame, and then determine the search scope. Secondly, the characteristic matching algorithm given in this paper are combined so as to track the moving vehicle effectively in the cases of bend,shelter and etc.Finally, we will conduct the simulation experiment and analyse the experimental data. Test results show that the vehicle tracking model based on the self-adaptive interactive model kalman filter tallys with the actual situation comparatively, it can reliably predict the trajectory of tracking vehicle. Finally, A vehicle tracking system is designed based on AIMM and feature fusion of using the LabVIEW development platform. Through this demo system, the feasibility and practicality of the algorithm are primarily validated.