The Research Of Moving Vehicle Target Recognition And Tracking Technology

The recognition and tracking technology of moving object has a broad application prospects in many fields, such as national economy, state military, social life and so on, which is one of the hot issues in the current research. It is based on image analysis, using some related methods of image recognition and tracking to achieve the purpose of object detection,recognition and tracking in image sequences.In this article, through comparing and analyzing various algorithms of image processing,target recognition and tracking, we have chosen a moving vehicle as research object and put forward a method to achieve the recognition and tracking of moving object in traffic scene.Firstly, in the image processing module, the methods of preprocessing and image segmentation have been studied and improved, through making compare with their results,finally we have selected the best method of them. In addition, in the process of image segmentation, combined with the Otsu algorithm and adaptive threshold segmentation method, which has made the image have a better segmentation effect and have a small amount of calculation. Secondly, in the target recognition module, we have studied two kinds of image matching methods based on pixel and feature. What’s more, we have matched the image based on MAD algorithm and SSDA algorithm, after programming and simulation experiments, we have determined to complete the object recognition and matching by SSDA algorithm owing to its higher efficiency. Finally, in the tracking module,a method of window centroid tracking and an algorithm using adaptive-neighborhood have been studied, which have completed target tracking in the window and improved the tracking speed and accuracy. Furthermore, the prediction algorithm also has been studied, so that the target tracking still can be realized in the case of short time occlusion.In this paper, through the program debugging and simulation of various processing methods and algorithms, we have determined the optimal method and algorithm, which hasgreatly improved the real-time property and stability of moving object’s recognition and tracking.