The Research Of The Moving Objects Tracking System

Moving object tracking in video sequences is the fundamental and challenging research topics in the area of computer vision at present.They have great research and application values in modern industry,military,navigation and spaceflight area.The moving object tracking system includes two important parts that are moving object detection and tracking.The main contributions in this dissertation are as follow:Three methods of moving object detection,including difference between adjacent frames,background subtraction and optical flow,are discussed.According to the characteristic of video,we use the segmentation algorithm based on change of multi-frame edges.Based on this algorithm,edge differences between a group of frames are used to draw the area of moving objects;then,background pixels are removed through setting up pixel-measuring window and the threshold;the areas of objects are set up by morphology operator.This method has the advantages of high-speed and convenient operating.The accuracy and stability of object tracking depend on the algorithm of moving object tracking to a great extent.Most of the existing approaches have to processe a lot of data and,therefore,they are very complicated.Two object tracking algorithms, tracking video object based on centroid and vector and Kalman filtering for tracking video objects based on motion estimation,are proposed in this dissertation.The basic idea of the former is as follows.After video objects segmentation,video objects are divided into some regions using region-growing method.Then with the centroid and the direction of the vector of each region in successive two frames of a video sequence, it is able to track multiple non-rigid moving objects fast,efficiently and automatically. The basic idea of the latter can be described as follows.Firstly,video objects are segmented and their centroids are calculated.Then,Kalman filtering is used to predict the object centroids in the next frame with the information of centroids and motion vectors in successive two frames in a video sequence.It is able to track multiple non-rigid moving objects fast,efficiently and automatically.The experimental results show that the proposed methods is robust to deal with the problems in multi-targets tracking such as new target entering,target leaving,scaling and deforming of non-rigid objects.