The Research On Moving Object Tracking In The Dynamic Background

The thesis is focused on the technique of the moving object tracking in the dynamic background, which is used in TV tracking system. Under these conditions, the aim of this thesis is to study image filtering, image segmentation, object-tracking algorithms and forecast-tracking algorithms systematically, and based on the available research work to summarize an applicable objects tracking scheme.For the image filtering, the comparison of some ready-made methods is included through the analysis of noise based on tracking system. The methods include neighborhood-averaging method, edge sharpening and median filter method. An improved median filter method with better real-time quality is also introduced in the thesis.The methods of image segmentation can be categorized into three types: threshold, district tracking and edge examination. The research for this part is mostly based on the threshold segmentation method. A statistics based method of Moment-algorithm is introduced to calculate the threshold value automatically. Another method based on the gray scale threshold is used to achieve better real-time quality.Through the analysis of window tracking algorithm which is usually based in static background, a solution is proposed to apply the algorithm to the condition of dynamic-background. An improved correlation-tracking algorithm with optimized principle is proposed to make the tracking system better real-time quality. Some other object-tracking solutions are also provided. Then, two tracking methods which is based image-segmentation and model-matching are provided to achieve the moving object tracking.Starting with the characteristics of motion objects, the importance of forecast tracking is discussed. A compositive algorithm is achieved based on the ready-made extrapolation prediction algorithms. Emulation results show that it is effective in the moving object forecast tracking.