| With the development of electronics industry, the hardware cost reductions and gradual improvement of a variety of video processing algorithms, video processing technology has reached the practical stage in the industrial, commercial and civil areas. Because the video processing technology has developed rapidly, virtual advertising system is used widely. For example, virtual advertising system can be widely used in various sports or arts programs, we put virtual ads into an open space of the arena or replace the existed ads with a new virtual ads and nobody can tell the difference between the real ads and virtual ads.This paper mainly studies the adaptive extraction of sports areas from video and camera motion tracking technology. Our aim is to improve the virtual advertising system adaptability and practicality. In addition, we designed a highly efficient video capture module. The main contributions of this thesis are as follows:Firstly, we design and implement a virtual advertisement system video capture module to capture real-time HD SDI video. In this paper we use DirectShow technology to implement video capture capabilities. DirectShow technology has the advantage that is the applications and data transmission, hardware differences, and synchronization issues are separated. In a word, the advantage is that the project development has nothing to with the video capture. Because of this advantage, the developers only need to focus on software implementation which results to the promotion of the efficiency and the wide applications.Secondly, we analyzed and proposed an adaptive extraction algorithm of sports areas, including the extraction of the main area, adaptive binary parameter and the accurate extraction of the field lines. The main area extraction method is to remove interferential scene information such as the information of the audience area in order to improve the image detection accuracy and detection speed. Adaptive binary parameter is used to get the reasonable threshold when we extract field lines. Using Hough transform and least-squares fitting a straight line method can achieve the accurate extraction of lines in the area.Finally, we analyzed field line pitch tracking methods when the camera was moved. In this paper, we achieve two motion tracking algorithms respectively based on natural feature point matching and based on Stadium models. However, the natural feature point method is time-consuming, so in the actual project, we use the latter method. This paper used two approaches to implement the latter method:one is a line detection method based on Hough space analysis, the other is feature point estimate method based on Kalman filtering. Compared the strengths and weaknesses of these two methods, we finally used the feature point estimate method for line tracking.An adaptive method of field extraction and motion tracking method had been applied to virtual advertising system well. The results had show that we also can get good practical results in different sport fields. |
