摘要
该文提出了一种基于材质特征的视觉跟踪算法。该算法利用背景差分法将需跟踪的样本从静止背景中提取出来,然后在在HIS色彩空间内,以样本像素的H、I、S分量为坐标,得到离散空间曲面∑0,经平滑处理,得到局部连续曲面∑1。对跟踪视频序列每一帧的每一像素计算其S分量到达曲面∑1的最近距离,从而判断该像素是否属于要跟踪的样本材质。根据所有属于该材质的像素点的空间坐标,可计算出样本出现的范围。实验证明该方法简单、有效,解决了视频跟踪中根据材质进行跟踪的问题。
A visual tracking algorithm based on the material feature is proposed. The image of sample material can be gained with the difference between the foreground and background. In the HIS color space, discrete space surface ∑ 0 will be resulted if H、I and S component of the sample pixels are regarded as coordinates. Local continuous surface ∑ 1 is resulted through smoothly han- dling. The distance between the S-component of the pixels in the detected image and the ∑1 will be regarded as the similar degree with which we identify the sample material and the detected material. The tracking scope can be found with the coordinate of the tracking material pixels. Experiment proved that the method is simple, effective, to solve the problem of visual tracking.
出处
《微计算机信息》
2009年第12期296-297,302,共3页
Control & Automation
基金
基金申请人:崔杜武
项目名称:基于社会和自然双进化的算法研究
基金颁发部门:国家自然科学基金委(60743009)