摘要
准确性和实时性是视频序列图像中运动目标跟踪算法研究的重要内容.为了克服传统的模板匹配跟踪算法运算量大、跟踪速度慢的缺点,提出了一种基于多分辨率的Kalman滤波快速跟踪算法.首先利用Kalman滤波的预测功能,预先估计出目标中心点坐标,然后在该坐标为中心的区域内进行多分辨率相关匹配,最终找到最佳匹配位置.该算法具有运算量小、跟踪速度快的优点.同时还采用了自适应更新记忆滤波算法解决发散问题,提高了跟踪精度.
Accuracy and real-time are important research contents of the moving object tracking algorithm for video sequence images. To overcome the large amount of computation and the weak overall anti-jamming capability of traditional template matching method, this paper presents a fast tracking algorithm based on multi-resolution Kalman filter. Firstly, take advantage of Kalman filterg prediction function to estimate the general object location and then conduct multi-resolution matching to find the best matching position within the location. This algorithm has small amount of computation and high tracking speed. And for the radiation problem, the self-adapting updating memory filter algorithm is proposed to improve the tracking accuracy.
出处
《应用科技》
CAS
2008年第12期7-10,共4页
Applied Science and Technology
基金
黑龙江省自然科学基金资助(AF200611).