摘要
本文在传统的Kalman滤波和Mean-Shift优化框架下提出了一种新的视频运动目标跟踪算法。融合色度直方图和梯度方向直方图,形成了一种新的综合直方图特征.构建运动目标图像区域的金字塔,采用Kalman滤波预测耦合Mean-Shift算法的框架,在尺度、位移空间内进行优化匹配搜索,确定最佳候选目标的位置信息。大量实验结果表明,本文提出的在滤波与优化算法框架下的运动目标跟踪算法,能够很好地解决运动目标的尺度伸缩、旋转和形变等难题,可以取得比基于传统直方图更好的稳定性和跟踪精度。
An improved moving object tracking algorithm is presented based on the framework of traditional Kalman filter and Mean-Shift optimization. Firstly, by the fusion of Oriented Grads and Colors histogram a new histograms of Oriented Grads ＆ Colors （HoGC） is proposed. Secondly, HoGC pyramid is constructed to more robustly characterize the multi-scale objects. Finally, by coupling kernel-based Mean-Shift algorithm with Kalman Filter, HoGC matching is optimized in terms of scale and displacement of the candidate object identified. Experiments demonstrate that the proposed HoGC is robust to tracking moving objects and is invariant to scale and deformation. The proposed tracking algorithm can improve the reliability and accuracy without losing the real-time performance.
出处
《河南大学学报（自然科学版）》
CAS
北大核心
2007年第6期629-634,共6页
Journal of Henan University:Natural Science
基金
国家自然科学基金项目(60672147)
高校杰出科研人才创新工程项目(2003KYCX003)
河南省自然科学基金项目(0411010400)
