摘要
提出了一种基于摄像机运动控制的实时运动对象检测与跟踪算法。该算法首先采用非参数核密度估计方法在复杂的动态背景条件下检测运动对象区域;然后由CamShift算法计算跟踪目标的位置,并采用Kalman滤波预测对象的运动信息来控制摄像机的运动,能够准确地跟踪对象并有效解决了背景中的部分遮挡问题。在SonyRZ25网络摄像机上完成了对象的实时检测与跟踪,对多个室内和室外场景的实验验证了所提出方法的实用性和有效性。
A novel algorithm for the real-time object detection and tracking based on camera movement control is proposed in this paper. The moving objects are detected by the non-parametric kernel density estimation, which can solve the problem with complex dynamic background. Then the CamShift algorithm is applied to find the center of the tracked objects. To deal with the camera control, a Kalman filter is introduced to predict the moving information. The experimental results with SonyRZ25 net camera show that this algorithm is robust and efficient.
出处
《电视技术》
北大核心
2008年第11期79-81,共3页
Video Engineering
基金
国家自然科学基金资助项目(60602012)
上海市教育发展基金会晨光计划项目(2007CG53)
上海大学优秀青年教师基金资助项目
