摘要
首先在通过Horn-Schunk方法得到的全局光流矢量图上运用阈值化、平滑、图像形态学操作等图像处理方式实现基于光流的静态背景下的运动目标检测;接着利用光流法得到的结果为卡尔曼滤波器提供初始值和观测向量,从而实现基于卡尔曼滤波器的运动目标检测。对上述两种跟踪方法给出了仿真结果并对两种跟踪的效果从跟踪轨迹和误差比两方面进行了比较,总结了两种跟踪算法存在的局限性和可能的改进方法。
This paper introduces the fundamental principles of optical flow at first, states the theory of calculating global optical flow with Horn-Schunk method, realizes the calculation of global optical flow. It uses image processing methods, such as thresholding algorithm, smoothing, image morphological operation on global optical flow vectogram to realize the detection and tracking of moving targets based on optical flow under static background. It also utilizes a set of freeway videos to carry out simulation experiment. At last, it summarizes the limitation and places to be improved of this method, and puts forward possible solutions.
出处
《电脑与电信》
2012年第9期32-34,50,共4页
Computer & Telecommunication
基金
湖南省科技计划项目:基于马尔科夫场的图像分割算法研究
项目编号:2011FJ3016
2012SK3116
益阳市科技计划项目
项目编号:2010JZ50
