基于主动视觉的运动目标检测跟踪方法

摘要

研究主动视觉运动目标检测跟踪系统.针对图像目标跟踪多为非连续动态过程,准确性差,通过混合高斯法建立背景模型,采用背景差分法,利用最大类间方差算法确定阈值,检测分割出运动目标,提出一种结合SURF算法的带宽自适应均值漂移跟踪算法实现目标跟踪,使用线程并行控制摄像机运动,确保跟踪目标在图像序列中的合适尺寸.实验表明,改进系统能够实现对场景中运动目标的准确检测,稳定跟踪,并能到达实时应用的要求.%In this paper we designed an active vision system for detecting and tracking moving object based on PTZ. For the object segmentation step we used the background difference method. Background model was established by mixture Gaussian algorithm, and Otsu algorithm was used to determine the threshold. We presented a new mean-shift algorithm to track. The bandwidth was adapted by SURF algorithm. The movement of camera was controlled by thread parallel technology to make sure to track the target in a suitable image size. Experiment shows that the system not only can detect the object exactly, but also is robust for tracking, and can reach the requirements of real-time applications.