基于PSO算法的运动目标跟踪研究

提出一种基于PSO的运动目标跟踪方法,并在TMS320DM642上实现.该方法首先针对采集到图像的Y分量通过帧间差分和背景差分相结合的方法建立背景模型,然后检测出前景运动目标,接着通过前景运动目标初始化目标模板,再对目标模板和待跟踪视频图像分别进行两次金字塔降采样,降低目标模板和待跟踪视频图像的分辨率.在顶层金子塔上采用粒子群优化算法对跟踪目标进行粗定位,在中间层和底层金字塔上采用钻石搜索方法对跟踪目标进行精确定位.在目标跟踪的过程中,目标模板随运动目标的变化而不断更新,实现对目标的实时连续性跟踪.该方法可以有效降低计算复杂度,提高搜索效率.

Study of Moving Target Tracking Based on PSO

This paper presents a moving target tracking method based on PSO, which is applied on TMS320DM642. First, according to the Y component of the collected image, this method establishes the background model by combining interframe difference with background difference and detects the prospect of the moving target. Then, it initializes the target template through the prospect of the moving target, and does two pyramid down-samplings for the target template and the video image which is to be tracked to reduce their resolution. In the pyramid＇s top-level, it adopts the particle swarm optimization algorithm to roughly locate the tracking target. In the middie layer and at the bottom of the pyramid, it uses the diamond search method to accurately locate the tracking target. In the process of target tracking, the target template constantly updates with the changes of the moving target so that it can achieve real-time and continuous target tracking. This method can effectively reduce the computational complexity and improve search efficiency.