针对视频监控的特点与跟踪目标的强机动性,提出了一种新的基于概率模型的目标跟踪框架,从目标表观模型、系统动态模型以及系统观测模型3个方面对当前标准的粒子滤波目标跟踪方法进行了改进.首先,考虑人眼细胞的分布特点,基于人眼分布结构建立目标表观模型来提高跟踪系统抵抗局部遮挡的能力;然后,建立基于自适应目标运动的系统动态模型,提高跟踪算法对快速机动目标的鲁棒性;最后,采用实时更新的系统观测模型,有效避免目标在遇到遮挡、光照变化、剧烈变形等情况下发生的跟踪漂移现象.实验结果表明,本文算法的正确跟踪率可达98%;平均跟踪误差小于6个像元.实验证明本文算法在保证系统跟踪精度要求的同时,具有计算量小、抗干扰能力强等特点.

According to characteristics of video surveillance and strong mobility of tracking targets,a novel target tracking framework based on a probability model was proposed.The current standard particle filtering target algorithm was improved based on a target appearance model,a systemic dynamic model,and a systemic observation model.Firstly,the target appearance model was established by taking the distribution of human eye cell into account to improve its resistance capability for partial occlusion of local occlusion.Then,the systemic dynamic model based on the adaptive target movement was built to improve the robustness of tracking framework for the fast moving target.Finally,the systemic observation model with online update was established to prevent the tracking shift when the target faced the occlusion,illumination changes,severe deformation,etc.,effectively. Experimental results show that the proposed algorithm achieves 98% of correct tracking rate,and the average tracking error is less than 6 pixels.The proposed method satisfies the video surveillance system requirements for stabilization,reliability,higher precision,less computing cost,as well as strong anti-jamming.