基于背景补偿引导的动态场景下目标跟踪算法

Target tracking algorithm based on background compensation in dynamic scene

针对动态场景的目标跟踪中,目标的尺度及姿态变化使得跟踪过程易出现误差累积导致跟踪失败的问题,以及可能出现的遮挡、粘连等复杂运动情况,提出了一种背景补偿引导的搜索及目标模型更新的策略,采用基于Random ferns的特征描述方法,实现动态场景下的运动目标跟踪。实验结果证明,该方法能对目标的姿态变化有较好地适应性,在遮挡、粘连现象存在的情况下,能在动态场景下较好的实现运动目标跟踪。

Aiming at the problems that in target tracking with dynamic scene, the scale and posture change of the target may result in error accumulation and lead to tracking failure, and there may exist some complex situations such as occlusion or adhering, this paper proposes a strategy of background compensation guided search and target model update ; the feature description method based on Random ferns is adopted to achieve the target tracking in dynamic scene. The experiment results show that the proposed method has good adapta- bility to target posture change and can achieve target tracking in dynamic scene, even under the conditions of occlusion and adhering.