模糊直方图模型的运动目标跟踪

为改善跟踪系统对跟踪场景中目标色度和光照变化鲁棒性，提出基于模糊直方图的目标模型建立方法。首先，在色度论域内定义色度模糊等级，根据模糊隶属度函数建立目标区域模糊直方图，由此降低目标直方图模型对色度等级阈值的敏感性。然后，利用模糊直方图模型进行反向投影，建立跟踪场景的概率分布图。最后，利用Camshift方法实现目标的识别、定位与跟踪。仿真实验结果表明:与传统方法相比，采用模糊直方图模型的跟踪方法对色度漂移等干扰具有更好的适应性，目标在顺光、侧光以及逆光环境下移动时，该方法能够完成目标的准确定位与跟踪，单帧平均跟踪时间与基本Camshift方法相当，单帧最大跟踪时间小于40 ms，满足电视跟踪等系统实时性要求。

Target tracking based on the fuzzy histogram model

To enhance the robustness of the tracking system to the hue drift and illumination variation, a new target model based on the fuzzy histogram is proposed in this paper. First, some fuzzy ranks were defined in the hue domain, and the fuzzy histogram model of the target area is built according to the fuzzy membership degree to reduce the sensibility to the hue thresholds. Then, the probability distribution image of the tracking-scene image based on the back-projection of the fuzzy histogram model was built. Finally, the Camshift tracking method was used to perform target recognition, location, and tracking. The simulation results show that the method has better adaptability to the interference caused by hue drift than conventional methods. Target tracking can still be completed by the method when the target moves in the illumination-varying condition, such as front, side, and back light. The average tracking computation time of a single frame of the method is similar to that of the basic Camshift method. The maximum tracking computation time of a single frame is less than 40 ms, which can satisfy the real-time request of tracking systems such as a television tracking system.