关于视频目标图像跟踪融合优化研究  

Fusion Optimization of Video Target Image Tracking

作　　者：李云彤[1] 徐海明[1] 

机构地区：[1]四川大学电气信息学院,四川成都610065

出　　处：《计算机仿真》2016年第3期433-437,共5页Computer Simulation

基　　金：四川省科研项目(2010YA001)

摘　　要：在目标图像跟踪融合优化时,由于背景特征参数容易受到运动速度和光照变化产生较大干扰,图像质量差。仅以固定权值进行描述,采用传统目标跟踪方法,由于概率权重变化趋势在这种干扰下呈现不稳定状态,无法准确地衡量变化趋势带来的影响,导致目标跟踪精度差、鲁棒性低的问题。提出采用多特征加权融合的目标跟踪算法。首先将LBP纹理特征和颜色特征融合起来建立目标模型和候选模型,然后通过未归一化的目标和背景直方图计算出每个特征在目标中的概率权重,并将概率权重引入巴氏系数的相似度量中,从而实现MWMF算法,最后采用仿真测试算法的性能。实验结果表明,改进算法提高了目标跟踪的精度,加快了目标跟踪的速度,并且具有较强的鲁棒性。The fusion optimization of target image tracking of background characteristic parameters is susceptible to movement speed and the larger interference illumination changes. Bacause the probability weights change trend under the disturbance shows unstable state, traditional target tracking methods with fixed weights cannot accurately measure the impact of the change trend, which results in low target tracking accuracy and robustness. A fusion opti- mization algorithm of target tracking based on multiple feature weighting is put forward. Firstly, the target model and candidate model are established by combining LBP texture features with color features, then through histogram nor- realization of target and background, the probability of each feature in the target weight is cMculated, and the proba- bility weighting pap coefficient is introduced to similarity measures, so as to realize MWMF algorithm. Finally, the performance of the algorithm is tested by simulation. The simulation results show that the improved algorithm en- hances the precision of target tracking, accelerates the target tracking, and has strong robustness.

关 键 词：目标跟踪 跟踪算法 特征融合 特征权值 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]