摘要
为解决单一特征目标跟踪鲁棒性较差的问题,提出一种基于颜色和空间信息的多特征融合目标跟踪算法。采用一种自适应划分颜色区间的方法提取目标颜色特征,利用空间直方图提取目标颜色的空间分布信息。在粒子滤波框架下将自适应颜色直方图和空间直方图相结合,在特征融合中引入特征不确定性度量方法,自适应调整不同特征对跟踪结果的贡献,提高算法的鲁棒性。仿真实验结果表明,该跟踪算法平均位置最小误差值仅6.967像素,而单一特征跟踪算法以及传统融合算法的跟踪误差达192.576像素和199.464像素。说明本文算法在跟踪准确性上优于单一特征跟踪算法及传统融合算法,具有更好的跟踪精度和更高的鲁棒性。
Object tracking using single feature often leads to a poor robustness.Aiming at this,an object tracking algorithm using multiple features fusion based on color and space information was presented.In order to enhance the important features,an adaptive method for choosing object color histogram was proposed to get an accurate color model of the object.Meanwhile,spatiograms were used to obtain spatial layout of these colors for the targets.These features were rationally fused in the framework of particle filter.The uncertainty measurement method was then introduced into features fusion to adjust the relative contributions of different features adaptively,and the robustness of the algorithm was significantly enhanced.Simulation experimental results show that the mean minimum location error of the proposed tracking algorithm is only 6.967 pixel,while that of the signal feature tracking algorithm and the traditional algorithm are 192.576 pixel and 199.464 pixel,respectively,which indicates that the proposed algorithm can track objects with better tracking accuracy and robustness.
出处
《应用光学》
CAS
CSCD
北大核心
2015年第5期755-761,共7页
Journal of Applied Optics
基金
国家自然科学基金(61175029
61473309)
