查询结果:   赵怀勋,姜莹礁.基于颜色和方向梯度特征融合的粒子滤波跟踪算法[J].计算机应用与软件,2014,31(6):260 - 263.
中文标题
基于颜色和方向梯度特征融合的粒子滤波跟踪算法
发表栏目
算法
摘要点击数
771
英文标题
PARTICLE FILTER TRACKING ALGORITHM BASED ON FUSING COLOUR and ORIENTATION GRADIENT FEATURES
作 者
赵怀勋 姜莹礁 Zhao Huaixun Jiang Yingjiao
作者单位
武警工程大学信息工程系 陕西 西安 710086     
英文单位
Department of Communication Engineering, College of Armed Police Force Engineering, Xi’an 710086, Shaanxi, China     
关键词
粒子聚类 带宽自适应 方向梯度 特征融合
Keywords
Particle clustering Bandwidth-adaptive Orientation gradient Feature fusion
基金项目
国家自然科学基金项目(61071217)
作者资料
赵怀勋,教授,主研领域:信息识别。姜莹礁,硕士生。[HJ2mm] 。
文章摘要
提出基于均值偏移的粒子滤波算法。该算法融合了颜色特征和方向梯度特征对人体行为实施跟踪。HSV颜色模型采用带宽自适应的改进策略,有效地实现了对尺度变化目标的跟踪;方向梯度特征的融合,提高了算法在复杂环境中应对光线变化和相似目标遮挡等情况的适应性。针对粒子退化问题,在特征融合的粒子滤波基础上引入具有聚类作用的Mean Shift算法,从而可以用较少数量的粒子达到较好的跟踪效果。仿真实验表明,该算法具有较好的准确性、鲁棒性与实时性。
Abstract
We present a mean shift-based particle filter algorithm in the paper. The algorithm fuses the colour feature and the orientation gradient feature to carry out human behaviour tracking. The HSV colour model adopts the bandwidth-adaptive improvement strategy, this realises the tracking on the targets with scale changes effectively. Orientation gradient features fusion improves the adaptability of the algorithm for coping with the light changing and the similar targets occlusion in complex environment. In light of particle degeneracy, the mean shift algorithm with clustering function is introduced based on the particle filter with feature fusion, so that it is able to use less number of particles to achieve better tracking performance. Simulation experimental results show that this algorithm has relatively good accuracy, robustness and real-time property.
下载PDF全文