摘要
为了提高目标检测的快速性与准确性,简化基于粒子滤波的目标跟踪算法中的直方图计算,提高检测和跟踪算法在基于DSP(数字信号处理器)的主动视觉系统上的运行速度,提出了一种基于DSP的机器人主动视觉系统.该系统通过改进的EMCV(embedded computer vision library)与启发式搜索方法,在DSP上实现了AdaBoost检测算法;利用增量式直方图计算算法实现粒子滤波中颜色直方图与边缘方向直方图的计算,将直方图融合到观测模型中,在DSP上实现并优化了目标跟踪算法.实验证实了该主动视觉系统中算法的快速性与系统的鲁棒性.
In order to improve target detection speed and accuracy, simplify particle filter based target tracking algorithm for histogram calculation, and improve the speed of detection and tracking algorithm in the DSP （digital signal processor）- based active vision systems, a DSP-based active vision system is proposed. By improved the EMCV （Embedded Computer Vision Library） and heuristic search methods, the system implements the AdaBoost detection algorithm on the DSP. And in that system, the color histogram and edge orientation histogram in particle filter are calculated by using the incremental histogram calculation algorithm, the histogram is integrated into the observation model, and the target tracking algorithm is optimized on the DSP. The experiment proves the rapidity and the robustness of the proposed algorithm in the active vision system.
出处
《机器人》
EI
CSCD
北大核心
2012年第3期354-362,共9页
Robot
基金
国家自然科学基金资助项目(61075027
90820304
91120011)
国家973计划资助项目(G2007CB311003
2009CB724002)
上海市研究生创新基金资助项目(JWCXSL1022)
河北省自然科学基金资助项目(F2010001106)
关键词
主动视觉
目标检测
目标跟踪
ADABOOST
增量式直方图
粒子滤波
DSP
active vision
target detection
target tracking
AdaBoost
incremental histogram
particle filter
DSP （digital signal processor）
