摘要
为了解决复杂场景中运动目标跟踪,特别是遮挡情况下对目标的连续、稳定跟踪的问题,提出一种基于决策主导的多模式融合跟踪算法。采用多层算法结构,以图像特征作为决策判据,自主控制算法流程;首先利用重心和改进的粒子滤波算法预测目标位置进行粗定位,而后用改进的SIFT特征匹配对目标精确定位。在保证跟踪性能的同时大大简化了算法的复杂度,提高了算法的实时性。实验表明,多模融合跟踪能够在目标发生旋转、缩放和有物体遮挡干扰的情况下,准确地提取目标,并保持连续稳定的跟踪,完全可以满足工程应用中实时性和鲁棒性的要求。
In order to solve the problem of moving target tracking in complex background, especially the continuous and robust target tracking under occlusion condition, a target tracking algorithm based on decision-leading and multipattern fusion is proposed. The muhilayer algorithm structure is adopted, and image feature is taken as the decision- making criteria, so the algorithm procedure can be self-controlled. Firstly, centroid and improved particle filter algorithm are used to predict the target position and coarsely locate the .target; then, the improved SIFT feature matching pair is adopted to precisely locate the target. Besides ensuring the tracking performance, the proposed method greatly simplifies the algorithm complexity, and improves its real-time performance. Experiment results indicate that with multi-pattern fusion, the proposed algorithm can accurately extract, and continuously and robustly track the target in the conditions of target rotation, scaling and occlusion. The proposed algorithm can fulfill the requirements of real-time and robustness in engineering applications.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2013年第3期487-493,共7页
Chinese Journal of Scientific Instrument
基金
国家自然科学基金(61172111)资助项目
关键词
图像处理
目标跟踪
决策主导
多模融合
粒子滤波
SIFT特征匹配
image processing
target tracking
decision-leading
multi-pattern fusion
particle filtering
SIFT feature matching
