摘要
论文提出了一种基于CMT(Consensus-based Tracking and Matching of Keypoints for Object Tracking)框架的目标跟踪算法。针对传统的CMT算法采用固定模型的缺点,该文引入时空上下文(Spatio-Temporal Context)跟踪算法,提出一种由跟踪器,检测器与融合器组成的目标跟踪算法。该算法将时空上下文跟踪算法作为跟踪器,使用Kalman滤波预测目标所在位置,减小在图像中特征点检测范围,将CMT中的特征点匹配算法作为检测器,使用融合器分析和评估跟踪器与检测器结果,决策出最终的结果,并执行有效的模型更新策略。
In this paper,an algorithm of object tracking based on CMT（Consensus-based Tracking and Matching of Keypointsfor Object Tracking）is proposed. The original CMT algorithm only has a fixed object model. To solve this problem,an improvedCMT algorithm which consist of tracker,detector and fusion is proposed. And spatio-temporal context（STC）tracking algorithm isintroduced as a tracker. Kalman filtering predicts object position for reducing the range of detection of feature points. The featurepoints matching algorithm of CMT is used as detector. Fusion analyses and evaluates the results of tracker and detector,and givesthe final result. At the same time,fusion runs an effective strategy of model update.
出处
《计算机与数字工程》
2017年第11期2143-2147,2168,共6页
Computer & Digital Engineering
