基于模型和区域相关融合的鲁棒跟踪算法

提出了一种模型部分匹配的相似度量方法,利用匹配的完整度、归一化的匹配点相对于模型的偏移量期望和方差的加权和作为相似性度量。在此基础上建立模型匹配算法,利用模型匹配和区域相关融合的方法进行目标跟踪。算法能够在摄像机静止和运动的情况下进行稳定的跟踪,能够适应目标的部分遮挡、光照和姿态的变化。利用参数化的模型能够在复杂的背景环境中实现实时的多目标跟踪。

Robust tracking algorithm based on model matching and region correlation fusion

In this paper,a similarity measurement of model partial matching is presented.The similarity measurement is based on the weighted sum of integrity of matching,normalized offset expectation and variance of matched point.A model matching algorithm and robust tracking algorithm based on model matching and region correlation fusion are presented.This method provides reliable object tracking on the situation of stationary camera and moving camera.Results show the method to be robust to occlusion,changes in illumination and object pose changes.By using parameterized model,the algorithm can provide real-time multi-object tracking in complex scenes.