基于SIFT特征匹配的运动目标检测及跟踪方法

Moving objects detection and tracking based on SIFT feature matching

摄像机运动情况下的运动目标检测及跟踪是视频监控中的热点问题。论文提出一种基于SIFT(Scale Invariant Feature Transform)特征匹配的运动目标检测和跟踪算法。在目标检测阶段,首先提取两帧带检测图像的SIFT特征点并进行特征匹配,然后计算两帧图像之间的几何变换矩阵,从而实现图像的几何对齐。再将几何对齐后的两幅图像进行差分,并在差分图像中寻找SAD最大值区域作为运动目标区域。在目标跟踪阶段,将已检测到的目标作为跟踪样本,与后检测到的目标区域进行SIFT特征匹配,结合论文提出的跟踪样本集更新机制实现目标跟踪。论文目标检测和跟踪均基于SIFT特征匹配方法且无需背景建模过程,以适用于实时应用。

Moving objects detection and tracking with moving camera is a hot issue in video surveillance.This paper proposes the moving target detection and tracking algorithm based on SIFT(Scale Invariant Feature Transform) feature matching. In objects detection stage, firstly extracting SIFT feature points from the two frames to be detected and matching the feature points, and then calculating the geometric transformation matrix between the two images, so as to aligning the images. Then differing the two aligned images and searching for the region with maximum SAD value in the difference image as the moving target region. In objects tracking stage, considering the detected object as the tracking sample and matching SIFT feature points with those of the currently detected target region, and combining with the proposed tracking sample set update mechanism to realize objects tracking. In this paper, both objects detection and tracking are on the basis of SIFT feature matching and without background modeling, in order to be suitable for real-time applications.