一种基于尺度不变特征点的运动目标检测算法

A Moving Object Detection Algorithm Based on Scale Invariant Feature Transform Feature Point

摘要: 针对PTZ（Pan/Tiit/Zoom）摄像机的运动目标检测，提出了一种基于尺度不变（Scale Invariant Feature Transform，SIFT）特征点的运动补偿解决方案．首先，采用相邻两幅图像特征点匹配的方法获得运动补偿，从而解决了PTZ摄像机的自主运动带来的问题；然后利用时间差分法得到运动目标的区域；晟后利用形态学中的腐蚀和膨胀方法去除噪声．测试结果验证了算法的实用性和有效性．

Abstract: For moving object detection of Pan/Tilt/Zoom(PTZ)camera,a motion compensation approach based on scale invariant feature transform(SIFT)feature point is presented.First,by matching the feature points of two adjacent images to make motion compensation,the problem caused by adaptive moving of PTZ camera is solved.Then,the area of moving object is retrieved with temporal difference.Finally,noise is removed with corrosion and inflation in morphology. Experiment results show that the presented approach is effective and practical in moving object detection.