SIFT特征匹配和差分相乘融合的运动目标检测  

Moving object detection by combining SIFT and differential multiplication

作　　者：王梅[1] 屠大维[1] 周许超[1] 

机构地区：[1]上海大学机电工程与自动化学院,上海200072

出　　处：《光学精密工程》2011年第4期892-899,共8页Optics and Precision Engineering

基　　金：国家自然科学基金资助项目(No.60774102;No.51075252);上海大学创新基金资助项目(No.A10-0109-09-015)

摘　　要：针对运动目标检测的难点问题,提出了一种结合尺寸不变特征变换(SIFT)和差分相乘算法的运动目标检测方法。首先,用SIFT特征匹配算法配准运动图像的旋转、缩放和平移量,利用SIFT匹配的稳定性和准确性,精确补偿运动摄像机下的背景图像。然后,用差分相乘方法,准确分割出运动目标的轮廓。最后,通过实拍视频序列的试验,证明算法的有效性和可行性。系列实验显示,连续4帧图像差分相乘的方法即能够较好地满足应用要求。实验结果表明,SIFT特征匹配和差分相乘融合的方法具有较好的鲁棒性和抗噪能力,对于摄像机运动、亮度变化、遮挡等影响因素具有较强的适应能力。For the difficulty of moving object detection by a moved camera,a method for detecting dynamic background caused by camera motion was proposed by combining a Scale Invariant Feature Transform（SIFT） and a differential multiplication.Firstly,the image registration based on SIFT features was applied to calculate transformation parameters,including translation factors,rotation angles and scaling coefficients,and to provide a robust matching for realizing the motion compensation precisely.Then,the object segmentation based on the differential multiplication was introduced to detect moving objects accurately.Finally,the experiments on the video sequences were performed to verify the robustness and validity of the method.Experiments show that instead of 6 or more frames,4 frames are enough for the differential multiplication in practical applications.The detection method is robust and denoising in dynamic scenes and has good adaptability to changing illumination,occlusion and camera movement.

关 键 词：尺度不变特征变换算法 差分相乘 动态背景 运动目标检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]