基于ASIFT算法特征匹配的研究

Research of feature matching based on ASIFT algorithm

针对SIFT算法对大角度视角变化下特征提取鲁棒性不强的弱点,引入了一种完全仿射不变的图像特征匹配算法—ASIFT。ASIFT算法不仅继承了SIFT算法的尺度、旋转和平移的不变性,并且在此基础上增加了两个空间特征描述参数:经度和纬度,从而定义出度量仿射形变的两个参量绝对倾斜t(absolute tilt)和过渡倾斜τ(transition tilt),模拟相机光轴变化,实现完全仿射不变。一种双分辨率(two-resolution)加速方法的提出,使ASIFT算法的复杂度约为SIFT的2倍。

SIFT algorithm for large angle viewing angle changes robust feature extraction is not strong,aiming at the weakness of SIFT algorithm,this paper introduced a completely affine invariant image feature matching algorithm,which named ASIFT. ASIFT algorithm not only inherits the scale,rotation and translation invariance of the SIFT algorithm,on this basis,two space characterization parameters,which are longitude and latitude are added. Thereby two parameters absolute transition t（ absolute tilt） and tilt transition τ（ transition tilt） are defined for measuring affine deformation. By simulating the change of the optical axis of the camera to achieve completely affine invariant. Two- resolution（ two-resolution） to accelerate the proposed method,and make the complexity of ASIFT algorithm is about 2 times of SIFT.