基于自适应阈值距离分段坐标系的形状上下文图像配准算法

摘要

Global structures are the relative invariant information of the remote sensing images before and after disasters. Feature points of these images locate in the global shapes. Shape Context adapts polar coordinate system. Discriminative property of this coordinate remain high in smaller distances while low in mid-long ones. This paper proposes a relative Fourier Shape Context based on self-adaptive threshold distance coordinate system (SAT-RFSC).SAT-RFSC uses distances between feature points to calculate self-adaptive threshold to get the distance coordinate system of the feature descriptor, thus significantly improve the discriminative property in mid-long distances. Relative Fourier shape context (RFSC) achieves rotation invariance. The affine transform matrix between reference and sensed images is evaluated by RANSAC method to achieve image registration. Experimental results show that the proposed algorithm has scale invariance and rotation invariance to a large range of angle changes, which has higher registration accuracy in comparison with original SIFT and traditional Relative Shape Context (RSC), thus can successfully accomplish the registration task for disaster remote sensing images.%灾害前后。感图像全局结构相对稳定不变，特征点遍布在全局形状中，形状上下文采用对数极坐标系，对小距离变化区分度高而中长距离区分度低，本文提出了一种基于自适应阈值距离分段坐标系的傅里《相对形状上下(SAT-RFSC)，SAT-RFSC使用特征点的距离，计算描述算子的自适应阈值(SAT)得到坐标系，这种自适应距离分段坐标系提高了特征描述符中长距离的区分度。相对傅里《形状上下文(RFSC)实现了算法的旋转不变性。使用RANSAC方法计算仿射变换矩阵实现配准。实验结果证明：本文提出的SAT-RFSC能实现大范围旋转角度变化图像配准。配准精度显著高于传统SIFT算法，和对数极坐标系和对数坐标系划分法相比，SAT具有更高的距离区分度。