异源景象匹配作为一种重要的辅助导航手段已经被广泛研究,但受到异源图像对之间非线性辐射失真和几何变形的影响,实现异源图像的匹配仍然是一项具有挑战性的任务。为了解决这些问题,提出了一种具备旋转和尺度不变性的异源景象匹配方法以同时估计异源图像对之间的旋转、尺度和位移变化。首先,基于图像局部的结构关系,利用局部自相似描述子进行特征描述以抵抗非线性辐射差异和局部变形的影响。再结合对数极坐标变换将图像整体的旋转和尺度变化正交展开并分别在笛卡尔坐标系上表示。最后,利用位移估计以及旋转和尺度估计的连续性,构造了一个五维特征描述符,并利用相位相关方法同时估计图像的旋转、尺度和位移变化量。在3种常见类型的异源图像匹配任务上进行的实验表明,文中的方法相较于当前其他先进的方法,在匹配正确率方面至少能提高4.5%,这突出了其在异源景象匹配领域的有效性。

Heterogeneous scene matching, as an important auxiliary navigation method, has been widely studied. However, due to the influence of nonlinear radiation distortion and geometric deformation between heterogeneous image pairs, achieving heterogeneous image matching remains a challenging task. To address these issues, a heterogeneous scene matching method with rotation and scale invariance is proposed to simultaneously estimate the rotation, scale, and displacement variations between heterogeneous image pairs. Firstly, based on the local structural relationships of the image, local self-similarity descriptors (LSS) are used for feature description to resist the influence of nonlinear radiation differences and local deformations. Combined with the logarithmic polar coordinate transformation, the overall rotation and scale changes of the image are orthogonally expanded and represented separately in the Cartesian coordinate system. Finally, by utilizing the continuity of displacement estimation, rotation, and scale estimation, a five-dimensional feature descriptor is constructed and using phase correlation method estimates the variation of image rotation, scale and displacement simultaneously. Experiments conducted on three common types of heterogeneous image matching tasks shows that the proposed method achieves a matching accuracy of at least 4.5% higher than existing state-of-the-art methods tech, that highlights its effectiveness in the field of heterogeneous scene matching.