A Multi-Level Scene Matching Algorithm Based on Adaptive Polar Transform

Abstract

In the scene matching aided-navigation system, there often exists rotation, scale difference and occlusion between actual image and reference image. According to the navigation system demands of robustness, instantaneity and precision, a multi-level algorithm based on adaptive polar transform is proposed for its robustness to rotation, scale difference and occlusion. In the matching process, the rectangle window is replaced by a circle window to eliminate the cut error caused by rotation, which improves the matching precision. To accelerate the matching procedure, a pixel-jump searching method combined with a multi-resolution imaging technique is used to reduce the computational load. Experiments illustrate the effectiveness of the proposed methods.