Missile-borne SAR image scene matching based on improved ORB algorithm

Abstract

Aiming at the application of SAR (Synthetic Aperture Radar) image scene matching technology in navigation guidance, an improved ORB (Oriented Fast and Rotated Brief) algorithm based on SURF (Speeded up Robust Feature) algorithm is proposed. Firstly, a multi-scale space is generated by using different sized box-shaped filter templates and integral images, and the stable extreme points with scale invariance are detected. Then the ORB descriptors are used to describe the feature points to get rotated binary descriptor and the Hanming distance is used to complete the initial match. Because of the mismatching, PROSAC algorithm is used to remove the mismatches of the initial matching points to improve the matching accuracy. This algorithm can effectively solve the defect that ORB does not have scale invariance and keep the rapidity of ORB algorithm and improve the matching accuracy. Experiment results verify the effectiveness and real-time of the proposed Improved ORB algorithm in SAR image matching.