Azimuth scaling for inverse synthetic aperture radar images with feature registration

摘要

This paper proposes a novel azimuth scaling algorithm for estimating the relative rotation angle of two continual say sub-aperture inverse synthetic aperture radar (ISAR) images of a rigid space ISAR target by using a feature registration of coordinate locations of interested points extracted from the images above. Specifically, we firstly extract sufficient interested points and feature descriptor vectors from two sub-aperture ISAR images by Scale Invariant Feature Transform (SIFT) and Speeded-Up Robust Features (SURF). The interested points are then matched by using the feature registration in a two-stage manner based on Euclid-Distance and Random Sample Consensus (RANSAC). A premise to the azimuth scaling for ISAR images, the rotation angle can be estimated by precisely linking the coordinate locations of the matched interested points, followed by a determination of the least value of a cost function to achieve the azimuth scaling for measuring the real size of the target. Simulated and real data experiments validate the proposal.