Robust control point detection for aerial synthetic aperture radar via a logarithmic quasi-random scale space framework

In this paper, a novel multi-scale method for robustly detecting control points from aerial synthetic aperture radar (SAR) data is proposed. A logarithmic quasi-random scale space framework is designed to decompose aerial SAR data into different scales, and a second-order moment analysis is performed at each scale to isolate control point candidates. Finally, the final set of control points are determined via a local Hessian trace extrema analysis across all scales. Preliminary results using AIRSAR aerial SAR data demonstrate the effectiveness of the proposed approach for automatically identifying semantically important control points in aerial SAR data compared to existing control point detection approaches.