Azimuth-Only NED Angle Estimation in Electronic Warfare: Optimization and Deep Segmentation Approaches

This paper addresses the problem of estimating both azimuth and elevation angles in the North-East-Down (NED) frame using only azimuth measurements obtained in the body frame of an airborne electronic warfare system. We propose two different approaches that exploit the maneuverability of the measurement platform to recover three dimensional angle of arrival information from limited input data. The first approach employs a non-convex optimization framework that directly utilizes time-varying azimuth measurements and platform orientation. The second approach reformulates the estimation problem as an image segmentation task, where azimuth measurements are transformed into 2D representations in which target angles correspond to curve intersections. Convolutional and transformer based deep segmentation models are then used to infer the target's spatial angles. Experimental results demonstrate that deep segmentation significantly outperforms the optimization based method in terms of estimation accuracy and robustness under varying noise and maneuver conditions.