Radar target altitude measurement - Evaluation of different optimization algorithms
Fabian Lundbäck, Victor Wareborn · 2013
When a flying target is detected by airborne radar it is common that specu-lar reflections occurs from smooth surfaces of the Earth, such as oceans and lakes. In the radar input signal, these reflective echoes are considered as pulses containing information about the detected targets altitude above the ground. In this project a curve fit to this input signal is applied, aligned with a Least-Squares Estimation, LSE. As a number of curve fits have been carried out, it is desired to find the minimum LSE-loss function value, i.e. the signal fit that gave the least-squares error. When this minimum is found, it is possible to determine the targets altitude. Because of a heavy computational burden some simplifications are made, meanwhile optimization methods, such as Gauss Newton with line search, GN, Steepest Descent with line search, SD, and Nu-merical Neighborhood search, NN, are applied to find local minima in a discrete search space composed by the LSE-loss function values. The results have shown that the method can fit curves for severe cases when pulses begin to interact,