Multi-Objective Detector and Tracker Parameter Optimization via NSGA-II

Modern tracking algorithms must engage a wide variety of targets. These targets vary in size, shape, intensity, and speed. While the targets change dependent upon application, oftentimes the tracking software remains predominantly constant. Rather, the tracking algorithm flexibility is achieved by user-defined parameters. Unfortunately even for experienced operators, these parameters may be difficult to tune resulting in suboptimal performance. This difficulty prompts the need for automated tuning software. To aid the operator in determining parameter values, this paper presents the novel application of non-dominated sort genetic algorithm II (NSGA-II) to determine optimal detector and tracker settings.