Generating Efficient Sensor Paths for Ground Target Search: A Performance Parameter and GIS-Driven Approach

In this work, we present a novel process to plan dynamically occurring search tasks for single, ground-based targets from an airborne platform equipped with a zoomable imaging sensor. To describe the underlying motivation, we first outline our use case in aerial multi-target tracking before presenting our approach to derive search plans. We achieve the goal of reducing the time to search by first generating a prior distribution of the target’s likely location based on background knowledge about the search area. This focuses search effort on parts within the area that have a higher likelihood of containing the target. To derive a mission and cell-specific detection function for the discretized search area, we further use sensor and platform specific parameters as well as the relative positions of target and platform. Additionally, we outline how perception performance models previously developed at our institute can be used to improve the results. We consecutively construct an optimal search problem from the prior distribution and detection function. The problem can be solved with well-established methods to find optimal search plans. Finally, we introduce how the resulting search density can be used to calculate a sensor path and show the process in an example application. We round out the paper with our conclusions and ideas for further improvement of the algorithm.