Unmanned aerial vehicles (UAVs) with imaging sensors are increasingly used for aerial detection of ground objects in various civil, commercial or military applications. Object detection involves the automated localization and classification of objects (e. g. vehicles) within the sensor-acquired image data. Detection performance represents a quantitative measure for the statistical correctness of object detection. A high detection performance is therefore essential for reliable detection and a prerequisite for a successful reconnaissance. A high detection performance is demanding on the performance of the perception chain, which consists of the imaging sensor, sensor data processing and algorithms for object detection. Various environmental conditions (e. g. brightness, visibility conditions) can negatively affect sensor data acquisition and the subsequent processing chain, which can ultimately deteriorate detection performance. To counteract this, a concept for sensor-model-based flight planning is developed and evaluated in this thesis. The goal is to generate local or global optimal reference flight trajectories for the UAV in order to increase the resulting detection performance. For this purpose, a sensor performance model is developed that maps the detection performance of a perception chain with respect to specific environmental conditions. This performance model is used in sensor-model-based flight planning, in particular to model the influence of photogrammetric conditions on the detection performance. Thereby, suitable UAV positions with high detection performance can be identified, which are the basis for the generation of the reference trajectories. Trajectory optimization is performed using two mathematical optimization methods to generate either local or global optimal reference trajectories, while taking into account perceptual, platform-specific, and mission-specific requirements. Finally, the concept is evaluated based on simulated aerial reconnaissance scenarios for ground based vehicle detection. In direct comparison with benchmark trajectories, the local optimal reference trajectories of the sensor-model-based flight planning yield detection performances that exceed the performance values of the benchmark trajectories. Even higher detection performances can be achieved from the global optimal reference trajectories, which are close to the theoretical maximum value.    «

Unmanned aerial vehicles (UAVs) with imaging sensors are increasingly used for aerial detection of ground objects in various civil, commercial or military applications. Object detection involves the automated localization and classification of objects (e. g. vehicles) within the sensor-acquired image data. Detection performance represents a quantitative measure for the statistical correctness of object detection. A high detection performance is therefore essential for reliable detection and a pr...    »