ABSTRACT Unmanned aerial vehicles (UAVs) have become indispensable in search and rescue (SAR) operations during emergencies such as natural disasters and maritime accidents, owing to their extended visual range and operational flexibility. However, UAV‐captured images of ground targets are often subject to false positives and false negatives due to suboptimal image quality and target occlusion. In target detection models, detection accuracy is primarily reflected by detection confidence. While improving detection algorithms remains a common approach, this paper emphasizes the significant impact of flight trajectory on detection performance. Specifically, it addresses the trade‐off between achieving a wide field of view at high altitudes and obtaining high‐resolution images at low altitudes. By analyzing image data collected through circular and oblique descending trajectories, we investigate the effects of flight altitude and viewing angles on detection confidence and establish a Gaussian‐based mathematical model to characterize its spatial distribution. To mitigate the probabilities of false negatives and false positives during detection, we propose an optimized flight trajectory planning method termed the V‐circular. This approach involves the UAV descending obliquely toward a potential target, conducting circular flights at the lowest altitude to capture high‐quality images from multiple viewing angles, and then ascending obliquely to rejoin its predefined trajectory. The V‐circular method not only enhances image quality and detection confidence but also enhances the stability of recognition performance. Simulation and experiment results demonstrate that, compared to conventional flight trajectories, the V‐circular significantly improves target detection accuracy and shortens SAR time. These advantages are particularly pronounced in missions demanding high precision. Consequently, the V‐circular presents an innovative solution for enhancing UAV rescue performance in complex environments.
He et al. (Sun,) studied this question.