A Review: Obstacle Perception Based on Panoramic Vision in Low-altitude Rotorcraft UAVs

In recent years, Low-altitude Rotorcraft Unmanned Aerial Vehicles (UAVs) have shown great potential in many tasks, such as aerial mapping, search and rescue, plant protection, and more. In unknown environments, UAVs face the risk of collisions with various unforeseen obstacles. For UAVs, the ability to effectively perceive and recognize obstacles from all directions is crucial for ensuring safe flight, and a variety of sensors have been employed for this purpose. Among these, panoramic vision-based obstacle perception systems offer significant advantages. This paper focuses on key technologies for obstacle detection based on panoramic vision in UAVs. It reviews the current state of research and applications both domestically and internationally, covering aspects such as hardware platforms, key algorithms (including model correction, distortion correction, feature matching, image stitching), panoramic imaging, and omnidirectional obstacle detection. The paper also identifies major bottlenecks and outlines future development directions. Based on this work, we will continue our research on autonomous omnidirectional obstacle detection and avoidance.