融合全景SLAM与目标识别的无人机自主飞行技术

Autonomous drone flight technology integrating panoramic SLAM and object recognition

无人机自主飞行可以使无人机更好地应对复杂的环境和任务,降低操作成本和飞行风险,对环境做出及时响应,实现更高效、安全、广泛的应用,在民用和军事领域都能发挥重要作用,为多个领域赋能。全景相机具有360°的全方位视角,获取的环境信息更加完善,相较于单目相机更有优势。为了实现无人机的自主飞行,基于全景相机进行同时定位与建图,利用SPHORB算子进行特征提取和匹配,从而获取无人机的位姿数据的同时对周围环境建图,获取环境信息。利用YOLOv5算法进行目标识别,获取目标地物在相机坐标系中位置,为无人机自主靠近/远离目标提供信息基础,使无人机具备更加强大的环境感知能力,有利于无人机避障飞行,更好地规划飞行路径,实现自主飞行。

Autonomous drone flight enables drones to better respond to complex environments and tasks,reducing operational costs and flight risks,responding promptly to environmental changes,and achieving more efficient,safe,and widespread applications.It plays a significant role in both civil and military domains,empowering various sectors.Panoramic cameras provide a 360-degree omnidirectional view,yielding more comprehensive environmental information compared to monocular cameras.To achieve autonomous drone flight,Simultaneous Localization and Mapping(SLAM)is performed using a panoramic camera,and the SPHORB operator is used for feature extraction and matching to obtain the drone's pose data and map the surroundings.The YOLOv5 algorithm is employed for object recognition,determining the positions of target objects in the camera coordinate system.This information serves as the basis for autonomous drone approaches and departures from targets,enhancing the drone's environmental awareness and aiding obstacle avoidance,as well as optimizing flight paths.