ROI加速的无人机自主精准降落系统

UAV autonomous precise landing system based on ROI acceleration

GPS信号在有遮挡的环境中信号差且定位精度低,无法满足无人机在特定区域精准定位的要求。而计算机视觉定位相较于GPS,受环境影响小且定位信息丰富。基于此,文中设计一种ROI加速的无人机自主精准降落系统,采用“H”与Aruco marker嵌套的降落标识在不同高度给无人机提供定位信息。基于深度学习的目标检测算法分割相机视野中的感兴趣区域(ROI),利用阈值分割的图像识别算法在ROI区域进行Aruco marker检测,滤除ROI区域外的图像区域,提升目标识别速度。根据marker二维像素点与三维坐标点的对应关系求解无人机与marker的相对位置。基于求解出的三轴位置设计位置PID控制器,将无人机与marker的三轴位置差转换为速度控制量。进行无人机自主降落的试验,随机统计20次的降落结果,得出18次的降落轴向误差在10 cm以内,说明所设计系统能够满足无人机降落精度要求。

The GPS signal has poor signal and low positioning accuracy in the occluded environment,which cannot meet the requirements of precise positioning of UAVs in specific areas.In comparison with GPS,computer vision positioning is less affected by the environment and has richer positioning information.On this basis,an ROI⁃accelerated autonomous precision landing system for UAVs is designed,which can use the landing signs embedded with"H"and Aruco marker to provide positioning information for UAVs at different heights.The object detection algorithm based on depth learning is used to segment the region of interest(ROI)in the camera′s field of vision,and the image recognition algorithm based on threshold segmentation is used to carry out Aruco marker detection in the ROI area,filter out the image area outside the ROI area,and improve the target recognition speed.The relative position of the UAV and the marker is calculated according to the corresponding relationship between the two⁃dimensional pixel points and the three⁃dimensional coordinate points of the marker.The position PID controller is designed according to the calculated three⁃axis position,and the three⁃axis position difference between the UAV and the marker is converted into the speed control quantity.The experiment of UAV autonomous landing is carried out,the results of 20 times landing are statistically analyzed,and axial error of 18 times landing is less than 10 cm,indicating that the designed system can meet the requirements of landing accuracy of UAV.