水平単眼カメラSLAMによる障害物計測の下方カメラ画像を用いた精度改善法

抄録

UAV has been attracting attention for the purpose of collecting information in disaster area. We are required to create a fast and accurate 3D map using the UAV. During autonomous flight, we create a 3D map and estimate position of the vehicle using SLAM (Simultaneous Localization and Mapping) because it is impossible to use the GPS in indoor environment. It is possible to navigate while using obstacle position by 3D SLAM. However, in a typical UAV configuration, since it is mainly observed by the front camera, detection of an object directly under the UAV is not always fast and accurate. We propose a method to detect obstacles under the UAV quickly by matching the information of the bottom cameras with the point cloud information obtained by SLAM. This aims for fast and robust navigation with cheap and lightweight equipment.