This research proposes tree detection and location methods using RGB-D data. The first proposed approach uses a simultaneous localization and mapping (SLAM) algorithm based on RGB-D image data to build a dense point cloud map. Through reducing the dimension of the point cloud map from 3D to 2D via a slicing method, and the Euler clustering algorithm, the system locates the approximate position of the trees around the camera within a certain range. Finally, when an approximate tree position is within the range of the depth camera, the system uses a merged depth map to detect and adjust the exact location of this tree in real time.

The second approach proposes an autonomous navigation algorithm to control an unmanned aerial vehicle (UAV) using a novel tree detection and navigation process. The navigation system controls the UAV to take off, rotate 360o to scan the surrounding scenes and navigate to the nearest tree by providing instructions from ROS to the PX4 flight controller. Kalman filtering improves the robustness and fault tolerance capability of the navigation by adjusting the relative position of the detected tree with respect to the camera.

From our 20 experiments, the proposed method has 100% correct tree detection rate in single tree scenes, 90.9% correct tree detection rate in multiple tree scenes with trees close together and 2.5 times faster calculation speed than prior research which only achieved an accuracy of 66%-89%.