Implementation of a Monocular ORB SLAM for an Indoor Agricultural Drone

Drones are increasingly being used in almost every major industry, including agriculture. Intelligent drone systems could enable precise agricultural. One of the critical uses of agricultural drone is to use in an automatic plant monitoring and inspecting. A drone must be tiny enough to fly between plants in order to capture images of plant trees or fruits in an indoor environment. Therefore, drone's payload is crucial because it limited onboard sensors weight. SLAM is necessary for autonomous navigation because it could provide all necessary information of drone navigation system without collisions. ORB SLAM was popular for monocular systems since it extracted ORB features from images to generate Visual Odometry. ORB SLAM generated the map as the output that can be used to estimate the position of the drone without the assistance of other sensors. In this research, monocular ORB SLAM system was proposed, explained and experimented in order to obtain and confirm the ORB SLAM performance for an indoor application. The experimental result showed that the ORB SLAM worked properly for generating a map of a tomato greenhouse and the output map could be used to estimate the drone position with some limitation.