Drone Navigation System for Autonomous Mosquito Sampling in Tree Canopies

Malaria, caused by a parasite infecting the female Anopheles mosquito, is a serious and potentially fatal disease common to tropical regions. Disease control programs rely on the capturing of mosquitoes from various vertical heights within tree canopies. To support such malaria control research efforts, the proposed solution aims to overcome the limitations of conventional methods that involve tree climbing and manual mosquito capturing. This paper presents the development of a novel drone navigation system designed to collect mosquito samples within tree canopies. Our solution builds upon existing technologies by implementing a 3D mapping algorithm using a stereo vision depth camera and object detection algorithm YOLOv7 to accurately identify perching sites in tree canopies. The developed drone navigation algorithm employs the obtained coordinates to plan suitable flight paths. We assessed the accuracy of the underlying pinhole camera model and conducted calibration of the depth camera to enhance depth accuracy. Additionally, we analyzed the YOLOv7 training configuration to minimize false positives in landing site detection. The results demonstrate the effectiveness of our solution in capturing mosquitoes at various vertical heights, providing valuable support for malaria control programs.