Cost-Effective Solution for Fallen Tree Recognition Using YOLOX Object Detection

Abstract

Tropical cyclones are the world’s deadliest natural disasters, especially causing tree death by pulling out or breaking the roots of trees, which has a great impact on the forest ecosystem and forest owners. To minimize additional damage, an efficient approach is needed to quickly grasp information on the location and distribution of fallen trees. There are several studies that try to detect fallen trees in the past, but most of the research requires huge costs and is difficult to utilize. This research focuses on resolving those problems. Unmanned aerial vehicle (UAV) is widely used for ground detection for those who need a cost-effective way while pursuing high-resolution images. To take this advantage, this research collects data mainly using a UAV with an auxiliary high-resolution camera. The collected data is used for training the YOLOX model, an object detection algorithm, which can perform an accurate detection within a remarkably short time period. Also, by using YOLOX as a detection model, a wide-range versatility is obtained, which means, the solution driven by this research can be utilized for every scenario where inexpensive, but highly reliable object detection result is needed. This research implements a visualization application that displays detection results, calculated by a trained model, in a client-friendly way. Fallen trees are recognized in images or videos, and the analyzed results are provided as web-based visualizations.