Stone Pine (Pinus Pinea L.) Detection from High-Resolution UAV Imagery Using Deep Learning Model

Abstract

The automatic detection and counting of trees are crucial for yield estimations, agricultural planning and forestry applications. Unmanned aerial vehicles (UAVs) and computer vision are increasingly used together for automatic image analysis since they can overcome time-consuming applications in a more precise, effective, and less error-prone manner. In this study, YOLOv4, a state-of-the-art deep learning-based object detection method, was employed for the extraction of stone pine trees from the images obtained from unmanned aerial vehicle (UAV). The YOLOv4 model was trained and validated on the images of stone pine in the created training and validation datasets. The trained model was used for stone pine detection in test images taken from the study area. The obtained results indicate that the YOLOv4 model was able to detect stone pine trees with mean average precision (mAP) of 98.96% and intersection over union (IoU) of 81.31%.