Auto Tree Counting for Oil Palm Plantation Based on Uav/Drone Orthophoto

Accurate tree counting in plantations is essential for effective management and resource allocation, yet traditional manual digitization methods are labor-intensive and time-consuming. This study presents an innovative approach to automate tree counting in oil palm plantations using UAV/drone orthophoto imagery and the Faster R- CNN algorithm implemented in Python. High-resolution images were captured using UAVs and pre-processed for clarity. The Faster R-CNN model was trained on annotated images with manually labeled bounding boxes. Leveraging a two-stage pipeline, the model generates region proposals and classifies objects within them. Performance evaluation using precision, recall, and Intersection over Union (IoU) metrics demonstrated the model's high accuracy, achieving an average precision (AP) of 75% and an average recall (AR) of 44%. The automated process significantly reduces labor costs and time compared to traditional methods. This study highlights the effectiveness of integrating UAV technology with machine learning for agricultural applications, providing a scalable solution adaptable to various plantations and tree species.