Optimizing drone-based pollination method by using efficient target detection and path planning for complex durian orchards

Abstract

Abstract Durian is a valuable tropical fruit whose pollination heavily relies on bats and nocturnal insects. However, environmental degradation and pesticide usage have reduced insect populations, leading to inefficient natural pollination. This study proposes an AI-powered drone-based pollination method for complex durian orchards, integrating improved object detection and optimized path planning. Specifically, we enhance the YOLO-v8 algorithm using the GhostNet lightweight network to reduce computational complexity while boosting detection precision. For path planning, we develop an Enhanced TSP (EN-TSP) algorithm based on a branch and bound strategy with least-cost optimization. Experimental results demonstrate that the proposed method improves detection accuracy by 5.85% and path efficiency by 26.89% compared to baseline algorithms. The novel use of GhostNet with YOLO-v8 enables superior detection of durian flowers under low-light and occluded conditions, while EN-TSP ensures globally optimal drone routes, reducing travel distance and improving operational reliability. This integrated solution advances smart agriculture by enabling scalable, efficient, and precise pollination, reducing labor costs and increasing durian yield and quality.