Flue-cured tobacco yield prediction by fusion of UAV LiDAR point cloud, orthophoto and leaf area density

Flue-cured tobacco is a high-value cash crop, and accurate yield prediction is essential for cultivation planning, procurement, and industry management. This study proposes a field-level yield estimation method integrating UAV LiDAR point clouds, orthophotos, and leaf surface density. First, UAV LiDAR and orthophoto data were acquired over the study areas. Tobacco field boundaries were then extracted from the orthophotos using a pre-trained U-Net model. Next, the average leaf area index of each field was estimated from classified LiDAR point clouds using the Beer–Lambert law. Leaf surface density was determined through field sampling and flue-curing experiments. These variables were subsequently integrated to estimate tobacco yield. The method was evaluated in 30 fields across three experimental regions in Shandong Province. Predicted yields closely agreed with measured values, achieving an RMSE of 5.8000 kg/mu, rRMSE of 3.3163%, MAE of 5.2557 kg/mu, MRE of 3.0051%, and R2 of 0.9755. The results demonstrate that the proposed method enables rapid data acquisition, reduces sensitivity to illumination and topographic variation, and provides accurate and reliable tobacco yield estimates for operational applications.