Generating Time-Series Crop Surface Models from Data Acquired by a UAV-Based Laser Scanner System

This study introduces a method for establishing a crop surface model using laser scanning technology from the integrated Velodyne VLP-16 scanner and DJI MATRICE M600 drone. The study also provides a process to create a 3D point cloud from the raw dataset. The point clouds are used to estimate plant height based on percentile analysis. The advantage of the proposed method is that the rice plant height is estimated without pre-determining the ground position. The 99th percentile rank is a suitable value to represent the ground position. The estimated plant height correlates with the directly measured plant height, specifically for areas within 100 to 200 of the field of view. The results are achieved with a coefficient of determination greater than 0.90 and an RMSE of less than 6.0 cm. Finally, time-series crop surface models are created. These models imply the change in rice plant height during the study period. The estimated growth rate is less than 1.6 cm/day, whereas the measured growth rate is less than 1.8 cm/day. This result confirms the ability to check the rice growth rate from the point cloud collected by the developed system.