Open AccessEstimation of Rice Plant Height from a Low-Cost UAV- Based Lidar Point Clouds
Author(s) -
Anh Thu Thi Phan,
Kazuhiko Takahashi
Publication year - 2021
Publication title -
international journal of geoinformatics
Language(s) - English
Resource type - Journals
ISSN - 2673-0014
DOI - 10.52939/ijg.v17i2.1765
Subject(s) - lidar , mean squared error , point cloud , remote sensing , canopy , environmental science , percentile , laser scanning , geography , mathematics , statistics , computer science , laser , physics , archaeology , optics , computer vision
UAV systems are considered effective tools to collect information regarding crops. In this study, the rice growth was observed by a small UAV-based LiDAR system from above. For developing the system, DJI S800 was chosen as a platform on which a non- survey-grade laser scanner HOKUYO UTM30LX-EW was mounted. Field experiments were carried out from late June to late early August 2017 in Nagaoka city, Niigata Prefecture, Japan. Percentile analysis is applied to locate the top and bottom positions of rice plants in three targeted areas. LIDAR-derived plant height is computed by taking the difference between the bottom and the rice plant's top. As a result, the LiDAR-derived canopy height well correlates to rice plant height (R2≥0.86; RMSE <6.0 cm). The small root means square error (RMSE =4.9 cm) is achieved with area 3. In the general case, the RMSE is 5.5 cm (R2=0.88). These results illustrate the capability of estimate plant height before the heading stage from UAV- based LiDAR point clouds without ground surface detection.