Improving the quality of LiDAR point cloud data for monitoring greenhouse environments

Authors: SI, GAOSHOUTONG - The Ohio State University; LING, PETER - The Ohio State University; KHANAL, SAMI - The Ohio State University; Zhu, Heping

Submitted to: Agronomy
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/8/2025
Publication Date: 9/16/2025
Citation: Si, G., Ling, P., Khanal, S., Zhu, H. 2025. Improving the quality of LiDAR point cloud data for monitoring greenhouse environments. Agronomy. 15. Article 2200. https://doi.org/10.3390/agronomy15092200.
DOI: https://doi.org/10.3390/agronomy15092200

Interpretive Summary: Substantial labor inputs are required for efficient monitoring of plant growth, nutrients, water, and other stress factors in greenhouse environments. This makes automation crucial for precise and economically viable plant production management. The emergence of small unmanned aerial systems (sUAS) presents a cost-effective opportunity for high-resolution plant health monitoring. In this research, a new method called data registration pipeline was developed to significantly improve the cloud data collection for 3-dimentional plant structures using a laser scanning sensor mounted on a sUAS in a greenhouse setting. The influence of varying orientations and altitudes of the laser sensor during flights on the plant detection accuracy was minimized by instantaneous corrections with the floor boundary registration algorithm. The system accuracy was validated with volume estimation of reference objects. Test results demonstrated that this advancement had great potential to enhance the understanding of the constraints and opportunities associated with the use of laser sensor-equipped sUAS for automatic plant monitoring tasks such as assessment of crop growth rate and health condition.

Technical Abstract: Automated crop monitoring, especially in controlled environments where high-value crops are grown, is imperative for enhancing future productivity. The availability of small unmanned aerial systems (sUAS) and cost-effective LiDAR sensors presents an opportunity to conveniently gather high-quality data for crop monitoring. LiDAR collected point cloud data, however, often encounters challenges such as occlusions and low point density that can be addressed by acquiring additional data from multiple flight paths. We evaluated the performance of using the Iterative Closest Point (ICP) algorithm for registering sUAS-based LiDAR point clouds collected in greenhouse environments. A two-step registration process was developed: first, the ground floor boundary was leveraged as a critical local subset for improving the initial position between target and source point clouds, and then the ICP algorithm was implemented globally to achieve a fine registration. The evaluation of point cloud registration performance included various metrics, including visualization, Root Mean Square Error (RMSE), volume estimation of reference objects, and the distribution of point cloud density. Results revealed that the point cloud registration performance was influenced by several factors such as overlap ratio, quality of the registration feature, and geometric distortion of point clouds. Both RMSE values and point cloud density showed improvement post-registration compared to single-view point clouds. The accuracy of volume estimation for reference objects significantly improved, exemplified by the basketball-on-pot, where the estimation error dropped from 72% to 6% post-registration. This study presents a promising approach to point cloud registration, eliminating the need for artificial reference objects, in a production greenhouse environment.