Volumetric Vegetation Monitoring from LiDAR Scans with Ground Estimation

Abstract

The full automation of the agricultural industry depends on an automated system for monitoring plant growth. The hyperparameters of environment control could be tuned for ideal plant growth using the data from the monitoring system. Several monitoring systems that identify the current condition of plants using computer vision techniques have been produced in an effort to accomplish this goal. However, different from the environmental factors that can be directly measured from specific sensors, such as humidity and temperature, variables such as growth rate or vegetation volume are not directly measured. The vegetation growth monitoring could be solved by post-processing thousands of plantation images and structure-from-motion, however, it is time-consuming and requires an exact measurement of camera poses. Therefore, we propose a LiDAR-based volumetric plant monitoring, which distinguish between ground and vegetation point cloud, and analyze the growth of plantation over time, in place of time-consuming process with a vision sensor.