Yurika TANIMOTO

A Study on Vegetation Coverage Measurement Using Point Cloud Data Obtained from Drone LiDAR

Kazuyoshi TAKAHASHI

This study examines a method for estimating vegetation coverage in agriculture using drone-mounted LiDAR technology.  Japan's agricultural sector faces severe labor shortages and an aging workforce. The outflow of people to urban areas and declining agricultural profitability have exacerbated this issue, making it difficult to maintain crop productivity and quality. To address these challenges, the adoption of smart agriculture is gaining attention, and remote sensing technology is being utilized to monitor crop growth.
Vegetation coverage is a key indicator of crop growth and is typically calculated using binary image processing. However, conventional image analysis is affected by weather and sunlight, leading to accuracy issues. This study proposes a new Canopy Layer Index (CL) using LiDAR point cloud data to estimate vegetation coverage more accurately.
The CL is determined by dividing the point cloud into vertical layers and identifying the thickness of the layer corresponding to the observed vegetation cover. The relationship between vegetation cover and CL obtained from the images was analyzed using linear, polynomial, and log regression. Linear regression showed high errors in the early and late stages, while polynomial regression showed unrealistic peaks. Log regression had the best fit, with errors ranging from 9.3% to 12.1%, improving from 7.1% to 7.8% when data from the early stages were excluded.
The low initial grass height was attributed to the growth stage of the rice plants. From germination to ear emergence, grass vigor is weak, so grass height tends to be low immediately after rice planting. On the other hand, as the harvest season approaches, grass height becomes constant, so plant coverage on subsequent measurement dates also tends to deviate from the estimated curve.
This study confirmed that estimation using LiDAR is effective and that accuracy can be improved by excluding data from the early stages. Future studies should improve the model to take growth stages into account .