Forested areas are among the most challenging environments for inventory, change detection, and forestry planning. Dense vegetation, limited accessibility, and reduced GPS performance beneath tree canopies make traditional surveying both time-consuming and labor-intensive.
LiDAR technology significantly simplifies fieldwork in these conditions. The DJI Zenmuse L1 provides a cost-effective airborne LiDAR solution that combines ease of field deployment with professional surveying capabilities, making laser scanning accessible to a wider range of users.
This project was completed through collaboration between DroneUA, Ukrhiprodor, and GEOinnovation+ to evaluate the effectiveness of LiDAR technology for forest inventory applications.
Project objective
- Compare airborne LiDAR results with conventional ground surveying and evaluate the potential of LiDAR for forest identification and inventory.
Equipment used
Aerial survey
- DJI Matrice 300 RTK with Zenmuse L1 LiDAR payload.
- DJI D-RTK Mobile Station 2.
Ground survey
- Leica GS16 GNSS receiver.
- Leica TS06 Plus total station.
Software
- DJI Terra.
- TerraSolid.
Survey area
- Area: 1.56 km².
- Terrain: Rugged landscape with ravines and dense forest vegetation.
Fieldwork schedule
- Drone survey: Approximately 1 hour, including control points.
- Ground survey: Approximately 2 days using a total station.
LiDAR flight parameters
Mission settings
- Flight altitude: 59 m.
- Mission type: Grid flight.
- Cross-track overlap: 50%.
Zenmuse L1 settings
- Scanning mode: Non-repetitive.
- Scanning frequency: 120 kHz.
- Returns: Up to 3 returns.
- Automatic RGB point cloud colorization enabled.
Data processing workflow
During each flight, the drone records nine datasets, including calibration information, LiDAR measurements, GNSS data, IMU data, RGB imagery for point cloud colorization, and base station information. These files are transferred to a workstation for processing.
Trajectory processing is performed in DJI Terra. For this project, processing required approximately 10 minutes, although actual processing time depends on dataset size and computer performance.
Compared with traditional photogrammetry, LiDAR produces a point cloud during acquisition, so software processing primarily refines the trajectory and spatial positioning rather than reconstructing geometry from images.
The point cloud can be processed using integrated RTK data or through PPK (Post-Processed Kinematic). PPK is recommended because it improves trajectory accuracy, smooths flight paths, and minimizes gaps between adjacent scan lines.
This project used RINEX observations from the System.NET base station network. The nearest reference station was located approximately 17 km from the survey area.
Processed point clouds can be exported in LAS, S3MB, and PLY formats for use in most point cloud processing applications.
Point cloud visualization in DJI Terra
RGB colorized point cloud
All points are displayed using true surface colors captured by the integrated RGB camera.
Height visualization
Points are displayed according to elevation using a color gradient from the lowest to the highest elevations.
Intensity visualization
Displays the intensity of laser reflections, which varies depending on surface material and reflectivity.
Multiple-return visualization
The Zenmuse L1 supports up to three laser returns, allowing measurements of both canopy surfaces and underlying terrain by recording multiple reflections from a single laser pulse.
Using DJI Terra's built-in measurement tools, users can determine tree height, canopy dimensions, and crop height. For larger datasets, automatic classification workflows significantly improve processing efficiency.
Point cloud classification
Point cloud classification was performed using TerraSolid by specialists from GEOinnovation+.
Classification assigns each point to a specific object class based on its geometric and spatial characteristics.
The workflow consisted of two stages:
- Automatic classification using advanced algorithms.
- Manual refinement to correct classification errors.
Five classes were created for this project:
- Ground.
- Vegetation.
- Buildings.
- Noise.
- Water.
The number of classes can be expanded or modified depending on project objectives.
Once classified, the point cloud can be used for automated analysis, terrain modeling, vegetation assessment, and other forestry applications.
You can explore the classified point cloud online here:
View the classified point cloud
