Forest is one of the most challenging types of terrain for conducting inventory, tracking changes, and planning forestry activities. Obtaining information about the terrain under vegetation is labor-intensive in this area due to poor accessibility, thick undergrowth, and limitations of GPS technology in obtaining information between trees. The use of LIDAR technology significantly facilitates fieldwork in such areas, and the cost-effective solution offered by DJI LiDAR Zenmuse L1 makes this technology more accessible to users due to its ease of use in field conditions and availability for most users.
As a company, DroneUA is open to collaboration and serves as a hub for bringing together various specialists. It collaborates with leading companies in the industry to provide objective information about the use of technology. This project was developed in collaboration with colleagues from Ukrhiprodor and GEOinnovation+.
Objectives:
- Compare the results of laser scanning with the results of ground surveying and analyze the possibilities for identification and inventory of forest areas.
Equipment:
Aerial surveying:
- Matrice300 RTK drone with Zenmuse L1 LiDAR payload
- GNSS receiver D-RTK mobile 2
Ground surveying:
- GNSS receiver Leica GS16
- Total station Leica TS06 plus
Software:
- DJI Terra
- Terrasolid
Object description:
- Area: 1.56 square kilometers
- Terrain type: rugged with ravines and dense forest vegetation
Work schedule:
- Fieldwork using the drone: 1 hour (with control points)
- Ground surveying with the total station: 2 days
Parameters for aerial surveying and laser scanning:
Flight mission settings:
- Drone flight altitude: 59m
- Mission type: grid (flight along a grid pattern)
- Cross-track overlap: 50% (overlap between adjacent flight lines for laser scanning data)
Payload settings:
- Scanning mode: non-repetitive
- Scanning frequency: 120kHz
- Number of returns: up to 3 reflections
- Point cloud colorization: enabled* (*Point cloud colorization is performed automatically using the RGB camera integrated into the Zenmuse L1 payload)
Office work:
Basic processing:
During the operation, the drone generates a set of 9 files (calibration data from modules, accumulated data from LiDAR, antennas, IMU, photos for tinting, and a base station file). These files need to be saved on a PC.
Trajectory processing is carried out using the DJI Terra software, and the processing time for this amount of information is 10 minutes. It should be noted that the processing speed depends on the size of the object and the computer's power. Compared to photogrammetric data processing methods, we obtain a point cloud much faster since it is generated during LiDAR operation, and the software only adjusts its spatial position.
The point cloud can be processed based on Real-Time Kinematic (RTK) data, which are already integrated into the files, or alternatively, Post-Processing Kinematic (PPK) can be performed. To improve the positioning of the point cloud, smooth the equipment's trajectory, and avoid "gaps" between adjacent data rows, we always recommend performing PPK.
For processing, we used RINEX data from the System.NET network of base stations, and the distance to the base station used was 17 km. To improve the accuracy of trajectory adjustment, it is desirable to place the base station closer to the survey area to increase accuracy.
As a result of the processing, we obtain a point cloud in LAS, S3MB, PLY formats, which are compatible with most point cloud classification or analysis software.
DJI Terra provides the following visualization options:
- Colorized point cloud (RGB): All points are displayed in surface color.
- Height gradation: Points are colored according to their height on a scale from lowest to highest.
- Intensity: Reflection intensity of the laser beam from the surface, depending on the surface type and color.
- Reflectance: The LiDAR used supports up to 3 returns, as the laser beam can penetrate some surfaces, allowing us to gather data about both the surface and objects on it (e.g., leaves and the ground).
At this stage, using the available measurement tools, the height of each tree, beam size, or even crop height can be measured. However, if we are exploring a large area, working with automatic classification algorithms is preferable, as it significantly improves efficiency.
Point Cloud Classification:
Point cloud classification was performed using the TerraSolid software by our colleagues from GEOinnovation+. Point cloud classification is the process of assigning points to different classes based on specified parameters and the type of object they represent.
Classification was performed in two stages:
- Automatic classification (using state-of-the-art algorithms)
- Manual classification (correcting errors from automatic classification)
For this example, we chose the following 5 classes:
- Ground surface
- Vegetation
- Buildings
- Noise
- Water
The number of classes (object definitions) may vary depending on the objective.
With this information, parameter analysis can be conducted, and even this process can be automated using specific algorithms.
You can view the classified point cloud online at the following link:
Link to classified point cloud
