Workflow for DJI Terra LiDAR terrain classification


The new firmware update allows DJI Terra to perform LiDAR point cloud terrain classification. In-depth guide to the terrain classification workflow, using sample data collected with the DJI Zenmuse L1.

A new firmware update allows DJI Terra to do terrain classification with LiDAR data.

In this way it is possible to classify the terrain/land so that it can be loaded into third party software for generating DEMs. Read our DEM vs DTM vs DSM blog for more details on this type of model.

Terrain classification in a point cloud is essential for creating a robust DEM, as it helps provide an accurate representation of the terrain and removes non-terrestrial features.

DJI Terra: LiDAR Terrain Classification – Workflow
In this example, point cloud data was acquired with the DJI Zenmuse L1 LiDAR sensor, which can be used with the DJI M300 series.

Thanks to the firmware update, the Point Cloud Processing section now contains Ground Point Type, as shown in the image, right center.

Enable Ground Point Type, which has four parameter settings: Ground Type; Maximum diagonal of the building; iteration angle; and iteration distance.

Terrain type

Choose between flat terrain, gentle slope or steep slope.

Flat terrain: suitable for areas with dense or flat buildings.
Gentle slope: suitable for areas such as common mountains and hills.
Steep slope: Suitable for areas with large elevation changes, such as high mountains or river valleys.
By changing the terrain type, Terra should also adjust the following iteration angle and iteration distance settings.

Building Max Diagonal

Maximum diagonal length of a building in top view. The default value is 20 m and the value range must be between 1 and 1,000 m. It can be measured using the Earth measuring tool. This helps Terra better identify the man-made structure and speeds up the classification process.

Iteration angle and iteration distance

Normally, it is not necessary to change the iteration angle and iteration distance settings, but adjusting them will help you fine-tune the terrain classification result and give you more freedom on terrain classification.

Iteration angle: parameter used to determine the ground point. When the angle between the point and the triangular surface falls within the configured value, the point is considered as a ground point. The default for Flat Terrain is 3°, Flat Slope is 6°, and Steep Slope is 10°. The range of values must be between 0.1° and 60°.
Iteration Distance: parameter used to determine the ground point. When the distance between the point and the triangular surface is within the configured value, the point is treated as a ground point. The default value for Flat Terrain is 0.3m, Flat Slope is 0.5m, and Steep Slope is 0.7m. The range of values should be between 0.01 and 10 m.
After configuring the Ground Point Type settings, press Start Processing to process the LiDAR point cloud data.

Click the Type button at the bottom of the screen, which changes the point cloud in this view, showing points that have been classified as terrain (yellow shading) versus unclassified points.

In the upper right corner, choose the Filter icon and hide or show certain categorized spots.

So this graph below shows both points on the ground and points that have not been classified, as displayed based on reflectivity (provides insights into the characteristics of the scanned objects or surfaces)…

…whereas this image shows only the terrain points, as viewed via the reflectivity setting.

Next, select the relevant LAS files which contain the classified ground point level…

…and import into third-party DEM-ready software.