Reverse Classification
Classify point cloud data using Digital Grid infrastructure vectors
Reverse Classification requires an Aura Pro subscription and Aura Desktop. It is not available in the web app.
Reverse Classification automatically classifies points in a point cloud based on their proximity to your Digital Grid infrastructure. Instead of manually classifying points, you use the poles, wires, and crossarms you've already modeled to tag nearby points with custom LAS classification codes. The output is a new COPC file with classified points ready for downstream workflows.
How It Works
The classification runs in two phases:
-
Pole classification — For each pole, the algorithm finds nearby points using spatial queries, clusters them with DBSCAN, and marks entire clusters that contain points close to the pole geometry. Ground-level points are filtered out to avoid false positives.
-
Wire and crossarm classification — Every point in the cloud is checked against all wire and crossarm segments. Points within the distance threshold of a segment are assigned that segment's classification code.
Priority order: wires and crossarms take precedence over poles when a point falls near both.
Prerequisites
Before running Reverse Classification you need:
- An Aura Pro subscription
- Aura Desktop installed (download)
- A COPC point cloud loaded in your project
- A Digital Grid model with poles, wires, and/or crossarms built on that point cloud
Running Reverse Classification
Open the Tool
Select Digital Grid → Reverse Classification from the tools menu.
Select a Point Cloud
In the settings panel, choose a point cloud layer from the Point Cloud Layer dropdown. Only layers loaded as local COPC files are available.
Configure Features
Each infrastructure type has a checkbox toggle and a class code input:
| Feature | Default Class Code |
|---|---|
| Poles | 150 |
| Primary Wires | 151 |
| Secondary Wires | 155 |
| Telecom Wires | 156 |
| Underbuilt Wires | 157 |
| Guy Wires | 154 |
| Service Wires | 153 |
| Crossarms | 152 |
Uncheck any feature types you don't want to classify. Class codes must be in the range 65–255 (LAS user-definable range).
Set Output Filename
Enter a name for the output file. The .copc.laz extension is added automatically. The output file is saved in the same directory as the input point cloud.
Run Classification
Click Start Classification. A progress bar shows the current status as the algorithm processes points. Classification can be cancelled if needed.
Distance Thresholds
Points are classified when they fall within these distances of the corresponding infrastructure geometry:
| Feature | Threshold |
|---|---|
| Poles | 0.2 m |
| Primary / Secondary / Telecom / Underbuilt / Service Wires | 0.3 m |
| Guy Wires | 0.2 m |
| Crossarms | 0.3 m |
For pole classification, DBSCAN clustering uses an epsilon of 0.2 m and a minimum of 3 samples. Points below 1 m above local ground are ignored to filter out ground clutter.
Output
The result is a new .copc.laz file containing all the original points with updated classification values. Points that match infrastructure geometry are assigned the configured class codes. All other points retain their original classification.
You can load the output file in Aura and color by classification to visually verify the results.
Tips
- Build a complete model first. The more infrastructure you've modeled (poles, wires, crossarms), the more comprehensive the classification.
- Review results visually. After classification, load the output in Aura and switch to the classification color mode to inspect the tagged points.
- Use custom class codes to integrate with your existing classification scheme. The defaults (150–157) are in the LAS user-definable range and won't conflict with standard ASPRS classes.
- Coordinate systems are handled automatically. The algorithm reprojects your Digital Grid coordinates (WGS84) to match the point cloud's native CRS.
Troubleshooting
No points are classified
- Verify your Digital Grid model overlaps with the point cloud spatially
- Check that infrastructure features are toggled on in the settings
- Ensure the point cloud has a valid coordinate reference system (CRS) — the algorithm needs this to align coordinates
Points appear in the wrong location
The input point cloud may have an incorrect or missing CRS. Use the CRS override feature to assign the correct EPSG code before running classification.
Classification takes a long time
Processing time scales with the number of points in the cloud and the number of infrastructure segments. For very large datasets, this is expected. Progress is reported every ~50,000 points.