Advanced AI Classification – Ground+, Forestry, and Utilities

Advanced AI Classification requires a base LP360 (Geospatial, Drone, or Land) license and at least one advanced AI cloud licenses - AI Ground+. AI Forestry, or AI Utilities - assigned to the user.  It does not require the Automatic AI Ground Classification license, which is still supported and available as AI Ground (Only).

LP360's Advanced AI Classification tools apply AI‑based workflows to LAS point cloud data to generate classified point clouds and depending on the workflow, extract point, vector, and polygon features and raster outputs. The tools use deep learning/machine learning AI models to classify your point cloud, extract feature layers and generate raster products.  LP360's AI models are generalized models.  You do not need to provide your own training data or build your own model.

This article outlines the differences between the three available advanced AI classification models and describes the common workflow between them.  Note if you are only interested in classifying the ground surface and not any above ground features, the AI Ground (Only) tool is the recommended option.

Advanced AI Classification Models

LP360 supports three advanced AI models.   

• Ground+ – Optimized for ground and terrain‑related classification, including classifying common above ground features. 
• Forestry – Advanced tree segmentation designed for vegetation, canopy, and tree‑level analysis.
• Utilities – Tower and wire extraction designed for power line workflows, where identifying wires, towers, and related infrastructure is required. 

Ground+

Classification Behavior

• Generates a Ground Class and Segments Above Ground Feature Classes. The AI model classifies ground points to form a ground surface.  It also classifies common above ground features and assigns them to the appropriate point classes (see full list below).  The AI segmented points will be merged with any existing points in each feature class.
• Has an Option to Preserve Class 2 (Ground). To preserve an existing ground class separate from the AI ground segmentation, choose the option to "Output AI Ground in dedicated class" and the AI ground points will be written to Class 23 (AI Ground).  If selected, the AI will not write any points in Class 2, typically a known good ground surface, allowing you to preserve your existing ground model separate from the AI surface.  Note the presence of an existing ground surfaces does not influence the results of the AI model; creating a separate AI ground class is solely for the user's convenience.  The AI ground points in Class 23 can be reviewed and then merged into to Class 2 using the Commit AI Ground button.
• Ignores Class 5 - High Vegetation When Classifying Ground.  The AI model will not reclassify any points in Class 5, typically 1st and intermediate returns, when classifying the ground surface.  It will reclassify these points when classifying above ground features.  This can help with identifying certain edge cases like storm drain covers or retaining wall edges, so it is good practice to move all 1st/Intermediate returns to Class 5 before submitting the data to the AI.  However, this step is optional and not required.
• Places All Vegetation in Class 3 - Low Vegetation.  The AI does not perform a height segmentation on any points identified as vegetation/canopy.  It places all these points on Class 3 (Low Vegetation).  Use LP360's Height Segmentation Point Cloud Task if you need to have vegetation points segmented by specific height bands.
• Appends To Existing Classes.  The AI will not overwrite points in existing classes other than Class 0/1 (Never Classified/Unclassified) and Class 5 (High Vegetation).  It will append any existing classes with the results of its segmentation.  For example, if you have building roofs already classified in Class 6 (Buildings) those point classifications will be preserved and any new building roof points segmented by the AI will be added to Class 6.

Point Classes Output (Class#)

• Unclassified (1)
• Ground (2)
• Low Vegetation (3)
• Building (6)
• Low Noise (7)
• Water (9)
• Bridge Deck (17)
• High Noise (18)
• AI Ground (23)
• Below Ground (25)
• Vehicles (27)
• Roof Objects (28)
• Walls (29)
• Fences (30)

Outputs

• Classified LAS only.
• No vector or raster layers are generated.

Utilities

Classification Behavior

• The AI Utilities model applies a generalized deep learning model to classify ground, canopy, wires, towers and related infrastructure (see full list below).  It assumes unclassified input and will overwrite all existing classes including Class 2 (Ground).  
• The workflow will also extract feature layers from the classified point cloud and produce a set of utility-related SHP files (see full list below) and add them back to LP360.

Learn more about the AI Utilities outputs

Point Classes Output (Class#)

• Unclassified (1)
• Ground (2)
• Low Vegetation (3)
• Low Noise (7)
• Guard Wire (13)
• Wires (14)
• Towers (15)
• High Noise (18)
• Insulators (33)
• Guy Wire (34)
• Jumper Wire (35)

Feature Layer Outputs

• Tower Location (3D Line Feature Layer)
• Wires (3D Line Feature Layer)
• Tower Connection (3D Line Feature Layer)
• Insulators (3D Line Feature Layer)
• Attachment Points (3D Point Feature Layer)

Prerequisites

Complete any initial geocoding, strip alignment/matching, accuracy assessment and debiasing as per standard procedures before submitting data to an AI classifier.  Noise removal and smoothing are not required prior to running an advanced AI classifier but may improve results for lower performance sensors with very high shot noise ('fuzzy' surfaces).

At least one of the advanced AI cloud add-on licenses - AI Ground+. AI Forestry, or AI Utilities - must be assigned to your user.

Learn more about Cloud add-on Subscription License Management

Working With Classified Data

It is important to understand how the advanced AI classifiers work with existing point classifications prior to submitting an advanced AI classification.  This will allow you configure your data appropriately depending on your workflow and desired results.  This is most important when running AI Ground+.  The following guidelines should be taken into consideration when preparing your data for advanced AI classification.

If you do not have any existing point classifications to preserve, no additional set-up is required.  Proceed to Step #2 below.

Ground+

• The Ground+ classifier will segment points in Class 0 (Never Classified) and Class 1 (Unclassified) and assign (write) them to the appropriate target classes as detailed in the AI extended class table (see Appendix at end of this article).  Any remaining unclassified points after segmentation will be written back to Class 1 (Unclassified).  
• The Ground+ classifier will segment the ground points and assign them to Class 2 (Ground).  You have the option to "Output AI Ground in dedicated class".  Checking this option will place the AI ground points into Class 23 (AI Ground).  This allows you to preserve an existing known good ground surface in Class 2 (Ground) and review the AI ground before committing it to your final ground surface.  You can use the Commit AI Ground button to quickly merge the advanced AI ground points back into Class 2 (Ground) when you have reviewed it.
• The Ground+ classifier will ignore points in Class 5 (High Vegetation) when classifying the ground.  It is a legacy practice to place 1st and Intermediate lidar returns (non-last returns) in Class 5 prior to running a ground classification and the AI will recognize this class set-up if it exists.  You are not required to move these returns to Class 5.  However, doing so will allow the AI to identify any 1st/Intermediate returns it segments as "ground" and place these separately in Class 24 (Temporary Ground) for later review.  This can be effective at detecting certain edge cases such as storm drain covers (1st returns that are from objects flush with the surrounding ground surface).
• All other standard classes will be preserved and appended to by the Ground+ classifier.  For example, if you have building roofs already classified in Class 6 (Buildings) those point classifications will be preserved and any new building roof points segmented by the AI will be added to Class 6.
• The Ground+ classifier does not perform a height segmentation on points identified as vegetation/canopy. It places all these points in Class 3 (Low Vegetation).  Any existing height segmentation applied to classes [3,4,5] will be lost during AI classification (all vegetation/canopy points will be placed in Class 3).  Use LP360's Height Segmentation Point Cloud Task after AI classification if you need to have vegetation points segmented by height bands.
• Any non-standard (custom) classes you want to preserve should be moved to Class 64 (User Definable) or higher.  The Ground+ classifier ignores all point classes 64+.

Forestry

• The Forestry classifier reclassifies and overwrites all point classes.  It should be run on unclassified points only.

Utilities

• The Utilities classifier reclassifies and overwrites all point classes.  It should be run on unclassified points only.

Executing from LP360 Desktop

To perform an Advanced AI Classification, open the AI Classification tab of the ribbon and select the AI tool you want to run in the AI Tools section of the ribbon.  This will open the Advanced AI Classification dialog. 

Input

Input LAS Layer(s)

The source LAS that will be classified.

Quick Set/Reset tools

The Available LAS Layers section has several tools to speed common settings. These tools are described in the table below.

Classification Settings

Under Classification Settings, select the AI workflow that best matches your use case:

• Ground+ - Enhanced Above Ground Feature Classifications
• Forestry – Advanced Tree Segmentation
• Utilities – Tower and Wire Extraction

Each workflow is optimized for a specific analysis and produces different outputs as described below. In the initial V2026.1 release you are limited to selecting one workflow to run. The ability to run combined workflows - for example Ground+ and Forestry - will be added in a future update.

Output

The tool automatically defines network paths within the project to save each of the output products. These paths can be reviewed and overwritten if desired but are typically recommended to be left as the defaults.

Depending on the workflow, outputs may include:

• New LAS Layer – A new layer for the classified point cloud.
• New Feature Layers – Extracted point, vector, or polygon feature layers.
• New Raster Layers – A new layer for rasterized outputs derived from point classification and analysis, such as a Canopy Height Model (CHM).

Cost Estimate

In the Cost Estimate panel:

• Review the total volume (GB) to be processed.
• Confirm the cost estimate in LP360 Points.
• Ensure your current balance has sufficient credits.

Job cost is based on total size, in GB, of the data being processed and billed in 10 GB steps.  Note that LP360 Points (LP$) covering five normal (< 10 GB) jobs are included each month with your subscription.  After your first five jobs in a month, extra jobs will be billed as follows:

Submit

Once you have configured the Advanced AI Classification job, press Submit to verify the setup and submit the CGC (Cloud Ground Classification) job to Job Manager for processing.

Monitor Progress Using Job Manager

You can monitor the progress of your Advanced AI Classification (CGC) job via Job Manager .  You will receive an email message once the AI processing is complete, notifying you to return to Job Manager . Open Job Manager , select your AI classification job and press Complete Job.  The newly generated LAS and derived outputs will now be added to your LP360 project and are ready for visualization and analysis.  New layers will be named with the data type/product and the date/time of the job processing.

How to Run Advanced AI Classification

1. Open a project in LP360 with at least one LAS Layer.
2. Select the desired Advanced AI Classification - AI Ground+ , AI Forestry , or AI Utilities tool on the AI Classification tab of the ribbon.
3. Select the LAS Layer(s) to be classified. Select the desired options.
4. Provide a name for the resulting LAS Layer, Feature Layer, and Raster Layer, as applicable.
5. Press Submit to create an AI Ground Classification (CGC = Cloud Ground Classification) job in the Job Manager . Applicable child jobs (XFR, and CLD) will be created as needed.
6. Once the job is Ready, press Complete Job to complete the job and add the AI classified point cloud, features, and rasters, as applicable, to your project.
7. Use the Ground cleanup tools in LP360 to refine the classification results.

Executing from LP360 Cloud using a Web Browser

Not Recommended.

To execute the Advanced AI Classification job from LP360 Cloud in a web browser, the LAS dataset must first be uploaded to the Cloud. Uploading data can be done using the Upload to Cloud tool on the LP360 Cloud of the ribbon within LP360 Desktop. 

Appendix - LAS V1.5/AI Class Table

When working with the AI classifiers in LP360, it is recommended to update your class table (class names) to the standard APRS LAS V1.5 Class Table with the extended LP360 AI classes as defined below.  The LP360 AI models use these classes for all segmentations and will not recognize if your data has custom classes that overlap these definitions.  For example, if you have a custom class named "Superstructure" assigned to Class 25, the AI will keep those points but add any "Below Ground" points to them, usually resulting in a compromised point class.  Any non-standard (custom) classes you want to preserve should be moved to Class 64 (User Definable) or higher before submitting to the AI classifier.  The advanced AI classifiers ignore all point classes 64+.

The ASPRS LAS V1.5 format defines all class names for Classes 0-22.  The LP360 extended AI class names are found from 23-35, including remaining Reserved classes 36-63 for future use.