RandLA-Net · YOLOv8 · MUTCD · Bridge Clearance

AI-Classified Mobile LiDAR for DOT Corridors

RandLA-Net classifies mobile LiDAR into 12 road-specific classes — road surface, guardrail, sign, curb, bridge deck, overhead wire — in minutes, not weeks.

The Problem

Where the time goes today

DOTs collect mobile LiDAR along hundreds of corridor-miles per year. Manual classification takes 200+ hours per 100 miles. By the time the data reaches engineers, the corridor has changed.

200+ hours manual classification per 100 corridor-miles

40+ hours additional for sign inventory extraction

30+ hours for guardrail inventory

Months from collection to engineering use

The Workflows

Transportation / DOT workflows on the platform

Twelve workflows tuned for DOT mobile LiDAR — semantic classification, asset inventory, and clearance / pavement analytics.

Workflow Name & what it automates Time saved
#211Road Corridor Classification
RandLA-Net: 12 DOT-specific classes from mobile LiDAR
#213Sign Inventory Extraction
YOLOv8 MUTCD classification + GPS coordinates
#214Bridge Clearance Measurement
Minimum clearance from LiDAR under bridge structures
#215Guardrail Inventory
Segment extraction + length + condition scoring
#212Pavement Condition
Roughness index proxy from LiDAR intensity analysis
#218Pavement Striping QA
Lane-marking presence and reflectivity audit from intensity raster
#220Overhead Utility Clearance
Vertical clearance to overhead conductors above corridor
#221Slope / Cut-Slope Stability
Roadside slope stability flagging from terrain derivatives
The Math

ROI vs. manual processing

Modeled on a state DOT classifying 500 corridor-miles annually with in-house GIS staff.

Metric Manual / current tooling GeoDataConverter
Annual classification cost$158,500$64,988
Net savings$93,512 (144% ROI)
Per-mile classification time2+ hours30 minutes
Sign inventory completenessSample-based100% wall-to-wall
Differentiator

Twelve DOT-specific point classes, not the generic ASPRS five

Workflow #211 uses RandLA-Net trained on DOT-collected mobile LiDAR to produce twelve practical classes — road surface, curb, sidewalk, guardrail, sign, traffic signal, light pole, bridge deck, retaining wall, vegetation, building, and overhead conductor — instead of the generic five ASPRS categories. The result is a classified point cloud you can hand to a pavement engineer or a sign-inventory analyst without rework.

Free 5-Mile Corridor Classification Pilot

Upload your mobile LiDAR. Receive AI-classified LAZ with 12 DOT-specific classes plus sign and guardrail inventory.

  • 5 corridor-miles classified
  • 12-class semantic segmentation
  • Sign inventory (MUTCD-typed)
  • Guardrail inventory with condition score