RandLA-Net classifies mobile LiDAR into 12 road-specific classes — road surface, guardrail, sign, curb, bridge deck, overhead wire — in minutes, not weeks.
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
Twelve workflows tuned for DOT mobile LiDAR — semantic classification, asset inventory, and clearance / pavement analytics.
| Workflow | Name & what it automates | Time saved |
|---|---|---|
| #211 | Road Corridor Classification RandLA-Net: 12 DOT-specific classes from mobile LiDAR | — |
| #213 | Sign Inventory Extraction YOLOv8 MUTCD classification + GPS coordinates | — |
| #214 | Bridge Clearance Measurement Minimum clearance from LiDAR under bridge structures | — |
| #215 | Guardrail Inventory Segment extraction + length + condition scoring | — |
| #212 | Pavement Condition Roughness index proxy from LiDAR intensity analysis | — |
| #218 | Pavement Striping QA Lane-marking presence and reflectivity audit from intensity raster | — |
| #220 | Overhead Utility Clearance Vertical clearance to overhead conductors above corridor | — |
| #221 | Slope / Cut-Slope Stability Roadside slope stability flagging from terrain derivatives | — |
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 time | 2+ hours | 30 minutes |
| Sign inventory completeness | Sample-based | 100% wall-to-wall |
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.
Upload your mobile LiDAR. Receive AI-classified LAZ with 12 DOT-specific classes plus sign and guardrail inventory.