Project aim

Measure road condition accurately, in real-world metres, with low-cost vehicle-mounted sensors rather than specialist survey equipment.

  • 50 kmRoad surveyed
  • 50Ground-truth measurements
  • 5Architecture components
  • IEEE T-ITSManuscript submitted
Illustrative map for Multi-Modal Road Assessment (TRECK) Illustrative graphic

Overview

As Principal Researcher at TRECK / Roadessy Lab, KNUST, Samuel led a five-component architecture that fuses camera, IMU and GPS data for metric-accurate road assessment. It was validated against 50 manually surveyed ground-truth measurements along 50 km of road, and the manuscript has been submitted to IEEE Transactions on Intelligent Transportation Systems.

The challenge

A camera alone sees defects but not their real size, GPS alone is too coarse to place them precisely, and an IMU drifts. Each sensor covers another's weakness, but only when they are synchronised and fused carefully.

Data & inputs

  • Vehicle camera
  • IMU
  • GPS
  • 50 surveyed ground-truth points

Approach

  1. 01

    Five-component architecture

    A five-component architecture brings the camera, IMU and GPS streams together into one measurement pipeline.

  2. 02

    Sensor fusion

    Camera, IMU and GPS streams are fused to recover metric scale and position.

  3. 03

    Field campaign

    Data were collected along 50 km of road.

  4. 04

    Ground-truth validation

    Outputs were compared with 50 manually surveyed measurements.

Results

The fused system produces metric-accurate road assessments validated against field survey.

Validation

Validated against 50 manually surveyed ground-truth measurements along 50 km of road.