Automated Road Quality Surveys
In collaboration with DFiD and the Department of Roads (DoR) Zanzibar, N/LAB is investigating methods combining remote sensing (drone/satellite imagery) and applied machine learning methods to automate the assessment of low volume road conditions. Traditionally such road quality surveys are conducted by driving a vehicle equipped with specialist equipment along all roads leading to non-trivial costs. High resolution imagery from drones and potentially satellites, however, provides an opportunity to automate this.
The proof-of-concept project, already in phase 2, aims to do this via the the application of advanced machine learning learnt and evaluated on ground-truths collected over all roads managed by the DoR and drone imagery.
Forming the first phase of the project, ground-truth data has been collected by the N/LAB team in conjunction with the DoR and includes:
- human in-car evaluations of road segment quality
- official measurements of road segment quality via a bump integrator
- measurements of road segment quality via the RoadLabPro mobile phone application recently commissioned by the World Bank
- video imagery of the roads and
- raw mobile sensor measurements.
Additionally as part of phase 1 drone imagery has been acquired. Phase 2 is currently ongoing.