Artificial Intelligence for Social Good | Automated Monitoring of Roads in Zambia: Pothole Detection

In this video Lighton Phiri [1] discusses a potential capstone research project with a group of undergraduate students he will be working with this year. The students are attached to the DataLab Research Group [2].
Screencast extracted from a weekly research progress meeting.
About the Project
High-quality roads have been identified as a crucial factor for social-economic development. This is especially the case for The Republic of Zambia where road transport is the most reliable and widely used mode of transportation.
Most Zambian roads are regularly reported to be characterised by potholes, posing a danger on both vehicles and safety of passengers. For instance, the March 2022 Ministerial Statement on "Rise in the Number of Road Traffic Crashes and Measures the Government is Putting in Place to Curb Road Crashes." reported on the death of seven (7) people on a Lusaka bound passenger bus from Nakonde. The Ministerial Statement further identified the "poor state of roads", characterised by deep potholes (National Assembly of Zambia).
According to the Roads and Road Traffic Act (Roads and Road Traffic Act), care and maintenance of roads is the responsibility of the "Director of Roads" and local authorities are responsible for care and maintenance of roads. While the specific details of monitoring of roads by Highway Authorities is not known, available information seems to indicate that this is a manual process that requires human intervention-the recently launched Road Safety Mobile Application by the Road Development Agency, in part, aims to enable road users to post pictures of defects (Mulowa).
A potential solution to the problem associated with monitoring of potholes could involve the use of data mining and Artificial Intelligence (AI) techniques for the detection and classification of potholes. Numerous studies have been conducted in the area of pothole detection, with studies focusing on support for real-time control for autonomous vehicles and offline data collection for road maintenance (Dhiman and Klette).
This project is about understanding how Highway Authorities monitor potholes and how the monitoring of potholes can be automated using Data Mining and AI techniques. In effect, the solution to be implemented will involve solving a “classification with location” problem (Murel and Kavlakoglu) that will be deployed as a Web-based mapping application to facilitate identification of specific locations of problematic portions of roads.
References
Dhiman, Amita, and Reinhard Klette. Pothole Detection Using Computer Vision and Learning. doi.org/10.1109/TITS.2019.293.... Accessed 11 Apr. 2024.
Mulowa, Anthony. “Press Statement. RDA And Ernkate Launch Road Safety Mobile Application.” Road Development Agency, 5 Jan. 2024, www.rda.org.zm/?p=2862.
Murel, Jacob, and Eda Kavlakoglu. “What Is Object Detection?” IBM, 3 Jan. 2024, www.ibm.com/topics/object-det....
National Assembly of Zambia. “Ministry of Transport and Logistics - On the Rise in the Number of Road Traffic Crashes and Measures the Government Is Putting in Place to Curb Road Crashes.” National Assembly of Zambia, Mar. 2022, www.parliament.gov.zm/node/10061.
Roads and Road Traffic Act. No. 11 of 2002, 2002, pp. 34-47, www.parliament.gov.zm/node/1492.
Video Timeline
00:00:00 Project Motivation: Ministerial Statement or Road Traffic Accidents
00:02:14 Dashcam Footage for Sample Road in Lusaka Zambia With Potholes
00:03:50 Zambian Legislation on Road Monitoring
00:06:00 Road Classification in Zambia
00:06:58 Who is Responsible for Monitoring of Roads in Zambia
00:08:15 Highway Authorities: Road Development Agency
00:08:57 Road Development Agency Road Safety Mobile App
00:10:43 Project Description: Pothole Detection and Classification
00:14:50 About Automatic Classification and Detection
[1] lightonphiri.org
[2] datalab.unza.zm/index.php/peo...
[3] datalab.unza.zm