Automated Road Inspection Using Videos for e-Marg

Context & Background

Maintaining rural road networks is critical for regional economic growth in India. The Ministry of Rural Development's e-Marg platform tracks road maintenance. Currently, inspections are done manually by engineers, which is slow, expensive, and subject to grading errors.

Problems to be Addressed

Road anomalies (potholes, structural cracks) must be identified across thousands of kilometers. Processing high-resolution video streams in real-time requires scalable and efficient deep learning architectures.

Aims and Objectives

1. Design automated computer vision models to detect road distress.
2. Integrate road grading algorithms into the government's e-Marg pipeline.
3. Train models to classify distress levels under varied light and weather.

Methodology

The system uses deep convolutional networks trained on video feeds captured from vehicle-mounted cameras. Potholes, cracks, and structural distress are classified using semantic segmentation. The automated inspection model assigns road quality grades directly linked to the e-Marg database.

Expected Outcomes

Integration with e-Marg platform, real-time distress detection software, and field testing reports with rural development agencies.