Road Quality Assessment for Safety and Adjustment of Driving Strategy

Improving Periodic Maintenance of Road Quality Through AI-Powered Video Analysis

Project Overview

This research project aims to revolutionize road maintenance through automated assessment of road quality using computer vision and deep learning techniques. By analyzing traffic videos, we can assess lane line paint quality, road surface conditions, and detect potholes and cracks to enable proactive maintenance strategies.

Key Objectives

  • Automated road quality assessment from video data
  • Real-time lane line detection and quality evaluation
  • Pothole and crack detection for maintenance scheduling
  • AR glasses integration for field assessment
  • Autonomous vehicle algorithm adaptation

Impact

  • Improved road safety through better maintenance
  • Cost-effective maintenance scheduling
  • Enhanced autonomous vehicle performance
  • Real-time road condition monitoring
  • Data-driven infrastructure management

Research Components

These research areas are pursued concurrently as interconnected parts of the overall project

Open Dataset V1 Creation

Development of a comprehensive open dataset using mobile cameras that provide video footage along with latitude and longitude coordinates. This enables road information mapping through APIs for accurate location-based analysis.

Mobile Camera Data Collection GPS Coordinate Mapping API Integration Data Annotation

Deep Learning Model Development

Creation and training of deep learning models for comprehensive analysis of road video data. These models will be trained on the annotated V1 dataset to detect and assess various road conditions.

Deep Learning Computer Vision TensorFlow/PyTorch Model Training

Dataset V2 with Head-Mounted Camera

Development of an enhanced dataset using a car equipped with head-mounted cameras for higher quality and more controlled data collection. This provides better perspective and consistency for model training.

Head-Mounted Cameras Vehicle Integration High-Quality Video Controlled Environment

Advanced Models and Algorithms

Development of sophisticated models and algorithms that leverage the V2 dataset for more accurate road quality assessment, including advanced computer vision techniques and machine learning approaches.

Advanced CV Algorithms Machine Learning Model Optimization Performance Enhancement

AR Glasses Integration

Implementation of real-time road quality assessment using AR glasses to support field workers and maintenance teams. This enables immediate assessment and better decision-making for road maintenance.

Augmented Reality Real-time Processing AR Glasses Integration Field Assessment Tools

Autonomous Vehicle Algorithm Adaptation

Development of algorithms that help autonomous vehicles adjust their driving strategies when encountering poor lane lines or road conditions, improving safety and performance in challenging road environments.

Autonomous Vehicles Adaptive Algorithms Safety Systems Real-time Adaptation

Technical Approach

Computer Vision

  • Lane line detection and quality assessment
  • Road surface condition analysis
  • Pothole and crack detection
  • Real-time video processing

Deep Learning

  • CNN-based road condition classification
  • Object detection for road defects
  • Transfer learning for efficiency
  • Model optimization and deployment

Data Processing

  • GPS coordinate integration
  • Video annotation and labeling
  • Data augmentation techniques
  • Quality control and validation

Project Support

This research project is supported by Dr. Shaif Chowdhury, Assistant Professor at Texas A&M University Kingsville.

View Project on GitHub

Get Involved

Interested in contributing to this research project? We welcome collaborations from students, researchers, and industry partners working in computer vision, autonomous vehicles, and road infrastructure.

Contact Dr. Chowdhury View on GitHub