Problem Statement

Road accidents caused by speed breakers and potholes result in an average of 9 fatalities daily. Contributing factors include poor visibility, inadequate signage, and driver inattention. This project addresses these safety concerns through proactive detection and driver notification.

Solution

This system reduces road accidents by alerting drivers in advance about upcoming speed breakers and potholes while simultaneously creating a comprehensive, crowdsourced database of road hazards.

Implementation Approach

  1. Crowdsourced Data Collection

    • Leverage smartphone sensors and geolocation data to identify speed breaker locations
    • Utilize accelerometer data from users’ devices to detect road anomalies in real-time

  2. Machine Learning Classification

    • Train an RNN model using TensorFlow to classify accelerometer data as speed breakers, potholes, or normal road conditions
    • Develop a custom dataset for training due to the absence of publicly available datasets

  3. Consensus-Based Validation

    • Require agreement from multiple devices (N) at a given geolocation (latitude, longitude) to confirm the presence of a speed breaker
    • This approach minimizes false positives and ensures data reliability

  4. Dynamic Database Creation

    • Continuously collect and aggregate data as users drive, mapping all speed breaker locations
    • Adapt to changing road conditions, particularly relevant in regions with dynamic infrastructure

  5. Real-Time Notification System

    • Alert users approaching validated speed breaker locations
    • Provide sufficient advance warning for drivers to adjust speed accordingly

Technology Stack

  • Backend: Django web framework
  • Frontend: HTML with JavaScript for data transmission and visualization
  • Machine Learning: TensorFlow for training RNN model to classify sensor data
  • Database: Geospatial database for storing location data

Key Features

  • Crowdsourced data collection for comprehensive coverage
  • Machine learning-based classification for accuracy
  • Real-time driver alerts for improved safety
  • Dynamic adaptation to changing road conditions

GitHub Repository