Auto-BlackBox-3D

Employing Black Box Mechanism for Investigation and Analysis of Road Accidents

Overview

This project presents a novel system for accident analysis using advanced sensor technology. The system gathers real-time acceleration, gyroscope, and GPS data to create detailed 3D models of vehicle orientation during accidents. By employing machine learning techniques and data visualization tools, the system enhances post-accident analysis and provides valuable feedback for improving vehicle design and safety.

Features

• Data Collection: Utilizes MPU6050 and GPS sensors to gather acceleration, gyroscope, and location data.
• Anomaly Detection: Employs autoencoders to detect anomalies in acceleration values.
• 3D Modeling: Uses Three.js to create real-time 3D models of vehicle orientation.
• Data Visualization: Implements Plotly.js for comprehensive data visualization.
• Location Tracking: Integrates OpenStreetMap API for live GPS tracking.

System Architecture

Methodology

1. Data Collection: Sensors capture acceleration, gyroscope, and GPS data.
2. Data Processing: NodeMCU processes and transmits data to the cloud.
3. 3D Modeling: Three.js visualizes vehicle orientation in 3D.
4. Anomaly Detection: Autoencoders identify anomalies in the data.
5. Data Visualization: Plotly.js generates graphs for data analysis.

Results

• 3D Model Rendering: Real-time visualization of vehicle orientation during accidents.
• Anomaly Detection: Identified 581 anomalies with a 94.99% accuracy rate.
• Data Visualization: Comprehensive graphs showing acceleration and gyroscope data.

Homepage of Flask Application

Hardware Connections

Viewing 3D Model

Anomalies

PCA Projection of Sensor Readings

Tracking Live Location