Designing a Multi-Modal Vehicle Data Capture System for Forensic Accident Analysis

Helmi, Wibowo; Ethys, Pranoto; Faris, Humami; Yoga, Pratama · 2025 · DOAJ

DOI: 10.1051/e3sconf/202562201004

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of obtaining reliable evidence for traffic accident investigations in Indonesia, where accident rates have risen significantly. Current investigative methods often rely on driver and witness statements, which are frequently unreliable due to fear, indifference, or memory lapses. While existing IoT-based vehicle monitoring systems can record location and speed, they often lack comprehensive audio and video capabilities. To bridge this gap, the authors designed a multi-modal vehicle data capture system intended to provide objective forensic evidence by recording speed, location, time, vehicle tilt, and internal cabin audio and video. The system was developed using a Raspberry Pi 4 Model B for visualization and image processing, and an ESP32 microcontroller for sensor management and data transmission. Key hardware components included a Beitian BE-220 GPS module for location tracking, an MPU6050 sensor for measuring tilt and acceleration, and cameras for visual recording. The system architecture allows data to be transmitted in real-time to a Firebase-hosted website while simultaneously storing it on an SD card. A critical feature of the design is an accident detection mechanism: when the MPU6050 sensor detects a vehicle tilt of 45 degrees or more, the system pauses data recording to preserve the specific data surrounding the incident, preventing data loss from subsequent vehicle movement or device damage. Testing was conducted on the Tegal-Pemalang toll road to evaluate sensor accuracy and system performance. The MPU6050 sensor demonstrated high precision, with error rates ranging from 0.028% to 0.123% across various tilt angles. The GPS sensor showed a 2% error rate, with latitude-longitude coordinate errors between 0.000661% and 0.001403%. During speed tests ranging from 10 km/h to 124 km/h, the system successfully recorded data, though some transmission failures occurred in areas with unstable signal strength. The accelerometer data captured significant vibration deviations, particularly on the Z-axis, correlating with road surface conditions and vehicle suspension dynamics. The study concludes that the developed system effectively captures comprehensive multi-modal data, offering a robust tool for traffic investigators and regulatory authorities. By providing objective audio-visual and telemetry evidence, the system can enhance the accuracy of accident analysis and serve as an educational tool to improve driver awareness. The authors note limitations regarding time discrepancies in video recording due to Raspberry Pi multitasking and suggest future improvements, including real-time cloud streaming, higher-resolution night-vision cameras, drowsiness detection features, and more robust housing designs.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• naturalistic crash near crash