Microcontroller-based camera with the sound source localization for automated accident detection

Abstract

This paper is on a microcontroller-based camera controller with sound source localization (SSL). With the rising frequency of highway accidents in Malaysia, there is a pressing need for a reliable detection system. The current approach, involving fixed-angled cameras, necessitates constant human monitoring, proving inefficient. To address this, the study introduces a hybrid camera system incorporating a camera for image capture and a microphone to detect collision sounds. By integrating a pan-tilt (PT) camera controller driven by time difference of arrival (TDOA) inputs, the system can swiftly move toward accident locations. The TDOA method is employed to convert sound arrival time differences into camera angles. The accuracy of the PT camera's rotation angle was analyzed based on the original sound source angle. As a result, this project produced an automated highway monitoring camera system that uses sound SSL to detect car crash sounds on highways. Its PT feature will help cover a large highway area and eliminate blind spots to capture possible accident scenes. The average inaccuracy of the experimental test of the pan and tilt angle of the camera is 19 and 23%, respectively. The accuracy of the pan tilt angle can be increased by adding more analog acoustic sensors.