Proof of Concept of Automatic Incident Detection System using Image Processing Technology

Abstract

This study presents the results of a proof-of-concept test for an automated incident detection system using image processing technology. A variety of simulated incidents, including stopped vehicles, pedestrians, fallen objects, and wrong-way driving, were tested on the Don Mueang 2 toll plaza elevated road, both during the daytime and nighttime. The system's performance was evaluated based on several variables, including Detection Rate (DR), Time to Detect (TTD), and False Alarm Rate (FAR). The results showed an average detection rate of 97.22% for various types of incidents. The average time for the system to detect incidents was 2.01 seconds, and the average false alarm rate was 1.69%. The detection performance was slightly lower at nighttime due to varying light conditions. However, due to distance limitations and safety measures at the tested site, it was not possible to test the system's performance for incidents involving high-speed vehicles. Overall, the results of this study demonstrate the potential of automated incident detection systems using image processing technology. The findings can be used to notify the selection of products and the development of incident management plans.