Abstract:
The automotive industry strives to increase the safety of road users using various advanced driver-assistance systems (ADASs). One of the ADAS solutions that should incre...Show MoreMetadata
Abstract:
The automotive industry strives to increase the safety of road users using various advanced driver-assistance systems (ADASs). One of the ADAS solutions that should increase the safety of drivers and road users approaching the vehicle from behind is the one that includes an alarm system when the driver leaves the vehicle if there is a possibility of a collision with objects comming from behind. In this paper, a new solution for this purpose is proposed. The solution consists of the YOLOv7 algorithm for detecting traffic participants in video sequences, an algorithm for tracking detected traffic participants, and a distance estimation algorithm needed for calculating the speed of the object coming from behind and the time to potential collision (TTC) with the driver’s door, which potentially triggers the alarm system. After the development and testing of the solution using a PC, the solution is implemented and tested on the Raspberry Pi 4 platform to test its performance in the case of the embedded platform with limited computational resources. The solution has been tested on a custom dataset which consists of 34 different video sequences, aiming to detect five types of objects (car, bus or truck, pedestrian, cyclist and motorcyclist). The solution achieves high performance in terms of alarm activation and is able to process 25 FPS, thus providing real-time operation capabilities.
Published in: 2024 International Conference on Software, Telecommunications and Computer Networks (SoftCOM)
Date of Conference: 26-28 September 2024
Date Added to IEEE Xplore: 23 October 2024
ISBN Information:
Electronic ISSN: 1847-358X