Design of a Real-Time-Integrated System Based on Stereovision and YOLOv5 to Detect Objects

Abstract

Real-time object detection represents a major part in the development of advanced driver assistance systems (ADAS). Pedestrian detection has become one of the most important tasks in the field of object detection due to the increasing number of road accidents. This study concerns the design and implementation of a Raspberry Pi 4-based embedded stereovision system to detect 80 object classes including persons and estimate 3D distance for traffic safety. Stereo camera calibration and deep learning algorithms are discussed. The study shows the system's design and a custom stereo camera designed and built using 3D printer as well as the implementation of YOLOv5s in the Raspberry Pi 4. The object detector is trained on the context object detection task (COCO) 2020 dataset and was tested using one of the two cameras. The Raspberry Pi displays a live video including bounding boxes and the number of frames per second (FPS).