Research on Pedestrian Detection Based on Jetson Xavier NX Platform and YOLOv4

Le résumé fourni par la source

With the rapid development of deep learning, object detection technology has been widely used in various fields. Pedestrian detection, as an important part of object detection, has also received attention and development. However, the development also brings a series of challenges. The rapid increase in information requires higher processing capabilities, and lightweight and real-time performance have become urgent issues while intelligent video surveillance tends to become routine. Therefore, obtaining accurate and real-time information flow while accurately locating pedestrian flow information, and quickly and accurately processing these information flows have become important issues. In response to the problems that may arise during pedestrian detection, this design uses the YOLOv4 algorithm in the software aspect, modifies it on the basis of the original algorithm, so that it can better achieve pedestrian detection while reducing the performance requirements on hardware, and ultimately achieve a high accuracy rate in real-time pedestrian detection. The NVIDIA Jetson Xavier NX platform is used in the hardware aspect. The powerful performance of Jetson Xavier NX can greatly meet the hardware performance requirements during pedestrian detection. In order to adapt to engineering requirements, the Jetson Xavier NX board is redesigned, and through comparative experiments, the help of the redesigned board for the experiment is verified. Finally, based on the modified YOLOv4 algorithm, the Jetson Xavier NX platform equipped with the redesigned board is used to achieve the design requirements for pedestrian detection through a series of experiments, and the research purpose of this article is achieved.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.