Tighte: A Model for Campus Security Target Tracking in Edge Intelligent Computing Architecture

• Published in: 2024 IEEE 10th International Conference on Edge Computing and Scalable Cloud (EdgeCom) pp. 110 - 115
• Main Authors: Zhou, Feng, Zhao, Xin, Liu, Jing, Zhang, Hongbing, Qi, Fuli, Zhou, Tongming, Ma, Jun
• Format: Conference Proceeding
• Language: English
• Published: IEEE 28.06.2024
• Subjects: Accuracy; Artificial intelligence; Computational modeling; Computer architecture; Deep learning; Edge computing; Machine learning; Public security; Security; Servers; Target tracking
• DOI: 10.1109/EdgeCom62867.2024.00025

Target tracking technology has been widely used in public security, human-computer interaction, and military fields. In application scenarios for campus security, we found that uneven light, blocked targets, and noisy video imaging were common problems when we conducted research on target tracking. In order to solve these common problems and improve the response speed of the model, we propose a target tracking model, Tighte (Target Tracking Modeling in Edge Computing Architecture), based on the improved DeepSORT (Deep Learning-based SORT) algorithm under the edge computing architecture. The Tighte model we proposed performs operations on edge nodes when performing target tracking. This technical architecture reduces the computing pressure of the central node server and the transmission load of the cloud edge network bandwidth. In experiments using the WiderPerson dataset, our proposed Tighte model achieved an accuracy of 71.23%, a precision of 78.97%, and 86 IDs. Therefore, the proposed Tighte model can provide strong technical support for target tracking business needs in campus security.