SmartCityVision: a privacy-preserving soft biometrics framework for pedestrian behaviour analytics in smart cities

Das, Sonjoy Ranjon; Patel, Preeti; Hassan, Bilal

London Met Repository

Tools

Lists

Das, Sonjoy Ranjon, Patel, Preeti and Hassan, Bilal (2026) SmartCityVision: a privacy-preserving soft biometrics framework for pedestrian behaviour analytics in smart cities. In: 14th International Conference on Frontiers in Intelligent Computing: Theory and Applications (FICTA 2026), 8 - 9 June 2026, London Metropolitan University, London (UK) / Online. (In Press)

Abstract
Documents
Details
Record

[+][-]

Abstract

The swift urbanization and rising population density have resulted in a rising demand of smart city environment intelligent and privacy conscious surveillance systems. More conventional surveillance methods tend to be based on physical biometric identifiers, including facial recognition, which provoke certain ethical issues and regulatory obstacles of personal data protection. This paper has recommended to these problems the SmartCityVision, a privacy preserving crowd analytics system, which employs soft biometric features and behavioural modeling to real time pedestrian analysis. The suggested system combines pedestrian detection through YOLOv8, DeepSORT multi-object tracking, and soft biometric attributes recognition to interpret pedestrian charac-teristics without using personally identifiable information. The PA-100K dataset is used to extract soft biometric features such as age group, gender, and clothing features whereas the JRDB-Act dataset is used to model behavioral interactions and group activities. A multi-modal fusion approach involves the integration of attribute data with spatio-temporal motion pattern to aid in estimating crowd density, tracking individual trajectories and recognizing group behavior. It has been shown that the performance attributes are reliably recognized and multi-object tracking is attainable by use of experimental results thus crowd monitoring is made to be effective. The proposed framework offers scalable and privacy protection to pedestrian behavior analytics in smart city surveillance frameworks.

Documents