Enhanced People Detection and Tracking in Smart Surveillance Using Probabilistic Modeling

VW Engineering International, Volume: 6, Issue: 2, 148-156

• Received: March 28, 2025
• Accepted: April 18, 2025
• Published online: April 26, 2025

Aravind B. B^*1, Sanjeev Kumar Shah², D. Abdul Jaleel³, G. Praburam⁴, Premalatha V⁵

^*1Department of Computer Science and Engineering (Data Science), Malla Reddy Engineering College, Hyderabad

²Department of Computer Science and Engineering, Uttaranchal Institute of Technology, Uttaranchal University, Dehradun

³Department of Computer Science & Technology Madanapalle Institute of Technology & Science, Madanapalle

⁴Department of Computer Science and Engineering, Viswam Engineering College, Madanapalle⁵Department of Computer Science and Engineering, SASI Institute of Technology and Engineering,

Abstract:  People detection and tracking play a crucial role in smart surveillance systems, enabling various applications such as behavior analysis, identity recognition, and anomaly detection. Accurately detecting and tracking individuals in dynamic environments remains challenging due to variations in lighting, occlusions, and background noise, presenting a new and innovative approach for robust people detection and tracking, incorporating multiple processing stages: foreground extraction, noise removal, people detection, and tracking with re-identification.

The proposed method begins by extracting the foreground region from the background image to isolate potential moving objects. Subsequently, a noise removal technique is applied to enhance the precision of detection by eliminating unwanted artifacts. Once the foreground is refined, individuals are detected and tracked across consecutive frames. A re-identification mechanism ensures the correct association of individuals, even in cases of occlusion or temporary disappearance. To evaluate the effectiveness of the approach, extensive experiments were conducted on the PETS2009 dataset, a widely recognized benchmark for multi-object tracking in surveillance applications, suggesting the achievement of higher accuracy in detecting and tracking multiple individuals by the proposed method, even in complex scenarios. Furthermore, the method enhances people’s counting capabilities, reducing errors commonly observed in traditional tracking techniques. The findings suggest that the proposed approach can be effectively integrated into real-time surveillance systems, improving automated monitoring and security applications.

Keywords: people detection, people tracking, conditional probability, people counting