International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 11 | Nov 2025
p-ISSN: 2395-0072
www.irjet.net
“A Comprehensive Review of YOLO-Based Real-Time Video Surveillance Systems for Intelligent Object Detection and Anomaly Monitoring” Prof. Dhage T. S.1, Dr. Bere S. S.2, Memane Swapnil Bapusaheb3 1,2 Assistant Professor, Department of Computer Engineering 3 Student, Department of Computer Engineering
1,2,3 Dattakala Group of Institution Faculty of Engineering. Swami-Chincholi Bhigwan -------------------------------------------------------------------------***-----------------------------------------------------------------------including operator fatigue, reduced vigilance, and a high Abstract— The rapid growth of urbanization and
potential for overlooked events, especially during long‑duration monitoring or low‑light/night‑time conditions [2][3].
increasing security concerns have amplified the need for intelligent video surveillance systems. Traditional CCTV monitoring relies heavily on human operators, which is time-consuming, error-prone, and ineffective under low-light conditions. This paper presents a comprehensive review of YOLO (You Only Look Once)based real-time video surveillance systems, highlighting their capabilities in automatic object detection, anomaly identification, and alert generation. YOLO, a state-of-the-art deep learning algorithm, enables simultaneous detection of multiple objects with high accuracy and minimal computational delay. The review explores recent advancements in pre-processing techniques, model optimization, and integration with alert mechanisms to improve the reliability and efficiency of surveillance systems. Additionally, the paper discusses current challenges, including night-time detection limitations, computational requirements, and dataset dependency, and identifies potential future enhancements such as facial recognition, thermal imaging, cloud analytics, and AI-based behavior analysis. The findings suggest that YOLO-based systems provide a robust, scalable, and cost-effective solution for modern security and surveillance applications.
The advent of deep learning and advanced computer vision techniques has significantly transformed how image and video data can be processed. Object detection, which involves identifying and localizing objects of interest in images or frames, has benefited from the shift from hand‑crafted features and sliding‑window classifiers to convolutional neural networks (CNNs) and single‑shot detectors [4]. Among these, the YOLO (You Only Look Once) family of algorithms has emerged as a highly influential approach. YOLO reframes object detection as a single regression problem rather than a multi‑stage pipeline of region proposals and classification, enabling faster and more efficient inference [5][6]. The original YOLO model demonstrated the ability to perform object detection at high speed (45 fps) while maintaining competitive accuracy on benchmarks [5][7]. When applied to video surveillance, YOLO’s real‑time object detection capability presents several clear advantages. A surveillance system tasked with detecting intrusions, loitering, unauthorized access, or the presence of harmful objects (such as weapons) must not only identify objects (people, vehicles, items) but also respond swiftly. The one‑shot detection paradigm of YOLO enables simultaneous detection of multiple object classes in a single pass through the neural network, thereby reducing latency and simplifying the data processing pipeline [8][9]. Moreover, preprocessing steps—such as video frame extraction, image denoising, grayscale conversion, resizing, and normalization—can further enhance the robustness of detection under challenging conditions (e.g., low light, shadows, or noisy frames) [10]. Integrating such preprocessing with a YOLO‑based detection backbone thus forms a promising architecture for smart surveillance systems with alert generation.
Keywords- YOLO, Object Detection, Video Surveillance, Real-Time Monitoring, Anomaly Detection, Intelligent Security Systems.
I.
INTRODUCTION
In recent years, the proliferation of video surveillance systems has become a cornerstone of security infrastructure in both public and private spaces. With growing urbanization, higher volumes of public events and increasing security threats, there is a rising demand for systems that can monitor wide‑area environments in real time and deliver proactive alerts when anomalous or suspicious behaviour occurs [1]. Traditional closed‑circuit television (CCTV) systems, while ubiquitous, largely depend on human operators to monitor large numbers of camera feeds. This reliance introduces considerable inefficiencies,
© 2025, IRJET
|
Impact Factor value: 8.315
However, the transition of YOLO‑based detectors into real‑world surveillance applications also presents notable challenges. Night‑time or low‑illumination conditions
|
ISO 9001:2008 Certified Journal
|
Page 95