Year: 2026 | Month: April | Volume: 13 | Issue: 4 | Pages: 101-108
DOI: https://doi.org/10.52403/ijrr.20260411
Surveillance System Using Deep Learning with Automatic Report Generation
Putti Venkata Siva Teja1, Sirla Sowmya1, Shaik Naseema Banu2, Tungala Nandu Sree3, Sangepu Venkat Sai4, Shaik Mohammed Irfan5, Kondapalli Devendranadh6
1,2,3,4,5,6Department of Information Technology,
Dhanekula Institute of Engineering & Technology, Vijayawada, Andhra Pradesh, India.
Corresponding Author: Putti Venkata Siva Teja
ABSTRACT
With the increasing deployment of surveillance cameras across cities and public spaces, the massive volume of generated video data has made manual monitoring time-consuming, inefficient, and prone to human error. To address these challenges, this paper presents an intelligent surveillance system that leverages deep learning techniques to automatically analyze video streams and generate structured textual reports. The proposed framework integrates multiple computer vision components, including YOLOv8 for detecting dangerous objects such as knives and blood stains, along with Convolutional Neural Networks (CNNs) for real-time face recognition and activity suspicious activity detection. A webcam is used to capture live video input, enabling continuous monitoring and visualization of the environment. OpenCV (Open-Source Computer Vision) is employed to efficiently process video frames and display detection results in real time. A key innovation of this system is the Automatic Report Generation module, which utilizes natural language generation to convert detected events into summaries containing precise details such as event snaps, location, and timestamps. Furthermore, the system incorporates a proactive security approach by integrating an ESP32-controlled hardware mechanism for remote access control and mobile alerts via a Flutter application. Experimental evaluations demonstrate high performance, with detection accuracies reaching 95.1% for unknown persons and 94.2% for weapons, alongside a rapid report generation response time of 25 ms. This end-to-end solution minimizes the need for continuous human supervision while significantly improving the reliability and responsiveness of surveillance operations in smart cities and industrial environments.
Keywords: Deep Learning, Intelligent Surveillance System, OpenCV, YOLOv8, Convolutional Neural Network (CNN), Suspicious activity Detection, Automatic Report Generation, ESP32.
[PDF Full Text]