Homography Based Real-Time Intrusion Detection and Multi-Person Tracking System

Abstract

This research proposes a video intrusion detection system for detecting person and raising alarms on their entry into unauthorized zones. Five metrics have been used for evaluating this system. These include precision of homography calibration, precision in zone detection, effectiveness in tracking multiple persons, alarm generation, and database logging of events. According to experiment results, all above mentioned metrics exhibit efficient performances. The highest reprojection error for homography calibration was observed to be 0.58 inches and average error is 0.23 inches which is precise enough to perform spatial mapping with a minimum zone size of 20 inches. Zone detection was found to be efficient enough in classifying the person into one of the three predefined zones based on coordinates

Introduction

The text explains the concept of surveillance systems and their role in modern security. Surveillance involves monitoring people or places to collect information, ensure safety, and maintain control, and is widely used in homes, workplaces, industries, and public areas. The most common system is CCTV, which records and transmits video for security purposes. However, traditional CCTV systems have major limitations, such as lack of automation, dependence on human monitoring, high cost due to multiple cameras, and poor ability to detect incidents automatically.

To improve this, pixel-based detection systems were introduced using tools like MediaPipe for pose detection and zone labeling. Although this approach improved automation, it still struggled with accuracy because changes in camera perspective often caused incorrect zone detection and false alarms.

To overcome these issues, the proposed idea introduces a planar homography-based detection system, which maps image points to real-world ground coordinates. This reduces distortion and improves the accuracy of zone detection, making surveillance more reliable and realistic.

The literature review highlights the importance of security systems in preventing intrusion and protecting property. Traditional sensor-based technologies such as Passive Infrared (PIR) sensors, microwave sensors, and photoelectric beam sensors are commonly used but have limitations like false alarms, inability to detect complex environments, and restricted coverage areas.

The text also notes that earlier computer vision models like R-CNN were too slow for real-time use, while YOLO (You Only Look Once) provides faster and more efficient object detection by processing images in a single pass, making it suitable for real-time surveillance applications.

Conclusion

The developed intrusion detection system has been able to fulfill the design requirements effectively and can be put into action for monitoring applications inside buildings. It was able to provide great accuracy in homography calibration, where the highest error in terms of the maximum reprojection error was just 0.58 inches and the average reprojection error was 0.23 inches, well under the 20-inch limit for minimum distance between any two zone boundaries, thereby ensuring reliability of spatial mapping throughout the area. As far as the function of detecting zones was concerned, there were no errors made by the algorithm, as it was able to classify each individual according to their coordinate locations correctly in all cases. The system worked smartly for the purpose of alerting by taking priority for restricting zone over warning zone, along with providing real-time updating of zones in terms of the highest-priority one for each tracked individual. Additionally, logging of all incidents in the form of data about the tracks made and images of corresponding frames was also carried out properly by the system.

References

[1] Robu.in, “PIR Sensor Working Principle”, 2020. [2] M. Baballe et al., “The Various Types of sensors used in the Security Alarm system. International Journal of New Computer Architectures and their Applications”, 2020. 10. 50-59. 10.17781/P002618. [3] REALPARS.com, “Photoelectric Sensor Explained (With Practical Examples)”, 2021. [4] S. Soni, “Concept of YOLOv1: The Evolution of Real-Time Object Detection”, Medium, 2023. [5] M. Kotthapalli et al., “YOLOv1 to YOLOv11: A Comprehensive Survey of Real-Time Object Detection Innovations and challenges”, arxiv.org, 2025. [6] Y. Zhang et al., \"ByteTrack: Multi-Object Tracking by Associating Every Detection Box,\" in Proc. European Conf. Computer Vision (ECCV), 2022. [7] S. Khan and M. Shah, “A Multiview Approach to Tracking People in Crowded Scenes Using a Planar Homography Constraint”, in Proc. European Conf. Computer Vision (ECCV), Lecture Notes in Computer Science, vol. 3954, 2006, pp. 133–146. [8] Y. Luo et al., \"A Review of Homography Estimation: Advances and Challenges,\" Electronics, vol. 12, no. 24, p. 4977, 2023. [9] R. Sundaresan, \"Pixel-Based Object Motion Detection and Tracking with a Moving Camera,\" M.Eng. thesis, Dept. Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA, USA, May 2020. [10] Eshel, Ran and Yael Moses. “Homography based multiple camera detection and tracking of people in a dense crowd,” in Proc. IEEE Conf. on Computer Vision and Pattern Recognition (2008): 1-8. [11] DigiMRO, “Zone System in Security: complete Guide to Smarter Protection”, 2025.

Copyright

Copyright © 2026 Khushpreet Sindhu, Karishma Rathore, Deepika Sahu, Prof. Durgeshwari Sahu. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.