ANPR-Based Secure Appointment Retrieval and Access Management System

Abstract

Managing visitor access in modern buildings has always been a challenge, especially when done manually. This paper describes ANPR-SARAM—an automated visitor access management platform designed for residential complexes, apartment buildings, and corporate offices. Rather than relying on security guards to manually record vehicle numbers, our system does it automatically using cameras and computer vision. We used YOLOv8, a state-of-the-art object detection model, along with CNNs and EasyOCR to detect vehicles, read their license plates in real time, and capture visitor photographs on arrival. All visitor information—including vehicle registration number, photo, entry time, and guest details—is stored on a secure, administrator-controlled server. The result is a smarter, faster, and more reliable way to handle building security that reduces human error and maintains a complete digital record of every visit.

Introduction

This work presents ANPR-SARAM, a complete cloud-based vehicle and visitor management system that uses Automatic Number Plate Recognition (ANPR) combined with modern web technologies to automate and secure gate-entry processes. The motivation comes from the limitations of traditional manual logbook-based visitor tracking, which is slow, error-prone, and lacks proper auditing.

The system uses a deep learning pipeline built on YOLOv8 for real-time license plate detection and EasyOCR (CRAFT + CRNN) for character recognition. To improve accuracy under real-world conditions such as poor lighting or occlusion, a three-tier OCR fallback mechanism is introduced. The system achieves high accuracy while maintaining real-time performance.

Beyond ANPR, ANPR-SARAM extends into a full visitor management platform, which is a key contribution missing in most existing research. It supports visitor pre-registration, secure 256-bit cryptographic QR codes, real-time Socket.IO notifications, JWT-based authentication, role-based access control, and automated email alerts, making it a complete end-to-end access control solution.

The architecture is built using a four-layer design: CV layer (YOLOv8 + OCR), application layer (Express.js APIs and security logic), data layer (MongoDB), and presentation layer (React dashboard). This modular design ensures scalability and independent deployment of components.

The system pipeline includes frame preprocessing, YOLOv8 detection, OCR recognition, visitor data registration, QR code generation, and secure check-in validation. Security is strengthened using TLS 1.3, Helmet.js, input validation, and rate limiting to prevent common web attacks.

Experimental results show strong performance across different lighting conditions on 400 test frames, along with low-latency API response and successful handling of concurrent users. Security testing confirms resistance to common OWASP-defined vulnerabilities.

Conclusion

This paper presented ANPR-SARAM, a system that started from a simple observation: buildings around us deserve smarter entry management, and the research community has not yet delivered a complete solution. We built one. After a systematic review spanning classical background subtraction methods [17], HOG features [21], deep CNN classifiers [26]–[29], the YOLO detector family [9], [40], [41], OCR sequence models [11], [12], and prior visitor management platforms [1], [2], [48]–[50]—we identified six gaps no single published system had addressed together. What we built achieves 89.8% full-plate recognition accuracy through a three-tier YOLOv8n and EasyOCR pipeline that drops OCR failure rates from 18% to 2.8%. The MERN-stack backend handles 80 concurrent users at under 90 ms P50 API latency. QR tokens backed by 256-bit cryptographic randomness with single-use consumption prevent replay attacks. Socket.IO delivers notifications in under 8 ms—two orders of magnitude faster than SMS. And for the first time in the visitor management literature, we ran formal OWASP ZAP penetration testing confirming protection against all eight tested attack vectors. Future work includes: (i) upgrading to YOLOv8m for improved nighttime accuracy; (ii) adding DeepFace or ArcFace facial verification at check-in; (iii) Redis Pub/Sub for federated multi-site deployment; (iv) differential-privacy CRNN training to mitigate PII exposure under the DPDP Act 2023; and (v) Grad-CAM visualization for targeted detection failure diagnosis.

References

[1] A. Mishra et al., \"IoT-based smart gate security system for residential buildings,\" in Proc. IEEE ICCCI, 2021, pp. 1–6. [2] A. Singh et al., \"Smart visitor management system using IoT and face recognition,\" in Proc. IEEE ICCCI, 2022, pp. 1–6. [3] Lubna, N. Mufti, and S. A. A. Shah, \"Automatic number plate recognition: A detailed survey,\" Sensors, vol. 21, no. 9, p. 3028, 2021. [4] A. Rakshe and N. Dongre, \"An efficient methodology of ANPR using deep learning,\" IJISAE, vol. 12, pp. 640–648, 2024. [5] S. W. Joshi et al., \"ANPR using YOLOv8,\" IJSRST, vol. 12, no. 2, pp. 1088–1097, 2025. [6] Y. LeCun, Y. Bengio, and G. Hinton, \"Deep learning,\" Nature, vol. 521, pp. 436–444, 2015. [7] A. Krizhevsky, I. Sutskever, and G. E. Hinton, \"ImageNet classification with deep CNNs,\" in Proc. NeurIPS, 2012. [8] J. Redmon et al., \"You only look once: Unified, real-time object detection,\" in Proc. IEEE CVPR, 2016, pp. 779–788. [9] G. Jocher, A. Chaurasia, and J. Qiu, \"Ultralytics YOLOv8,\" GitHub, Jan. 2023. [10] J. Kittenplon, \"EasyOCR: Ready-to-use OCR with 80+ languages,\" JaidedAI GitHub, 2020. [11] Y. Baek et al., \"CRAFT: Character region awareness for text detection,\" in Proc. IEEE CVPR, 2019, pp. 9365–9374. [12] B. Shi, X. Bai, and C. Yao, \"An end-to-end trainable neural network for sequence recognition (CRNN),\" IEEE Trans. PAMI, vol. 39, no. 11, pp. 2298–2304, 2017. [13] L. Laroca et al., \"An efficient layout-independent ANPR system,\" IET Intell. Transp. Syst., vol. 15, no. 4, pp. 483–503, 2021. [14] Z. Cao et al., \"Survey of license plate detection techniques,\" IEEE Access, vol. 9, pp. 102–118, 2021. [15] Ren and Lan, \"VMMR with modified AlexNet (MMR),\" custom dataset, 98.7%, 2020. [16] J. Prokaj and G. Medioni, \"Persistent tracklet fusion,\" in Proc. IEEE CVPR, 2009. [17] ] M. Piccardi, \"Background subtraction techniques: A review,\" in Proc. IEEE SMC, 2004, pp. 3099–3104. [18] A. A. Siddiqui et al., \"Bag of SURF features for vehicle classification,\" custom dataset, 94.8%, 2021. [19] V. Petrovi? and T. Cootes, \"Feature analysis for vehicle recognition,\" custom dataset, 93%, 2019. [20] R. G. Wijnhoven and P. H. N. de With, \"Fast training of object detection using stochastic gradient descent,\" in Proc. IEEE ICPR, 2010. [21] N. Dalal and B. Triggs, \"Histograms of oriented gradients for human detection,\" in Proc. IEEE CVPR, 2005, pp. 886–893. [22] Y. Zhou et al., \"Unified CNN framework for vehicle detection,\" DAVE dataset, 79.1%, 2020. [23] Semi-automatic training-and-evaluation VMMR, custom vehicle dataset, 97.3%, 2022. [24] J. Wei, \"CNN robustness and training characteristics,\" arXiv, 2020. [25] Y. LeCun et al., \"Handwritten digit recognition with a back-propagation network,\" in Proc. NeurIPS, 1989. [26] A. Krizhevsky, I. Sutskever, and G. E. Hinton, \"ImageNet classification with AlexNet,\" in Proc. NeurIPS, 2012. [27] X. Luo et al., \"Deep CNN with extended AlexNet layers for vehicle recognition,\" custom dataset, 97.51%, 2021. [28] H. J. Lee et al., \"SqueezeNet for vehicle recognition,\" custom dataset, 96.3% at 108 ms, 2021. [29] F. N. Iandola et al., \"SqueezeNet: AlexNet-level accuracy with 50× fewer parameters,\" arXiv:1602.07360, 2016. [30] B. Yang et al., \"Large-scale car dataset for fine-grained categorisation (CompCar),\" in Proc. IEEE CVPR, 2015. [31] C. Szegedy et al., \"Going deeper with convolutions (GoogLeNet),\" in Proc. IEEE CVPR, 2015. [32] M. Sabri et al., \"CNN classification of car images using Keras and TensorFlow,\" custom dataset, 88%, 2020. [33] S. Prabu et al., \"CNN-based vehicle classification (3 classes),\" custom dataset, ~99%, 2021. [34] M. Abed et al., \"CNN real-time vehicle classification (4 types),\" custom dataset, 97%, 2022. [35] S. Ren, K. He, R. Girshick, and J. Sun, \"Faster R-CNN,\" IEEE Trans. PAMI, vol. 39, no. 6, pp. 1137–1149, 2017. [36] J. Redmon and A. Farhadi, \"YOLO9000: Better, faster, stronger,\" in Proc. IEEE CVPR, 2017. [37] J. Redmon and A. Farhadi, \"YOLOv3: An incremental improvement,\" arXiv:1804.02767, 2018. [38] A. Bochkovskiy et al., \"YOLOv4: Optimal speed and accuracy,\" arXiv:2004.10934, 2020. [39] C.-Y. Wang et al., \"YOLOv7: Trainable bag-of- freebies,\" in Proc. IEEE CVPR, 2023. [40] S. Luo and J. Liu, \"Improved YOLOv5m + LPRNet for license plate recognition,\" IEEE Access, vol. 10, pp. 93692–93700, 2022. [41] T. Mustafa and M. Karabatak, \"Real time car model and plate detection using deep learning,\" IEEE Access, vol. 12, pp. 107616–107632, 2024. [42] T. M. Al-Hasan et al., \"Enhanced YOLOv8-based ANPR for Qatar,\" Technologies, vol. 12, no. 9, p. 164, 2024. [43] M. Safran et al., \"Multistage YOLOv8+CNN for smart parking,\" J. Sensors, 2024. [44] R. A. Rasoul et al., \"ANPR systems: Formats and challenges,\" IJCA, 2021. [45] A. Zafar et al., \"Enhancing ANPR accuracy: A comprehensive survey,\" IEEE Access, 2022. [46] R. Smith, \"An overview of the Tesseract OCR engine,\" in Proc. IEEE ICDAR, 2007. [47] K. Ahmed et al., \"Douglas-Peucker and connected- component OCR for ANPR,\" IJCA, 2020. [48] N. A. Abd Rahman et al., \"Secure parking with ANPR + QR,\" in Proc. IEEE ICDCECE, 2022. [49] V. Mehta et al., \"QR-based visitor management with SMS,\" in Proc. IEEE ICCUBEA, 2020. [50] F. Anuar and N. Lingas, \"Smart campus QR + YOLOv3,\" AIP Conf. Proc., 2023. [51] OWASP Foundation, \"OWASP API Security Top10—2023,\" 2023. [52] M. Jones et al., \"JSON Web Token (JWT),\" RFC 7519, IETF, May 2015. [53] N. Dragoni et al., \"Microservices: Yesterday, today, and tomorrow,\" Springer, 2017. [54] S. Roy and K. Indukuri, \"Security analysis of JWT in healthcare REST APIs,\" in Proc. IEEE ICDCS, 2021. [55] V. Abramova and J. Bernardino, \"NoSQL databases:MongoDB vs. Cassandra,\" in Proc. ICSEA, 2013. [56] G. P. Cuaton and Y. Su, \"Philippines LPR: Systematic literature review,\" JARDCS, 2021.

Copyright

Copyright © 2026 P. Mallikarjuna, Rajesh Yamparala (Asst.Professor), P. Paavan Siddhartha, P. Madhu Babu. 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.