Design and Implementation of the "Deep Eye Intelligent Control" Campus Security Management System

Kunyu Xie; Guilu Jiang; Hongda Yan; Jiahao Zhang; Linqing Wang; Kun Teng

doi:10.54097/49hjey20

Authors

Kunyu Xie School of Engineering Machinery, Shandong Jiaotong University, Jinan, 250357, China
Guilu Jiang School of Engineering Machinery, Shandong Jiaotong University, Jinan, 250357, China
Hongda Yan School of Engineering Machinery, Shandong Jiaotong University, Jinan, 250357, China
Jiahao Zhang School of Engineering Machinery, Shandong Jiaotong University, Jinan, 250357, China
Linqing Wang School of Engineering Machinery, Shandong Jiaotong University, Jinan, 250357, China
Kun Teng School of Engineering Machinery, Shandong Jiaotong University, Jinan, 250357, China

DOI:

https://doi.org/10.54097/49hjey20

Keywords:

Campus Security, Target Detection, Behavior Recognition, Intelligent Early Warning, Multi-modal Data Analysis

Abstract

To solve the problems of low efficiency, poor real-time performance and high labor cost in traditional campus security management, this paper designs and implements an intelligent campus security control system named "Shenmou Zhikong". The system integrates the improved YOLOv3-tiny target detection model, AAGC-LSTM spatiotemporal behavior recognition algorithm and multi-modal data collaborative analysis technology, and adopts the architecture of "video perception-algorithm analysis-multi-terminal linkage" to realize accurate identification and hierarchical early warning of abnormal behaviors, dangerous goods (such as knives and open flames) and traffic risks (such as fake license plate vehicles and speeding vehicles) in campus scenarios. To improve the adaptability of the system in complex weather, wavelet fusion algorithm is used to enhance the detection accuracy in foggy days. At the same time, memory mapping technology is adopted to realize continuous video frame analysis, and 5G network is used to support concurrent processing of 200 cameras. The experimental results show that the system has a foggy day detection accuracy (mAP) of 82%, a license plate detection accuracy of 94.47%, and a fight recognition response time of 0.8 seconds, which can improve the efficiency of security incident handling by 60%. It forms a full-chain security closed loop of "real-time monitoring-intelligent analysis-accurate disposal".

Downloads

Download data is not yet available.

References

[1] D P, K., & B, S. (2025). FARE RNET: Fashion Cloth Retrieval via Egret Swarm Optimization Based Convolutional Attention Module Integrated ResNet. International Journal of Computational Intelligence Systems, 19(1), 29.

[2] Dai, J., Feng, Y., Zhu, Y., et al. (2025). SAD GCN: Semantic aware deformable graph convolutional networks for human abnormal behavior recognition. Signal, Image and Video Processing, 19(16), 1377.

[3] Valarmathi, V., & Sudha, S. (2025). Violence and panic abnormal behaviour recognition in crowded environment using spatio temporal aggregator based machine learning model. Signal, Image and Video Processing, 19(15), 1302.

[4] Munoz, A., Thomas, N., Vapsi, A., et al. (2025). Veri Car: towards open world vehicle information retrieval. Neural Computing and Applications, 37(20), 1–39.

[5] Zheng, B., Zhou, J., Hong, Z., et al. (2025). Vehicle Recognition and Driving Information Detection with UAV Video Based on Improved YOLOv5 DeepSORT Algorithm. Sensors, 25(9), 2788.

[6] Shirani, A., & Rastin, N. (2025). Attention enhanced DenseNet 121 for histopathological classification of oral cancer. Signal, Image and Video Processing, 19(17), 1433.

[7] Sarkar, S., Mukherjee, S., Paul, S., et al. (2025). Modified Convolutional Block Attention Module Based Custom Deep Learning Model for the Classification of Skin Diseases. ES General, 11.

[8] Daniel, L., & Subhradeep, R. (2023). Using information theory to detect model structure with application in vehicular traffic systems. IFAC PapersOnLine, 56(3), 367–372.

[9] G.S.R., S., Prashant, D., & Kumar, S. D. (2022). Vehicle detection and classification with spatio temporal information obtained from CNN. Displays, 75.

[10] Kumar, D. D., & Prakash, S. S. (2022). Optimized Convolutional Neural Network for Road Detection with Structured Contour and Spatial Information for Intelligent Vehicle System. International Journal of Pattern Recognition and Artificial Intelligence, 36(6).