校园安全在当今社会的公共安全监测中扮演着不可或缺的角色，直接涉及到师生的生命安全、财产保障，以及教育环境的稳定与进步。然而，随着校园环境的开放和人员结构的复杂化，校园安全问题日益突显，呈现出多样化和复杂化的挑战。为了有效应对这些挑战，迫切需要高效智能的监控系统来保障校园的安全。在视频监控方面，目标检测面临诸多困难，包括特征提取难、漏检、误检以及检测效率低等问题，这些挑战直接影响着监控系统的准确性和效率。针对这些挑战，本文提出了一系列新颖的解决方案，主要贡献如下：

       (1)针对校园安全场景的复杂多样性、特征提取方面，提出一种增强式自适应自演化特征提取算法EASE-YOLO(An enhanced adaptive self-evolving feature extraction algorithm, EASE-YOLO)。首先，基于YOLOv5框架，融入SE-ResNet注意力机制，增加通道关注机制，提升模型对关键特征的感知能力；其次，又引入了改进的BiFPN加强特征融合技术，显著的提升了目标检测的准确性和对复杂场景的适应能力；然后，自适应锚框机制与多任务学习框架进一步优化了数据适应性和特征泛化；最后，算法借助自演化学习策略，实现了模型的持续优化与自我更新。这些模块之间的协同作用促进了特征提取过程中的自我演化能力，从而不断提高模型的预测准确性。通过实验结果表明，改进的模型相比原来的模型，精度提升了4.9%。

       (2)为了有效监测校园内的骑行行为、提高安全性，提出了基于改进目标检测的多尺度特征融合行为检测算法MSFF-YOLO(Multi-scale feature fusion behavior detection algorithm based on improved object detection，MSFF-YOLO)。首先，针对传统数据样本提取能力有限的问题，本方法在骨干网络中引入了ShuffleNetV2轻量级网络，以实现高效率的特征提取过程。其次，为了提升不同尺度特征的融合效果，SPP（Spatial Pyramid Pooling）模块改进为SimSPPF（Simplified Spatial Pyramid Pooling - Fusion）模块，进一步增强了网络在多尺度特征处理中的性能。实验结果表明，MSFF-YOLO相比YOLOv5的精度提升了4.8%，为安全行为监测提供保障。

       (3)综合以上研究，设计开发了系统的仿真验证平台。在该平台上进行仿真测试，以验证所提出算法的有效性。仿真测试通过多种场景和数据集进行，确保算法在不同环境下的可靠性和准确性。

       此外，仿真验证平台还提供了可视化结果展示功能，帮助用户直观理解算法的工作原理和性能表现。这些可视化结果不仅有助于校园安全的预测和预警，还为进一步的优化和改进提供了宝贵的数据支持和实践依据。

Campus safety plays an indispensable role in today's public security monitoring, directly involving the life safety of teachers and students, property protection, and the stability and progress of the educational environment. However, with the opening of campus environments and the complexity of personnel structures, campus security issues have become increasingly prominent, presenting diverse and complex challenges. To effectively address these challenges, there is an urgent need for efficient and intelligent monitoring systems to ensure campus safety. In terms of video surveillance, target detection faces numerous difficulties, including challenges in feature extraction, missed and false detections, and low detection efficiency, which directly impact the accuracy and efficiency of monitoring systems. In response to these challenges, this paper proposes a series of innovative solutions with the following main contributions:

(1) For the complex diversity of campus safety scenarios, an enhanced adaptive self-evolving feature extraction algorithm, EASE-YOLO, is proposed. Initially, based on the YOLOv5 framework, the SE-ResNet attention mechanism is integrated to enhance the channel attention mechanism, thereby improving the model's perception of critical features. Furthermore, an improved BiFPN feature fusion technique is introduced to significantly enhance target detection accuracy and adaptability to complex scenes. The adaptive anchor box mechanism and multi-task learning framework further optimize data adaptability and feature generalization. Finally, with the aid of a self-evolving learning strategy, the algorithm achieves continuous optimization and self-updating of the model. The synergy between these modules promotes the self-evolving capability in the feature extraction process, thereby continuously improving the model's predictive accuracy. Experimental results show that the improved model has increased its accuracy by 4.9% over the original.

(2) In order to effectively monitor cycling on campus and improve safety, MSFF-YOLO(Multi-scale feature fusion behavior detection algorithm based on improved object) is proposed detection, MSFF-YOLO). Firstly, to solve the problem that the traditional data sample extraction ability is limited, ShuffleNetV2 lightweight network is introduced into the backbone network in this method to achieve an efficient feature extraction process. Secondly, in order to improve the Fusion effect of features of different scales, the SPP (Spatial Pyramid Pooling) module is improved to SimSPPF (Simplified Spatial Pyramid Pooling - Fusion) module. The performance of the network in multi-scale feature processing is further enhanced. The experimental results show that the accuracy of MSFF-YOLO is improved by 4.8% compared with YOLOv5, which provides guarantee for safety behavior monitoring.

(3) Based on the above research, a system simulation and verification platform was designed and developed. Simulation tests were conducted on this platform to verify the effectiveness of the proposed algorithms. The simulation tests were carried out across various scenarios and datasets to ensure the reliability and accuracy of the algorithms in different environments.

Furthermore, the simulation verification platform provides visualized results to help users intuitively understand the working principles and performance of the algorithms. These visualized results not only aid in the prediction and warning of campus safety but also provide valuable data support and practical basis for further optimization and improvement.

[1]Zhang Q, Sun H, Wu X, et al. Edge video analytics for public safety: A review[J]. Proceedings of the IEEE, 2019, 107(8): 1675-1696.

[2]Anwar M Z, Kaleem Z, Jamalipour A. Machine learning inspired sound-based amateur drone detection for public safety applications[J]. IEEE Transactions on Vehicular Technology, 2019, 68(3): 2526-2534.

[3]中国安防行业“十四五”发展规划（2021-2025）[J].中国安防,2021(06):22-38.

[4]O'Malley L J. Campus Safety: Student Perceptions and The Impact on Student Engagement[M]. Hofstra University, 2019.

[5]Mahajan M, RajeshwariBajre R, Sharma L, et al. Computer Vision Based Campus Surveillance[C]//2020 2nd International Conference on Advances in Computing, Communication Control and Networking (ICACCCN). IEEE, 2020: 1013-1018.

[6]Hamad A H. Smart campus monitoring based video surveillance using haar like features and k-nearest neighbour[J]. International Journal of Computing and Digital Systems, 2021, 10.

[7]铁富珍.基于改进光流法的视频监控中人群异常行为检测算法[J].现代电子技术,2024,47(07):45-48.

[8]Pan H, Zhang H, Lei X, et al. Hybrid dilated faster RCNN for object detection[J]. Journal of Intelligent & Fuzzy Systems, 2022, 43(1): 1229-1239.

[9]邵延华,张铎,楚红雨等.基于深度学习的YOLO目标检测综述[J].电子与信息学报,2022,44(10):3697-3708.

[10]Zou Z, Chen K, Shi Z, et al. Object detection in 20 years: A survey[J]. Proceedings of the IEEE, 2023, 111(3): 257-276.

[11]Ren S, Li J, Tu T, et al. Towards efficient video detection object super-resolution with deep fusion network for public safety[J]. Security and Communication Networks, 2021, 2021: 1-14.

[12]Zhang Q, Sun H, Wu X, et al. Edge video analytics for public safety: A review[J]. Proceedings of the IEEE, 2019, 107(8): 1675-1696.

[13]Liu W, Cao J, Zhu Y, et al. Real-time anomaly detection on surveillance video with two-stream spatio-temporal generative model[J]. Multimedia systems, 2023, 29(1): 59-71.

[14]Gupta A, Anpalagan A, Guan L, et al. Deep learning for object detection and scene perception in self-driving cars: Survey, challenges, and open issues[J]. Array, 2021, 10: 100057.

[15]龚晓蓉, 白玉. 基于深度学习的视频图像目标检测方法[J]. 经纬天地, 2021(2): 85-88+92.

[16]Liu, W., Cao, J., Zhu, Y. et al. Real-time anomaly detection on surveillance video with two-stream spatio-temporal generative model. Multimedia Systems 29, 59–71 (2023).

[17]Liu Y, Liu D, Wang B, et al. Mob-YOLO: A Lightweight UAV Object Detection Method[C]//2022 International Conference on Sensing, Measurement & Data Analytics in the era of Artificial Intelligence (ICSMD). IEEE, 2022: 1-6.

[18]孙好, 董兴法, 王军等. 基于改进YOLOV4-tiny轻量化校内行人目标检测算法[J]. 计 算机工程与应用, 2022: 1-12

[19]Tong Y, Ma H, Zhang S, et al. Research on object detection in campus scene based on faster R-CNN[C]//Journal of Physics: Conference Series. IOP Publishing, 2022, 2203(1): 012050..

[20]Bansal A, Singh J, Verucchi M, et al. Risk ranked recall: Collision safety metric for object detection systems in autonomous vehicles[C]//2021 10th Mediterranean Conference on Embedded Computing (MECO). IEEE, 2021: 1-4.

[21]陈婷,朱熟康,高涛,等.基于自适应融合的实时车辆检测[J].同济大学学报(自然科学版),2024,52(04):532-540.

[22]王莹,田莹.基于改进YOLOv5s的复杂环境行人检测模型[J/OL].微电子学与计算机,2024(03):29-36.

[23]潘烨新,杨哲.基于多级特征双向融合的小目标检测优化模型[J/OL].计算机应用:1-8.

[24]Song H, Zhu J, Jiang Y. Two-stage merging network for describing traffic scenes in intelligent vehicle driving system[J]. IEEE Transactions on Intelligent Transportation Systems, 2021, 23(12): 25509-25520.

[25]孙福临,李振轩,粱允泉,等.基于改进YOLOv5算法和边缘设备的电动车违规载人检测[J].现代计算机,2023,29(08):1-11.

[26]Yang R, Pan Z, Jia X, et al. A novel CNN-based detector for ship detection based on rotatable bounding box in SAR images[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2021, 14: 1938-1958.

[27]李玺,高佳,励雪巍等.结构自适应自演化的高级机器学习方法研究与进展[J].中国基础科学,2022,24(03):47-53.

[28]Lu L, Liu C, Zhang C, et al. Architecture for Self-Evolution of 6G Core Network Based on Intelligent Decision Making[J]. Electronics, 2023, 12(15): 3255.

[29]Wang H, Guo J, Ma X, et al. Online Self-Evolving Anomaly Detection in Cloud Computing Environments[J]. arXiv preprint arXiv:2111.08232, 2021.

[30]Bhusal N, Lohani S, You C, et al. Spatial mode correction of single photons using machine learning[J]. Advanced Quantum Technologies, 2021, 4(3): 2000103.

[31]Baz M. SEHIDS: Self evolving host-based intrusion detection system for IoT networks[J]. Sensors, 2022, 22(17): 6505.

[32]Park S, Kim G, Oh Y, et al. Self-evolving vision transformer for chest X-ray diagnosis through knowledge distillation[J]. Nature communications, 2022, 13(1): 3848.

[33]Zhou H, Greenwood D, Taylor S. Self-supervised monocular depth estimation with internal feature fusion[J]. arXiv preprint arXiv:2110.09482, 2021.

[34]Zhao Z Q, Zheng P, Xu S, et al. Object detection with deep learning: A review[J]. IEEE transactions on neural networks and learning systems, 2019, 30(11): 3212-3232.

[35]Zaidi S S A, Ansari M S, Aslam A, et al. A survey of modern deep learning based object detection models[J]. Digital Signal Processing, 2022, 126: 103514.

[36]Shetty A K, Saha I, Sanghvi R M, et al. A review: Object detection models[C]//2021 6th International Conference for Convergence in Technology (I2CT). IEEE, 2021: 1-8.

[37]Bansal A, Singh J, Verucchi M, et al. Risk ranked recall: Collision safety metric for object detection systems in autonomous vehicles[C]//2021 10th Mediterranean Conference on Embedded Computing (MECO). IEEE, 2021: 1-4.

[38]Samadzadegan F, Dadrass Javan F, Ashtari Mahini F, et al. Detection and recognition of drones based on a deep convolutional neural network using visible imagery[J]. Aerospace, 2022, 9(1): 31.

[39]邵延华,张铎,楚红雨等.基于深度学习的YOLO目标检测综述[J].电子与信息学报,2022,44(10):3697-3708.

[40]陈幻杰, 王琦琦, 杨国威, 等. 多尺度卷积特征融合的SSD 目标检测算法[J]. 计算机科学与探索, 2019, 13(6): 1049-1061.

[41]Ye T, Zhang J, Li Y, et al. CT-Net: An efficient network for low-altitude object detection based on convolution and transformer[J]. IEEE Transactions on Instrumentation and Measurement, 2022, 71: 1-12.

[42]Samadzadegan F, Dadrass Javan F, Ashtari Mahini F, et al. Detection and recognition of drones based on a deep convolutional neural network using visible imagery[J]. Aerospace, 2022, 9(1): 31.

[43]Li Z, Yang L, Zhou F. FSSD: feature fusion single shot multibox detector[J]. arXiv preprint arXiv:1712.00960, 2017.

[44]陈泽, 叶学义, 钱丁炜等. 基于改进Faster R-CNN的小尺度行人检测[J]. 计算机工程, 2020, 46(09): 226-232+241.

[45]Amjoud A B, Amrouch M. Object detection using deep learning, CNNs and vision transformers: a review[J]. IEEE Access, 2023.

[46]Kim W, Kerle N, Gerke M. Mobile augmented reality in support of building damage and safety assessment[J]. Natural Hazards and Earth System Sciences, 2022, 16(1): 287-298.

[47]Sharma S, Engel D. Mobile augmented reality system for object detection, alert, and safety[J]. Electronic Imaging, 2023, 35(12).

[48]Anwar M Z, Kaleem Z, Jamalipour A. Machine learning inspired sound-based amateur drone detection for public safety applications[J]. IEEE Transactions on Vehicular Technology, 2019, 68(3): 2526-2534.

[49]Girshick R. Fast r-cnn[C]//Proceedings of the IEEE international conference on computer vision. 2015: 1440-1448.

[50]Kaur R, Singh S. A comprehensive review of object detection with deep learning[J]. Digital Signal Processing, 2023, 132: 103812.

[51]Sirisha U, Praveen S P, Srinivasu P N, et al. Statistical analysis of design aspects of various YOLO-based deep learning models for object detection[J]. International Journal of Computational Intelligence Systems, 2023, 16(1): 126.

[52]Peng Y, Rios A, Kavuluru R, et al. Chemical-protein relation extraction with ensembles of SVM, CNN, and RNN models[J]. arXiv preprint arXiv:1802.01255, 2018.

[53]Sahoo A, Barik A, Samantaray S, et al. Prediction of sedimentation in a watershed using RNN and SVM[C]//Communication Software and Networks: Proceedings of INDIA 2019. Springer Singapore, 2021: 701-708.

[54]陈泽, 叶学义, 钱丁炜等. 基于改进 Faster R-CNN 的小尺度行人检测[J]. 计算机工程,2020, 46(09): 226-232+241.

[55]Ren S, He K, Girshick R, et al. Faster r-cnn: Towards real-time object detection with region proposal networks[J]. Advances in neural information processing systems, 2015, 28.

[56]陈幻杰, 王琦琦, 杨国威, 等. 多尺度卷积特征融合的 SSD 目标检测算法[J]. 计算机科学与探索, 2019, 13(6): 1049-1061.

[57]REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 779-788

[58]Feng D, Harakeh A, Waslander S L, et al. A review and comparative study on probabilistic object detection in autonomous driving[J]. IEEE Transactions on Intelligent Transportation Systems, 2021, 23(8): 9961-9980.

[59]Jiang P, Ergu D, Liu F, et al. A Review of Yolo algorithm developments[J]. Procedia computer science, 2022, 199: 1066-1073.

[60]Chen W, Huang H, Peng S, et al. YOLO-face: a real-time face detector[J]. The Visual Computer, 2021, 37: 805-813.

[61]Sun Z, Li P, Meng Q, et al. An improved YOLOv5 method to detect tailings ponds from high-resolution remote sensing images[J]. Remote Sensing, 2023, 15(7): 1796.

[62]邱天衡,王玲,王鹏等.基于改进YOLOv5的目标检测算法研究[J].计算机工程与应用,2022,58(13):63-73.

[63]Li B, Jiao B, Chou C H, et al. Self-evolution cascade deep learning model for high-speed receiver adaptation[J]. IEEE Transactions on Components, Packaging and Manufacturing Technology, 2020, 10(6): 1043-1053.

[64]Zhu XK, yu LSC, Wang X, Zhao Q. TPH-YOLOv5: Improved YOLOv5 based on transformer prediction head for object detection on drone-captured scenarios. In: Proceedings of the IEEE International Conference on Computer Vision; October 2021; Montreal, BC, Canada. IEEE (2021). p. 2778–88.

[65]于慧伶, 霍镜宇, 张怡卓. 融合多层特征 SENet 和多尺度宽残差的高光谱图像地物分类[J]. 实验室研究与探索, 2020, 39(7): 28-34.

[66]Zhang H, Du Q, Qi Q, et al. A recursive attention-enhanced bidirectional feature pyramid network for small object detection[J]. Multimedia tools and applications, 2023, 82(9): 13999-14018.

[67]Yang X, He Y, Wu J, et al. EBiDA-FPN: enhanced bi-directional attention feature pyramid network for object detection[J]. Journal of Electronic Imaging, 2024, 33(2): 023013-023013.

[68]Zhang J, Zhang H, Liu B, et al. Small object intelligent detection method based on adaptive recursive feature pyramid[J]. Heliyon, 2023, 9(7).

[69]Ma N, Zhang X, Zheng H T, et al. Shufflenet v2: Practical guidelines for efficient cnn architecture design[C]//Proceedings of the European conference on computer vision (ECCV). 2018: 116-131.

[70]Yan B, Fan P, Lei X, et al. A real-time apple targets detection method for picking robot based on improved YOLOv5[J]. Remote Sensing, 2021, 13(9): 1619.

[71]Ji W, Pan Y, Xu B, et al. A real-time apple targets detection method for picking robot based on ShufflenetV2-YOLOX[J]. Agriculture, 2022, 12(6): 856.

[72]Zhou, Q.; Liu, H.; Qiu, Y.; Zheng, W. Object Detection for Construction Waste Based on an Improved YOLOv5 Model. Sustainability 2023, 15, 681.

[73]Luo H, Wei J, Wang Y, et al. An improved lightweight object detection algorithm for YOLOv5[J]. PeerJ Computer Science, 2024, 10: e1830.

[74]Ren J, Wang Z, Zhang Y, et al. YOLOv5-R: lightweight real-time detection based on improved YOLOv5[J]. Journal of Electronic Imaging, 2022, 31(3): 033033-033033.

[75]Venkatesvara Rao N, Venkatavara Prasad D, Sugumaran M. Real-time video object detection and classification using hybrid texture feature extraction[J]. International Journal of Computers and Applications, 2021, 43(2): 119-126.

[76]Lin Y C, Kuan T W, Tseng S P, et al. Implementation of Campus Pedestrian Detection Using YOLOv5[C]//International Conference on Genetic and Evolutionary Computing. Singapore: Springer Nature Singapore, 2023: 447-455.