臺灣博碩士論文加值系統

自動視訊監控系統過去以人力為主，是安全巡防人員實地監控區域安全，現在以網路為主，運用數位化網路串連監控設備進行遠端監控，受惠於機器學習的快速發展，自動化視訊監控系統已朝向以智慧為主演進，整合計算機視覺技術來進行異常活動偵測。本論文目的係提出自動視訊監控系統架構以Apache Spark Streaming串流平台為基礎，整合大資料處理與機器學習、深度學習技術以解決異常活動偵測的問題與解高負載問題，並具負載平衡以及即時偵測等效能。使用方法以基礎伺服器核心架構利用 Apache Kafka，連結三台Intel Core i7與1080Ti GPU電腦主機為分散式運算環境，做為伺服節點之間的通訊中介軟體。使用UCF-Crime資料集作為測試與訓練，還設立了教異常活動的異常場景，異常指數是經過sigmoid所產生，數值範圍0~1，越高分代表越異常，發生異常活動的那瞬間異常值為0.6。實驗結果：網路延遲的實驗中發現兩個Broker可以稍微降低延遲，開啟越多的Topic會提高硬體的負擔，5個Topic的延遲時間相比1個Topic高達4倍左右。開啟多核心可以大幅提高效能，當Topic數量逐漸成長，效能提升幅度也越來越明顯，數據結果證明監控系統能達到低延遲、高吞吐的目標。本研究貢獻有二：1.自動化監控系統能夠做到即時目標偵測。2.運用Kafka降低自動化監控系統的延遲。

In the past, automatic surveillance system has changed from human-based system, which on-side security patrolled critical region, to network-based system, which utilized connected IP cameras for remote monitoring. With the advanced development of machine learning and deep learning technology, automatic surveillance system has changed towards intelligence-based with emerging computer vision technology for abnormal activity detection. The study based on apache spark streaming platform to integrate big data processing, machine learning and deep learning to design the automatic surveillance system. The proposed system is not only to solve the problem of abnormal activity detection, but also to solve the problem of high load. The designed system is excepted to have load balancing, and real-time detection. The architecture of the basic server uses Apache Kafka to connect three Intel Core i7 and 1080Ti GPU computer hosts as a distributed computing environment as the communication intermediary software between the server nodes. We use the UCF-Crime data set for testing and training. There is also an additional scene in the classroom. The abnormality index is generated by sigmoid. The value range is 0~1. The higher the score, the more abnormal. The abnormal value at the moment of abnormal activity is 0.6. In the network delay experiment, it is found that two Brokers can slightly reduce the delay. The more Topics that are turned on will increase the burden on the hardware. The delay time of 5 Topics is as high as about 4 times compared with that of 1 Topic. Enabling multiple cores can greatly improve performance. As the number of topics gradually grows, the performance improvement rate becomes more and more obvious. The above experimental data results prove that the monitoring system achieves the goal of low latency and high throughput.

誌謝
摘要
ABSTRACT
目錄
圖目錄
表目錄
1. 前言
1.1研究背景
1.2研究動機
1.3 研究目的
1.4論文架構
2.文獻探討
2.1 目標偵測
2.2 異常活動辨識
2.3 視訊串流中介層介紹
2.4 自動化視訊監控系統架構設計
2.5 Kafka 串流介紹
3. 系統架構與原理方法
3.1 視訊資料收集層(Video Data Curation Layer)
3.2 視訊資料處理層(Video Data Processing Layer)
3.3 視訊資料分析層(Video Data Analytic Layer)
3.4 視訊資料表現層(Video Data Presentation Layer)
3.5 模組化架構
3.6 異常行為偵測
4. 系統實作與實驗分析
4.1系統環境
4.2異常活動辨識實驗
4.2.1 數據集 32
4.2.2 效果分析
4.3 視訊監控網路自動化傳輸效能最佳化
4.3.1 實驗環境
4.3.2 效能指標
4.3.3 實驗結果與分析
5.結論與未來工作
5.1結論
5.2未來工作
參考文獻
附錄
自傳


圖目錄
圖2.1 R-CNN模型
圖2.2 Fast R-CNN模型
圖2.3 SSD模型架構
圖2.4Yolo模型
圖2.5 anchor box 示意圖
圖2.6 Yolov2使用的新網路架構darknet-19
圖2.7 Darknet-53
圖2.8 YOLOv4架構
圖2.9 C3D network
圖2.10 Deep MIL Ranking Model
圖2.9 Kafka 模型
圖3.1自動監控系統架構圖
圖4.1教室異常活動影片測試結果
圖4.2系統拓樸圖
圖4.3 Broker 1
圖4.4 Broker 2
 
表目錄
表4.1資料及比較表
表4.2異常偵測時間偵測

[1]Ching-Tang Fan, Yuan-Kai Wang, and Huang, Cai-Ren Huang, “Heterogeneous Information Fusion and Visualization for a Large-Scale Intelligent Video Surveillance System”, IEEE Transactions on Systems, Man, and Cybernetics: Systems, 47(4), 593–604,2017.
[2]Alberto Pérez, Pablo Chamoso, Victor Parra, and Antonio Juan Sánchez, “Ground Vehicle Detection through Aerial Images Taken by a UAV”. IEEE 17th International Conference., pp. 1-6.,2014.
[3]Nassim Ammour, Haikel Alhichri, Yakoub Bazi, Bilel Benjdira, Naif Alajlan, and Mansour Zuair, “Deep Learning Approach for Car Detection in UAV Imagery”, Remote Sensing, 9(4), 312 ,2017.
[4]Kaiming He, Xiagyu Zhang, Shaoqing Ren, and Jian Sun, “Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition”. Springer Cham In European Conference on Computer Vision, pp.346-361,2014.
[5]Wei Liu, Dragomir Anguelov, Dumitru Erhan, Christian Szegedy, Scott Reed, Cheng-Yang Fu, and Alexander C. Berg, “Ssd: Single shot multibox detector”, Springer Cham In European Conference on Computer Vision ,pp. 21-37,2016.
[6]Joseph Redmon, Santosh Divvala, Ross Girshick, and Ali Farhadi, “You Only Look Once: Unified, Real-Time Object Detection”, In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 779-788, 2016.
[7]Joseph Redmon, Ali Farhadi, , “YOLO9000: Better, Faster, Stronger”, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 7263-7271, 2017.
[8]Joseph Redmon, Ali Farhadi, “YOLOv3: An Incremental Improvement”, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR),2018.
[9]Chien-Yao Wang, Hong-Yuan Mark Liao, Yueh-Hua Wu, Ping-Yang Chen, Jun-Wei Hsieh, and I-Hau Yeh., “CSPNet: A new backbone that can enhance learning capability of cnn’, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshop (CVPR Workshop), 2020.
[10]Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun., “Spatial pyramid pooling in deep convolutional networks for visual recognition”, IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 1904–1916, 2015.
[11]Shu Liu, Lu Qi, Haifang Qin, Jianping Shi, and Jiaya Jia, “ Path aggregation network for instance segmentation”, In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 8759–8768, 2018.
[12]Varun Chandola, Arindam Banerjee, and Vipin Kumar, “Anomaly detection: A survey”, ACM computing surveys (CSUR), vol. 41, no. 3, pp. 1-58, 2009.
[13]Norimchi Ukita, Yusuke Moriguchi, and Norihiro Hagita, “People re-identification across non-overlapping cameras using group features”, Computer Vision and Image Understanding ,Volume 144, 2016.
[14]Tal Hassner, Yossi Itcher, Orit Kliper-Gross, “Violent flows: Real-time detection of violent crowd behavior”, IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops, 2012.
[15]Naoki Abe, Bianca Zadrozny, John Langford, “Outlier detection by active learning”, Proceedings of the 12th ACM SIGKDD international conference on Knowledge discovery and data mining , pp. 504–509,2006.
[16]Waqas Sultani, Chen Chen, Mubarak Shah, “Real-World Anomaly Detection in Surveillance Videos”, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 6479-6488,2018.
[17]Du Tran, Lubomir Bourdev, Rob Fergus, Lorenzo Torresani, and Manohar Paluri, “Learning spatiotemporal features with 3d convolutional networks”, Proceedings of the IEEE international conference on computer vision, pp. 4489-4497, 2015.
[18]Weishan Zhang, Liang Xu, Pengcheng Duan, Wenjuan Gong, Qinghua Lu, and Su Yang, “A video cloud platform combing online and offline cloud computing technologies”. Pers. Ubiquitous Compute, P1099–1110, 2015.
[19]Chuanping Hu, Gengjian Xue, Lin Mei, Li Qi, Jie Shao, Yanfeng Shang, and Jian Wang, “ Building an intelligent video and image analysis evaluation platform for public security”, In Proceedings of the 14th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS), pp. 1–6, 2017.
[20]Wu Chao, and Xin Ming Jun, “Multi-agent based distributed video surveillance system over IP”, In Proceedings of the International Symposium on Computer Science and Computational Technology, pp. 97–100,2008.
[21]M. Anwar Hossain., “Framework for a Cloud-Based Multimedia Surveillance System.”, International Journal of Distributed Sensor Networks, Volume 10,2014.
[22]Haitao Zhang, Jin Yan, and Yue Kou, “ Efficient online surveillance video processing based on spark framework.”, In Proceedings of the International Conference on Big Data Computing and Communications, pp. 309–318, 2016.
[23]Azher Uddin, Aftab Alam, Nguyen Anh Tu, Md Siyamul Islam ,and Young-Koo Lee, “SIAT: A Distributed Video Analytics Framework for Intelligent Video Surveillance”, Multidisciplinary Digital Publishing Institute, Volume 11, 2019.
[24]Open-sourcing Kafka, LinkedIn’s distribute message queue, January, 2011
[25]Mark Hamilton, Sudarshan Raghunathan, Akshaya Annavajhala, Danil Kirsanov, Eduardo Leon, Eli Barzilay, Ilya Matiach, Joe Davison, Maureen Busch, Miruna Oprescu, Ratan Sur, Roope Astala, Tong Wen, and ChangYoung Park, “Flexible and scalable deep learning with MMLSpark”, Proceedings of the 4th International Conference on Predictive Applications and APIs, pp. 11–22, 2018.
[26]Stuart Andrews, Ioannis Tsochantaridis, and Thomas Hofmann, "Support vector machines for multiple-instance learning," Advances in neural information processing systems, vol. 15, pp. 577-584, 2002.
[27]http://www.svcl.ucsd.edu/projects/anomaly/UCSD_Anomaly_Dataset.tar.gz
[28]http://vision.eecs.yorku.ca/research/anomalous-behaviour-data/
[29]http://www.cse.cuhk.edu.hk/leojia/projects/detectabnormal/Avenue_Dataset.zip
[30]http://mha.cs.umn.edu/Movies/Crowd-Activity-All.avi
[31]https://aliensunmin.github.io/project/dashcam
[32]https://ankitshah009.github.io/accident_forecasting_traffic_camera
[33]https://github.com/MoonBlvd/tad-IROS2019/tree/master/datasets
[34]https://pan.baidu.com/s/1RfNjeW0Rjj6R4N7beSTYrA
[35]https://github.com/MoonBlvd/Detection-of-Traffic-Anomaly/tree/master/dataset
[36]https://svip-lab.github.io/dataset/campus_dataset.html
[37]https://visionlab.uncc.edu/download/summary/60-data/477-ucf-anomaly-detection-dataset
[38]https://github.com/dexXxed/abnormal-event-detection