整合Camshift及DeepSort的行人追蹤辨識系統

以作者查詢圖書館館藏

、以作者查詢臺灣博碩士

、以作者查詢全國書目

、勘誤回報

、線上人數：57

、訪客IP：18.191.150.173

姓名

李明陽(Ming-Yang Lee) 查詢紙本館藏

畢業系所

資訊工程學系在職專班

論文名稱

整合Camshift及DeepSort的行人追蹤辨識系統
(Pedestrian tracking and recognition system integrating Camshift and DeepSort)

相關論文

★ 整合GRAFCET虛擬機器的智慧型控制器開發平台	★ 分散式工業電子看板網路系統設計與實作
★ 設計與實作一個基於雙攝影機視覺系統的雙點觸控螢幕	★ 智慧型機器人的嵌入式計算平台
★ 一個即時移動物偵測與追蹤的嵌入式系統	★ 一個固態硬碟的多處理器架構與分散式控制演算法
★ 基於立體視覺手勢辨識的人機互動系統	★ 整合仿生智慧行為控制的機器人系統晶片設計
★ 嵌入式無線影像感測網路的設計與實作	★ 以雙核心處理器為基礎之車牌辨識系統
★ 基於立體視覺的連續三維手勢辨識	★ 微型、超低功耗無線感測網路控制器設計與硬體實作
★ 串流影像之即時人臉偵測、追蹤與辨識─嵌入式系統設計	★ 一個快速立體視覺系統的嵌入式硬體設計
★ 即時連續影像接合系統設計與實作	★ 基於雙核心平台的嵌入式步態辨識系統

檔案

[Endnote RIS 格式]

[Bibtex 格式]

[相關文章]

[文章引用]

[完整記錄]

[館藏目錄]

至系統瀏覽論文 (2027-9-26以後開放)

摘要(中)

目前相關的行人檢測追蹤辨識系統，因為精準度要求越來越高，就有可能會需要使用到目前較流行的CNN深度學習演算法提高精準度，但相對的對硬體的需求也必須要提高，才不會造成延遲的情況發生，所以一般都建議要搭配一定等級的GPU(Graphics Processing Unit)裝置，才較能達到即時性的速度，此情況可能導致在資源沒那麼充裕的地方，因無法添購GPU裝置，導致較進階的行人檢測追蹤安全系統設置困難，進而影響社會安全之類的提升。故本論文提出結合CamShift和DeepSort的整合式行人追蹤辨識系統，此系統既無需使用GPU裝置加速，也保有一定的追蹤及時性，以CamShift搭配DeepSort的卡爾曼濾波和適時執行Re-ID追蹤行人，此結合CamShift及DeepSort的整合式行人追蹤辨識系統，可因適時調整使用CNN深度學習演算法，增加無GPU裝置的追蹤即時性。實驗結果顯示本研究提出的方法，可以在只使用CPU(Central Processing Unit)的情況下，也能達到不錯的行人追蹤即時性，證明可以使用較少硬體資源也可以有較佳的追蹤辨識效率。

摘要(英)

At present, the relevant pedestrian detection, tracking and identification systems may need to use the more popular CNN deep learning algorithm to improve the accuracy because of the increasingly high accuracy requirements, but the relative demand for hardware must also be improved. , it will not cause delay, so it is generally recommended to match a certain level of GPU(Graphics Processing Unit) device to achieve instant speed. The purchase of additional GPU devices makes it difficult to set up a more advanced pedestrian detection and tracking security system, which in turn affects the improvement of social security. Therefore, this paper proposes an integrated pedestrian tracking and identification system combining CamShift and DeepSort. This system does not need to use GPU device acceleration, but also maintains a certain tracking timeliness. CamShift is combined with DeepSort′s Kalman filter and timely execution of Re-ID to track pedestrians. This integrated pedestrian tracking and identification system combined with CamShift and DeepSort can adjust the use of CNN deep learning algorithm in a timely manner to increase the real-time tracking of devices without GPU. The experimental results show that the method proposed in this study can achieve a good real-time pedestrian tracking under the condition of only using the CPU(Central Processing Unit), which proves that it can use less hardware resources and have better tracking and identification efficiency.

關鍵字(中)

★ 整合式
★ 追蹤辨識
★ 卡爾曼濾波
★ 卷積神經網絡
★ 物件偵測

關鍵字(英)

論文目次

摘　要 I
Abstract II
誌謝 III
目錄 IV
圖目錄 VI
表目錄 VIII
第一章緒論 1
1.1 研究背景 1
1.2 研究目的 2
1.3 論文架構 3
第二章技術回顧 4
2.1 引言 4
2.2 物件追蹤原理 4
2.2.1 CamShift物件追蹤技術 4
2.2.2 卡爾曼濾波(Kalman filter) 6
2.3 卷積神經網路(Convolutional Neural Network，CNN) 8
2.3.1 YOLO物件偵測技術 10
2.4 DeepSort多目標跟蹤演算法 12
2.4.1 匈牙利演算法(Hungarian algorithm) 14
2.4.2 Re-ID特徵提取技術 16
第三章整合式行人追蹤辨識系統設計 18
3.1 整合式行人追蹤辨識系統架構 18
3.2 MIAT系統設計方法論 19
3.2.1 IDEF0階層式系統設計 19
3.2.2 Grafcet離散事件建模 21
3.3 整合式行人追蹤辨識系統架構IDEF0及Grafcet 24
3.3.1 YOLO目標偵測模組IDEF0及Grafcet 27
3.3.2 CamShift追蹤模組IDEF0及Grafcet 28
3.3.3 DeepSort追蹤補強模組IDEF0及Grafcet 30
第四章實驗 33
4.1 實驗環境 33
4.2 CamShift追蹤模組影像強化識別介紹 36
4.3 IoU匹配校正整合關鍵模組流程介紹 38
4.4 行人追蹤辨識準確度實驗 39
4.5 行人追蹤辨識執行速度實驗 45
4.6 實驗結果整體比較與討論 48
第五章結論與未來研究展望 50
5.1 結論 50
5.2 未來展望 51
參考文獻 52

參考文獻

[1] S. Shirmohammadi and A. Ferrero, "Camera as the instrument: the rising trend of vision based measurement," IEEE Instrumentation & Measurement Magazine, vol. 17, no. 3, pp. 41-47, Jun. 2014.
[2] X. Hou, Y. Wang, L. Chau, "Vehicle Tracking Using Deep SORT with Low Confidence Track Filtering," IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS), pp. 1-6, 2019 16th.
[3] E. N. Kajabad, S. V. Ivanov, N. Ramezanzade, "Customer Detection and Tracking By Deep Learning and Kalman Filter Algorithms," International Conference on Electrical, Communication, and Computer Engineering (ICECCE), pp. 1-6, 2020.
[4] X. Wang, M. Liu, D. S. Raychaudhuri, S. Paul, Y. Wang, A. K. Roy-Chowdhury, "Learning Person Re-Identification Models From Videos With Weak Supervision," IEEE Transactions on Image Processing, vol. 30, pp. 3017-3028, 2021.
[5] O. Nouar, G. Ali, and C. Raphael, "Improved Object Tracking with CAMSHIFT Algorithm," IEEE International Conference on Acoustics, Speech and Signal Processing, pp. 1-10, 2006.
[6] Y. Cheng, "Mean shift, mode seeking, and clustering," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 17, pp. 1-6, August 1995.
[7] R. Kalman, "A New Approach to Linear Filtering and Prediction Problems," Journal of Basic Engineering, vol. 82, no. Series D, pp. 35–45, 1960.
[8] A. Krizhevsky, I. Sutskever, and G. E. Hinton, "ImageNet Classification with Deep Convolutional Neural Networks," Communications of the ACM, Vol. 60, Issue. 6, pp. 84-90, 2017.
[9] K. Simonyan, and A. Zisser man, "Very Deep Convolutional Networks for Large Scale Image Recognition," Proceedings of the IEEE, Vol. 86, Issue. 11, pp. 2278-2324, 1998.
[10] Y. Zhang, Z. Chen, B. Wei, "A Sport Athlete Object Tracking Based on Deep Sort and Yolo V4 in Case of Camera Movement," IEEE International Conference on Computer and Communications (ICCC), pp. 1312-1316, 2020 6th.
[11] N. Wojke, A. Bewley, and D. Paulus., "Simple Online and Realtime Tracking with a Deep Association Metric," CoRR, abs/1703.07402, 2017.
[12] A. Bewley, G. Zongyuan, F. Ramos, and B. Upcroft, "Simple online and realtime tracking," ICIP, pp. 3464–3468, 2016.
[13] S. Ren, K. He, R. Girshick, and J. Sun, "Faster R-CNN: Towards real-time object detection with region proposal networks," Neural Information Processing Systems, 2015.
[14] C. T. Ginalih, A. S. Jatmiko, R. Darmakusuma, "Simple Application of Kalman Filter On a Moving Object in Unity3D," International Conference on Interactive Digital Media (ICIDM), pp. 1-3, 2020 6th.
[15] S. Chewi, F. Yang, A. Ghosh, A. Parekh and K. Ramchandran, "Matching Observations to Distributions: Efficient Estimation via Sparsified Hungarian Algorithm," Annual Allerton Conference on Communication, Control, and Computing (Allerton), pp. 368-375, 2019 57th.
[16] N. Gheissari, T.B. Sebastian, R. Hartley, "Person Reidentification Using Spatiotemporal Appearance," CVPR, 2006
[17] E. Ristani, C. Tomasi, "Features for multi-target multi-camera tracking and re-identification," Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 6036–6046, 2018.
[18] C. H. Chen, M. Y. Lin, and X. C. Guo, "High level modeling and synthesis of smart sensor networks for Industrial Internet of Things," Computers & Electrical Engineering, vol. 61, no. Supplement C, pp. 48-66, July 2017.
[19] R. Mayer, IDEF0 function modeling - Areconstruction of the original Air Force Wright Aeronautical Laboratory Technical Report - AFWAL-TR-81-4023 (the IDEF0 Yellow Book), Texas: College Station, 1992.
[20] R. David, "Grafcet: a powerful tool for specification of logic controllers," IEEE Transactions on Control Systems Technology, vol. 3, pp. 253-268, 1995.
[21] X. Zhang, L. Rai, "Multi-Feature Extraction and Fusion Method for Pedestrian Re-identification," IEEE 3rd International Conference on Electronic Information and Communication Technology (ICEICT), pp. 743-747, 2020.
[22] A. G. Howard, M. Zhu, B. Chen, D. Kalenichenko, W. Wang, T. Weyand, M. Andreetto, and H. A. Mobilenets, "Efficient convolutional neural networks for mobile vision applications," arXiv preprint arXiv:1704.04861, 2017.
[23] Z. Wang, L. Zheng, Y. Liu, and S. Wang, "Towards real-time multiobject tracking," arXiv preprint arXiv:1909.12605, 2019.
[24] S. Kumar, Vishal, P. Sharma, N. Pal, "Object tracking and counting in a zone using YOLOv4, DeepSORT and TensorFlow," 2021 International Conference on Artificial Intelligence and Smart Systems (ICAIS), pp. 1017-1022, 2021.

指導教授

陳慶瀚

審核日期

2022-9-28

推文