基於卷積神經網路的無人機即時目標偵測與追蹤

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：9

、訪客IP：13.58.247.31

姓名

楊少瑜(Shao-Yu Yang) 查詢紙本館藏

畢業系所

資訊工程學系

論文名稱

基於卷積神經網路的無人機即時目標偵測與追蹤
(Convolutional Neural Network-Based Real-Time Object Detection and Tracking for UAV)

相關論文

★ 影片指定對象臉部置換系統	★ 以單一攝影機實現單指虛擬鍵盤之功能
★ 基於視覺的手寫軌跡注音符號組合辨識系統	★ 利用動態貝氏網路在空照影像中進行車輛偵測
★ 以視訊為基礎之手寫簽名認證	★ 使用膚色與陰影機率高斯混合模型之移動膚色區域偵測
★ 影像中賦予信任等級的群眾切割	★ 航空監控影像之區域切割與分類
★ 在群體人數估計應用中使用不同特徵與回歸方法之分析比較	★ 以視覺為基礎之強韌多指尖偵測與人機介面應用
★ 在夜間受雨滴汙染鏡頭所拍攝的影片下之車流量估計	★ 影像特徵點匹配應用於景點影像檢索
★ 自動感興趣區域切割及遠距交通影像中的軌跡分析	★ 基於回歸模型與利用全天空影像特徵和歷史資訊之短期日射量預測
★ Analysis of the Performance of Different Classifiers for Cloud Detection Application	★ 全天空影像之雲追蹤與太陽遮蔽預測

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 (2028-7-16以後開放)

摘要(中)

近年來，無人機作為機器人產業中的一個熱門應用領域，得到了蓬勃的發展。無人機，也被稱為無人飛行載具（Unmanned Aerial Vehicle，UAV），是一種能夠在無需駕駛員在機上進行操控的情況下飛行的飛行載具。它可以通過自主控制飛行或者遠端遙控的方式進行操作，這使得無人機在許多領域中廣泛應用，包括物流運輸、搜救救援、環境監測等。
然而，實現無人機的自主飛行並確保其在複雜環境下的安全與效能仍然面臨著挑戰。其中之一是物體偵測和目標追蹤的問題。為了在無人機上實現高效的物體偵測和目標追蹤，本論文提出了一種基於ROS（Robot Operating System）的系統，並將其應用於Tello無人機上。
在本系統中，我們選擇了經過剪枝的YOLOv4架構作為物體偵測模型，這種架構在保持偵測準確性的同時，能夠提供更快的運行速度。同時，我們選擇了SiamMask作為目標追蹤模型，它是一種基於單目標追蹤的方法，能夠實現即時的目標追蹤。此外，我們引入了PID模組用於計算誤差並調整飛行動作，在本系統中，PID模組根據目標物體的位置計算誤差，並通過調整無人機的飛行動作，實現對目標物體的穩定追蹤。
通過飛行實驗的驗證，我們證明了本系統在日常環境中的可行性。經過剪枝的YOLOv4模型提供了高效的物體偵測能力，能夠在即時環境中實現快速的目標檢測。同時，SiamMask模型能夠實現目標的連續追蹤和定位。PID模組能夠準確計算誤差並適應不同的飛行情況，使無人機能夠穩定地追蹤目標物體。

摘要(英)

In recent years, UAVs have seen significant growth as a popular application in the robotics industry. Also known as UAVs or Unmanned Aerial Vehicles, these flying vehicles can operate without a pilot on board. They are widely used in various fields such as logistics, search and rescue, and environmental monitoring.
However, achieving autonomous flight and ensuring safety and efficiency in complex environments remain challenging. Object detection and target tracking are among the key issues. To address these challenges, this study proposes a ROS-based system applied to the Tello UAV.
The system incorporates a pruned YOLOv4 architecture for efficient object detection with faster runtime speeds. Additionally, SiamMask, a single-object tracking method, enables real-time target tracking. A PID module is introduced to compute errors and adjust flight attitude, allowing stable target tracking by adapting to different flight conditions.
Through flight experiments, the system′s feasibility in real-world environments has been validated. The pruned YOLOv4 model provides efficient object detection, while the SiamMask model enables continuous target tracking and localization. The PID module accurately calculates errors and adjusts the UAV′s attitude control signals.
In summary, this study proposes a ROS-based system for efficient object detection and target tracking on UAVs. The system′s effectiveness in everyday environments contributes to advancing UAV technology.

關鍵字(中)

★ 無人機
★ 深度學習
★ ROS
★ 卷積神經網路
★ 剪枝網路
★ 目標追蹤網路
★ PID控制

關鍵字(英)

★ UAV
★ deep learning
★ ROS
★ convolutional neural network
★ pruned network
★ target tracking network
★ PID control

論文目次

摘要 I
Abstract II
致謝 III
圖目錄 VI
表目錄 VIII
第一章緒論 1
1.1 研究背景與動機 1
1.2 研究目的 3
1.3 論文架構 4
第二章文獻回顧 5
2.1 四旋翼基本概念 5
2.2 ROS 9
2.2.1 ROS節點環境架構 9
2.3 PID 12
2.4 Batch normalization 14
2.5 YOLOv4 15
第三章研究方法 16
3.1 研究流程設計 16
3.2 實驗資料集 17
3.2.1 COCO 2014 17
3.3 實驗設備與實施方法 19
3.4 物件偵測 20
3.4.1 Darknet training 21
3.4.2 Pruning stage 21
3.4.3 Sparsity training 22
3.4.4 Channel cutting 24
3.4.5 Layer cutting 25
3.4.6 Fine-tune 26
3.5 物件追蹤 26
3.5.1 SiamMask 26
3.6 ROS通訊系統的節點圖 28
3.7 PID控制 29
3.7.1 Roll、Yaw選擇 32
第四章實驗結果 33
4.1 設備環境與參數設置 33
4.2 驗證指標 34
4.2.1精確率(Precision)、召回率(Recall) 34
4.2.2 F1 score 35
4.2.3 BFLOPs 36
4.2.4 mAP 36
4.3 Darknet基礎訓練 37
4.4 Pruning Results 38
4.4.1 Sparsity training Results 38
4.4.2 Channel cutting Results 40
4.4.3 Layer cutting Results 42
4.4.4 Final Model Tuning 43
4.5 無人機實驗設計 45
4.5.1實驗場地 46
4.5.2 PID設定之參數 48
4.5.3無人機控制分析 49
4.5.4無人機實驗場景資料集 53
第五章結論與未來研究方向 55
參考文獻 57

參考文獻

[1] J. Gray, "Attention Drone Geeks! Here′s Some Answers You′ve Been Looking For," [Online]. Available: https://thelocalbrand.com/attention-drone-geeks-some-answers/.
[2] Amazon, "Amazon Prime Air," [Online]. Available: https://www.amazon.com/Amazon-Prime-Air/b?node=8037720011.
[3] Federal Aviation Administration, "Operation and Certification of Small Unmanned Aircraft," [Online]. Available: https://www.federalregister.gov/documents/2016/06/28/2016-15079/operation-and-certification-of-small-unmanned-aircraft-systems#h-33.
[4] Multirotor Aircraft - Wikipedia, [Online]. Available: https://zh.wikipedia.org/zh-tw/%E5%A4%9A%E8%BB%B8%E9%A3%9B%E8%A1%8C%E5%99%A8.
[5] M .Quigley, B. Gerkey, K. Conley, J. Faust, T. Foote, J. Leibs, E. Berger, R. Wheeler, and A. Ng, "ROS: An Open-Source Robot Operating System," in Proceedings of the ICRA Workshop on Open Source Software, 2009.
[6] ROS, [Online]. Available: http://www.ros.org/.
[7] ROS Wiki, [Online]. Available: http://wiki.ros.org/ROS/Concepts.
[8] J.M. O′Kane, "A Gentle Introduction to ROS," Create Space Independent Publishing Platform, pp. 117-118, 2013.
[9] I.Baranov, "ROS 101: Intro to the Robot Operating System," [Online]. Available: https://www.clearpathrobotics.com/2014/01/how-to-guide-ros-101/.
[10] Wikipedia, "PID Controller," [Online]. Available: http://en.wikipedia.org/wiki/PID_controller.
[11] S.Ioffe and C. Szegedy, "Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift," arXiv preprint arXiv:1502.03167, 2015.
[12] A.Bochkovskiy, K. Wong, and A. Farhadi, "YOLOv4: Optimal Speed and Accuracy of Object Detection," arXiv preprint arXiv:2004.10934, 2020.
[13] K.He, X. Zhang, S. Ren, and J. Sun, "Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 37, no. 9, pp. 1904-1916, 2015.
[14] S.Liu, L. Qi, H. Qin, J. Shi, and J. Jia, "Path Aggregation Network for Instance Segmentation," in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 8759-8768, 2018.
[15] N.Wang, J. Li, Z. Wang, F. Yang, and X. Zhu, "CSPNet: A New Backbone That Can Enhance Learning Capability of CNN," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 43, no. 9, pp. 2823-2835, 2020.
[16] Z.Li, C. Peng, G. Yu, X. Zhang, Y. Deng, and J. Sun, "C3: Lightweight and Efficient Network for Object Detection," in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 12826-12835, 2020.
[17] S.Yun, D. Han, S. Oh, S. Chun, J. Choe, and Y. Yoo, "CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Features," in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 6023-6032, 2020.
[18] T.-Y. Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Dollár, and C. L. Zitnick, "Microsoft COCO: Common Objects in Context," in European Conference on Computer Vision (ECCV), 2014, pp. 740-755.
[19] Tello, [Online]. Available: https://www.ryzerobotics.com/zh-tw/tello-edu?site=brandsite&from=landing_page.
[20] YOLOv4 Baseline Training, [Online]. Available: https://github.com/AlexeyAB/Darknet.
[21] TNTWEN, "Pruned-OpenVINO-YOLO," [Online]. Available: https://github.com/TNTWEN/Pruned-OpenVINO-YOLO.
[22] P.Zhang, Y. Zhong, and X. Li, "SlimYolov3: Narrower, Faster, and Better for Real-Time UAV Applications," in 2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW), IEEE, Oct 2019.
[23] Q.Wang, L. Zhang, L. Bertinetto, et al., "Fast Online Object Tracking and Segmentation: A Unifying Approach," in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1328-1338, 2019.
[24] G.Koch, R. Zemel, and R. Salakhutdinov, "Siamese Neural Networks for One-Shot Image Recognition," in ICML Deep Learning Workshop, February 2015.
[25] J.Bromley and Y. LeCun, "Signature Verification Using a ′Siamese′ Time Delay Neural Network," in Advances in Neural Information Processing Systems, vol. 6, pp. 737-744, 1993.
[26] ROS rqt, [Online]. Available: https://index.ros.org/p/rqt/.
[27] Murtazahassan, "Drone-Face-Tracking," [Online]. Available: https://github.com/murtazahassan/Drone-Face-Tracking.
[28] Z.Liu, J. Li, Z. Shen, G. Huang, S. Yan, and C. Zhang, "Learning Efficient Convolutional Networks Through Network Slimming," in Proceedings of the IEEE International Conference on Computer Vision, pp. 2736-2744, 2017.
[29] Willow Garage - Wikipedia, [Online]. Available: https://en.wikipedia.org/wiki/Willow_Garage.

指導教授

鄭旭詠(Hsu-Yung Cheng)

審核日期

2023-7-24

推文