對於許多應用（如監控和災害響應），無人機的態勢感知能力是必要的。在這些應用中，實時人員檢測和動作辨識在準備有效響應方面起著至關重要的作用。傳統動作辨識準確率不高，無法解決不確定性問題，因此本文采用深度學習的方法。首先由 OpenPose 進行實時姿態估計，然後使用深度SORT 演算法跟踪多人場景。最後，基於從 OpenPose 檢測到的單幀關節，訓練深度神經網絡以辨識每個人的動作。深度學習模型的預測通過指標（準確度、精確度、召回率、精確召回曲線、F1 分數）進行評估。本文提出了一種無人機監控系統，可以自動檢測暴力和受傷的人。重點討論環境和行動的不確定性。環境不確定性包括視角、光線和障礙物。動作的不確定性包括動作的模糊性和多人的重疊。綜上所述，採用姿態估計和動作辨識的無人機監控系統可以有效辨識暴力和受傷人員，幫助人們快速獲得幫助。

For many applications (such as surveillance and disaster response), situational awareness of unmanned aerial vehicles is essential. In these applications, people detection and posture recognition in real-time plays a crucial role in preparing an effective response. The accuracy of traditional posture recognition is not high, and it cannot solve the uncertainty problem, so this paper uses the method of deep learning. First, the real-time pose estimation is performed by the OpenPose network, then the deep Sort algorithm is used for tracking multi-person scenes. Finally, based on a single frame of joints detected from OpenPose, the deep neural network is trained to recognize each person’s postures. The deep learning model’s predictions are evaluated via the metrics (accuracy, precision, recall, P-R curve, F1 score). This paper proposes a drone surveillance system, which can automatically detect violent and injured people. Focus on discussing the uncertainty of the environment and postures. Environmental uncertainty includes viewing angles, light, and obstacles. The uncertainty of the posture includes the ambiguity of the posture and the overlap of multiple people. In conclusion, the drone surveillance system using pose estimate and posture recognition can effectively identify violent and injured people to help people get support quickly.

[1] X. Hu and D. Li, “Research on a Single-Tree Point Cloud Segmentation Method Based on UAV Tilt Photography and Deep Learning Algorithm,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing., vol. 13, pp. 4111-4120, Jul. 2020.
[2] Z. Zheng, T. R. Bewley, and F. Kuester, “Point Cloud-Based Target-Oriented 3D Path Planning for UAVs,” in Proc. International Conference on Unmanned Aircraft Systems, Athens, Greece, 2020, pp. 790-798.
[3] J. Williams, S. Jiang, M. O’Brien, G. Wagner, E. Hernandez, M. Cox, A. Pitt, R. Arkin, and N. Hudson, “Online 3D Frontier-Based UGV and UAV Exploration Using Direct Point Cloud Visibility,” in Proc. IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems, Karlsruhe, Germany, 2020, pp. 263-270.
[4] P. Chen, “Cotton Leaf Area Index Estimation Using Unmanned Aerial Vehicle Multi-Spectral Images,” in Proc. IEEE Geoscience and Remote Sensing Society, Yokohama, Japan, 2019, pp. 6251-6254.
[5] I. Ahmed, M. Eramian, I. Ovsyannikov, W. van der Kamp, K. Nielsen, H. Sudhakar Duddu, A. Rumali, S. Shirtliffe, and K. Bett, “Automatic Detection and Segmentation of Lentil Crop Breeding Plots From Multi-Spectral Images Captured by UAV-Mounted Camera,” in Proc. IEEE Winter Conference on Applications of Computer Vision, Waikoloa, HI, USA, 2019, pp. 1673-1681.
[6] M. Hota, S. Rao B, and U. Kumar, “Power Lines Detection and Segmentation In Multi-Spectral Uav Images Using Convolutional Neural Network,” in Proc. IEEE India Geoscience and Remote Sensing Symposium, Ahmedabad, India, 2020, pp. 154-157.
[7] D. B. Abeywickrama, C. Cirstea, and S. D. Ramchurn, “Model Checking Human-Agent Collectives for Responsible AI,” in Proc. IEEE International Conference on Robot and Human Interactive Communication, New Delhi, India, 2019, pp. 1-8.
[8] W. Lina and J. Ding, “Behavior detection method of OpenPose combined with Yolo network,” in Proc. International Conference on Communications, Information System and Computer Engineering, Kuala Lumpur, Malaysia, 2020, pp. 326-330.
[9] Z. Cao, G. Hidalgo, T. Simon, S. -E. Wei and Y. Sheikh, “OpenPose: Realtime Multi-Person 2D Pose Estimation Using Part Affinity Fields,” IEEE Transactions on Pattern Analysis and Machine Intelligence., vol. 43, no. 1, pp. 172-186, Jan. 2021.
[10] S. Qiao, Y. Wang, and J. Li, “Real-time human gesture grading based on OpenPose,” in Proc. International Congress on Image and Signal Processing, BioMedical Engineering and Informatics, Shanghai, China, 2017, pp. 1-6.
[11] E. D’Antonio, J. Taborri, E. Palermo, S. Rossi, and F. Patanè, “A markerless system for gait analysis based on OpenPose library,” in Proc. IEEE International Instrumentation and Measurement Technology Conference, Dubrovnik, Croatia, 2020, pp. 1-6.
[12] Y. Hirasawa, N. Gotoda, R. Kanda, K. Hirata, and R. Akagi, “Promotion System for Home-Based Squat Training Using OpenPose,” in Proc. IEEE International Conference on Teaching, Assessment, and Learning for Engineering, Takamatsu, Japan, 2020, pp. 984-986.
[13] C. Huang and M. H. Nguyen, “Robust 3D Skeleton Tracking based on OpenPose and a Probabilistic Tracking Framework,” in Proc. IEEE International Conference on Systems, Man and Cybernetics, Bari, Italy, 2019, pp. 4107-4112.
[14] B. Rao, B. Mulloth, and A. J. Harrison, “Integrating AI Capabilities into Existing Technology Platforms: Drones as a Case in Point,” in Proc. Portland International Conference on Management of Engineering and Technology, Portland, OR, USA, 2019, pp. 1-6.
[15] M. Yousef, F. Iqbal, and M. Hussain, “Drone Forensics: A Detailed Analysis of Emerging DJI Models,” in Proc. International Conference on Information and Communication Systems, Irbid, Jordan, 2020, pp. 66-71.
[16] J. K. W. Lan and F. K. W. Lee, “Drone Forensics: A Case Study on DJI Mavic Air 2,” in Proc. International Conference on Advanced Communication Technology, PyeongChang, Korea (South), 2021, pp. 291-296.
[17] D. A. Hamdi, F. Iqbal, S. Alam, A. Kazim, and Á. MacDermott, “Drone Forensics: A Case Study on DJI Phantom 4,” in Proc. International Conference on Computer Systems and Applications, Abu Dhabi, United Arab Emirates, 2019, pp. 1-6.
[18] V. M. Respall, S. Sellami, and I. Afanasyev, “Implementation of Autonomous Visual Detection, Tracking and Landing for AR.Drone 2.0 Quadcopter,” in Proc. International Conference on Developments in eSystems Engineering, Kazan, Russia, 2019, pp. 477-482.
[19] G. E. Setyawan, W. Kurniawan, and A. C. L. Gaol, “Linear Quadratic Regulator Controller (LQR) for AR. Drone's Safe Landing,” in Proc. International Conference on Sustainable Information Engineering and Technology, Lombok, Indonesia, 2019, pp. 228-233.
[20] I. Astaburuaga, A. Lombardi, B. La Torre, C. Hughes, and S. Sengupta, “Vulnerability Analysis of AR.Drone 2.0, an Embedded Linux System,” in Proc. Annual Computing and Communication Workshop and Conference, Las Vegas, NV, USA, 2019, pp. 666-672.
[21] A. Bewley, Z. Ge, L. Ott, F. Ramos, and B. Upcroft, “Simple online and realtime tracking,” in Proc. IEEE International Conference on Image Processing, Phoenix, AZ, USA, 2016, pp. 3464-3468.
[22] N. Wojke, A. Bewley, and D. Paulus, “Simple online and realtime tracking with a deep association metric,” in Proc. IEEE International Conference on Image Processing, Beijing, China, 2017, pp. 3645-3649.
[23] Z. Zheng, J. Zhao, and Y. Li, “Research on Detecting Bearing-Cover Defects Based on Improved YOLOv3,” IEEE Access, vol. 9, pp. 10304-10315, Jan. 2021.
[24] A. Azwan, A. Razak, M. F. M. Jusof, A. N. K. Nasir, and M. A. Ahmad, “A multiobjective simulated Kalman filter optimization algorithm,” in Proc. IEEE International Conference on Applied System Invention, Chiba, Japan, 2018, pp. 23-26.
[25] A. Titoriya and S. Sachdeva, “Breast Cancer Histopathology Image Classification using AlexNet,” in Proc. International Conference on Information Systems and Computer Networks, Mathura, India, 2019, pp. 708-712.
[26] S. B., A. Lesle A., B. Diwakar, K. R., and G. M., “Evaluating Performance of Deep Learning Architectures for Image Classification,” in Proc. International Conference on Communication and Electronics Systems, 2020, pp. 917-922.
[27] G. Cheng, C. Ma, P. Zhou, X. Yao and J. Han, "Scene classification of high-resolution remote sensing images using convolutional neural networks," in Proc. International Geoscience and Remote Sensing Symposium, 2016, pp. 767-770.