Cecilia Dinesh,
Anas Fodkar,
Amey Shelar,
Tushar Nandy,
Nilamabri G Narkar,
- Student, Department of Computer Engineering, Xavier Institute of Engineering, Mahim, Mumbai, Maharshtra, India
- Student, Department of Computer Engineering, Xavier Institute of Engineering, Mahim, Mumbai, Maharshtra, India
- Student, Department of Computer Engineering, Xavier Institute of Engineering, Mahim, Mumbai, Maharshtra, India
- Student, Department of Computer Engineering, Xavier Institute of Engineering, Mahim, Mumbai, Maharshtra, India
- Faculty Guide, Department of Computer Engineering, Xavier Institute of Engineering, Mahim, Mumbai, Maharshtra, India
Abstract
Traditional unmanned aerial vehicle (UAV) systems primarily rely on Global Positioning System (GPS)-based navigation and conventional computer vision techniques. However, these approaches face significant challenges in GPS-denied environments, such as dense urban areas, indoor spaces, and disaster-affected regions where GPS signals may be weak, unavailable, or unreliable. In addition, limitations in real-time processing, object detection accuracy, and environmental perception can reduce UAV effectiveness in dynamic and unpredictable situations. This paper presents JATAYU, a smart autonomous UAV system designed to operate reliably without GPS by integrating deep learning-based human detection with stereo vision for real-time depth estimation. The proposed system employs a dual-camera configuration to estimate the distance between the UAV and detected objects accurately. The perception unit continuously processes visual information and communicates with the flight control system, enabling the UAV to make informed navigation decisions and maintain reliable operation in complex environments. The proposed system achieves an average human detection accuracy of 92.3%, demonstrating its effectiveness in identifying people under challenging operating conditions. The system is specifically designed for high-risk applications where affordability, reliability, and real-time performance are essential. By combining deep learning, stereo vision, and autonomous flight control, JATAYU provides robust perception and navigation capabilities while reducing dependence on GPS infrastructure. Furthermore, the system is optimized for embedded hardware, making it suitable for deployment on resource-constrained UAV platforms. Its lightweight and cost-effective design makes JATAYU a versatile solution for applications such as surveillance, disaster management, search-and-rescue missions, and emergency response. Overall, the proposed UAV demonstrates reliable real-time human detection and navigation capabilities in GPS-denied environments.
Keywords: UAV, STEREO vision, deep learning, human detection, autonomous drone
[This article belongs to International Journal on Drones ]
References
- 1. Barr H, Levy D, Rosenfeld A, Maksimov O, Kraus S. Advising agent for supporting human–multi-drone team collaboration. arXiv. 2025;arXiv:2502.17960. doi:10.48550/arXiv.2502.17960.
- Mur-Artal R, Montiel JMM, Tardós JD. ORB-SLAM: A versatile and accurate monocular SLAM system. IEEE Transactions on Robotics. 2015;31(5):1147–1163. doi:10.1109/TRO.2015.2463671.
- Engel J, Schöps T, Cremers D. LSD-SLAM: Large-scale direct monocular SLAM. In: European Conference on Computer Vision (ECCV); 2014. p. 834–849.
- Geiger A, Lenz P, Urtasun R. Are we ready for autonomous driving? The KITTI vision benchmark suite. In: IEEE Conference on Computer Vision and Pattern Recognition (CVPR); 2012. p. 3354–3361.
- Wang J, Gao X, Shen S. Visual odometry and mapping for aerial robots. Robotics and Autonomous Systems. 2017;87:118–134. doi:10.1016/j.robot.2016.10.015.
- Huynh-The T, Pham Q-V, Nguyen T-V, da Costa DB, Kim D-S. RF-UAVNet: High-performance convolutional network for RF-based drone surveillance systems. IEEE Access. 2022;10:1618–1630. doi:10.1109/ACCESS.2021.3139555.
- Liu Z, Ding K, Xu Q, Song Y, Yuan X, Li Y. Scene images and text information-based object location of robot grasping. IET Cyber-Systems and Robotics. 2022. doi:10.1049/csy2.12049.
- Achtelik M, Weiss S, Siegwart R. Stereo vision and UAV navigation in GPS-denied environments. In: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS); 2011. p. 1–6.
- Scharstein D, Szeliski R. A taxonomy and evaluation of dense two-frame stereo correspondence algorithms. International Journal of Computer Vision. 2002;47(1–3):7–42. doi:10.1023/A:1014573219977.
International Journal on Drones
| Volume | 02 | |
| Issue | 02 | |
| Received | 26/05/2026 | |
| Accepted | 04/08/2026 | |
| Published | 20/08/2026 | |
| Publication Time | 86 Days |