A vision-based navigation system for Unmanned Aerial Vehicles (UAVs)

Al-Kaff, Abdulla; Armingol, José María; de La Escalera, Arturo · 2019 · Crossref

DOI: 10.3233/ica-190601

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of providing robust, autonomous navigation for Unmanned Aerial Vehicles (UAVs) in both indoor and outdoor environments. The authors identify limitations in existing navigation systems, noting that Global Positioning System (GPS) signals are often unreliable due to signal obstruction or absence indoors, while stereo cameras and RGB-D sensors suffer from baseline constraints or environmental sensitivity. To overcome these issues, the study proposes a vision-based navigation system relying primarily on lightweight, low-power monocular cameras. The system is designed to handle complex tasks in real-time by integrating three core components: Pose Estimation, Navigation Guidance, and Visual Servoing. The methodology centers on a monocular vision algorithm for pose estimation. The process begins with image preprocessing, utilizing histogram equalization and Gamma Correction to normalize illumination and contrast. For feature detection, the authors employ a combination of SIFT and FREAK descriptors, selected for their high accuracy and low computational time compared to alternatives like SURF. Keypoints are matched between consecutive frames using a Brute-Force algorithm, with outliers removed via RANSAC. The system then extracts frame-to-frame and world-to-frame homographies to estimate the UAV’s 3D position and orientation. Navigation guidance involves an efficient strategy for following and swapping waypoints. Finally, visual servoing is implemented using a Fuzzy Logic Controller (FLC) that fuses visual data with inputs from Inertial Measurement Units (IMU) and GPS to stabilize flight and manage uncertainties. The proposed algorithms were validated through real flights in various indoor and outdoor settings, accounting for different visual conditions such as illumination and texture. The pose estimation results were benchmarked against high-precision reference systems, including a VICON motion capture system for indoor tests and Differential GPS (DGPS) for outdoor tests. The experimental results demonstrated that the system achieves significant improvements in accuracy and robustness compared to previous state-of-the-art methods. Specifically, the SIFT-FREAK combination proved effective in maintaining reliable feature tracking, and the homography-based approach successfully localized the UAV in 3D space. The significance of this work lies in its contribution to the field of autonomous UAV navigation, particularly for Micro Aerial Vehicles (MAVs) with strict payload and power constraints. By demonstrating that a monocular camera-based system can provide reliable navigation in GPS-denied environments, the paper offers a viable alternative to heavier or more expensive sensor suites. The integration of fuzzy logic control further enhances the system's ability to handle environmental disturbances, making it suitable for complex civil and academic applications such as rescue operations, inspections, and data collection.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• driverless ads