Evaluation of Image Matching Methods for Visual Odometry on UAVs

Gašper Spagnolo; Luka Čehovin Zajc; Matej Dobrevski

cs.CVAug 19, 2026

Evaluation of Image Matching Methods for Visual Odometry on UAVs

Authors: Gašper Spagnolo, Luka Čehovin Zajc, Matej Dobrevski

Organizations: Faculty of Computer and Information Science, University of Ljubljana, Večna pot 113, Ljubljana, Slovenia.

Abstract

Unmanned aerial vehicles (UAVs) are becoming a powerful tool for many environmental monitoring and transport applications. Yet, their reliance on Global Navigation Satellite System (GNSS) technology for navigation makes them susceptible to catastrophic failures in scenarios where the positioning signal is unavailable or disrupted. This work explores Visual Odometry (VO) as a crucial navigation component. Recently, numerous deep-learning-based methods for image matching have been proposed that are yet to be implemented in a fully-fledged VO system. In this paper, we evaluate recent state-of-the-art image matching methods for the task of VO for UAV position tracking, with a downwards-facing camera, on our synthetic dataset, and find that while the best results are generated by the recent RoMa matcher, SIFT features can outperform some recent state-of-the-art.

Figures & tables

Figure 1: Visual odometry for UAV position estimation. Overlapping images from the drone are matched, and a relative camera transformation is estimated based on matches. Finally, local transformations between images are aggregated.

Figure 2: Overview of Ljubljana sequence in the dataset. The sequences contain images of diverse urban, suburban, agricultural, and forestry scenery.

	Ljubljana		Graz		Venice
Method	Min.	Max.	Min.	Max.	Min.	Max.
SIFT+LRT	889	24K	1186	25K	88	32K
SP+SG	8	1980	2	1825	45	1866
SP+LG	0	5641	1	4738	124	4935
RoMa	10K	10K	10K	10K	10K	10K
MASt3R	1257	2304	1300	2304	885	2278

Table 1: Number of matched keypoints.

	Ljubljana		Graz		Venice
Method	$<8$	$<100$	$<8$	$<100$	$<8$	$<100$
SIFT+LRT	0	0	0	0	0	1
SP+SG	0	2	2	3	0	24
SP+LG	2	2	3	3	0	0
RoMa	0	0	0	0	0	0
MASt3R	0	0	0	0	0	0

Table 2: Failures, as the number of frames where a method found less than 8 or 100 matches.

	Ljubljana		Graz		Venice
Method	Ess.	Hom.	Ess.	Hom.	Ess.	Hom.
SIFT+LRT	0.01	0.01	0.10	0.09	0.03	0.03
SP+SG	0.24	0.33	0.20	0.49	0.22	0.07
SP+LG	0.04	0.04	0.11	0.36	0.10	0.06
RoMa	0.01	0.01	0.10	0.09	0.02	0.01
MASt3R	0.09	0.02	0.20	0.10	0.14	0.03

Table 3: Average absolute error in the estimated direction of the translation vector [degrees].

	Ljubljana		Graz		Venice
Method	Ess.	Hom.	Ess.	Hom.	Ess.	Hom.
SIFT+LRT	0.03	0.05	0.03	0.03	0.09	0.04
SP+SG	0.71	0.25	0.66	0.24	1.26	0.10
SP+LG	0.22	0.12	0.15	0.11	0.64	0.10
RoMa	0.03	0.04	0.06	0.03	0.07	0.03
MASt3R	1.04	0.11	0.77	0.06	1.16	0.07

Table 4: Average absolute error in the estimated rotation about the camera Z axis [degrees].

Panel of Figure 3 — Figure 3: Visualizations of the estimated trajectories, using the estimated essential matrix to calculate the rotation and the translation. Red - small number of matched keypoints, blue - large number (normalized per sequence).

Figure 4: A comparison of the matched keypoints by different methods on sequential frames with large rotation from the Graz sequence. SIFT+LRT, RoMa and MASt3R have no problems finding matches. Since RoMa and MASt3R match all (matchable) pixels, the keypoints cover a rectangular region. SuperPoint, SuperGlue and LightGlue are sensitive to rotation, and produce relatively few matches.

Panel of Figure 5 — Figure 5: Visualizations of the estimated trajectories, using the estimated homography to calculate the rotation and the translation. Red - small number of matched keypoints, blue - large number (normalized per sequence).

Visual Odometry

Visual odometry estimates camera or robot motion from successive images, enabling navigation and mapping when external positioning is unavailable.

See all 55 papers

Robust Visual SLAM for UAV Navigation in GPS-Denied and Degraded Environments: A Multi-Paradigm Evaluation and Deployment Study
May 5, 2026Prasoon Kumar, Akshay Deepak, Sandeep KumarUAV Navigation GNSS-Denied Localization
National Institute of Technology Patna, Bihar, India · Central Research Laboratory, Bharat Electronics Ltd., Ghaziabad, Uttar Pradesh, India
DAPEVO: Deep Adaptive Patch Frame-Event Visual Odometry
Sep 25, 2026Luca Gandolfi, Simone Nascivera, Roberto Pellerito +3Visual Odometry RGB-Event Sensor Fusion
Robotics and Perception Group, University of Zurich, Switzerland · Bocconi University, Italy
ROFT-VINS: Robust Feature Tracking-based Visual-Inertial State Estimation for Harsh Environment
Mar 19, 2026Sanghyun Park, Soohee HanVisual-Inertial Odometry Visual Object Tracking
Department of Convergence IT Engineering, POSTECH, Pohang, 37673, Korea · Department of Electrical Engineering and of Convergence IT Engineering, POSTECH, Pohang, 37673, Korea

May 5, 2026J/K move · Enter open · S save

Robust Visual SLAM for UAV Navigation in GPS-Denied and Degraded Environments: A Multi-Paradigm Evaluation and Deployment Study

Prasoon Kumar, Akshay Deepak, Sandeep Kumar

National Institute of Technology Patna, Bihar, India · Central Research Laboratory, Bharat Electronics Ltd., Ghaziabad, Uttar Pradesh, India

UAV Navigation GNSS-Denied Localization Visual SLAM

Open paper AskPDF arXiv

Reliable localization in GPS-denied, visually degraded environments is critical for autonomous UAV opera- tions. This paper presents a systematic comparative evaluation of five V-SLAM systems ORB-SLAM3, DPVO, DROID-SLAM, DUSt3R, and MASt3R spanning classical, deep learning, recurrent, and Vision Transformer (ViT) paradigms. Experiments are conducted on curated sequences from four public benchmarks (TUM RGB-D, EuRoC MAV, UMA-VI, SubT-MRS) and a custom monocular indoor dataset under five controlled degradation conditions (normal, low light, dust haze, motion blur, and combined), with sub-millimeter Vicon ground truth. Results show that ORB-SLAM3 fails critically under severe degradation (62.4% overall TSR; 0% under dense haze), while learning-based methods remain robust: MASt3R achieves the lowest degraded ATE (0.027 m) and DUSt3R the highest tracking success (96.5%). DPVO offers the best efficiency robustness trade-off (18.6 FPS, 3.1 GB GPU memory, 86.1% TSR), making it the preferred choice for memory-constrained embedded platforms. Embedded deployment analysis across NVIDIA Jetson platforms provides actionable guidelines for SLAM selection under SWaP-constrained UAV scenarios.

Abstract

Visual Odometry

Explore similar work

Robust Visual SLAM for UAV Navigation in GPS-Denied and Degraded Environments: A Multi-Paradigm Evaluation and Deployment Study

DAPEVO: Deep Adaptive Patch Frame-Event Visual Odometry

ROFT-VINS: Robust Feature Tracking-based Visual-Inertial State Estimation for Harsh Environment