Accuracy potential of visual localization exploiting high-end street-level imagery
Accuracy potential of visual localization exploiting high-end street-level imagery
This research introduces the FHNW Muttenz dataset, a large-scale outdoor resource covering a contiguous 10 km street network mapped in two campaigns with 5 mm co-registration accuracy. The dataset provides high-resolution reference imagery and query sequences from four cameras, enabling systematic evaluation of sparse feature-based visual localization. Results show that with accurately georeferenced scenes and pre-calibrated cameras, median pose errors are 1-5 cm in translation and 0.03-0.1 degrees in rotation, approaching survey-grade GNSS accuracy while providing full orientation information.
The authors present the FHNW Muttenz dataset, a publicly available resource covering a 10 km street network with ground-truth poses accurate to 6-9 mm in position and well below 0.1 degrees in orientation, used to demonstrate that sparse feature-based visual localization can achieve survey-grade accuracy comparable to GNSS positioning under specific test conditions.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
position accuracy
6–9
View original evidence
r cameras in five scenes. The two campaigns are co-registered to each other with an accuracy of 5 mm. Co-registration of the query images to the reference imagery yields ground-truth poses accurate to 6-9 mm in position and well below 0.1{\deg} in orientation. All images are georeferenced with an accuracy of about 2.5 cm. Second, we use this dataset to systematically evaluate sparse feature-based visual localization
Open source S4orientation accuracy
<0.1
View original evidence
r cameras in five scenes. The two campaigns are co-registered to each other with an accuracy of 5 mm. Co-registration of the query images to the reference imagery yields ground-truth poses accurate to 6-9 mm in position and well below 0.1{\deg} in orientation. All images are georeferenced with an accuracy of about 2.5 cm. Second, we use this dataset to systematically evaluate sparse feature-based visual localization
Open source S4georeferencing accuracy
≈2.5
View original evidence
r cameras in five scenes. The two campaigns are co-registered to each other with an accuracy of 5 mm. Co-registration of the query images to the reference imagery yields ground-truth poses accurate to 6-9 mm in position and well below 0.1{\deg} in orientation. All images are georeferenced with an accuracy of about 2.5 cm. Second, we use this dataset to systematically evaluate sparse feature-based visual localization
Open source S4translation error
1–5
View original evidence
with a modular pipeline built from established components, using the high-resolution reference imagery directly as the scene representation. For daytime imagery and pre-calibrated cameras, the median pose errors with respect to the reference imagery are 1-5 cm in translation and 0.03-0.1{\deg} in rotation. Given an accurately georeferenced scene representation, visual localization not only approaches the accuracy of
Open source S5rotation error
0.03–0.1
View original evidence
with a modular pipeline built from established components, using the high-resolution reference imagery directly as the scene representation. For daytime imagery and pre-calibrated cameras, the median pose errors with respect to the reference imagery are 1-5 cm in translation and 0.03-0.1{\deg} in rotation. Given an accurately georeferenced scene representation, visual localization not only approaches the accuracy of
Open source S5
- dataset: Available Artifact link Open source S6
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
What remains unknown
Not established in the collected evidence: Environment, Control, Data origin.
Dataset licensing and commercial-use rights are not established here.
Reported performance applies to the described task. It does not establish general autonomy or deployment readiness.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
r cameras in five scenes. The two campaigns are co-registered to each other with an accuracy of 5 mm. Co-registration of the query images to the reference imagery yields ground-truth poses accurate to 6-9 mm in position and well below 0.1{\deg} in orientation. All images are georeferenced with an accuracy of about 2.5 cm. Second, we use this dataset to systematically evaluate sparse feature-based visual localization
Open source S4
with a modular pipeline built from established components, using the high-resolution reference imagery directly as the scene representation. For daytime imagery and pre-calibrated cameras, the median pose errors with respect to the reference imagery are 1-5 cm in translation and 0.03-0.1{\deg} in rotation. Given an accurately georeferenced scene representation, visual localization not only approaches the accuracy of
Open source S5
survey-grade GNSS positioning but also provides full orientation information. The dataset is publicly available at https://fhnw-muttenz-vl-dataset.github.io/
Open source S6
Implications for data suppliers
RoboSignal interpretation and collection questions, not statements of buyer demand.
- Confirm the required data type and collection setting with the buyer; this source does not establish a complete collection specification.
- Compare the reported units and scope before using these quantities in a budget. Recording hours, sensor-hours and trajectories are different measures.
- Validate demand and acceptance criteria with a buyer before scaling. Publication, popularity and a research result do not establish a purchase commitment.
Source：arXiv Robotics — research abstracts · arxiv.org