Omnidirectional Vision Based Topological Navigation
Citations
Visual-inertial navigation, mapping and localization: A scalable real-time causal approach
Large-Scale 6-DOF SLAM With Stereo-in-Hand
Visual teach and repeat for long-range rover autonomy
Vast-scale Outdoor Navigation Using Adaptive Relative Bundle Adjustment
Vision-based topological mapping and localization methods
References
Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography
A note on two problems in connexion with graphs
Object recognition from local scale-invariant features
A mathematical theory of evidence
Fundamentals of digital image processing
Related Papers (5)
Frequently Asked Questions (14)
Q2. How do the authors include colour information in the descriptor vector?
To include colour information in the descriptor vector, the authors compute the colour invariants, based on generalised colour moments (equation 1), over the column segment.
Q3. Why do the authors use the DCT instead of Fourier?
In their method, because it is computationally less intensive and gives real output values, the authors choose to use the seven first coefficients of the discrete cosine transform (DCT), instead of Fourier.
Q4. How many correct matches are found without including erroneous ones?
Using their rotation reduced and colour enhanced algorithm, the authors see that up to 25 correct matches are found without including erroneous ones.
Q5. Why is the robotic wheelchair more economically feasible?
Because of the increased independence of the users the cost of personal helpers is reduced, making the robotic wheelchair even more economically feasible.
Q6. What is the feature to use to characterise the intensity profile along the column segment?
To characterise the intensity profile along the column segment, the best features to use are those obtained through the Karhunen-Lòeve transform (PCA).
Q7. What are the main assumptions that are used to determine the homing of landmarks?
They conclude that every visual homing method based solely on bearing angles of landmarks like this one, inevitably depends on basic assumptions such as equal landmark distances, isotropic landmark distribution or the availability of an external compass reference.
Q8. What is the popular approach to the topological map building problem?
Very popular are various probabilistic approaches of the topological map building problem. [40] for instance use Bayesian inference to find the topological structure that explains best a set of panoramic observations, while [45] fit hidden Markov models to the data.
Q9. What is the way to characterise the column segments?
The authors characterise the extracted column segments with a descriptor that holds information about colour and intensity properties of the segment.
Q10. What is the key technique used to extract local regions in images?
These techniques extract local regions in each image, and describe these regions with a vector of measures which are invariant to image deformations and illumination changes.
Q11. What is the way to find the new feature position in a small search space?
In the image sequence, visual features move only a little from one image to the next, which enables to find the new feature position in a small search space.
Q12. How many times did the same experiment show an average homing accuracy?
Repeated similar experiments showed an average homing accuracy of 11 cm, with a standard deviation of 5 cm, after a homing distance of around 3 m.
Q13. How does the user of the system give the instruction to go towards a certain goal?
How the user of the system, for instance the wheelchair patient, gives the instruction to go towards a certain goal is highly dependent on the situation.
Q14. What is the way to use a laser scanner?
Time-of-flight laser scanners are widely applicable, but are expensive and voluminous, even when the scanning field is restricted to a horizontal plane.