EXPERIMENTAL EVALUATION OF USER INTERFACES FOR VISUAL INDOOR NAVIGATION

Andreas Möller ✽, Matthias Kranz ❖, Stefan Diewald ✽, Luis Roalter ✽, Robert Huitl ✽,

Tobias Stockinger ❖, Marion Koelle ❖, Patrick Lindemann ❖ !

✽ Technische Universität München, Germany ❖ Universität Passau, Germany

VISION-BASED NAVIGATION

Send query image to server

Database of images with known position

Return position and orientation of most similar

reference image

■ Advantages □ No infrastructure □ Centimeter-level accuracy (Schroth et al. 2011)

■ But: query images impact localization quality □ Image distinctiveness □ Motion blur □ Pose

MOTIVATION

✘✔✘✘

■ Advantages □ No infrastructure □ Centimeter-level accuracy (Schroth et al. 2011)

■ But: query images impact localization quality □ Image distinctiveness □ Motion blur □ Pose

MOTIVATION

✘✔✘✘□ Traditional user interfaces usually require a high degree of accuracy, e.g. maps (Kray et al. 2003) or Augmented Reality (Liu et al. 2008)

■ User interface concept for visual localization that copes with inaccuracy, and UI elements to improve query images

■ First experimental evaluation

MAIN CONTRIBUTION

Augmented Reality (AR)

Virtual Reality(VR)

USER STUDY

3 Experiments

Navigation Time

Distraction

AR/VR

METHOD !12 Participants

!Wizard of Oz

Accuracy Perception Preferences

EffectivenessUI

ELEMENTS

RESEARCH QUESTIONS

EXPERIMENT 1: VR/AR COMPARISON■ Task: Navigate in building with AR and VR mode ■ Simulation of varying localization accuracy ■ Hypotheses: VR is faster, seems more accurate

and is more popular

AR VR

Live video Panorama

EXPERIMENT 1: VR/AR COMPARISON■ Task: Navigate in building with AR and VR mode ■ Simulation of varying localization accuracy ■ Hypotheses: VR is faster, seems more accurate

and is more popular

AR VR

Live video Panorama

■ AR: users were slower in error conditions ■ VR: no differences between conditions

m:ssuntil destination

(average)

EXPERIMENT 1: VR/AR COMPARISON

2:393:04 AR

VR

Navigation time

EXPERIMENT 1: VR/AR COMPARISONGuidance quality

3 VR

1 AR -3 = worst

3 = best

position error

2 VR

1 AR

orientation error

VR 2.5

AR 2

no errors

EXPERIMENT 1: VR/AR COMPARISONUser preferences

VR 50%

AR 33%

Undecided 17%

„Carrying the phone was convenient“

2 VR

0 AR -3 = strongly disagree

3 = strongly agree

■ Hypothesis: indicator increases average number of features visible in the image

■ 3 random appearances of indicator during navigation task

EXPERIMENT 2: FEATURE INDICATOR

EXPERIMENT 2: FEATURE INDICATOR

Features per frame(average)

% of frameswith >150 features

42

8.1%

101

20.7%

Effectiveness

without FI

with FI

EXPERIMENT 3: OBJECT HIGHLIGHTING■ Hypothesis: Soft border leads to less distraction

than Frame ■ Evaluation on Likert Scale

EXPERIMENT 3: OBJECT HIGHLIGHTING

Soft Border

1„Aroused my attention“

„Distracted during navigation task“

!Frame

3

1 Frame

-1 Soft

Border

-3 = strongly disagree3 = strongly agree

AR

FI

DISCUSSION■ VR as primary visualization ■ AR and indicators improve localization ■ Automatic switching between VR and AR ■ Future Work: live system, env. transformations

AR VR

+

accurate inaccurate

after (re-)localization navigation location estimate

too unreliable

Location Estimate

SUMMARY■ Novel UI for visual localization ■ Faster & more popular than AR ■ Increases perceived and

system localization accuracy

Contact: andreas.moeller@tum.de www.eislab.net

Contact: andreas.moeller@tum.de www.eislab.net

REFERENCES■ Slide 2: Measurement image: MS Office Clipart ■ Slide 4: Paper References:

Schroth, Georg, et al. "Mobile visual location recognition." Signal Processing Magazine, IEEE 28.4 (2011): 77-89. Kray, Chris, et al. "Presenting route instructions on mobile devices." Proc. of the 8th Intl. Conf. on Intelligent User Interfaces (IUI), ACM (2003), 117–124.Liu, A., et al. "Indoor wayfinding: Developing a functional interface for individuals with cognitive impairments." Disability & Rehabilitation: Assistive Technology 3, 1-2 (2008): 69–81. !!

■ All other photos and graphics: own material by Andreas Mölleror TU München or Universität Passau

■ Please cite this work as follows: Andreas Möller, Matthias Kranz, Stefan Diewald, Luis Roalter, Robert Huitl, Tobias Stockinger, Marion Koelle, and Patrick A. Lindemann. 2014. Experimental evaluation of user interfaces for visual indoor navigation. In Proceedings of the 32nd annual ACM conference on Human factors in computing systems (CHI '14). ACM, New York, NY, USA, 3607-3616. !■ If you use BibTex: @inproceedings{Moller:2014:EEU:2611222.2557003,! author = {M\"{o}ller, Andreas and Kranz, Matthias and Diewald, Stefan and Roalter, Luis and Huitl, Robert and Stockinger, Tobias and Koelle, Marion and Lindemann, Patrick A.},! title = {Experimental Evaluation of User Interfaces for Visual Indoor Navigation},! booktitle = {Proceedings of the 32Nd Annual ACM Conference on Human Factors in Computing Systems},! series = {CHI '14},! year = {2014},! isbn = {978-1-4503-2473-1},! location = {Toronto, Ontario, Canada},! pages = {3607--3616},! numpages = {10},! publisher = {ACM},! address = {New York, NY, USA},!}