Visual landmarks are an approach to use camera images to correct the inaccurate position calculated by smartphone sensors. The resulting approach is mostly the play with probabilities, angles, and the match between detected doors and known doors from the floor plans.
However, the accuracy of the traditional computer vision approach for the detection was limited. We are working on an AI method for door detection, enabling accurate positioning with regards to performance.
See more:
- Miroslav Opiela, Doors as Visual Landmarks for Indoor Positioning (IPIN 2025 conference, WCAL paper)
- Lucia Hajduková. Camera-based indoor localization accuracy improvement, 2021, master thesis
- Poster