Mapping Solution for Programmable Intelligent Spaces Based on Computer Vision

This paper presents a mapping approach to be applied in intelligent spaces based on computer vision. By using cameras installed in the environment, the three-dimensional pose of mobile robots can be recovered through visual odometry, that provides a global and more accurate localization without cumulative errors. This enables the creation of high-quality maps that better represent the environment, ensuring invariance concerning the robot's reference frame and effectively addressing the kidnapped robot problem without compromising the map's integrity. The result is a robust mapping system capable of supporting more complex navigation tasks.