A Computer Vision System for Attention Mapping in SLAM based 3D Models
This work addresses the need for automated attention studies in real-world environments, offering potential advancements in human factors technologies, though it appears incremental by combining existing methods like SLAM and descriptor matching for a new application.
The paper tackles the problem of pervasive mapping and monitoring of human attention by developing a computer vision system that enables full 3D recovery of gaze pointers and human view frustums directly into an automatically computed 3D model in real-time, using RGB-D SLAM and descriptor matching for 3D modeling and automated annotation of regions of interest.
The study of human factors in the frame of interaction studies has been relevant for usability engi-neering and ergonomics for decades. Today, with the advent of wearable eye-tracking and Google glasses, monitoring of human factors will soon become ubiquitous. This work describes a computer vision system that enables pervasive mapping and monitoring of human attention. The key contribu-tion is that our methodology enables full 3D recovery of the gaze pointer, human view frustum and associated human centred measurements directly into an automatically computed 3D model in real-time. We apply RGB-D SLAM and descriptor matching methodologies for the 3D modelling, locali-zation and fully automated annotation of ROIs (regions of interest) within the acquired 3D model. This innovative methodology will open new avenues for attention studies in real world environments, bringing new potential into automated processing for human factors technologies.