Eye tracking metrings in perceptiopn and visual attention research for human-infrastructure interaction

To ensure the safety and longevity of infrastructure, on-site inspections are required by federal law [1]. Based on current modalities, there is a gap between the knowledge (experiences, bias, and dynamicity vs preprogrammed and static) that each inspector (human or UAV) possesses. The aim of this work is to use eye-tracking to capture and examine the process of salient feature detection and implicit human perception during the visual inspection (VI) and structural health monitoring (SHM) processes. This study focuses on eye-tracking metric analysis by understanding gaze pattern and pupillary responses to anticipate visual attention of the observer. Eye-tracking data can accurately track what human is looking at, and which part of a structure they are specifically focusing on. In general, human gaze anticipation can be predicted based on fixation count, fixation length, saccadic movement and pupil gradient induced by an eye response. These eye tracking metrics will be useful in learning how a human eye behaves during VI/SHM processes. To this end, we conducted a pilot study for data collection and analysis on crack damage detection to understand human cognition for infrastructure monitoring. We investigated which of the eye tracking metrics in literature are useful for crack damage detection in attempt to better inspect infrastructure and its surrounding environment. By enabling a sense-making UAV through AI support, this method will facilitate information sharing and sense-making during the inspection process.