Visual attention predictive model of built colonial heritage based on visual behaviour and subjective evaluation

Although physiological measurements, subjective evaluation and other methods have been applied to visual attention research, architects still lack a systematic quantitative classification method when assessing the visual attention to built colonial heritage. This study aimed to investigate the relationship between people's visual behaviour and subjective evaluation when observing built colonial heritage and to construct a prediction model based on eye-movement metrics and subjective evaluation indicators to distinguish the visual attention to built colonial heritage. This study recorded data from 54 participants while observing five scenes of built colonial heritage, and the results showed that participants had different visual behaviours and subjective evaluations when viewing built colonial heritage in different scenes. And visual attention to built colonial heritage was negatively correlated with the average saccades peak velocity and average saccades amplitude and positively correlated with the average pupil diameter; visual attention was correlated with 12 subjective evaluation indicators. The eye-movement metrics and subjective evaluation indicators with correlation to visual attention were used as input variables to construct a prediction model of visual attention to built colonial heritage based on the BP neural network. Different built colonial heritage's low, middle and high visual attention were identified with high accuracy (74.46%). This quantitative method can help architects to measure the visual attention to built colonial heritage to develop conservation and renewal strategies.