Human perception of location and space forms the basis upon which the interaction with location-based services (LBS) takes place. Our work aims to develop a shared awareness and common understanding of location and space,between machines and their users by building upon research into the numerical representation of the visual perception of space. Different structures in buildings like rooms, hallways and doorways form different, corresponding patterns in these representations. Thanks to recent advances in the field of deep learning with neural networks, it now seems possible to explore the idea of automatically learning these recurring structures. This article presents a complete framework: starting from the collection of isovist measures along geospatial trajectories on indoor floor plans,over statistical data analysis, the unsupervised extraction of meaningful structure, up to the training of models that generalize to different environments. We show that isovist measures do reflect the recurring structures found in different buildings, that these recurring patterns are encoded in the data in a way that unsupervised machine learning can identify them andthat the identified structures are meaningful as they represent human relatable concepts.Furthermore, we propose to use cluster similarity analysis as a promising concept for quantifying visual perception similarity.