A Crowdsourced Model of Landscape Preference

摘要: The advent of new sources spatial data and associated information (e.g. Volunteered Geographic Information (VGI)) allows us to explore non-expert conceptualisations space, where the number participants extent coverage encompassed can be much greater than is available through traditional empirical approaches. In this paper we such prism landscape preference or scenicness. VGI in form photographs particularly suited task, volume images has been suggested as a simple proxy for preference. We propose another approach, which models aesthetics based on descriptions some 220000 collected large project UK, more 1.5 million votes related perceived scenicness these crowdsourcing project. use image build features supervised machine learning algorithm. Features include most frequent uni- bigrams, adjectives, presence verbs perception adjectives from "Landscape Adjective Checklist". Our results not only qualitative relating terms but model our predict 52% variation scenicness, comparable typical using useful are 800 unigrams, Checklist" weighting term.