City safety perception model based on street images using machine learning and image processing techniques

Abstract

Abstract: Safety perception measurement has been a subject of interest in many cities of the world. This importance is due to its social relevance, and to its influence on many of the economic activities that take place in a city. The methods and procedures presented in this work make use of image processing and machine learning techniques to model citizen's safety perception using visual information of city street images. Even though people safety perception is a subjective topic, results show that it is possible to find out common patterns given a limited geographical and sociocultural context, and based on people judgment of the visual appearance of a street image. Technics based on Support Vector Machines and Neural Networks are presented. The exposed models along with ranking methods are used to predict how safe a given street of Bogotá City is perceived. Results suggest that the obtained models can detect different patterns, where a common visual feature of a street or an urban environment, is linked to an activity or street condition that has a significant influence on their associated safety perception. This feature makes the proposed models an alternative tool for decision makers concerning urban planning, safety, and public health policies, as well as a collective memory associated with a particular urban environment.