Visualization of urban roadway surface temperature by applying deep learning to infrared images from mobile measurements

Abstract

Urban heat islands (UHIs) have been worsening, and Tokyo, Japan, is among the worst globally. The urban thermal environment requires measurement to formulate effective countermeasures. This study proposes a method for detecting roadways from infrared images of captured by a moving automobile and using deep learning to extract roadway surface temperatures from the detected roadways. Additionally, a roadway surface temperature map of Tokyo was created from 17,000 infrared images covering a route of 37 km and was then used to identify hotter and cooler areas of the city in order to validate the proposed methodology. The surface temperatures were high on wide roadways with a high sky view factor, but were lower and less variable in street canyons. Roadways perpendicular to the solar azimuth had lower temperatures due to shading by buildings. The accuracy of deep learning to detect roadway was imperfect, but a significant improvement in analytical efficiency was achieved. The method could also be applied to three-dimensional evaluation of urban surface temperatures by extracting surface temperatures for various city components, such as buildings and vegetation.