Prediction of riverside greenway landscape aesthetic quality of urban canalized rivers using environmental modeling

Abstract

Canalized rivers are often the product of urbanization. They are an important component of urban blue-green infrastructure whose visual landscape quality affects the aesthetics of a city. Existing research on urban waterways pays little attention to urban canalized rivers and lacks a quantitative interpretation of their visual quality. This study aims to simulate human visual perception and establish quantifiable landscape features for urban riverside greenways. Using a deep learning model to build connections between landscape features and the aesthetic preference of riverside spaces, we establish the best predictive model for the aesthetic quality of urban canalized riverside greenways. We use panoramic photos of rivers in central Beijing and image semantic segmentation technology to extract physical features of landscape composition. We use an ordinary least squares (OLS) regression model and a random forest (RF) model to develop a predictive model for riverside aesthetic quality. The RF model was found to be the most accurate for predicting the aesthetic quality score of the channelized riverside. Trees and bridges are the most important input factors that affect the output score of the RF model. Trees are positively correlated with aesthetic quality, while bridges are negatively correlated. In practice, this model can be used for batch evaluation and aesthetic quality prediction of urban canalized riverside landscapes to help practitioners improve analysis efficiency and guide the management and planning of riverside landscapes.