Automated Sustainability Assessment of Natural World in Infrastructure Projects: A Machine Learning Approach

Infrastructures are problematic in nature. Considering the size, lifespan and importance, infrastructures greatly impact their surrounding environment and nearby communities. Hence, the concept of sustainability has turned into an indispensable part of evaluating infrastructure’s performance. However, today’s infrastructures overall grade is impressively low, even in highly developed countries. This paper focuses on bringing computational tools (e.g., machine learning and machine vision) to environmental sustainability assessment of infrastructures in order to automate assessing the extent of sustainability based on Envision rating system for sustainable infrastructure. The machine learning conceptual model that developed in this study, concentrates on Natural World category of Envision rating system, and uses aerial images from Google Earth’s database as inputs. After qualification and classification of images, three proposed machine learning models (Siting, Land and Water, and Biodiversity) assess the performance of infrastructures upon given criteria. Then, an overall score based on the weights of criteria grants to the infrastructure’s project. Considering the concept of automation, although time-consuming, error prone processes and human dependency of current assessing methods turns into accelerated, accurate and automated ones, but the fact that reaching to a sufficient level of autonomy is highly challenging should not be neglected.

کلیدواژه ها:

Automation ، Environmental Assessment ، Sustainability ، Infrastructure ، Envision ، Machine Learning ، Machine Vision