A Machine Learning Classification to Identify Houses Suitable for Electric Vehicle Charging from Remotely Sensed Imagery

James Flynn, Giannetti

Over the past decade, Deep Convolutional Neural Networks (DCNN’s) have emerged as a powerful tool for the classification of remotely sensed imagery. In this multi-disciplinary paper, we demonstrate a novel application of machine learning in the field of remote sensing by developing a workflow to survey urban areas for residential properties suitable for electric vehicle charging. A fine-tune transfer learning approach is presented as a new method for analysing remotely sensed image data. A unique dataset comprised of Google Street View images sourced from multiple UK towns and cities is used to train can compare three neural networks and represents the first attempt to classify residen- tial driveways from streetscape imagery using machine learning. When testing the full workflow on two urban areas the full system achieves accuracies of 87.2% and 89.3% respectively. This proof of concept demonstrates a promising new application of deep learning in the field of remote sensing, geospatial analysis, and urban planning.

Published Date: 2021-09-08;