Large-Scale Solar Panel Mapping from Aerial Imagery 

Up-to-date maps of installed solar photovoltaic panels are a critical input for policy and financial assessment of solar distributed generation. Such maps are not available for large areas. We train a deep convolutional network with 12 square kilometers of training data that are manually labeled, which reliably maps solar panels in imagery covering 200 square kilometers in two cities. We use 0.3 meter resolution aerial images with RGB bands. Each image below is of size 40,000 * 30,000 pixels. Red polygons in images below are detected solar panels.

Boston, MA

San Francisco, CA

Network output for an area in San Francisco (detections are marked in transparent red)

Please refer to the following paper for details.

Jiangye Yuan, Hsiu-Han Yang, Olufemi Omitaomu, and Budhendra Bhaduri, Large-Scale Solar Panel Mapping from Aerial Images Using Deep Convolutional Networks, IEEE International Workshop on Big Spatial Data , 2016. [pdf]