Automatic Solar Panel Detection from Remote Sensing Images Using Deep Convolutional Networks

The invention relates to image analysis and more specifically to a method for identifying objects within an image such as solar panels. Remote sensing imagery provides a complete and cost-effective data source for mapping semantic objects. However, there is no automatic method that works reliably on large images with complex scenes, because of large variations of object appearances. The developed technique is capable of automatically identifying installed solar panels from aerial view images. Experiments show that the technique produces accurate results on large, real-world datasets. This technique can be applied to identifying other objects as well.

---

---