Abstract:
Automatic building extraction from aerial and satellite imagery is highly challenging due to extremely large variations of building appearances. Moreover, the detection of homes with complex structures patterns from images taken from aerial and satellite view is challenging. To attack this problem, high resolution images were taken and then downsize the image to small images. Then design a Convolutional network with a final stage that integrates activations from multiple preceding stages for pixel-wise prediction, and introduce the signed distance function of building boundaries as the output representation, which has an enhanced representation power. The resultant images are then upsized back to single image. We leverage abundant building footprint data available from INRIA Project to compile training data. The trained network achieves superior performance on datasets that are significantly larger and more complex than those used in prior work, demonstrating that the proposed method provides a promising and scalable solution for automating this labor-intensive task.
