Improved Deep Fully Convolutional Network with Superpixel-Based Conditional Random Fields for Building Extraction

Fully convolutional network (FCN) modeling is a recently developed technique that is capable of significantly enhancing building extraction accuracy; it is an important branch of deep learning and uses advanced state-of-the-art techniques, especially with regard to building segmentation. In this paper, we present an enhanced deep convolutional encoder-decoder (DCED) network that has been customized for building extraction through the application of superpixel-based conditional random fields (SCRFs). The improved DCED network, with symmetrical dense-shortcut connection structures, is employed to establish the encoders for automatic extraction of building features. Our network's encoders and decoders are also symmetrical. To further reduce the occurrence of falsely segmented buildings, and to sharpen the buildings' boundaries, an SCRF is added to the end of the improved DCED architecture. Experimental results indicate that the proposed approach exhibits competitive quantitative and qualitative performance, effectively alleviating the salt-and-pepper phenomenon and retaining the edge structures of buildings. Compared with other state-of-the-art methods, our method demonstrably achieves the optimal final accuracies.