Segmentation of Roads in Satellite Images using specially modified U-Net CNNs
This work addresses the problem of automated road mapping from satellite imagery for urban planning and navigation applications, representing an incremental improvement over existing methods.
The paper tackles road segmentation in satellite images by developing a classifier using specially modified U-Net CNNs with sliding windows and data augmentation, achieving state-of-the-art performance with improved mean F-score metrics.
The image classification problem has been deeply investigated by the research community, with computer vision algorithms and with the help of Neural Networks. The aim of this paper is to build an image classifier for satellite images of urban scenes that identifies the portions of the images in which a road is located, separating these portions from the rest. Unlike conventional computer vision algorithms, convolutional neural networks (CNNs) provide accurate and reliable results on this task. Our novel approach uses a sliding window to extract patches out of the whole image, data augmentation for generating more training/testing data and lastly a series of specially modified U-Net CNNs. This proposed technique outperforms all other baselines tested in terms of mean F-score metric.