Very High Resolution Land Cover Map** of Urban Areas at Global Scale with Convolutional Neural Networks
Authors:
Thomas Tilak,
Arnaud Braun,
David Chandler,
Nicolas David,
Sylvain Galopin,
Amélie Lombard,
Michaël Michaud,
Camille Parisel,
Matthieu Porte,
Marjorie Robert
Abstract:
This paper describes a methodology to produce a 7-classes land cover map of urban areas from very high resolution images and limited noisy labeled data. The objective is to make a segmentation map of a large area (a french department) with the following classes: asphalt, bare soil, building, grassland, mineral material (permeable artificialized areas), forest and water from 20cm aerial images and…
▽ More
This paper describes a methodology to produce a 7-classes land cover map of urban areas from very high resolution images and limited noisy labeled data. The objective is to make a segmentation map of a large area (a french department) with the following classes: asphalt, bare soil, building, grassland, mineral material (permeable artificialized areas), forest and water from 20cm aerial images and Digital Height Model. We created a training dataset on a few areas of interest aggregating databases, semi-automatic classification, and manual annotation to get a complete ground truth in each class. A comparative study of different encoder-decoder architectures (U-Net, U-Net with Resnet encoders, Deeplab v3+) is presented with different loss functions. The final product is a highly valuable land cover map computed from model predictions stitched together, binarized, and refined before vectorization.
△ Less
Submitted 12 May, 2020;
originally announced May 2020.
