We propose an end-to-end framework for the dense, pixelwise classification of satellite imagery with convolutional neural networks (CNNs). In our framework, CNNs are directly trained to produce classification maps out of the input images. We first devise a fully convolutional architecture and demonstrate its relevance to the dense classification problem. We then address the issue of imperfect training data through a two-step training approach: CNNs are first initialized by using a large amount of possibly inaccurate reference data, then refined on a small amount of accurately labeled data. To complete our framework we design a multi-scale neuron module that alleviates the common trade-off between recognition and precise localization. A series of experiments show that our networks take into account a large amount of context to provide fine-grained classification maps.

New challenges in remote sensing impose the necessity of designing pixel classification methods that, once trained on a certain dataset, generalize to other areas of the earth. This may include regions where the appearance of the same type of objects is significantly different. In the literature it is common to use a single image and split it into training and test sets to train a classifier and assess its performance, respectively. However, this does not prove the generalization capabilities to other inputs. In this paper, we propose an aerial image labeling dataset that covers a wide range of urban settlement appearances, from different geographic locations. Moreover, the cities included in the test set are different from those of the training set. We also experiment with convolutional neural networks on our dataset.

No abstract

This is a preprint, to read the final version please go to IEEE Geoscience and Remote Sensing Magazine on IEEE XPlore. Airborne and spaceborne hyperspectral imaging systems have advanced in recent years in terms of spectral and spatial resolution, which makes data sets produced by them a valuable source for land-cover classification. The availability of hyperspectral data with fine spatial resolution has revolutionized hyperspectral image classification techniques by taking advantage of both spectral and spatial information in a single classification framework. The ECHO (Extraction and Classification of Homogeneous Objects) classifier, which was proposed in 1976, might be the first spectral-spatial classification approach of its kind in the remote sensing community. Since then and especially in the latest years, increasing attention has been dedicated to developing sophisticated spectral-spatial classification methods. There is now a rich literature on this particular topic in the remote sensing community, composing of several fast-growing branches. In this paper, the latest advances in spectral-spatial classification of hyperspectral data are critically reviewed. More than 25 approaches based on mathematical morphology, Markov random fields, segmentation, sparse representation, and deep learning are addressed with an emphasis on discussing their methodological foundations. Examples of experimental results on three benchmark hyperspectral data sets, including both wellknown long-used data and a recent data set resulting from an international contest, are also presented. Moreover, the utilized training and test sets for the aforementioned data sets as well The work of Pedram Ghamisi is supported by the "High Potential Program" of Helmholtz-Zentrum Dresden-Rossendorf.

We propose a convolutional neural network (CNN) model for remote sensing image classification. Using CNNs provides us with a means of learning contextual features for large-scale image labeling. Our network consists of four stacked convolutional layers that downsample the image and extract relevant features. On top of these, a deconvolutional layer upsamples the data back to the initial resolution, producing a final dense image labeling. Contrary to previous frameworks, our network contains only convolution and deconvolution operations. Experiments on aerial images show that our network produces more accurate classifications in lower computational time.

Abstract-While initially devised for image categorization, convolutional neural networks (CNNs) are being increasingly used for the pixelwise semantic labeling of images. However, the proper nature of the most common CNN architectures makes them good at recognizing but poor at localizing objects precisely. This problem is magnified in the context of aerial and satellite image labeling, where a spatially fine object outlining is of paramount importance.Different iterative enhancement algorithms have been presented in the literature to progressively improve the coarse CNN outputs, seeking to sharpen object boundaries around real image edges. However, one must carefully design, choose and tune such algorithms. Instead, our goal is to directly learn the iterative process itself. For this, we formulate a generic iterative enhancement process inspired from partial differential equations, and observe that it can be expressed as a recurrent neural network (RNN). Consequently, we train such a network from manually labeled data for our enhancement task. In a series of experiments we show that our RNN effectively learns an iterative process that significantly improves the quality of satellite image classification maps.

No abstract

The ultimate goal of land mapping from remote sensing image classification is to produce polygonal representations of Earth's objects, to be included in geographic information systems. This is most commonly performed by running a pixelwise image classifier and then polygonizing the connected components in the classification map. We here propose a novel polygonization algorithm, which uses a labeled triangular mesh to approximate the input classification maps. The mesh is optimized in terms of an 1 norm with respect to the classifiers's output. We use a rich set of optimization operators, which includes a vertex relocator, and add a topology preservation strategy. The method outperforms current approaches, yielding better accuracy with fewer vertices.