Learning to semantically segment high-resolution remote sensing images

Abstract

Land cover classification is a task that requires methods capable of learning high-level features while dealing with high volume of data. Overcoming these challenges, Convolutional Networks (ConvNets) can learn specific and adaptable features depending on the data while, at the same time, learn classifiers. In this work, we propose a novel technique to automatically perform pixel-wise land cover classification. To the best of our knowledge, there is no other work in the literature that perform pixel-wise semantic segmentation based on data-driven feature descriptors for high-resolution remote sensing images. The main idea is to exploit the power of ConvNet feature representations to learn how to semantically segment remote sensing images. First, our method learns each label in a pixel-wise manner by taking into account the spatial context of each pixel. In a predicting phase, the probability of a pixel belonging to a class is also estimated according to its spatial context and the learned patterns. We conducted a systematic evaluation of the proposed algorithm using two remote sensing datasets with very distinct properties. Our results show that the proposed algorithm provides improvements when compared to traditional and stateof-the-art methods that ranges from 5 to 15% in terms of accuracy.

Dynamic Multicontext Segmentation of Remote Sensing Images Based on Convolutional Networks

“…Also, for these two datasets, we considered as baselines: (i) Fully Convolutional Networks (FCN) [17]. In this case, the pixelwise architectures proposed by [41] were converted into fully convolutional network and exploited as baseline. (ii) Deconvolutional networks [18], [19].…”

“…For the Coffee [41] and the GRSS Data Fusion [42] datasets, we employed the same protocol of [41]. Specifically, for the former dataset, we conducted a five-fold cross-validation to assess the performance of the proposed algorithm.…”

“…Although the presented approach is proposed to select automatically the best patch size, in training time, avoiding lots of experiments to adjust such size (as done in several works [8], [26], [41]), in this section, the patch size range is analyzed in order to examine the robustness of the method.…”

Semantic labeling in very high resolution images via a self-cascaded convolutional neural network

“…Based on CNNs, many patch-classification methods are proposed to perform semantic labeling (Mnih, 2013;Mostajabi et al, 2015;Paisitkriangkrai et al, 2016;Nogueira et al, 2016;Alshehhi et al, 2017;Zhang et al, 2017). These methods determine a pixel's label by using CNNs to classify a small patch around the target pixel.…”

GAN-Assisted Two-Stream Neural Network for High-Resolution Remote Sensing Image Classification

“…The process of resizing the input data would not only cause the loss of information but would also make it difficult to obtain a direct pixel-to-pixel classification result. For the traditional CNN, we obtained one result for one patch [36], which represented the classification result for one pixel surrounded by that patch, rather than a result at the same size as the patch and each location represented the type of corresponding pixel in the input panchromatic data. Therefore, we revised our residual net in the main line and placed feature maps from the convolution layers directly to the classifier, instead of flatting the feature maps into 1-D structures in advance.…”