Semantic segmentation requires methods capable of learning high-level features while dealing with large volume of data. Toward such goal, convolutional networks can learn specific and adaptable features based on the data. However, these networks are not capable of processing a whole remote sensing image, given its huge size. To overcome such limitation, the image is processed using fixed size patches. The definition of the input patch size is usually performed empirically (evaluating several sizes) or imposed (by network constraint). Both strategies suffer from drawbacks and could not lead to the best patch size. To alleviate this problem, several works exploited multicontext information by combining networks or layers. This process increases the number of parameters, resulting in a more difficult model to train. In this paper, we propose a novel technique to perform semantic segmentation of remote sensing images that exploits a multicontext paradigm without increasing the number of parameters while defining, in training time, the best patch size. The main idea is to train a dilated network with distinct patch sizes, allowing it to capture multicontext characteristics from heterogeneous contexts. While processing these varying patches, the network provides a score for each patch size, helping in the definition of the best size for the current scenario. A systematic evaluation of the proposed algorithm is conducted using four high-resolution remote sensing data sets with very distinct properties. Our results show that the proposed algorithm provides improvements in pixelwise classification accuracy when compared to the state-of-the-art methods.

Urban forests are an important part of any city, given that they provide several environmental benefits, such as improving urban drainage, climate regulation, public health, biodiversity, and others. However, tree detection in cities is challenging, given the irregular shape, size, occlusion, and complexity of urban areas. With the advance of environmental technologies, deep learning segmentation mapping methods can map urban forests accurately. We applied a region-based CNN object instance segmentation algorithm for the semantic segmentation of tree canopies in urban environments based on aerial RGB imagery. To the best of our knowledge, no study investigated the performance of deep learning-based methods for segmentation tasks inside the Cerrado biome, specifically for urban tree segmentation. Five state-of-the-art architectures were evaluated, namely: Fully Convolutional Network; U-Net; SegNet; Dynamic Dilated Convolution Network and DeepLabV3+. The experimental analysis showed the effectiveness of these methods reporting results such as pixel accuracy of 96,35%, an average accuracy of 91.25%, F1-score of 91.40%, Kappa of 82.80% and IoU of 73.89%. We also determined the inference time needed per area, and the deep learning methods investigated after the training proved to be suitable to solve this task, providing fast and effective solutions with inference time varying from 0.042 to 0.153 minutes per hectare. We conclude that the semantic segmentation of trees inside urban environments is highly achievable with deep neural networks. This information could be of high importance to decision-making and may contribute to the management of urban systems. It should be also important to mention that the dataset used in this work is available on our website.

Flooding is the world's most costly type of natural disaster in terms of both economic losses and human causalities. A first and essential procedure toward flood monitoring is based on identifying the area most vulnerable to flooding, which gives authorities relevant regions to focus. In this letter, we propose several methods to perform flooding identification in high-resolution remote sensing images using deep learning. Specifically, some proposed techniques are based upon unique networks, such as dilated and deconvolutional ones, whereas others were conceived to exploit diversity of distinct networks in order to extract the maximum performance of each classifier. The evaluation of the proposed methods was conducted in a high-resolution remote sensing data set. Results show that the proposed algorithms outperformed the state-of-the-art baselines, providing improvements ranging from 1% to 4% in terms of the Jaccard Index.

Over the past decade, Convolutional Networks (ConvNets) have renewed the perspectives of the research and industrial communities. Although this deep learning technique may be composed of multiple layers, its core operation is the convolution, an important linear filtering process. Easy and fast to implement, convolutions actually play a major role, not only in ConvNets, but in digital image processing and analysis as a whole, being effective for several tasks. However, aside from convolutions, researchers also proposed and developed non-linear filters, such as operators provided by mathematical morphology. Even though these are not so computationally efficient as the linear filters, in general, they are able to capture different patterns and tackle distinct problems when compared to the convolutions. In this paper, we propose a new paradigm for deep networks where convolutions are replaced by non-linear morphological filters. Aside from performing the operation, the proposed Deep Morphological Network (DeepMorphNet) is also able to learn the morphological filters (and consequently the features) based on the input data. While this process raises challenging issues regarding training and actual implementation, the proposed DeepMorphNet proves to be able to extract features and solve problems that traditional architectures with standard convolution filters cannot.