In the research of automatic interpretation of remote sensing images, semantic segmentation based on deep convolutional neural networks has been rapidly developed and applied, and the feature segmentation accuracy and network model generalization ability have been gradually improved. However, most of the network designs are mainly oriented to the three visible RGB bands of remote sensing images, aiming to be able to directly borrow the mature natural image semantic segmentation networks and pre-trained models, but simultaneously causing the waste and loss of spectral information in the invisible light bands such as near-infrared (NIR) of remote sensing images. Combining the advantages of multispectral data in distinguishing typical features such as water and vegetation, we propose a novel deep neural network structure called the multispectral semantic segmentation network (MSNet) for semantic segmentation of multi-classified feature scenes. The multispectral remote sensing image bands are split into two groups, visible and invisible, and ResNet-50 is used for feature extraction in both coding stages, and cascaded upsampling is used to recover feature map resolution in the decoding stage, and the multi-scale image features and spectral features from the upsampling process are fused layer by layer using the feature pyramid structure to finally obtain semantic segmentation results. The training and validation results on two publicly available datasets show that MSNet has competitive performance. The code is available: https://github.com/taochx/MSNet.

High-resolution satellite images contain valuable road semantic information, but the occlusion of vegetation and buildings and the sparse distribution and heterogeneous appearance of roads limit the accuracy of road extraction models. In this article, we propose a novel method for extracting roads using an ensemble learning model with a postprocessing stage. The network weights and biases of our proposed deep learning model are transmitted through the random combination of layers of different submodels during forward and backward propagation. In the gradient descent process, a superior loss function is designed to solve the problem of class imbalance caused by road sparseness, and more attention is given to hard classification samples to extract narrow and covered roads. In addition, we solve road disconnection issues in the results obtained with the neural network by extracting and analyzing the geometric structures and feature points of the roads. Experiments on two challenging datasets of remote sensing imagery show that the proposed method performs better than other models and can extract road information from complex scenes.

Super-resolution (SR) techniques based on deep learning have a pivotal role in improving the spatial resolution of images. However, remote sensing images exhibit ground objects characterized by diverse types, intricate structures, substantial size discrepancies, and noise. Simultaneously, variations in imaging mechanisms, imaging time, and atmospheric conditions among different sensors result in disparities in image quality and surface radiation. These factors collectively pose challenges for existing SR models to fulfill the demands of the domain. To address these challenges, we propose a novel cross-sensor SR framework (ConvFormerSR) that integrates transformers and convolutional neural networks (CNNs), catering to the heterogeneous and complex ground features in remote sensing images. Our model leverages an enhanced transformer structure to capture long-range dependencies and high-order spatial interactions, while CNNs facilitate local detail extraction and enhance model robustness. Furthermore, as a bridge between the two branches, a feature fusion module (FFM) is devised to efficiently fuse global and local information at various levels. Additionally, we introduce a spectral loss based on the remote sensing ratio index to mitigate domain shift induced by cross-sensors. The proposed method is validated on two datasets and compared against existing state-of-the-art SR models. The results show that our proposed method can effectively improve the spatial resolution of Landsat-8 images, and the model performance is significantly better than other methods. Furthermore, the SR results exhibit satisfactory spectral consistency with high-resolution (HR) images.

Deep learning technology has achieved great success in the field of remote sensing processing. However, the lack of tools for making deep learning samples with remote sensing images is a problem, so researchers have to rely on a small amount of existing public data sets that may influence the learning effect. Therefore, we developed an add-in (LabelRS) based on ArcGIS to help researchers make their own deep learning samples in a simple way. In this work, we proposed a feature merging strategy that enables LabelRS to automatically adapt to both sparsely distributed and densely distributed scenarios. LabelRS solves the problem of size diversity of the targets in remote sensing images through sliding windows. We have designed and built in multiple band stretching, image resampling, and gray level transformation algorithms for LabelRS to deal with the high spectral remote sensing images. In addition, the attached geographic information helps to achieve seamless conversion between natural samples, and geographic samples. To evaluate the reliability of LabelRS, we used its three sub-tools to make semantic segmentation, object detection and image classification samples, respectively. The experimental results show that LabelRS can produce deep learning samples with remote sensing images automatically and efficiently.