Road extraction from high-resolution remote sensing images has long been a focal and challenging research topic in the field of computer vision. Accurate extraction of road networks holds extensive practical value in various fields, such as urban planning, traffic monitoring, disaster response and environmental monitoring. With rapid development in the field of computational intelligence, particularly breakthroughs in deep learning technology, road extraction technology has made significant progress and innovation. This paper provides a systematic review of deep learning-based methods for road extraction from remote sensing images, focusing on analyzing the application of computational intelligence technologies in improving the precision and efficiency of road extraction. According to the type of annotated data, deep learning-based methods are categorized into fully supervised learning, semi-supervised learning, and unsupervised learning approaches, each further divided into more specific subcategories. They are comparatively analyzed based on their principles, advantages, and limitations. Additionally, this review summarizes the metrics used to evaluate the performance of road extraction models and the high-resolution remote sensing image datasets applied for road extraction. Finally, we discuss the main challenges and prospects for leveraging computational intelligence techniques to enhance the precision, automation, and intelligence of road network extraction.

Fully convolutional networks (FCNs) have been widely applied for dense classification tasks such as semantic segmentation. As a large number of works based on FCNs are proposed, various semantic segmentation models have been improved significantly. However, duplicated upsampling and deconvolution operations in the FCNs will lead to information loss in semantic segmentation tasks and to problems such as ignoring the relationship between pixels and pixels and the lack of spatial consistency. In this study, we propose a parallel fully convolutional neural network that integrates holistically-nested edge detection (HED) network to capture image edge information, improving semantic segmentation performance. We carry out comprehensive experiments and achieve a better result on the PASCAL VOC 2012, PASCAL-Context and Cityscapes, comparing the results with some existing semantic segmentation methods.

The accurate detection and extraction of roads using remote sensing technology are crucial to the development of the transportation industry and intelligent perception tasks. Recently, in view of the advantages of CNNs in feature extraction, its related road extraction methods have been proposed successively. However, due to the limitation of kernel size, they perform less effectively at capturing long-range information and global context, which are crucial for road targets distributed over long distances and highly structured. To deal with this problem, a novel model named RoadFormer with a Swin Transformer as the backbone is developed in this paper. Firstly, to extract long-range information effectively, a Swin Transformer multi-scale encoder is adopted in our model. Secondly, to enhance the feature representation capability of the model, we design an innovative bottleneck module, in which the spatial and channel separable convolution is employed to obtain fine-grained and globe features, and then a dilated block is connected after the spatial convolution module to capture more integrated road structures. Finally, a lightweight decoder consisting of transposed convolution and skip connection generates the final extraction results. Extensive experimental results confirm the advantages of RoadFormer on the Deepglobe and Massachusetts datasets. The comparative results of visualization and quantification demonstrate that our model outperforms comparable methods.

Infrared and visible image registration is a very challenging task due to the large geometric changes and the significant contrast differences caused by the inconsistent capture conditions. To address this problem, this paper proposes a novel affine and contrast invariant descriptor called maximally stable phase congruency (MSPC), which integrates the affine invariant region extraction with the structural features of images organically. First, to achieve the contrast invariance and ensure the significance of features, we detect feature points using moment ranking analysis and extract structural features via merging phase congruency images in multiple orientations. Then, coarse neighborhoods centered on the feature points are obtained based on Log-Gabor filter responses over scales and orientations. Subsequently, the affine invariant regions of feature points are determined by using maximally stable extremal regions. Finally, structural descriptors are constructed from those regions and the registration can be implemented according to the correspondence of the descriptors. The proposed method has been tested on various infrared and visible pairs acquired by different platforms. Experimental results demonstrate that our method outperforms several state-of-the-art methods in terms of robustness and precision with different image data and also show its effectiveness in the application of trajectory tracking.

Cross-view geo-localization via scene matching is crucial in unmanned aerial vehicle (UAV) systems in global navigation satellite system (GNSS) denial environment. However, images in the same scene may undergo geometric distortion and occlusion due to difference in capture viewpoint, time and platform. The existing methods mainly extract consistent features between images by CNNs, while ignoring the semantic distribution and structural information of the objects. Aiming at addressing this issue, we introduce a semantic-aware graph convolutional network (SeGCN). To improve consistent representation of object features from different viewpoints, potential semantic features are inferred via cross-attention of image context. Then, for exploring the structural information of objects, SeGCN performs graph convolution on graph structures constructed from the same semantic features. Finally, the composite features generated by SeGCN and backbone are utilized for scene matching. Comprehensive experiments conducted on the University-1652 and SUES-200 benchmarks establish that the proposed approach attains the highest levels of accuracy in both localization and navigation tasks. Furthermore, we conducted localization simulation experiments on our real UAV datasets, confirming the effectiveness of SeGCN in real world application scenarios.

Infrared and visible image fusion is to combine the information of thermal radiation and detailed texture from the two images into one informative fused image. Recently, deep learning methods have been widely applied in this task; however, those methods usually fuse multiple extracted features with the same fusion strategy, which ignores the differences in the representation of these features, resulting in the loss of information in the fusion process. To address this issue, we propose a novel method named multi-modal feature self-adaptive transformer (MFST) to preserve more significant information about the source images. Firstly, multi-modal features are extracted from the input images by a convolutional neural network (CNN). Then, these features are fused by the focal transformer blocks that can be trained through an adaptive fusion strategy according to the characteristics of different features. Finally, the fused features and saliency information of the infrared image are considered to obtain the fused image. The proposed fusion framework is evaluated on TNO, LLVIP, and FLIR datasets with various scenes. Experimental results demonstrate that our method outperforms several state-of-the-art methods in terms of subjective and objective evaluation.