Deep learning (DP) based road extraction methods often produce fragmented results. Direct optimization of end-to-end DP-based methods requires the design of more complex network structures to enhance the model’s adaptability in complex scenarios. To tackle this challenge, this paper adopts a novel approach that treats the discrepancy (the road breakage part) between road prediction and ground-truth as the extraction target, thereby constructing a highly efficient and lightweight semantic segmentation network, termed the Road Patch Extraction Network (RPE-Net). RPE-Net includes multi-directional striped residual (MDSR) encoder, multi-directional striped pooling (MDSP) units, and multi-directional striped decoder (MDSD). The structure is similar to LinkNet34, but the overall size of the network parameters is just 1.56MB, which is 1/50 of LinkNet34. The post-processing datasets used for training can be semi-automatically generated by the algorithm, and only a small amount of manual intervention can be used for training. A large number of experiments showt hat the post-processing method proposed in this paper has extremely high speed and generalization ability.

The completeness of road extraction is very important for road application. However, existing deep learning (DP) methods of extraction often generate fragmented results. The prime reason is that DP-based road extraction methods use square kernel convolution, which is challenging to learn long range contextual relationships of roads. The road often produce fractures in the local interference area. Besides, the quality of extraction results will be subjected to the resolution of remote sensing (RS) image. Generally, an algorithm will produce worse fragmentation when the used data differs from the resolution of the training set. To address these issues, we propose a novel road extraction framework for RS images, named the Multi-Scale and Multi-Direction Feature Fusion Network (MSMDFF-Net). This framework comprises three main components: the Multi-Directional Feature Fusion (MDFF) Initial Block, the Multi-Scale Residual (MSR) encoder, and the Multi-Directional Combined Fusion (MDCF) decoder. Firstly, according to the road’s morphological characteristics, we develop a strip convolution module with a direction parameter (SCM-D). Then, to make the extracted result more complete, four SCM-D with different directions are used to MDFF-Initial Block and MDCF-decoder. Finally, we incorporate an additional branch into the ResNet encoding module to build MSR-encoder for improving the generalization of the model on different resolution RS image. Extensive experiments on three popular datasets with different resolution (Massachusetts, DeepGlobe, and SpaceNet datasets) show that the proposed MSMDFF-Net achieves new state-of-the-art results. The code will be available at https://github.com/wycloveinfall/MSMDFF-NET.

This paper investigates the problem of building extraction from very high resolution (VHR) satellite imagery. Deep learning methods are deemed as emerging trends for solving this problem due to their increasingly prominent extraction effects. However, most state-of-the-art deep learning-based building segmentation methods produce pixel-level segmentation masks rather than accurate polygon-level extraction results required in real-world applications. This paper introduces a Harris function based active contour network (HACNet) that extracts polygon-level masks directly from satellite imagery. Our proposed HACNet not only leverages the full potential of active contour methods, but also successfully integrates the Harris function into the polygon contour evolution of the active contour models. The incorporation of the Harris function concentrates the attention of the neural network more on the vicinity of corners, thereby significantly improving the performance of building contour extraction, particularly in scenes with high curvature and noisy boundaries. Experiments on two public datasets (namely Vaihingen, Bing Huts) demonstrate the effectiveness and superiority of our model over other outstanding deep learning methods for the building extraction.

Using deep learning to extract roads from satellite images is one of the most popular methods. However, the existing encoder-decoder-based deep networks usually produce fragmented roads, due to the complex spatial and color characteristics of the road. In this paper, motivated by the road multi-scale information, we proposed a multi-scale and multi-direction feature fusion network (MSMDFF-Net) to reduce the fragmentation of road extraction results. The proposed method mainly consists of three processes: 1) In the initial stage, the image details from different directions were transmitted; 2) At different encoding stages, the multi-scale information of the image was fused; 3) In the decoding process, the matching modules of road characteristics were used to up-sample the feature map. Extensive experiments on the popular datasets (LSVD and Deep-Globe datasets) demonstrate that the MSMDFF-Net has higher accuracy and generalization performance with less fragmentary road results.

In order to make the best use of the available data in the remote sensing database, this paper focuses on an important topic of using the complementary information of multi-source remote sensing data, that is, road network extraction based on fusion technology with synthetic aperture radar (SAR) and optical images. Starting with the line segments achieved from the road segmentation maps, a decision-level fusion method which mainly includes two stages is proposed in this paper. The first stage is fusing based on the geometric overlapping rules. In the second stage, a road network extraction approach that takes into account both the contextual information and evidence theory is presented. The experiments on TerraSAR-X and WorldView-4 images showed that our proposed method had an excellent performance in terms of the completeness and quality of the road extraction.

In order to speed up the update of existing road maps, it is crucial to develop a more efficient road extraction method from remote sensing images. In recent years, deep learning techniques have been widely used for road extraction applications. Among current CNN-based deep networks for road extraction, few works study the shape of the convolution kernel, and the remote contextual features dependency relationship is not fully utilized. In view of these problems, an improved DLinkNet is proposed in this paper. Firstly, a convolutional layer which fuses information of multiple scales is used to replace the InitBlock in the front of the network. Secondly, instead of using the D-Block, the DenseRe-Block is applied to the center structure of DLinkNet. The experimental results show that the improved network has a higher IoU score than DLinkNet when extracting roads from optical images in both city and mountain town areas.

Due to different imaging mechanisms, the registration of optical and Synthetic Aperture Radar (SAR) image is a very challenging task. Many optical and SAR registration methods have been proposed. But most of them are for low-to-medium resolution images, and less for high-resolution images. Therefore, this paper proposes a high-resolution optical and SAR image registration method using local self-similar descriptor based on edge feature. Firstly, a Gauss-Gamma bi-windows algorithm is used to extract the edge intensity maps of the images respectively. Its function is to eliminate the non-linear gray-scale difference between SAR and optical images, and also to avoid the interference of isolated speckle noise on feature point extraction. Then, local self-similar descriptor is extracted on the edge intensity map, and descriptor matching is performed using Euclidean distance. Finally, the fast sample consensus algorithm is used to eliminate mismatched point pairs. The experimental results can effectively resist speckle noise and radiation differences, and obtain pixel-level registration accuracy.

This paper proposes a semi-empirical microwave model of methane emissions (CH4) based on the biogeochemical processes from rice paddy. By exploiting the mechanism processes of methane production, oxidation and emission, a combined microwave model is developed to predict methane emissions from rice paddy. Simultaneously, the main influencing emissions factors, which concluded soil, underlying water body, vegetation, climate and management, are analyzed in the present model. During the whole growth season, the multi-polarization backscattering coefficients of rice are measured by the ground-based radar scatterometer (GBRS), and relevant parameters are observed in the rice fields. Moreover, experiments of the emissions samples are conducted on conventional static box, and the results are compared to the semi-empirical microwave model and Denitrification-Decomposition (DNDC) simulation model, respectively, which shows the good extend performance is developed from experience model into mechanism model based on active microwave remote sensing data.

Interferometric synthetic aperture radar (InSAR) has become the primary means to obtain digital elevation models(DEM) on the earth's surface, including several key steps such as removal of flatten effect, filter processing, and phase unwrapping. Due to atmospheric interference, etc., the distribution of azimuth and range to phase quality on SAR images is uneven. Thus, this paper proposes a method based on weighted quality graph to guide phase unwrapping, which considers the difference of the contribution weight of spatial noise to the range and azimuth of the quality graph. The weighting method is used to eliminate the error of the range direction and the azimuth direction, form a new quality graph, and guide the phase unwrapping. In order to solve the problem of slow speed of traditional methods, this paper introduces the method of heapsort to improve the speed of phase unwrapping.

This paper introduces a method of road extraction using dual-temporal high-resolution synthetic aperture radar (SAR) images. Firstly, multiplicative Duda operators are applied to detect line features. Then, coherence and backscattering coefficient are combined to distinguish road from river and shadow and the coefficient of variation is used to remove heterogeneous areas. Next, road is segmented via path opening and thresholding. At last, a novel thinning and gap connection approach is proposed to gain the thinned and more complete road map. The experiments were test on TerraSAR-X images and results showed that the proposed method improved considerably the results of road extraction compared with approach using single-temporal SAR image.

The inclination direction monitor of the power transmission tower can issue an early-warning for the risk of the tower collapse, as well as the collapse direction. In this paper, we proposed a new method to retrieve the inclination direction of the power transmission tower using geocoding. First, the backscattering coefficient was utilized to separate the power transmission tower from the background, and figure up those coefficients along the tower direction to extract the main body of the tower. Second, the whole image was traversed by a$3^{\ast}3$template window to refine and mark the most possible learning tower points. Finally, a distance-Doppler model was built to retrieve geographical coordinates of those marked points. According to the geographical coordinate of the tower point at different time, we got tower inclination direction. Compared with the actual measurement results of the study area, the retrieval result has a 0.0025° error, and the direction is north-east, consistent with the actual direction.

In order to study the coherence characteristics of low-backscattering objects such as roads and rivers, we introduce a coherence estimation approach based on clustering method. When the approach is applied to multi-temporal high-resolution TerraSAR-X images in urban areas, the results show that the coherence of roads is higher than that of rivers and shadows. This indicates that coherence features can be used to distinguish roads and water bodies. Further, this paper proposes a road detection method of synthetic aperture radar (SAR) images based on path operators and support vector machine (SVM), which combines the backscattering and coherence characteristics of the roads. Experimental results show that coherence features can be used for road detection.

This paper reports on an experiment conducted at the wind-wave tank in UESTC microwave chamber to characterize the C- and X-band radar return from water surfaces covering oil films when observed at different incidences. The measurements of Normalized Radar Cross Section (NRCS) with bi-objective calibration technique were carried out for full polarization and various wind speeds. Comparisons are performed with clean sea water and diverse oil spills. From this data set we validate the Two-Scale Model (TSM) as a calculating method to simulate the backscatter coefficients of oil spill surface. The applicability of experimental results to SAR image extraction is discussed.

In this paper, a method for road detection based on Duda and path operators has been presented. The roads are represented as slender dark regions with constant width and reflectance in the high-resolution SAR images. The path operators (path openings and closings) were performed as morphological filters in retaining linear structures. However, the filters were not sensitive to the width of linear feature. Focused on the limitation of the method, a preprocessing procedure using Duda operators was introduced before adopting the method of morphological profiles with path operators. When the modified method was applied on RADARSAT-2 datasets for different areas, the research results show that the completeness and correctness are over 70% for road detection from SAR images.