Search

Image interpretation algorithms based on deep learning are becoming increasingly important in land cover information acquisition. We propose CSSwin-unet, a network designed for the semantic segmentation of remote sensing images. CSSwin-unet is based on Swin-unet, follows the U-shaped codec structure of U-Net, but utilizes Swin transformer blocks with superior global modelling capabilities to constitute the codec. In addition, we design a parallel branch in the encoder with a context aggregation module (CAM) to enhance contextual information extraction and alleviate the semantic ambiguity problem resulting from occlusion. To address the problem of semantic information mismatch between codecs and improve the model's ability to extract spatial information, we constructed a space extraction module (SEM) in the skip connections, which replaces the direct copying of encoder features in Swin-unet. To reduce information loss during the downsampling process and strengthen the segmentation capacity of the network, we designed a feature shrinkage module (FSM) in the downsampling session. We conducted comprehensive ablation experiments on a dataset we produced ourselves and compared the results with other advanced methods. The test results showed significant improvement, with mIoU, mF1, and OA values improving by 2.83%, 2.47%, and 2.05%, respectively, compared to the second best performing model, Swin-unet. The above results prove the excellent performance of CSSwin-unet. [ABSTRACT FROM AUTHOR]

The evolution of land-use spatial conflict is a comprehensive reflection of regional development and a common phenomenon in urbanization. The research on land-use spatial conflicts can provide theoretical basis and practical support for spatial control. The present study employs Tianjin as the research object, and ecological risk assessment model was improved based on landscape pattern analysis by selecting appropriate indicators that may bring conflict risk problems. The spatial conflict measurement model of land-use based on the coordination of human-land-landscape elements was preliminarily constructed. The spatial and temporal evolution of land-use spatial conflict degree, and spatial autocorrelation from 2000 to 2020 was preliminarily analyzed. Results showed that (1) the conflict measurement level of the model was verified by Google Earth remote sensing image platform and field survey, and the results were relatively consistent with the actual situation, which proves that this method was feasible. (2) The areas with low intensity of land-use conflict were mainly distributed in remote mountainous areas with less human activities, while the areas with high comprehensive conflict index were mainly distributed in the central urban area with more active human activities and urbanization, the urban fringe area, the water-land ecotone, and some landscape transition areas. (3) There was a close relationship between land use spatial conflict and land-use change. Areas with severe and frequent land-use changes were often areas with noticeable land-use conflict changes. In the period of high land-use change intensity and frequent land-use mode conversion, it was also a period of a large area where regional land-use conflict was enhanced. (4) The overall level of conflict showed an average trend year by year, and the high conflict fluctuates greatly with first rising and then falling, but the change range was small. Moderate conflict mainly showed a decreasing trend, with an extensive change range. The early stage of the mild conflict was relatively stable, and the later stage changed significantly. [ABSTRACT FROM AUTHOR]

With the further development of the construction of “smart mine,” the technology of mine unmanned truck has developed rapidly. Road data are an important prerequisite for the mature application of mining and transportation dispatching system and unmanned driving, so it is particularly important to extract the road in open-pit mines timely and accurately. Different from urban roads, mine roads have no clear road edge and more background interference. In view of the above problems, a convolutional neural network called RATT-UNet (R: residual connection; ATT: attention), which combines residual connection, attention mechanism, and U-Net, is proposed to extract mine road from high-resolution remote sensing images. The advantages of the proposed method are listed as follows: first, the well-designed RATT unit combining residual connection and attention mechanism is used to construct the proposed neural network, which improves the network’s perception ability of detailed features. Second, the abundant skip connections and residual connections can improve the information flow, allowing a better network to be designed with fewer parameters. Third, a composite loss function based on structural similarity is proposed, which effectively alleviates the noise and edge blurring phenomenon and improves segmentation quality. Finally, we perform postprocessing optimization operations on the road extraction results. Experimental results demonstrate that the proposed RATT-UNet outperforms all the comparing network models in terms of the quality of extraction results and evaluation indicators in the mining road extraction task. [ABSTRACT FROM AUTHOR]

In view of the many shortcomings of traditional copper ore detection methods, this paper proposes a new method for copper ore content detection, which is to detect copper ore content through spectral information. And we propose an error compensation extreme learning machine model. Using the error compensation extreme learning machine and the spectral information of the copper ore can quickly and accurately identify the copper content in the copper ore. Finally, the satellite remote sensing data were simulated, and the copper content distribution in the mining area was obtained. This provides guidance for future mining, reclamation and ore prospecting. [ABSTRACT FROM AUTHOR]

• A remote sensing inversion method for saline and alkaline land based on D-TELM. • The advantages of this method are high accuracy. • The different hidden layers of D-TELM have different activation functions. In order to obtain the distribution of saline and alkaline land in Wuyuan County and take appropriate improvement measures, it is necessary to invert the distribution map accurately. This paper introduces an analysis method for saline and alkaline land, which is used to analyze the salt content (‰), organic matter content (g/kg), and PH value. First of all, thirteen saline and alkaline land samples of Wuyuan County were collected in the field. Then, the spectral data was acquired by the SVC HR-1204 portable spectrometer, and the salt content, organic matter content, and PH value data were acquired by chemical analysis. Taking different hidden layers with different activation functions as the innovation point, the TELM is improved to propose D-TELM algorithm. Finally, D-TELM was used to construct three model of the relationship between the spectral data and the salt content, organic matter content, and PH value data, and the inversion and distribution maps were performed. The experimental result shows that most of the D-TELM models are better than the TELM models. According to the principle of minimum standard deviation, the inversion model of salt content uses the dsig-sin model, the inversion model of organic matter content uses the tanh-dsig model, and the inversion model of pH value uses the dsig-sin model. [ABSTRACT FROM AUTHOR]