Semantic segmentation for remote sensing images (RSIs) is widely applied in geological surveys, urban resources management, and disaster monitoring. Recent solutions on remote sensing segmentation tasks are generally addressed by CNN-based models and transformer-based models. In particular, transformer-based architecture generally struggles with two main problems: a high computation load and inaccurate edge classification. Therefore, to overcome these problems, we propose a novel transformer model to realize lightweight edge classification. First, based on a Swin transformer backbone, a pure Efficient transformer with mlphead is proposed to accelerate the inference speed. Moreover, explicit and implicit edge enhancement methods are proposed to cope with object edge problems. The experimental results evaluated on the Potsdam and Vaihingen datasets present that the proposed approach significantly improved the final accuracy, achieving a trade-off between computational complexity (Flops) and accuracy (Efficient-L obtaining 3.23% mIoU improvement on Vaihingen and 2.46% mIoU improvement on Potsdam compared with HRCNet_W48). As a result, it is believed that the proposed Efficient transformer will have an advantage in dealing with remote sensing image segmentation problems.

Crop disease seriously affects production because of its highly destructive property. Wheat under different levels of disease infection should be treated by various chemical strategies to enable a precision plant protection. Therefore, a fast and robust algorithm for wheat yellow rust disease severity determination is highly desirable for its sustainable management. The recent use of remote sensing and deep learning is drawing increasing research interests in wheat yellow rust severity detection at leaf level. However, little reviews take field-scale rust severity detection into account by using UAV multispectral images and deep learning networks. As a result, by the means of UAV multispectral images, a real-time yellow rust detection algorithm named Efficient Dual Flow UNet (DF-UNet) to detect different levels of yellow rust is designed and proposed in this paper to meet practical requirements. First, pruning strategy is utilized to realize a lightweight structure. Second, the Sparse Channel Attention (SCA) Module is designed to increase the receptive field of the network and enhance the ability to distinguish each category. Third, by fusing SCA, a novel dual flow branch model with segmentation and ranking branch based on UNet is proposed to accomplish yellow rust severity determination at field scale. The comparative results show that the proposed method reduces more than half computation load and achieves the highest overall accuracy score among other state-of-the-art deep learning models. It is convinced that the proposed DF-UNet can pave the way for automated yellow rust severity detection at farmland scales in a robust way.

Crop disease is widely considered as one of the most pressing challenges for food crops, and therefore an accurate crop disease detection algorithm is highly desirable for its sustainable management. The recent use of remote sensing and deep learning is drawing increasing research interests in wheat yellow rust disease detection. However, current solutions on yellow rust detection are generally addressed by RGB images and the basic semantic segmentation algorithms (e.g., UNet), which do not consider the irregular and blurred boundary problems of yellow rust area therein, restricting the disease segmentation performance. Therefore, this work aims to develop an automatic yellow rust disease detection algorithm to cope with these boundary problems. An improved algorithm entitled Ir-UNet by embedding irregular encoder module (IEM), irregular decoder module (IDM) and content-aware channel re-weight module (CCRM) is proposed and compared against the basic UNet while with various input features. The recently collected dataset by DJI M100 UAV equipped with RedEdge multispectral camera is used to evaluate the algorithm performance. Comparative results show that the Ir-UNet with five raw bands outperforms the basic UNet, achieving the highest overall accuracy (OA) score (97.13%) among various inputs. Moreover, the use of three selected bands, Red-NIR-RE, in the proposed Ir-UNet can obtain a comparable result (OA: 96.83%) while with fewer spectral bands and less computation load. It is anticipated that this study by seamlessly integrating the Ir-UNet network and UAV multispectral images can pave the way for automated yellow rust detection at farmland scales.

Real-time coal flow monitoring is crucial for efficient coal mine transportation. Traditional vision acquisition devices generate two-dimensional images that are not accurately identified in the complex underground coal mine environment, such as dust, water mist, and low light. In this paper, we propose a conveyor belt coal flow detection method that integrates laser scanning and binocular vision to address this problem. Our proposed method has several advantages over traditional approaches. Firstly, we calibrate the binocular camera using Zhang’s calibration method to enhance the accuracy of the system. Secondly, we extract the centerline of the laser stripe using the grayscale center of gravity method, which improves the system’s performance in complex environments. Thirdly, we calculate the cross-sectional area of the material accurately using the trapezoidal area accumulation method, and visualize it in two dimensions based on a single frame, while the point cloud data from multiple consecutive frames are visualized in a 3D model at a realistic scale. Finally, we use the continuous multi-frame cross-sectional area to calculate the current conveyor flow, and apply the BP neural network to establish an energy-saving optimization model for the belt conveyor. We also design a PLC fuzzy controller based on fuzzy control algorithms to adjust the belt’s operating speed intelligently according to the coal flow size, achieving energy-saving operation and intelligent control of the belt conveyor. Our experiments show that our method can accurately obtain coal volume and control the belt speed of the conveyor in real-time, making it an innovative and practical solution for coal flow detection and energy-saving operation in underground coal mines.

In the actual production of coal mines, the transmission needs of existing underground applications cannot be met due to a lack of strategies and customized equipment for underground 5G application scenarios, which causes increased underground data processing delay and low transmission efficiency. To solve the problem above, the mobile edge computing (MEC) technology based on the 5G wireless base station is studied, and underground 5G communication capabilities is improved by edge caching and dynamic resource allocation according to the actual situation of coal mines. The experimental result shows that under the premise of maintaining the rated power and transmit power of the existing base station, the average delay of executing tasks is 15ms, which is 50% lower than the average delay of all local execution methods. The average delay is reduced by 37.5% than all MEC execution methods. At the same time, the uplink rate of a single base station can reach 1Gbps and the downlink rate can reach 1.5 Gbps. Our method can significantly improve the reliability of mining 5G communication systems and the rational allocation of resources. Keywords: mining 5G communication; edge computing; resource allocation; deep learning; underground coal mines.

Hyperspectral sensors provide an opportunity to capture the intensity of high spatial/spectral information and enable applications for high-level earth observation missions, such as accurate land cover mapping and target/object detection. Currently, convolutional neural networks (CNNs) are good at coping with hyperspectral image processing tasks because of the strong spatial and spectral feature extraction ability brought by hierarchical structures, but the convolution operation in CNNs is limited to local feature extraction in both dimensions. In the meanwhile, the introduction of the Transformer structure has provided an opportunity to capture long-distance dependencies between tokens from a global perspective; however, Transformer-based methods have a restricted ability to extract local information because they have no inductive bias, as CNNs do. To make full use of these two methods’ advantages in hyperspectral image processing, a dual-flow architecture named Hyper-LGNet to couple local and global features is firstly proposed by integrating CNN and Transformer branches to deal with HSI spatial-spectral information. In particular, a spatial-spectral feature fusion module (SSFFM) is designed to maximally integrate spectral and spatial information. Three mainstream hyperspectral datasets (Indian Pines, Pavia University and Houston 2013) are utilized to evaluate the proposed method’s performance. Comparative results show that the proposed Hyper-LGNet achieves state-of-the-art performance in comparison with the other nine approaches concerning overall accuracy (OA), average accuracy (AA) and kappa index. Consequently, it is anticipated that, by coupling CNN and Transformer structures, this study can provide novel insights into hyperspectral image analysis.