Semantic segmentation is an important task for the interpretation of remote sensing images. Remote sensing images are large in size, contain substantial spatial semantic information, and generally exhibit strong symmetry, resulting in images exhibiting large intraclass variance and small interclass variance, thus leading to class imbalance and poor small-object segmentation. In this paper, we propose a new remote sensing image semantic segmentation network, called CAS-Net, which includes coordinate attention (CA) and SPD-Conv. In the model, we replace stepwise convolution with SPD-Conv convolution in the feature extraction network and add a pooling layer into the network to avoid the loss of detailed information, effectively improving the segmentation of small objects. The CA is introduced into the atrous spatial pyramid pooling (ASPP) module, thus improving the recognizability of classified objects and target localization accuracy in remote sensing images. Finally, the Dice coefficient was introduced into the cross-entropy loss function to maximize the gradient optimization of the model and solve the classification imbalance problem in the image. The proposed model is compared with several state-of-the-art models on the ISPRS Vaihingen dataset. The experimental results demonstrate that the proposed model significantly optimizes the segmentation effect of small objects in remote sensing images, effectively solves the problem of class imbalance in the dataset, and improves segmentation accuracy.

To address the problem of ignoring unpaved roads when planning off-road emergency rescue paths, an improved A* algorithm that incorporates road factors is developed to create an off-road emergency rescue path planning model in this study. To reduce the number of search nodes and improve the efficiency of path searches, the current node is classified according to the angle between the line connecting the node and the target point and the due east direction. Additionally, the search direction is determined in real time through an optimization method to improve the path search efficiency. To identify the path with the shortest travel time suitable for emergency rescue in wilderness scenarios, a heuristic function based on the fusion of road factors and a path planning model for off-road emergency rescue is developed, and the characteristics of existing roads are weighted in the process of path searching to bias the selection process toward unpaved roads with high accessibility. The experiments show that the improved A* algorithm significantly reduces the travel time of off-road vehicles and that path selection is enhanced compared to that with the traditional A* algorithm; moreover, the improved A* algorithm reduces the number of nodes by 16.784% and improves the search efficiency by 27.18% compared with the traditional 16-direction search method. The simulation results indicate that the improved algorithm reduces the travel time of off-road vehicles by 21.298% and improves the search efficiency by 93.901% compared to the traditional A* algorithm, thus greatly enhancing off-road path planning.

One of the key conditions for achieving superior regional growth is ensuring the harmonious development of both the layout and functions of territorial space. Territorial space, which includes production space, living space, and ecological space, serves as a critical system and venue for economic, cultural, and social activities in a region. The harmonized growth of production–living–ecological space (PLES) is essential for attaining sustainable development goals. Research on PLES offers a fresh perspective on promoting sustainable development of the spatial domain and the sustainable use of resources. However, studies on PLE functions in ecologically fragile areas are lacking. Therefore, in this study, which adopts a PLES perspective, land-use data are used to classify land according to the dominant functions of production, ecology, and living. Integration of point-of-interest (POI) data with socio-economic data was established to spatially describe indicators at the grid level and create a scoring system for PLES indicators in ecologically fragile areas. Finally, the entropy weight method, holistic assessment methods, coupling coordination degree model (CCDM), and geodetector were employed to explore the coupling coordination relationships and factors influencing PLESs in the Turpan–Hami Basin from 2010 to 2020. The results indicate that the Turpan–Hami Basin consists predominantly of potential ecological space, mainly in the central and northern regions, which are characterized by the Gobi Desert and bare rock landforms. Over the past decade, the PLES framework has seen a notable rise in the allocation of residential and ecological areas, whereas the portion dedicated to production spaces has diminished. The overall coupling coordination degree (CCD) of PLES in the Turpan–Hami Basin is at a coordinated level and gradually increasing. The most significant impact on the degree of PLES coupling coordination is exerted by population and natural factors. The research findings provide theoretical support for the sustainable utilization of resources in the Turpan–Hami Basin and other ecologically fragile areas while also offering scientific evidence to promote the coupling coordination of PLES, thereby contributing to high-quality regional development.

Lentinus edodes sticks are susceptible to mold infection during the culture process, and manual identification of infected sticks is heavy, untimely, and inaccurate. Aiming to solve this problem, this paper proposes a method for identifying infected Lentinus edodes sticks based on improved ResNeXt‐50(32 × 4d) deep transfer learning. First, a dataset of Lentinus edodes stick diseases was constructed. Second, based on the ResNeXt‐50(32 × 4d) model and the pretraining weight of the ImageNet dataset, the influence of pretraining weight parameters on recognition accuracy was studied. Finally, six fine‐tuning strategies of the fully connected layer were designed to modify the fully connected layer of ResNeXt‐50(32 × 4d). The experimental results show that the recognition accuracy of the method proposed in this paper can reach 94.27%, which is higher than the Vgg16, GoogLeNet, ResNet50, and MobileNet v2 models by 8.47%, 6.49%, 4.68%, and 9.38%, respectively, and the F1‐score can reach 0.9422. The improved method proposed in this paper can reduce the calculation pressure and overfitting problem of the model, improve the accuracy of the model in the identification of Lentinus edodes stick mold diseases, and provide an effective solution for the selection of diseased sticks.

Lentinula edodes logs are susceptible to sundry bacteria contamination during the culture process, and the manual identification of contaminated logs is difficult, untimely, and inaccurate. Aiming to solve this problem, this paper proposes a method for the identification of contaminated Lentinula edodes logs based on the deep learning model Ghost–YOLOv4. Firstly, a data set of Lentinula edodes log sundry bacteria contamination was constructed. Secondly, in view of the problems that the YOLOv4 network parameters are too large and that the detection speeds of Lentinula edodes log videos are slow, the backbone feature extraction network was replaced with a lightweight network, GhostNet, and the YOLOv4 enhancement feature extraction network PANet and the Yolo Head modules were completed. The modification of the network reduced the number of parameters of the network and improved the detection speed of the network. Finally, the feature extraction network introduced the migration learning pre-training model, which reduced the computational pressure and overfitting problems of the model and further improved the performance of the Ghost–YOLOv4 network. Not only did the constructed Ghost–YOLOv4 ensure the accuracy of the identification and detection of Lentinula edodes log sundry bacteria contamination, but it also had better results in detection speed and real-time performance, and it provides an effective solution for the lightweight deployment of a target detection model on embedded equipment in culture sheds.

Aiming at the shortage of sufficient continuous parameters for using models to estimate farmland soil organic carbon (SOC) content, an acquisition method of factors influencing farmland SOC and an estimation method of farmland SOC content with Internet of Things (IOT) are proposed in this paper. The IOT sensing device and transmission network were established in a wheat demonstration base in Yanzhou Distict of Jining City, Shandong Province, China to acquire data in real time. Using real-time data and statistics data, the dynamic changes of SOC content between October 2012 and June 2015 was simulated in the experimental area with SOC dynamic simulation model. In order to verify the estimation results, potassium dichromate external heating method was applied for measuring the SOC content. The results show that: 1) The estimated value matches the measured value in the lab very well. So the method is feasible in this paper. 2) There is a clear dynamic variation in the SOC content at 0.2 m soil depth in different growing periods of wheat. The content reached the highest level during the sowing period, and is lowest in the flowering period. 3) The SOC content at 0.2 m soil depth varies in accordance with the amount of returned straw. The larger the amount of returned straw is, the higher the SOC content.