To determine both the size of a satellite antenna and the thermal deformation of its surface shape, a novel high-accuracy close-range photogrammetric technique is used in this study. The method is also applied to assess the performance of the antenna in orbit. The measurement principle and solution method of close-range photogrammetry were thoroughly investigated, and a detailed measurement test scheme was developed. A thermal deformation measurement of the surface shape of a satellite antenna was then carried out. The results show that the measurement error using close-range photogrammetry was smaller than 0.04 mm, which meets the accuracy requirement. Thanks to the high accuracy, it was discovered that both the surface shape and the rib precision of the satellite antenna deteriorate with decreasing temperature. The accuracy of the surface shape and ribs was lowest when the temperature node was −60 °C. The maximum root mean square errors (RMSEs) reached 0.878 mm and 0.761 mm, respectively. This indicates that the surface shape deformation error of the antenna caused by high and low temperatures is relatively high. However, the requirement for the technical design index (RMSE ≤ 1 mm for the surface shape accuracy of the antenna) is still met. Furthermore, for temperature differences of 40 °C and 80 °C, the measured RMSEs for the surface shape deformation were 0.216 mm and 0.411 mm, respectively. Overall, the technical design indicators (RMSE ≤ 0.3 mm and RMSE ≤ 0.5 mm, respectively) for the surface shape deformation of the antennas are met.

The crack detection of concrete bridges is an important link in the safety evaluation of bridge structures, and the rapid and accurate identification and detection of bridge cracks is a prerequisite for ensuring the safety and long-term stable use of bridges. To solve the incomplete crack detection and segmentation caused by the complex background and small proportion in the actual bridge crack images, this paper proposes a coarse–fine combined bridge crack detection method of “double detection + single segmentation” based on deep learning. To validate the effect and practicality of fine crack detection, images of old civil bridges and viaduct bridges against a complex background and images of a bridge crack against a simple background are used as datasets. You Only Look Once V5(x) (YOLOV5(x)) was preferred as the object detection network model (ODNM) to perform initial and fine detection of bridge cracks, respectively. Using U-Net as the optimal semantic segmentation network model (SSNM), the crack detection results are accurately segmented for fine crack detection. The test results showed that the initial crack detection using YOLOV5(x) was more comprehensive and preserved the original shape of bridge cracks. Second, based on the initial detection, YOLOV5(x) was adopted for fine crack detection, which can determine the location and shape of cracks more carefully and accurately. Finally, the U-Net model was used to segment the accurately detected cracks and achieved a maximum accuracy (AC) value of 98.37%. The experiment verifies the effectiveness and accuracy of this method, which not only provides a faster and more accurate method for fine detection of bridge cracks but also provides technical support for future automated detection and preventive maintenance of bridge structures and has practical value for bridge crack detection engineering.

Evapotranspiration (ET) is a critical component of the hydrological cycle, and it has a decisive impact on the ecosystem balance in arid and semi-arid regions. The Yinchuan Plain, located in the Gobi of Northwest China, has a strong surface ET, which has a significant impact on the regional water resource cycle. However, there is a current lack of high-resolution evapotranspiration datasets and a substantial amount of time is required for long-time series remote sensing evapotranspiration estimation. In order to assess the ET pattern in this region, we obtained the actual ET (ET_a) of the Yinchuan Plain between 1987 and 2020 using the Google Earth Engine (GEE) platform. Specifically, we used Landsat TM+/OLI remote sensing imagery and the GEE Surface Energy Balance Model (geeSEBAL) to analyze the spatial distribution pattern of ET over different seasons. We then reproduced the interannual variation in ET from 1987 to 2020, and statistically analyzed the distribution patterns and contributions of ET with regard to different land use types. The results show that (1) the daily ET_a of the Yinchuan Plain is the highest in the central lake wetland area in spring, with a maximum value of 4.32 mm day⁻¹; in summer, it is concentrated around the croplands and water bodies, with a maximum value of 6.90 mm day⁻¹; in autumn and winter, it is mainly concentrated around the water bodies and impervious areas, with maximum values of 3.93 and 1.56 mm day⁻¹, respectively. (2) From 1987 to 2020, the ET of the Yinchuan Plain showed an obvious upward and downward trend in some areas with significant land use changes, but the overall ET of the region remained relatively stable without dramatic fluctuations. (3) The ET_a values for different land use types in the Yinchuan Plain region are ranked as follows: water body > cultivated land > impervious > grassland > bare land. Our results showed that geeSEBAL is highly applicable in the Yinchuan Plain area. It allows for the accurate and detailed inversion of ET and has great potential for evaluating long-term ET in data-scarce areas due to its low meteorological sensitivity, which facilitates the study of the regional hydrological cycle and water governance.

A high-quality dataset is a basic requirement to ensure the training quality and prediction accuracy of a deep learning network model (DLNM). To explore the influence of label image accuracy on the performance of a concrete crack segmentation network model in a semantic segmentation dataset, this study uses three labelling strategies, namely pixel-level fine labelling, outer contour widening labelling and topological structure widening labelling, respectively, to generate crack label images and construct three sets of crack semantic segmentation datasets with different accuracy. Four semantic segmentation network models (SSNMs), U-Net, High-Resolution Net (HRNet)V2, Pyramid Scene Parsing Network (PSPNet) and DeepLabV3+, were used for learning and training. The results show that the datasets constructed from the crack label images with pix-el-level fine labelling are more conducive to improving the accuracy of the network model for crack image segmentation. The U-Net had the best performance among the four SSNMs. The Mean Intersection over Union (MIoU), Mean Pixel Accuracy (MPA) and Accuracy reached 85.47%, 90.86% and 98.66%, respectively. The average difference between the quantized width of the crack image segmentation obtained by U-Net and the real crack width was 0.734 pixels, the maximum difference was 1.997 pixels, and the minimum difference was 0.141 pixels. Therefore, to improve the segmentation accuracy of crack images, the pixel-level fine labelling strategy and U-Net are the best choices.

Lakes play a crucial role in regional economic development and ecological construction. The variation in lake water depth has a direct impact on local economic activities, such as agriculture, livestock farming, and fisheries, as well as the stability of hydrological conditions and water ecology. In response to the lack of unified evaluation in the application of remote sensing water-depth estimation models for inland lakes, this study systematically compares the performance of numerical models and machine learning models for water-depth estimation in inland lakes. A machine learning-based water-depth estimation model construction methodology suitable for inland lakes is proposed. This study introduces an innovative approach by integrating machine learning techniques with multispectral remote sensing data, improving the accuracy and applicability of water-depth estimation models for inland lakes. The results show the following: (1) The machine learning models based on random forest (RF), BP neural networks (BP), and AdaBoost demonstrate better performance (R² = 0.88, 0.72, and 0.61; MAE = 0.12 m, 0.24 m, and 0.31 m; RMSE = 0.32 m, 0.48 m, and 0.57 m) compared to the multi-band logarithmic ratio (MLR) model (R² = 0.59; MAE = 0.32 m; RMSE = 0.58 m); (2) the machine learning water-depth estimation model constructed based on this methodology exhibits improved precision (R² = 0.92, 0.89, and 0.80; MAE = 0.11 m, 0.17 m, and 0.25 m; RMSE = 0.25 m, 0.30 m, and 0.41 m). This suggests that the methodology is more suitable for the estimation of water depth in medium- and small-sized lakes; (3) The machine learning model developed in this study, combined with multispectral remote sensing imagery, achieves the accuracy required for the evaluation of water depths for practical water resources. This model enables the rapid acquisition of high-precision underwater three-dimensional topographic maps, providing more accurate and timely hydrological data support for lake water resource management.

A hyperspectral image classification method based on a mixed structure with a 3D multi-shortcut-link network (MSLN) was proposed for the features of few labeled samples, excess noise, and heterogeneous homogeneity of features in hyperspectral images. First, the spatial–spectral joint features of hyperspectral cube data were extracted through 3D convolution operation; then, the deep network was constructed and the 3D MSLN mixed structure was used to fuse shallow representational features and deep abstract features, while the hybrid activation function was utilized to ensure the integrity of nonlinear data. Finally, the global self-adaptive average pooling and L-softmax classifier were introduced to implement the terrain classification of hyperspectral images. The mixed structure proposed in this study could extract multi-channel features with a vast receptive field and reduce the continuous decay of shallow features while improving the utilization of representational features and enhancing the expressiveness of the deep network. The use of the dropout mechanism and L-softmax classifier endowed the learned features with a better generalization property and intraclass cohesion and interclass separation properties. Through experimental comparative analysis of six groups of datasets, the results showed that this method, compared with the existing deep-learning-based hyperspectral image classification methods, could satisfactorily address the issues of degeneration of the deep network and “the same object with distinct spectra, and distinct objects with the same spectrum.” It could also effectively improve the terrain classification accuracy of hyperspectral images, as evinced by the overall classification accuracies of all classes of terrain objects in the six groups of datasets: 97.698%, 98.851%, 99.54%, 97.961%, 97.698%, and 99.138%.