Scene classification, aiming to identify the land-cover categories of remotely sensed image patches, is now a fundamental task in the remote sensing image analysis field. Deep-learning-model-based algorithms are widely applied in scene classification and achieve remarkable performance, but these high-level methods are computationally expensive and time-consuming. Consequently in this paper, we introduce a knowledge distillation framework, currently a mainstream model compression method, into remote sensing scene classification to improve the performance of smaller and shallower network models. Our knowledge distillation training method makes the high-temperature softmax output of a small and shallow student model match the large and deep teacher model. In our experiments, we evaluate knowledge distillation training method for remote sensing scene classification on four public datasets: AID dataset, UCMerced dataset, NWPU-RESISC dataset, and EuroSAT dataset. Results show that our proposed training method was effective and increased overall accuracy (3% in AID experiments, 5% in UCMerced experiments, 1% in NWPU-RESISC and EuroSAT experiments) for small and shallow models. We further explored the performance of the student model on small and unbalanced datasets. Our findings indicate that knowledge distillation can improve the performance of small network models on datasets with lower spatial resolution images, numerous categories, as well as fewer training samples.

The combination of lidar and digital photography provides a new technology for creating a high-resolution 3D digital outcrop model. The digital outcrop model can accurately and conveniently depict the surface 3D properties of an outcrop profile, making up for the shortcomings of traditional outcrop research techniques. However, the advent of digital outcrop poses additional challenges to the 3D spatial analysis of virtual outcrop models, particularly in the interpretation of geological characteristics. In this study, the detailed workflow of automated interpretation of geological characteristics of fractures and cavities on a 3D digital outcrop texture model is described. Firstly, advanced automatic image analysis technology is used to detect the 2D contour of the fractures and cavities in the picture. Then, to obtain an accurate representation of the 3D structure of the fractures and cavities on the digital outcrop model, a projection method for converting 2D coordinates to 3D space based on geometric transformations such as affine transformation and linear interpolation is proposed. Quantitative data on the size, shape, and distribution of geological features are calculated using this information. Finally, a novel and comprehensive automated 3D quantitative characterization technique for fractures and cavities on the 3D digital outcrop texture model is developed. The proposed technology has been applied to the 3D mapping and quantitative characterization of fractures and cavities on the outcrop profile for the Dengying Formation (second member), providing a foundation for profile reservoir appraisal in the research region. Furthermore, this approach may be extended to the 3D characterization and analysis of any point, line, and surface objects derived from outcrop photos, hence increasing the application value of the 3D digital outcrop model.