Soil salinity due to irrigation diversion affects regional agriculture, and the development of soil composition estimation models for the dynamic monitoring of regional salinity is important for salinity control. In this study, we evaluated the performance of hyperspectral data measured using an analytical spectral device (ASD) field spec standard-res hand-held spectrometer and satellite sensor visible shortwave infrared advanced hyperspectral imager (AHSI) in estimating the soil salt content (SSC). First derivative analysis (FDA) and principal component analysis (PCA) were applied to the data using the raw spectra (RS) to select the best model input data. We tested the ability of these three groups of data as input data for partial least squares regression (PLSR), principal component regression (PCR), and multiple linear regression (MLR). Finally, an estimation model of the SSC, Na⁺, Cl⁻, and SO₄²⁻ contents was established using the best input data and modeling method, and a spatial distribution map of the soil composition content was drawn. The results show that the soil spectra obtained from the satellite hyperspectral data (AHSI) and laboratory spectral data (ASD) were consistent when the SSC was low, and as the SSC increased, the spectral curves of the ASD data showed little change in the curve characteristics, while the AHSI data showed more pronounced features, and this change was manifested in the AHSI images as darker pixels with a lower SSC and brighter pixels with a higher SSC. The AHSI data demonstrated a strong response to the change in SSC; therefore, the AHSI data had a greater advantage compared with the ASD data in estimating the soil salt content. In the modeling process, RS performed the best in estimating the SSC and Na⁺ content, with the R² reaching 0.79 and 0.58, respectively, and obtaining low root mean squared error (RMSE) values. FDA and PCA performed the best in estimating Cl⁻ and SO₄²⁻, while MLR outperformed PLSR and PCR in estimating the content of the soil components in the region. In addition, the hyperspectral camera data used in this study were very cost-effective and can potentially be used for the evaluation of soil salinization with a wide range and high accuracy, thus reducing the errors associated with the collection of individual samples using hand-held hyperspectral instruments.

Addressing the critical challenges of complex semantic segmentation in coal mine areas, this study proposes a precise recognition method based on Semantic-SAM. We developed the SAM-SEF (SAM-based Semantic Enhancement Framework) framework, which innovatively integrates the zero-shot segmentation capability of Semantic-SAM with specialized deep learning models. Through a “segmentation-classification-update” strategy, accurate segmentation is achieved. The framework initially employs Semantic-SAM to obtain category-agnostic multi-granularity segmentation results, followed by semantic understanding of the segmented regions using specialized deep learning models, and finally generates precise segmentation results through a category mapping mechanism. Experiments evaluated 18 deep learning models, including ResNet series, Vision Transformer series, classical convolutional neural networks, and lightweight networks. Results indicate that the SAM-SEF-ENet-B0 model, based on the SAM-SEF framework, performs most prominently in segmenting Open-pit Coal Mining Sites (OCMS) and Composite Coal-related Sites (CCS), achieving IoUs of 93.68% and 81.52%, respectively, and an mIoU of 87.60%, significantly outperforming traditional segmentation models. Additionally, a semi-automatic annotation tool based on Semantic-SAM was developed, reducing traditional manual annotation time by approximately 60% through pre-segmentation and interactive editing functions. A multi-scale evaluation system based on the 33% and 66% percentiles of target area was established, revealing a clear positive correlation between target scale and segmentation performance, with small-scale target segmentation being the primary challenge (IoU of 0.79 for OCMS and 0.67 for CCS in small scales). Systematic experiments validated the necessity of selecting Semantic-SAM, showing that Semantic-SAM has significant advantages over SAM (mIoU of 54.03%) and SAM2 (mIoU of 45.33%) in semantic understanding capability, architectural adaptability, and computational efficiency. The study also found that classification networks based on the SAM-SEF framework achieved superior performance in segmentation tasks compared to specially designed segmentation networks, demonstrating the unique advantages of decoupling semantic segmentation tasks into “general segmentation + specialized classification” strategies in specific task domains. This research not only provides a new technical solution for remote sensing monitoring of coal mine areas but also offers valuable insights for the development of remote sensing image analysis, particularly demonstrating the effectiveness of a “classification-driven segmentation” paradigm where lightweight classifiers operating on purified targets from a frozen foundation model can outperform traditional end-to-end segmentation networks in specific, complex domains.