To address issues such as low efficiency in multi-source feature fusion and poor characterization of soil boundary features during soil element extraction in non-close-range areas, a military geological soil classification method based on spatial-frequency feature adaptive fusion network (SFAM-Net) is proposed. The spatial and spectral attention mechanisms are introduced in the encoder-decoder architecture, and a dual-branch feature extraction module is constructed to capture both spatial and spectral characteristics of soil in imagery. Features are extracted through the fusion of convolutional block attention modules, and an adaptive fusion mechanism is adopted to optimize the distribution of feature weights. Experimental results demonstrate that SFAM-Net can effectively distinguish three typical soil types, namely rocky soil, hard soil and ordinary soil, and their boundary information. The mean intersection over union (mIoU) and mean pixel accuracy (mPA) are 71.16% and 84.60% respectively, which are 5.46 percentage points and 8.11 percentage points higher than the best result of the baseline model. It is verified that the accuracy of military geological soil classification are improved by using the method.
PUNDIRS K, GARGR D. Development of rule based approach for assessment of off-road trafficability using remote sensing and ancillary data[J]. Quaternary International, 2021,575/576:308-316.
LUY, HEK, XUH R, et al. Remote-sensing interpretation for soil elements using adaptive feature fusion network[J]. IEEE Transactions on Geoscience and Remote Sensing, 2023,61:No.4505515.
[10]
HANW, LIJ, WANGS, et al. Geological remote sensing interpretation using deep learning feature and an adaptive multisource data fusion network[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022,60:No.4510314.
PADMAPRIYAJ, SASILATHAT. Deep learning based multi-labelled soil classification and empirical estimation toward sustainable agriculture[J]. Engineering Applications of Artificial Intelligence, 2023,119:No.105690.
[14]
JALAPURS, PATILN. An integrated deep learning framework for soil type classification[J]. SN Computer Science, 2025,6(3):251.
[15]
WOO S, PARKJ, LEEJ Y, et al. CBAM: convolutional block attention module[C]∥Proceedings of the European Conference on Computer Vision 2018. Cham, Switzerland: Springer, 2018:3-19.
[16]
郭宏伟.顾及军事地质要素的越野路径规划研究[D].郑州:信息工程大学,2022:8-15.
[17]
CHENL C, ZHUY K, PAPANDREOUG, et al. Encoder-decoder with atrous separable convolution for semantic image segmentation[C]∥Proceedings of the European Conference on Computer Vision 2018. Cham, Switzerland: Springer, 2018:833-851.