The organic matter content of saline soils is an important biochemical indicator for evaluating the effectiveness of saline land improvement. Therefore, monitoring soil organic matter content rapidly and accurately is the key to realizing accurate monitoring of the degree of saline soil improvement. Hyperspectral image technology allows rapid inversion of soil organic matter content.Using measured hyperspectral data and soil organic matter content of saline-amended soils as raw data and extraction of spatial information by GLCM to form texture features fused with hyperspectral reflectance mapping. Construction of a convolutional network model and inversion of soil organic matter content in saline-amended soils and comparison with hyperspectral reflectance data processing. The results show that compared to other data processing methods, the best prediction model is inverted using the inversion of the texture features extracted with GLCM fused with the hyperspectral reflectance maps, the method of image and spectrum fusion, with is 0.9869, which is a 9.85% enhancement compared to the hyperspectral reflectance inversion method, and is 9.7192, which is a decrease of 65.9210 compared to the hyperspectral reflectance inversion method.