Abstract: The extraction of river regions from remote sensing images is of great theoretical and practical significance for grasping the hydrological characteristics,environmental protection and development of a certain region. The ground objects in remote sensing images are complex and diverse. In order to obtain high-precision river area information and make up for the limited applicability of single-characteristic river extraction,a river region extraction method that integrates texture,spectral and shape features is proposed. Firstly,the infrared channel of the image is selected as the input,the gray level co-occurrence matrix is calculated according to the texture feature,and the angular secondorder moment map in the statistics of the gray level co-occurrence matrix is selected as the processing object. Then,maximum inter-class variance（OTSU）calculation is conducted according to the diagonal second moment map of the spectral characteristics to obtain a binary image. Then,the connected domain is marked,and a specific filter is constructed according to the geometric features to filter out the non-river noise,and finally a high-precision river basin map is generated. Based on the remote sensing images of Taishan City obtained by the GF-1satellite,the intercepted images include different types of river basins such as forest land,urban,mountainous,and cultivated land. This method and other algorithms for river extraction is used and then the performances are compared. These include the OTSU algorithm commonly used in remote sensing image segmentation,the SWT algorithm proposed by some scholars in recent years,and the U-Net algorithm in the fully convolutional neural network segmentation method. The results show that the method in this paper has obvious advantages in terms of accuracy and completeness. It is not easily affected by changes in the surrounding environment of the river,and has strong robustness. In particular,it has high practicability for extracting the global river basin of a certain region. The method in this paper comprehensively takes into account the texture,grayscale and geometric characteristics of the river,and carries out a comprehensive and effective feature description,which can maintain the integrity of the river basin to the greatest extent and can effectively suppress the noise adhesion phenomenon that occurs in other algorithms. It can effectively improve the accuracy and completeness of the river basin extraction.