The invention discloses a remote sensing image cultivated land information extraction method based on a convolutional neural network, and the method comprises the following steps: A, data preprocessing: carrying out the processing of remote sensing image data through ENVI5.3 software, including radiometric calibration, ortho-rectification, image fusion, information enhancement and image cutting, and finally obtaining a remote sensing image with high resolution; B, constructing of a remote sensing image library; C, establishing and training of a convolutional neural network model; D, extracting and classifying of remote sensing image information; E, spatial statistical analysis on cultivated land information. According to the invention, landsat8 image data of a certain area is used as a data source, ENVI5.3 software is used for preprocessing the image data, the time-consuming and labor-consuming problem of visual interpretation can be solved, remote sensing image data can be accurately recognized and classified, a feasible technical method is provided for research and application about remote sensing image extraction and classification in the future, the cultivated land management efficiency in China is improved, and agricultural monitoring is more scientific and modernized.