Abstract: In order to solve the problem of modifying the host image when the traditional image watermarking algorithm protects the copyright of remote sensing image, and to improve the robustness of remote sensing image watermarking algorithm, a remote sensing image zero watermarking algorithm based on bi-dimensional empirical mode decomposition (BEMD) and discrete Fourier transform (DFT) is proposed. Firstly, BEMD is used to decompose the host image to get the intrinsic mode function and margin, and the residual information is divided into 8×8 blocks DFT, and the high frequency coefficients are constructed into coefficient images. Then Arnold transform is used to scramble the watermark image to improve the security of the watermark image. Then the size of the watermark image is adjusted to solve the problem of image size matching between the host image and the watermark image in the process of watermark embedding. Finally, the coefficient image decomposed by DFT and the watermark image are constructed into feature image by XOR operation, and the construction of zero watermark in remote sensing image is realized. The experimental results of a large number of remote sensing images and watermark images show that the proposed algorithm can effectively resist all kinds of attacks such as filtering, sharpening, histogram equalization, Gamma transform, noise, scaling and so on. In the case of 12 attacks and 60 parameters, 95.0% of the extracted watermark images have a normalized correlation value of more than 0.990 with the original watermark image, compared with the traditional remote sensing image watermarking algorithm. The algorithm ensures the harmlessness of the host image, so that the feature image saved in the notary office for copyright protection verification will not cause any damage to the host image, and has better robustness.