Abstract: In order to better evaluate the diffusion range and temperature rise of thermal drainage, it is necessary to study the background temperature extraction method based on the monitoring and predication data of thermal drainage of nuclear power plants. Taking the Qinshan nuclear power plants as an example, this paper proposes a more accurate and faster new method, temperature gradient method, based on the commonly used background temperature extraction method from remote sensing imagery. The method calculates the divergence value of the critical temperature gradient and the temperature spatial gradient, overlays them to obtain the outermost temperature envelope containing the main heat source, and subsequently identifies the extraction range for the background temperature before computing it. Compared with the traditional method, this method directly calculates the background temperature by using the temperature gradient. The calculation principle is clear, the physical significance is unequivocal, and no additional data is required, which improves the accuracy and convenience of background temperature extraction. It can be applied to the background temperature extraction of remote sensing image of temperature drainage in coastal nuclear power plants or even numerical simulation data.