Abstract—The optimal data selection technique mainly studies which images and the contents of the images should be selected for processing in the process of natural resource survey using optical remote sensing satellites, so as to obtain the best processing effect. This paper realizes a data selection method for optical remote sensing satellite images. The method uses the image semantic segmentation technology to achieve refined cloud detection, uses the various indicator comprehensively evaluating the image quality, uses the regional coverage technology to form an optimal data selection and organization scheme. It can quickly select a set of optimal data sets that can cover the target area and ensure image quality from the massive candidate data according to the time and space range specified by the user. The method is applied to the optimal data selection software module of optical satellite for fine classification of forest land. The results show that the algorithm can effectively replace the manual visual data selection method and provide technical support for various types of natural resources survey tasks