The Using of Satellite Images to Evaluate Vegetation Processes Based on Cloud Detection

This paper presents a solution to the problem of vegetativity prediction based on time series models. At the same time, special attention is paid to the development of a filter of images with increased cloudiness. The classification accuracy of distorted images at above 95% based on convolutional neural networks is obtained. The normalized difference vegetation index prediction error using an ensemble of neural networks is less than 0.1. It is shown how the developed algorithms can be used for differential fertilizer application. Expressions for calculating normalized difference vegetation index and techniques for smoothing indices in case of missing data and filtering cloud images are presented. The results obtained in this paper can be useful for specialists engaged in processing remote sensing data from space, and the algorithms for filtering cloud images can be used in solving other applied problems, such as fire monitoring.