Time synchronization of the data of a training sample and the data of recognized objects in classification problems in the course of interpretation of space images of the Earth’s surface

The classification problem is considered under the conditions when there is temporal non-coordination in the data of a training sample and the data of the recognized object. A modification of the conventional classification scheme is proposed that presumes that an additional stage is involved, whose goal is to temporally match the data of the training sample to the current observation moment. The feature vectors are recalculated at the current observation moment by calculating their conditional expectations connecting the realizations of feature vectors at the time instants taking place in the training sample with their values for the current observation moment. By the example of processing and analyzing space images, it is demonstrated that the use in the course of classification of the proposed procedure of synchronizing the data of the training sample and the recognized object can substantially improve the reliability of classification results.