Satellite cloud image cloud amount calculation method based on multi-dimensional dense connection convolutional neural network

Abstract

The invention provides a satellite cloud image cloud amount calculation method based on a multi-dimensional dense connection convolutional neural network. The method comprises the following steps: step 1) training of the multi-dimensional dense connection neural network model structure: learning of the multidimensional dense connection neural network is performed through the marked sample (X, Y) so as to obtain the optimal model parameters; and the output layer of the network model uses the supervised learning mode to realize cloud classification learning; step 2) classification of the satellite image cloud: the satellite image is divided into small blocks of each pixel size to act as the input data of the neural network and the output of the whole neural network is obtained, final classification is performed through the output value and the type of the cloud is judged according to the maximum output value; and step 3) calculation of the cloud amount in the satellite image: the total cloud amount on the cloud image is calculated by using the spatial correlation method according to the type of the cloud in the step 1). The beneficial effects are that the cloud image features are extracted well, the detection accuracy of the multi-spectral cloud image is improved and the generalization ability is greatly improved in comparison with that of other models.

Description

technical field [0001] The invention belongs to the technical field of cloud image processing, and in particular relates to a cloud amount calculation method for satellite cloud images based on a multi-dimensional densely connected convolutional neural network. Background technique [0002] Clouds cover more than 50% of the earth's surface and are one of the important meteorological and climatic elements. In order to obtain accurate cloud amount distribution, it is necessary to detect and classify clouds on satellite cloud images first, and then perform cloud amount calculation on the basis of cloud classification. At present, international satellite cloud calculation methods mainly include ISCCP method, CLAVR method, APOLLO method, etc. The ISCCP algorithm assumes that the observed radiation value only comes from either cloud or clear sky. Yes, if the difference between the two is greater than the maximum change range of the clear sky radiation itself, it is determined tha...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to overcome the deficiencies of the above-mentioned background technology, to provide a method for calculating the cloud amount of satellite cloud images based on multi-dimensional densely connected convolutional neural networks, which overcomes the defects of insufficient utilization of cloud image features by traditional neural networks, and overcomes the traditional deep enhancement method. In learning, the number of convolutional neural network layers is deepened, and the calculation parameters are large, and the signal transmission is slow.

Images