Precipitation intensity estimation method based on deep learning

Abstract

The invention discloses a precipitation intensity estimation method based on deep learning. The method comprises: acquiring meteorological satellite data and precipitation data respectively; accordingto the acquired meteorological satellite data, cutting a required estimation region out from the meteorological satellite data, correcting and storing the estimation region in an array form; according to the obtained rainfall data, resampling to a required spatial resolution, and classifying the rainfall data to obtain rainfall intensity labels of different levels; converting the rainfall intensity label into single-channel images with only background and different levels of rainfall intensity, cutting the single-channel images into different sizes, and respectively taking the single-channelimages as input and labels of a rainfall intensity estimation model; establishing a precipitation intensity estimation model based on deep learning; training to obtain an optimal model; testing new meteorological satellite data, and generating a complete rainfall intensity estimation result; and superposing the generated rainfall intensity estimation result to the shp terrain file. According to the method, the corresponding rainfall intensity can be accurately estimated, and high-precision rainfall intensity estimation is realized.

Description

technical field [0001] The invention relates to a method for estimating precipitation intensity based on deep learning, which belongs to the technical field of remote sensing image processing. Background technique [0002] In recent years, with the continuous development of computer vision technology, machine learning technology has begun to be applied to traditional remote sensing weather forecasting and weather monitoring industries to improve the accuracy of forecasting and monitoring. Among them, the target detection and segmentation technology can divide each pixel in the picture into its own category, and use these technologies to help weather forecasters improve forecasting efficiency and forecasting accuracy, and solve the problem that the time resolution of existing precipitation estimation algorithms is not high. The problem of non-real-time release of precipitation estimation products can strengthen the monitoring of disaster weather, so it has broad application p...

AI Technical Summary

Problems solved by technology

PERSIAN-SDAE uses infrared and water vapor channels as input data to build a multi-layer automatic noise reduction encoder to detect whether precipitation has occurred and the amount of precipitation through automatic feature extraction. Information that is useful for precipitation estimation is obtained between adjacent pixels

PERSIANN-CNN uses a convolutional neural network to establish a precipitation estimation model, using infrared and microwave data as input. Although it has more realistic and reliable detection and precipitation estimation results than PERSIANN-CCS and PERSIANN-SDAE, it does not make full use of infrared With the high-dimensional feature information of the water vapor band, and the multi-layer pooling will lose the corresponding feature information, it is unnecessary to use the infrared and water vapor bands as input and then perform feature fusion

Images