Method and device for performing vegetation parameter remote sensing retrieval in neural network system

Abstract

The invention provides a method and device for performing vegetation parameter remote sensing retrieval in a neural network system. The method comprises the steps as follows: training data sets between vegetation parameters and earth surface reflectance of different types of earth surfaces are obtained according to simulation models of earth surface reflectance of different types of earth surfaces; a neutral network between the vegetation parameters and earth surface reflectance of different types of earth surfaces is established; training data sets between earth surface reflectance and apparent reflectance under different meteorological conditions are obtained according to atmospheric radiation models; a neutral network between the earth surface reflectance and the apparent reflectance under different atmospheric conditions is established; multiple neutral networks are combined to form a neural network system; and remote sensing data obtained after radiometric calibration or atmospheric correction is subjected to vegetation parameter remote sensing retrieval in the neural network system. According to the method and the device, multiple different neutral networks can be combined to form the neural network system, compared with the prior art, the data analog capability is stronger, and an obtained result is more accurate.

Description

technical field [0001] The invention relates to the field of image analysis, in particular to a method and device for remote sensing inversion of vegetation parameters by a neural network system. Background technique [0002] Through data analysis and processing of satellite remote sensing images, some vegetation parameters that cannot be represented by ground measurements can be obtained on a macro scale. For example, estimating the leaf area index (LAI), that is, the single-sided area sum of all leaves per unit area, can well characterize the evapotranspiration and soil and material exchange processes in the soil-vegetation-atmosphere transmission process, which is also a vegetation function. An important driver of the model. [0003] As a new method of remote sensing image analysis and processing, neural network has been used to estimate LAI of vegetation land types from remote sensing data. In the prior art, the neural network is used to estimate the LAI, which has the...

AI Technical Summary

Problems solved by technology

[0006] The present invention provides a vegetation parameter remote sensing inversion method and device of a neural network system to solve the technical problem in the prior art that it is impossible to perform high-precision nonlinear simulations on different vegetation structure types.

Images