CN-122023153-A - Method for realizing vegetation index topography correction based on adaptive parameters

CN122023153ACN 122023153 ACN122023153 ACN 122023153ACN-122023153-A

Abstract

The invention discloses a method for realizing vegetation index topography correction based on self-adaptive parameters, which comprises the steps of obtaining remote sensing topography images to be corrected, loading a vegetation image correction model, realizing remote sensing topography image correction by the vegetation image correction model through a spatial distribution diagram of the self-adaptive parameters, inputting the remote sensing topography images into the vegetation image correction model, and outputting corrected vegetation index images, wherein the spatial distribution diagram of the self-adaptive parameters is output by the self-adaptive parameter model. According to the technical scheme, the terrain correction can be realized by only utilizing the self-wave band data of the remote sensing image under the condition of no DEM assistance through the self-adaptive parameter adjustment mechanism, the application range can be further expanded, the synchronous elimination of the principal shadows and the falling shadows is realized, the index values of similar vegetation in different illumination areas are ensured to be consistent, meanwhile, the apparent reflectivity and the earth surface reflectivity data subjected to atmosphere correction are adapted, and the application scene universality is realized.

Inventors

ZHANG YONG
HE YAO
XU JING
CAO SHUANGHE
ZHAO JIAN
WANG DEHONG
ZHAO YASONG
LIU XIANGBIN
REN WENLONG
ZHAO LE

Assignees

中国电建集团贵州电力设计研究院有限公司

Dates

Publication Date: 20260512
Application Date: 20251230

Claims (9)

1. The method for realizing vegetation index topography correction based on the adaptive parameters is characterized by comprising the following steps: acquiring a remote sensing topographic image to be corrected; Loading a vegetation image correction model, wherein the vegetation image correction model realizes remote sensing terrain image correction through a spatial distribution diagram of self-adaptive parameters; inputting the remote sensing topographic image into the vegetation image correction model, and outputting a corrected vegetation index image; the spatial distribution map of the adaptive parameter is output by the adaptive parameter model.
2. The method for implementing vegetation index topography correction based on adaptive parameters according to claim 1, wherein the construction method of the adaptive parameter model comprises: Acquiring multi-source remote sensing data, preprocessing the multi-source remote sensing data, extracting characteristic data in each sample area, and constructing a sample characteristic set; Defining adaptive parameters The adaptive parameters The method is used for basic regulation of vegetation index balance of yin-yang slopes; Associating a sample feature set with corresponding adaptive parameters Constructing a feature-label training data set; defining an adaptive parameter model, training the adaptive parameter model by adopting a feature-label training data set, wherein the adaptive parameter model is used for generating adaptive parameters according to image features 。
3. The method for implementing vegetation index topography correction based on adaptive parameters according to claim 2, wherein the characteristic data comprises NIR band reflectance, red band reflectance, slope direction, sun incidence angle cosine value, vegetation type probability, the sample area comprises sunny slope, own shadow area, and falling shadow area, wherein the NIR band reflectance and Red band reflectance respectively comprise sunny slope and cloudy slope data, respectively expressed as: 、 And 、。
4. The method for implementing vegetation index topography correction based on adaptive parameters according to claim 2, wherein the method for extracting the feature data includes a principal image extraction and a falling image extraction; The ghost extraction is used for evaluating the anti-topographic influence; the falling shadow extraction means that after the topographic shadow is extracted from the panoramic image characteristics, the result of the extraction of the original shadow is subtracted from the topographic shadow extraction result.
5. The method for implementing vegetation index terrain correction based on adaptive parameters according to claim 2, wherein the spatial distribution map of the adaptive parameter model output adaptive parameters refers to: Inputting the full-image characteristic data into the self-adaptive parameter model, and predicting pixel by pixel to obtain the self-adaptive parameter Thereafter, adaptive parameters are generated by image processing Is a spatial distribution map of (c).
6. A method for implementing vegetation index topography correction based on adaptive parameters according to claim 3, wherein the adaptive parameters are defined In this case, different adaptive parameters are set for different derived models Comprises the following steps: Based on RVI derivative model, adaptive parameters The calculation method of (1) is expressed as follows: , wherein, , ; Adaptive parameters based on NDVI derived model The calculation method of (1) is expressed as follows: 。
7. The method for implementing vegetation index topography correction based on adaptive parameters of claim 2, wherein the source of the multi-source telemetry data comprises: remote sensing image data comprising NIR and Red core bands; the terrain auxiliary data is used for extracting gradient and slope direction; reflectance data, including apparent reflectance and earth's surface reflectance after atmospheric correction.
8. The method for implementing vegetation index topography correction based on adaptive parameters according to claim 2, wherein the preprocessing includes outlier detection and removal, data normalization and missing value processing.
9. The method for realizing vegetation index topography correction based on adaptive parameters according to claim 1, wherein the vegetation image correction model is a dual-version correction model, comprising an RVI derivative version and an NDVI derivative version; The vegetation image correction model suitable for the RVI derivative version is SRESEVI-A, is applied to vegetation coverage monitoring of a high-density area, and is expressed as follows: ; The vegetation image correction model applicable to the NDVI derivative version is SRESEVI-B, and is applied to the evaluation of the vegetation growth vigor in a low-density area, and is expressed as follows: 。

Description

Method for realizing vegetation index topography correction based on adaptive parameters Technical Field The invention relates to the technical field of machine learning combined with remote sensing image processing and vegetation monitoring, in particular to a method for realizing vegetation index topography correction based on self-adaptive parameters. Background The vegetation indexes (such as NDVI and RVI) are core indexes for remotely monitoring the vegetation growth conditions, are obviously influenced by terrains in mountain areas, have uneven solar radiation distribution caused by gradient and slope direction differences, have high direct radiation received by a sunny slope and high vegetation indexes, have underestimation on vegetation indexes in shadow slopes (shadow areas) and falling shadow areas, form a phenomenon of 'same object and different spectrum', and seriously influence the accuracy of vegetation estimation. The current mainstream vegetation index implementation terrain correction method has obvious defects that firstly, a correction method based on DEM (such as C correction and SCS+C correction) needs to rely on auxiliary data of a digital elevation model, cannot be applied to a data missing area, can only partially eliminate the influence of a ghost, is basically ineffective to the ghost correction, secondly, a traditional wave band combination model (such as SEVI and TAVI) is difficult to adapt to local complex terrain by introducing fixed adjustment parameters through global fitting, the correction precision is limited, thirdly, the existing method has low sensitivity to data of Surface Reflectivity (SR), the data source correction effect after the atmospheric correction is not obvious, and the degree of automation is low due to the fact that parameters are set by multiple manual experiences. Along with the wide application of remote sensing technology in the fields of ecological monitoring, agricultural investigation and the like, the demands for the correction of vegetation indexes in complex terrain areas are increasingly urgent. The existing method can not simultaneously meet the core requirements of 'no auxiliary data is needed, the original shadow and the falling shadow are considered, multiple data sources are adapted, and local high precision' are met, so that the interpretation precision of mountain vegetation remote sensing information is limited. Although partial researches try to improve the terrain resistance through band combination optimization, the problem of vegetation index balance of yin-yang slopes is not solved from the mathematical mechanism, and the correction effect still has room for improvement. Therefore, a method for realizing terrain correction based on mathematical mechanism, self-adaptive adjustment and adaptation of vegetation indexes of multiple data sources is needed, and the current technical bottleneck is broken through. Disclosure of Invention In order to achieve the above object, the present application provides a method for implementing vegetation index topography correction based on adaptive parameters, comprising the following steps: acquiring a remote sensing topographic image to be corrected; loading a vegetation image correction model, wherein the vegetation image correction model realizes remote sensing terrain image correction through a spatial distribution diagram of the self-adaptive parameters; inputting a remote sensing terrain image into the vegetation image correction model, and outputting a corrected vegetation index image; wherein the spatial distribution map of the adaptive parameter is output by the adaptive parameter model. The construction method of the self-adaptive parameter model comprises the following steps: Acquiring multi-source remote sensing data, preprocessing the multi-source remote sensing data, extracting characteristic data in each sample area, and constructing a sample characteristic set; Defining adaptive parameters Adaptive parametersThe method is used for basic regulation of vegetation index balance of yin-yang slopes; Associating a sample feature set with corresponding adaptive parameters Constructing a feature-label training data set; Defining an adaptive parameter model, training the adaptive parameter model by using a feature-label training data set, wherein the adaptive parameter model is used for generating adaptive parameters according to image features 。 The characteristic data comprise NIR wave band reflectivity, red wave band reflectivity, gradient, slope direction, sun incidence angle cosine value and vegetation type probability, the sample area comprises a sunny slope, a present shadow area and a falling shadow area, the NIR wave band reflectivity and the Red wave band reflectivity respectively comprise sunny slope and cloudy slope data, and the data are respectively expressed as:、 And 、。 The method for extracting the characteristic data comprises the steps of principal image extraction and falli