CN-121982518-A - Remote sensing space-time data fusion method based on physical constraint and meteorological drive

CN121982518ACN 121982518 ACN121982518 ACN 121982518ACN-121982518-A

Abstract

The invention relates to a remote sensing space-time data fusion method based on physical constraint and meteorological drive, and belongs to the technical field of data processing. The method comprises the steps of performing three-step operation of regression modeling, spatial filtering and residual error compensation on a high-spatial resolution remote sensing image and a high-temporal resolution remote sensing image, performing preliminary space-time fusion on a fine resolution image and a coarse resolution image, establishing a weather driving model between weather variable change and short-term response of a vegetation index, extracting a time sequence of a sample region from high-temporal resolution data, performing nonlinear curve fitting on the time sequence by using a Gaussian model, establishing a weather growth model, performing parameter solving and optimization on the model to obtain a region representative weather model of the crop type, and generating a final high-spatial-temporal resolution image sequence by using weather constraint and weather driving correction on a preliminary space-time fusion result. The method can generate a high space-time resolution remote sensing product with lower noise, more stable climatic features and more credible characteristics.

Inventors

LU KE
JI YAO
Zan Youming
ZHANG JIAN
WANG JIE

Assignees

浪潮光音卫星技术(山东)有限公司

Dates

Publication Date: 20260505
Application Date: 20251223

Claims (9)

1. A remote sensing space-time data fusion method based on physical constraint and meteorological drive is characterized by comprising the following steps: s1, acquiring a high-spatial-resolution remote sensing image, a high-temporal-resolution remote sensing image and contemporaneous weather product data, preprocessing, and calculating a vegetation index WDRVI to carry out surface coverage classification; S2, modeling by meteorological driving: constructing a meteorological driving factor based on meteorological analysis data, and establishing a mapping relation between meteorological variable change and vegetation index short-term response to form a meteorological driving model for describing the influence of short-time meteorological disturbance on the vegetation index; s3, performing physical modeling: Selecting a typical sample land block of a target crop according to the earth surface coverage classification, extracting a time sequence of a sample region from high-time resolution data, performing nonlinear curve fitting on the time sequence by using a Gaussian model, establishing a weathered growth model, performing parameter solving and optimizing on the model to obtain a region representative weathered model of the crop type, extracting a key weathered period from a fitted curve by using the weathered model, establishing a weathered constraint model by using the fitted weathered curve as a priori constraint function of the time sequence, and performing space-time popularization on the weathered constraint model to generate a pixel-level weathered constraint function; S4, performing preliminary space-time fusion on the fine resolution image and the coarse resolution image by using the high-spatial resolution remote sensing image and the high-temporal resolution remote sensing image through three steps of regression modeling, spatial filtering and residual error compensation; S5, calculating deviation indexes of the preliminary space-time fusion result and the predicted value of the weather model curve, detecting abnormality, correcting a normal result by using a weather constraint function, correcting a large deviation or abnormal result by using a weather constraint function, calculating weather weight of a weather driving factor, calculating short-term change of a vegetation index according to the weight, performing weather driving correction on the preliminary space-time fusion result, and fusing the weather constraint and the weather driving correction result to generate a final high space-time resolution image sequence.
2. The remote sensing spatiotemporal data fusion method based on weather constraints and weather driving according to claim 1, wherein step S1 weather product data utilizes ERA5 product weather data.
3. The remote sensing space-time data fusion method based on the weather constraint and the weather driving according to claim 1, wherein the construction process of the weather driving model in the step S2 is as follows: 1) Based on historical data, researching influence degree of meteorological driving factors on vegetation indexes to obtain sensitivity coefficients of meteorological factors ; 2) Weather weight calculation: , Wherein the method comprises the steps of For the growth season sensitivity factor, determining according to the calculated vegetation index value, if the vegetation index is larger than the fixed threshold value, the growth season is assumed to be entered, if At the stage of growing season If it is not in the growing season , According to the overall quality of the remote sensing data, when the overall quality of the remote sensing data is higher Taking larger value, when the remote sensing data quality is higher overall Taking a smaller value; 3) Predicting short-term increment of vegetation index by short-term change of vegetation index Expressed as a linear combination of the amount of change in a plurality of weather factors, each weather factor corresponding to a regression coefficient indicating the sensitivity of the variable to the vegetation index,
4. Wherein, the Describing the strength of the effect of the meteorological variable on the vegetation status, The deviation of the meteorological variable from the reference state is described as the variation of the kth meteorological variable from the mean value of the climate over years at the time t.
5. The method for remote sensing spatiotemporal data fusion based on climatic constraints and meteorological drive according to claim 1, wherein the climatic growth model and the climatic curve in step S3 The calculation formula of (2) is as follows: , wherein a represents peak amplitude and reflects the growth intensity of crops, b represents peak time and corresponds to the key period of crop growth, c represents curve width parameter and reflects the duration of growing season, and t represents time variable for describing the change of vegetation index along with time.
6. The remote sensing space-time data fusion method based on physical constraint and meteorological drive according to claim 1, wherein in step S3, the parameter solving and optimizing adopts nonlinear least square fitting parameters, so that model curves are optimally matched with observed data, a group of parameters are stored for each sample land parcel, and the weighted average is carried out on the land parcel areas of the same type, so that the area representative physical model of the crop type is obtained.
7. The method for remote sensing spatiotemporal data fusion based on climatic constraints and meteorological drive according to claim 1, wherein in step S3, a climatic curve obtained by fitting is obtained As a priori constraint function of time series, for any pixel If the time sequence variation deviates from the curve obviously, the constraint correction is carried out by the following constraint function: , Wherein the method comprises the steps of For the weatherconstrained weight coefficient, the judgment is carried out according to the overall quality of the remote sensing image, if the image quality is poor, the coefficient value is larger, and WDRVI obs (x, y, t) represents the result after primary fusion.
8. The remote sensing spatiotemporal data fusion method based on physical constraint and meteorological drive according to claim 1, wherein the spatiotemporal popularization process of the step S3 is as follows: 1) Parameterized result preparation: acquiring a Gaussian model parameter set obtained by sample area fitting: i=1,2,...,N, where N is the number of sample regions; Building a time-sequential climate vector for each pixel: , Wherein the method comprises the steps of Is used as an index of the accumulated temperature, Is a water stress factor; 2) Calculating the grid space distance, namely, the pixel index is The target pixel is The calculation formula of the relative spatial distance is as follows: ; 3) Calculating a climate similarity distance: first, the climate variables are standardized, and the formula is as follows: , Wherein the method comprises the steps of Represent the first The average value of the individual climate variables is calculated, Represent the first The standard deviation of the individual climate variables, And then calculating the Euclidean distance of the climate, wherein the more similar the climate conditions are, the smaller the Euclidean distance of the climate is, and the formula is as follows: , 4) Joint weight calculation: Introducing climate bandwidth parameters The scope of action of controlling climate similarity and the calculation of climate weights are given by the following formula: , Introducing spatial bandwidth parameters The spatial weights are calculated as follows: , And combining the climate weight and the space weight to obtain a final combined weight, wherein the formula is as follows: ; 5) Space-time popularization formula of the climatic parameters: To the target pixel Its physical parameters And (3) pushing: ; 6) Substituting the generalized weathered parameters into a weathered curve formula: , I.e. a, b, c with parameters at (x 0 ,y 0 ).
9. The remote sensing space-time data fusion method based on the weather constraint and the weather drive according to claim 1, wherein the data fusion result after the weather drive correction in the step S5 is as follows: , representing the final fusion result; representing the index result after the climatic constraint; △ representing the vegetation index variation caused by meteorological driving; And the weather correction weight coefficient is represented, the judgment is carried out according to the overall quality of the remote sensing image, and the coefficient value is larger when the image quality is poorer.

Description

Remote sensing space-time data fusion method based on physical constraint and meteorological drive Technical Field The invention relates to a remote sensing space-time data fusion method based on physical constraint and meteorological drive, and belongs to the technical field of data processing. Background The remote sensing image is used as an important means for earth observation and is widely applied to the fields of agricultural monitoring, ecological environment assessment, resource investigation, global change research and the like. The existing remote sensing data has complementarity in spatial resolution and time resolution, namely, high spatial resolution images (such as Landsat, sentinel-2 and GF series) can provide fine spatial textures and ground feature information, but the revisitation period is longer, the influence of cloud coverage is obvious, the time continuity is insufficient, high time resolution images (such as MODIS and VIIRS) can provide continuous time sequence data, but the spatial resolution is lower, and the requirements of fine agriculture and regional ecological monitoring are difficult to meet. In order to solve the contradiction, a plurality of space-time fusion methods (Spatio-temporal data fusion methods) are proposed in the academy, and remote sensing products with high spatial resolution and high time resolution are generated by fusing different source images. Typical methods include STARFM (SPATIAL AND sample ADAPTIVE REFLECTANCE Fusion Model) for predicting high resolution images by similar pixel weighting, ESTARFM (Enhanced STARFM) for introducing dual phase images based on STARFM for improving prediction accuracy under complex ground surfaces, and FSDAF (Flexible Spatiotemporal Data Fusion) for improving adaptability to heterogeneous ground surfaces based on image decomposition. However, most of the above methods are based on the assumption of linear variation, i.e. it is believed that the pixel spectral variation trend can be approximately linear between two known phases. This assumption is often not true in practical agriculture and ecological environments. For example, the growth process of crops has obvious climatic stage, the remote sensing indexes (NDVI, EVI and the like) of the crops show typical nonlinear change rules (rapid growth-plateau-decay), the vegetation dynamics are also strongly driven by meteorological conditions (temperature, precipitation, illumination and the like), the vegetation index change often has mutation under extreme climatic events (drought, flood and high temperature) and is difficult to accurately describe through a linear model, and the existing method usually only depends on the remote sensing images per se, lacks the utilization of external driving factors, and leads to insufficient prediction accuracy under complex space-time environments. Thus, the existing space-time fusion method still has the following disadvantages in agricultural and ecological monitoring applications: (1) Ignoring crop weather rules, resulting in deviations in the generated time-series images at critical growth stages; (2) The weather factor driving is not considered, the prediction is inaccurate under the extreme weather condition, and the robustness is insufficient; (3) The lack of uncertainty quantization does not allow for an effective assessment of the reliability of the fusion result. Disclosure of Invention The invention aims to overcome the defects of the existing remote sensing space-time fusion method, and provides a remote sensing space-time data fusion method based on the physical constraint and the meteorological drive. The technical scheme adopted by the invention is as follows: a remote sensing space-time data fusion method based on physical constraint and meteorological drive comprises the following steps: s1, acquiring a high-spatial-resolution remote sensing image, a high-temporal-resolution remote sensing image and contemporaneous weather product data, preprocessing, and calculating a vegetation index WDRVI to carry out surface coverage classification; S2, modeling by meteorological driving: constructing a meteorological driving factor based on meteorological analysis data, and establishing a mapping relation between meteorological variable change and vegetation index short-term response to form a meteorological driving model for describing the influence of short-time meteorological disturbance on the vegetation index; s3, performing physical modeling: Selecting a typical sample land block of a target crop according to the earth surface coverage classification, extracting a time sequence of a sample region from high-time resolution data, performing nonlinear curve fitting on the time sequence by using a Gaussian model, establishing a weathered growth model, performing parameter solving and optimizing on the model to obtain a region representative weathered model of the crop type, extracting a key weathered period from a fitted curve by us