CN113222010A - Method and device for fusing earth surface reflectivity images - Google Patents

Method and device for fusing earth surface reflectivity images Download PDF

Info

Publication number
CN113222010A
CN113222010A CN202110506228.8A CN202110506228A CN113222010A CN 113222010 A CN113222010 A CN 113222010A CN 202110506228 A CN202110506228 A CN 202110506228A CN 113222010 A CN113222010 A CN 113222010A
Authority
CN
China
Prior art keywords
surface reflectivity
time
landsat
modis
image
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202110506228.8A
Other languages
Chinese (zh)
Other versions
CN113222010B (en
Inventor
宋金玲
杨磊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Normal University
Original Assignee
Beijing Normal University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Normal University filed Critical Beijing Normal University
Priority to CN202110506228.8A priority Critical patent/CN113222010B/en
Publication of CN113222010A publication Critical patent/CN113222010A/en
Application granted granted Critical
Publication of CN113222010B publication Critical patent/CN113222010B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/251—Fusion techniques of input or preprocessed data

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses a method and a device for fusing earth surface reflectivity images. The method comprises the following steps: constructing an MODIS earth surface reflectivity difference image data set and an MODIS earth surface reflectivity difference background field for all available MODIS earth surface reflectivity images of the target area for many years; predicting to obtain a Landsat surface reflectivity difference image at a moment adjacent to the known moment, and decomposing all MODIS surface reflectivity difference images from the known moment to the prediction moment according to a Bayesian unmixing theory; deducing to obtain all Landsat surface reflectivity difference images from the known moment to the prediction moment by constructing a linear regression model; and finally, fusing the Landsat surface reflectivity image at the known moment and all the Landsat surface reflectivity difference images from the known moment to the prediction moment to obtain the Landsat surface reflectivity image at the prediction moment. The method can effectively solve the problem of cloud pollution, and can be used for data fusion work in a large area.

Description

Method and device for fusing earth surface reflectivity images
Technical Field
The invention relates to the technical field of image processing, in particular to a method and a device for fusing earth surface reflectivity images, which can be applied to dynamic change monitoring research of vegetation remote sensing.
Background
The existing spatio-temporal data fusion method can be divided into three categories: a fusion algorithm based on unmixing, a fusion algorithm based on a weight function and a fusion algorithm based on dictionary pair learning. The method comprises the following steps that a unmixing-based fusion algorithm decomposes a coarse resolution pixel through a linear spectrum mixing theory so as to obtain an estimation value of a fine pixel; the fusion algorithm based on the weight function is to estimate the pixel value of the fine image by combining the information of all input images through the weight function; the dictionary-based learning fusion algorithm models the relationship between observed coarse-fine image pairs, establishes the corresponding relationship between high-resolution and coarse-resolution images according to the structural similarity of the images, and then predicts the unobserved fine images to capture the main characteristics in the prediction, such as the change of the land cover type.
The earth surface reflectivity images to be fused include: MODIS (moderate-resolution imaging spectrometer) earth surface reflectance image and Landsat (terrestrial resource satellite) earth surface reflectance image. Wherein, the MODIS image has a time resolution of 8 days, and the spatial resolution can be 250m, 500m and 1000 m; the Landsat image has a temporal resolution of 16 days, and its spatial resolution is 30 m. The spatial resolution, i.e. the ground resolution, refers to the ground dimension of the smallest target that can be resolved by the remote sensing instrument, i.e. the size of the ground range corresponding to one pixel on the remote sensing image. If the ground range corresponding to one pixel of the Landsat image is 30m by 30m, the spatial resolution is 30 m. Therefore, the MODIS image is generally considered to be a coarse resolution image and the Landsat image is a fine resolution image, compared with the Landsat image.
The existing three classical models for fusing earth surface reflectivity images are: a spatio-Temporal Adaptive reflection Fusion Model (STARFM), an Enhanced spatio-Temporal Adaptive reflection Fusion Model (ESTARFM), and a Flexible spatio-Temporal Data Fusion Model (FSDAF). Wherein the content of the first and second substances,
the STARFM fusion model obtains a regression relationship between the Landsat image and the MODIS image through at least one period of paired images, and predicts the Landsat surface reflectivity image by combining the MODIS surface reflectivity image according to the regression relationship. If the image element in the MODIS image is a "pure" image element, the STARFM fusion model assumes that the changes in the reflectance ratio of MODIS and Landsat are consistent and comparable, i.e., in the case that only one type of land cover is included in a MODIS image, the image element changes obtained from the MODIS image can be directly added to the image element in the Landsat image to obtain a predicted value; when the MODIS image pels are mixed with different land cover types, the STARFM fusion model uses a function to predict Landsat image pels that gives higher weight to pure MODIS image pixels based on information from neighboring Landsat image pixels. Specifically, the STARFM fusion model assumes that at time T1 there is a pair of high resolution Landsat and low resolution MODIS image pairs, and that at time T0 there is a single MODIS image. Firstly, re-projecting and re-sampling the MODIS image at the time of T1 to be the same as the Landsat image; then, pixels similar to the central pixel are searched in the Landsat image at the time T1, the pixels are pixels which have high spectral similarity with the central pixel and are close to the central pixel, the reflection rate of the central pixel is the integration of the contributions of all the pixels around the central pixel, and the similar pixels are given higher weight; and finally, carrying out re-projection and re-sampling on the MODIS image at the T0 moment in the same way, and predicting to obtain the Landsat image at the T0 moment through the weighting system in the previous step.
However, the assumption of the STARFM fusion model is not valid in heterogeneous landscapes, and the weight function in the STARFM fusion model is empirical. When the types of surface coverage are more complex, the fusion accuracy of the STARFM fusion model is lower. In addition, the STARFM fusion model is only applicable when there is data and the data quality is good, and if the MODIS image at the prediction time contains noise, the prediction result is invalid, so that the STARFM fusion model is not suitable for mass production of surface reflectance data sets in a large area.
The ESTARFM fusion model is an improvement on a STARFM fusion model, and improves the capture capability of the STARFM fusion model on spatial heterogeneity by calculating different conversion coefficients of homogeneous and heterogeneous pixels in a prediction process, thereby improving the reflectivity fusion precision of a research area with large spatial dimension change. And the method defaults to input two Landsat images for prediction, combines more spectral information at different time, and can more accurately measure the similarity between Landsat pixels and MODIS pixels. Specifically, the ESTARFM fusion model respectively needs two pairs of Landsat images and MODIS images at tm and tn moments and an MODIS image at a predicted moment tp, and firstly, the MODIS images at the three moments are preprocessed, including reprojection and resampling, so that the predicted Landsat images have the same row number and column number and the same pixel size; and then, obtaining the contribution of similar image elements around the central image element to the reflectivity of the central image element by using a similar image element searching strategy and a weight distribution system which are the same as those of the STARFM fusion model. Compared with a STARFM fusion model, the ESTARFM fusion model is improved in that a conversion coefficient is introduced, different pixels have different coefficients, so that the ESTARFM fusion model can be well represented in a heterogeneous region on the earth surface, and the conversion coefficients of similar pixels are calculated and added into a weight distribution system; and finally, inputting the resampled STARFM image at the tp moment into the weight distribution system to obtain the final Landsat image at the tp moment.
However, the ESTARFM fusion model is computationally very time consuming and therefore consumes a significant time cost when fusing large areas; when the time span of the two input images of the ESTARFM fusion model is large, the precision of the fusion result is not high; in addition, the ESTARFM fusion model cannot well solve the problem of cloud pollution, and if the MODIS image at the prediction time has cloud coverage, the prediction result is invalid.
The FSDAF fusion model is a method based on unmixing, a weight function and spatial interpolation in combination. The method only needs one Landsat image and two MODIS images, and the reflectivity changes of the Landsat image and the two MODIS images are estimated by solving a linear mixing equation. For the land cover mutation, the spatial interpolation can be captured as long as the land cover mutation is shown in the MODIS image. The FSDAF fusion model firstly classifies by utilizing Landsat images at known moments to obtain the proportion of each end member component in an MODIS pixel; then estimating the time change of each object type in the MODIS image; interpolating the MODIS image at the prediction time by using Thin Plate Spline interpolation (TPS) to obtain the spatial variation of the Landsat image; and finally, distributing errors by using a residual error distribution system, and then predicting the final result of the Landsat image by using the adjacent similar pixel. Specifically, the FSDAF fusion model only needs the MODIS and Landsat image pair at the time t1 and the MODIS image at the time t2 to predict the Landsat image at the time t 2. Firstly, clustering Landsat images at the time t1 to obtain various ground feature classification images; then, subtracting the MODIS image at the time t2 from the MODIS image at the time t1, and obtaining the time change of each ground category by combining land use classification images; obtaining a spatial prediction of the Landsat image at time t2 by using a TPS interpolation method, and assuming that the prediction is closest to a real result; and (3) performing difference calculation on the time prediction result and the change of each land type to obtain a residual error, and distributing the residual error to the space prediction image to obtain the final Landsat image at the t2 moment.
However, the accuracy of the fusion result of the FSDAF fusion model depends on the accuracy of the land use classification. The fusion effect is poor in the region with strong heterogeneity, namely the region with complex ground surface coverage type; when there is a cloud at the time of prediction, an accurate prediction result cannot be obtained, and therefore, the method is also not suitable for use in a large area.
In summary, the existing fusion earth surface reflectivity image model has many errors in the acquired data due to the incomplete principle and the theory of basis. On one hand, the existing fusion model mostly does not obtain good fusion effect in a heterogeneous region of the earth surface; on the other hand, the existing fusion model aims at the remote sensing satellite data without cloud coverage and with good quality, and the problem of cloud coverage cannot be effectively solved.
Disclosure of Invention
In view of the above, it is a primary object of the present invention to provide a method and apparatus for fusing earth surface reflectivity images, which overcome or at least partially solve the above-mentioned problems of the prior art.
According to a first aspect of the present invention, there is provided a method of fusing earth surface reflectivity images, comprising:
collecting all available MODIS earth surface reflectivity images of the target area for many years, and collecting Landsat earth surface reflectivity images of one or more available known moments of the target area;
performing pairwise difference on the collected MODIS surface reflectivity images at adjacent moments to construct a multi-year MODIS surface reflectivity difference image data set, and constructing an MODIS surface reflectivity difference background field according to the multi-year average value of the MODIS surface reflectivity difference image data set;
predicting to obtain a Landsat earth surface reflectivity difference image at a time adjacent to the known time according to the Landsat earth surface reflectivity image at the known time, the MODIS earth surface reflectivity image corresponding to the known time and the MODIS earth surface reflectivity image at the time adjacent to the known time; and
decomposing all MODIS surface reflectivity difference images from the known moment to the prediction moment according to the MODIS surface reflectivity difference background field and the current year-predicted MODIS surface reflectivity difference image data set according to a Bayesian unmixing theory to obtain Bayesian decomposition values of all MODIS surface reflectivity difference images from the known moment to the prediction moment;
according to the Landsat surface reflectivity difference image at the adjacent moment with the known moment and the Bayesian decomposition values of all MODIS surface reflectivity difference images from the known moment to the prediction moment, all Landsat surface reflectivity difference images from the known moment to the prediction moment are obtained by constructing a linear regression model and deducing;
and fusing the Landsat surface reflectivity image at the known moment and all the Landsat surface reflectivity difference images from the known moment to the prediction moment to obtain the Landsat surface reflectivity image at the prediction moment.
According to a second aspect of the present invention, there is provided an apparatus for fusing earth surface reflectivity images, comprising:
the image collecting unit is used for collecting all available MODIS earth surface reflectivity images of the target area for many years and collecting Landsat earth surface reflectivity images of one or more available known moments of the target area;
the MODIS difference data set construction unit is used for carrying out pairwise difference at adjacent moments on the collected MODIS earth surface reflectivity images to construct a multi-year MODIS earth surface reflectivity difference image data set, and constructing an MODIS earth surface reflectivity difference background field according to the multi-year average value of the MODIS earth surface reflectivity difference image data set;
the Landsat difference value prediction unit is used for predicting to obtain a Landsat difference value image of the earth surface reflectivity of the time adjacent to the known time according to the Landsat earth surface reflectivity image of the known time, the MODIS earth surface reflectivity image corresponding to the known time and the MODIS earth surface reflectivity image of the time adjacent to the known time;
the Bayesian unmixing unit is used for decomposing all MODIS surface reflectivity difference images from the known time to the prediction time according to the MODIS surface reflectivity difference background field and the current year-predicted MODIS surface reflectivity difference image data set and obtaining Bayesian decomposition values of all MODIS surface reflectivity difference images from the known time to the prediction time;
the Landsat difference derivation unit is used for deriving and obtaining all Landsat surface reflectivity difference images from the known time to the prediction time by constructing a linear regression model according to the Landsat surface reflectivity difference images at the adjacent time with the known time and Bayesian decomposition values of all MODIS surface reflectivity difference images from the known time to the prediction time;
and the image fusion unit is used for fusing the Landsat surface reflectivity image at the known moment and all the Landsat surface reflectivity difference images from the known moment to the prediction moment to obtain the Landsat surface reflectivity image at the prediction moment.
According to a third aspect of the present invention, there is provided an electronic apparatus comprising: a processor, a memory storing computer executable instructions which, when executed by the processor, implement the aforementioned method of fusing a surface reflectance image.
According to a fourth aspect of the present invention, there is provided a computer readable storage medium storing one or more programs which, when executed by a processor, implement the aforementioned method of fusing earth surface reflectance images.
The invention has the beneficial effects that:
the method includes the steps that a MODIS earth surface reflectivity difference image data set for years is constructed by carrying out pairwise difference on collected MODIS earth surface reflectivity images at adjacent moments, so that information of the MODIS earth surface reflectivity images changing along with time is obtained, and an MODIS earth surface reflectivity difference background field is constructed according to the multi-year average value of the MODIS earth surface reflectivity difference image data set; decomposing all MODIS surface reflectivity difference images from the known moment to the prediction moment according to a Bayesian unmixing theory, and using an MODIS surface reflectivity difference background field as prior information for restricting the decomposition value of a mixed pixel; and finally, deducing to obtain all Landsat earth surface reflectivity difference images from the known moment to the prediction moment by constructing a linear regression model, and fusing the Landsat earth surface reflectivity images at the known moment to obtain the Landsat earth surface reflectivity image at the prediction moment.
On one hand, the MODIS earth surface reflectivity difference background field is used as the pixel with poor prior information optimization quality, so that the abnormal value caused by cloud pollution can be corrected by improving the weight (sigma) of the prior information even in the cloud coverage area, the cloud pollution problem can be effectively solved, and an accurate prediction result can be obtained when the MODIS image quality is poor.
On the other hand, the MODIS surface reflectivity difference image data set construction mechanism enables the changed information to be captured in time, and is different from the existing fusion model which only utilizes two MODIS images.
In addition, because the invention does not use a search strategy of similar pixels in a fusion model such as STARFM, ESTARFM and the like, the Bayesian unmixing is directly carried out by taking an MODIS coarse resolution pixel as a unit, the multi-year average value of an MODIS surface reflectivity difference image data set is used as prior information for constraint, the obtained Bayesian unmixing value is closer to the actual situation, the boundary of the pixel is more uniform, the patch trace of the MODIS pixel boundary can not be reserved, and the higher image quality can be reserved.
Drawings
Various advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention. Also, like reference numerals are used to refer to like parts throughout the drawings. In the drawings:
FIG. 1 is a schematic flow chart of a method for fusing earth surface reflectance images according to one embodiment of the present invention;
FIG. 2 is a block diagram of an apparatus for fusing earth's surface reflectance images according to one embodiment of the present invention;
FIG. 3 is a block diagram of an electronic device in accordance with one embodiment of the present invention.
Detailed Description
Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. These embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be embodied in various forms and should not be limited to the embodiments set forth herein.
Fig. 1 is a schematic flow chart showing a method for fusing earth surface reflectivity images according to an embodiment of the present invention, and referring to fig. 1, the method of the embodiment of the present invention includes the following steps S110 to S160:
step S110, collecting all available MODIS surface reflectance images of the target area for many years, and collecting one or more available Landsat surface reflectance images of the target area at known time.
MODIS is an important optical remote sensing instrument carried on terra and aqua satellites, and is the only on-board instrument on the satellites, which directly broadcasts real-time observation data to the world through an x wave band, can receive the data free of charge and can be used for no compensation. MODIS image data is relatively easy to collect, and all available MODIS surface reflectance images of the target area over many years (e.g., within five years) can be collected.
The Landsat image data is difficult to collect, and can be downloaded to one or more available Landsat surface reflectivity images at known moments in a target area by using a vegetation index extraction tool by means of a Landsat remote sensing image data service system provided by a computer network information center of the Chinese academy of sciences.
All available MODIS raw image data and a small amount of available Landsat raw image data at a known time in the target region for many years may be collected in advance by step S110.
And step S120, carrying out pairwise difference on the collected MODIS earth surface reflectivity images to construct a multi-year MODIS earth surface reflectivity difference image data set, and constructing an MODIS earth surface reflectivity difference background field according to the multi-year average value of the MODIS earth surface reflectivity difference image data set.
In a longer time sequence, because the obtainable Landsat original image data with good quality is very rare, and the MODIS original image data is relatively easy to obtain, the MODIS original image data can be used for constructing a surface reflectivity difference image data set.
The construction of the MODIS surface reflectivity difference image dataset is divided into two parts: on one hand, the obtained MODIS original image data is subjected to pairwise difference of adjacent moments to construct an MODIS earth surface reflectivity difference image data set. For example, 46 sets of MODIS surface reflectivity images with a time resolution of 8 days can be acquired in one year, and then two sets of the images at adjacent times are subjected to difference to obtain images with reflectivity changes every 8 days, that is, 45 sets of MODIS surface reflectivity difference images can be obtained in one year. On the other hand, similar to the above idea of constructing the annual difference image data set, the MODIS earth surface reflectivity difference image data set for a plurality of years (generally five years) can be averaged, specifically, the 45 scene change images corresponding to the same time are respectively averaged for a plurality of years, so as to construct an MODIS earth surface reflectivity difference background field, and obtain the prior information of the annual reflectivity change in the target region.
And step S130, predicting to obtain a Landsat surface reflectivity difference image of the time adjacent to the known time according to the Landsat surface reflectivity image of the known time, the MODIS surface reflectivity image corresponding to the known time and the MODIS surface reflectivity image of the time adjacent to the known time.
In step S130, the landform surface reflectance difference image is predicted.
The Landsat image data itself has a spatial resolution of 30m and a temporal resolution of 16 days, and in order to keep consistent with the MODIS image data with a temporal resolution of 8 days, the reflectivity change of the Landsat image data at a known moment for the next 8 days is predicted through the following steps of (i) - (v) to obtain a Landsat earth surface reflectivity difference image at a moment adjacent to the known moment:
firstly, establishing a linear regression relationship between the MODIS earth surface reflectivity image and the Landsat earth surface reflectivity image at a known moment to obtain a spatial prediction of the Landsat earth surface reflectivity at a moment adjacent to the known moment.
Resampling all MODIS image data to the same spatial resolution as Landsat image data using bilinear interpolation, and then establishing a linear regression relationship between the MODIS image and the Landsat image at a known time:
Figure BDA0003058552820000101
wherein the content of the first and second substances,
Figure BDA0003058552820000102
the spatial prediction of Landsat surface reflectivity representing the time adjacent to the known time is obtained by performing bilinear interpolation resampling on MODIS surface reflectivity images of the time adjacent to the known time, xij、yijIn an image of 30m resolutionCoordinates of picture elements, λ being a certain band, F1 (x)ij，yijAnd lambda) are Landsat surface reflectivity images at known time, and p and q are the slope and intercept of the two image fitting respectively.
And secondly, assuming that the change of the same type of ground objects in a short time is the same, obtaining the time prediction of the Landsat ground surface reflectivity at the time adjacent to the known time, wherein a residual exists between the time prediction and the real image.
The method comprises the following steps of firstly utilizing high-precision land utilization data of 30m, such as data rasterized from a 1:10 ten thousand land utilization vector diagram provided by Chinese academy of sciences remote sensing and digital earth research to obtain the end member area components of each ground feature type in an MODIS coarse resolution pixel:
fc(xi，yi)＝Nc(xi，yi)/m (2)
wherein N isc(xi，yi) Is represented in a MODIS pixel (x)i，yi) In the category c, the number of fine resolution pixels of the ground object, m represents the number of Landsat pixels contained in one MODIS pixel, fc(xi，yi) Is prepared by mixing the following components in proportion;
for any band λ, the time prediction of the change in surface reflectivity of Landsat is:
ΔC(xi，yi，λ)＝C2(xi，yi，λ)-C1(xi，yi，λ) (3)
substituting (2) into (3) to obtain:
Figure BDA0003058552820000111
assuming that the same ground object changes in a short time, there are:
Figure BDA0003058552820000112
wherein, C2(xi，yiλ) MODIS surface reflectance image of the λ band at a time adjacent to the known time, C1(xi，yiλ) is the MODIS surface reflectance image at a known time of the λ band, Δ C (x)i，yiλ) is C2(xi，yiλ) and C1(xi，yiλ) difference; l is the number of surface feature classes; Δ F (c, λ) is the variation of c-type ground objects in the λ band;
Figure BDA0003058552820000113
is a temporal prediction of the landform reflectivity at times adjacent to the known time.
In the above process, it is assumed that the same change of the same feature in a short time does not satisfy the actual situation, and in the actual situation, the change differs even if the same feature is used, and therefore, the time prediction is performed
Figure BDA0003058552820000114
The difference from the real image is not exactly equivalent to Δ C (x)i，yiλ), there is a deviation between the two, i.e. a residual R (x)i，yi，λ)：
Figure BDA0003058552820000115
In addition, (x)i，yi) Is a certain pixel in the MODIS coarse resolution image; (x)ij,yij) Is a certain pixel in the Landsat fine resolution image. i represents the position index of the image element in the MODIS image, and j represents the position index of the image element in the Landsat image. Such as: the first picture element in the MODIS image can be represented as (x)1,y1) The MODIS image element can comprise a plurality of Landsat image elements, and the first Landsat image element can be expressed as (x)11,y11) It represents the first Landsat pel in the first MODIS pel. And thirdly, assuming that the result of the spatial prediction is closer to a true value, and obtaining the error of the temporal prediction relative to the spatial prediction.
The invention assumes that the result of the spatial prediction is closer to the true value, and obtains the time prediction
Figure BDA0003058552820000116
With respect to spatial prediction
Figure BDA0003058552820000121
The error of (2) is:
Figure BDA0003058552820000122
as can be seen, error E (x)ij，yijλ) is calculated in units of one MODIS pixel, and the residual R (x)i，yiLambda) is calculated in units of one Landsat pixel, which are different in fine scale.
And fourthly, introducing a weight distribution coefficient to distribute the error and the residual error:
ω(xij，yij，λ)＝E(xij，yij，λ)×I(xij，yij，λ)+R(xi，yi，λ)×(1-I(xij，yij，λ)] (8)
wherein, ω (x)ij,yijλ,) is the weight distribution coefficient, I (x)ij，yijAnd lambda) is a homogeneity coefficient,
Figure BDA0003058552820000123
representing the ratio of the number of the image elements with the same type of the ground objects and the same center in a sliding window to the total number of the image elements in the window, when the k-th image element and the center image element (x) of the sliding windowij，yij) Are of the same class, Ik1, otherwise, Ik＝0；
Normalizing the weight distribution coefficients:
Figure BDA0003058552820000124
predicting the Landsat surface reflectivity difference image based on the weight distribution coefficient and the residual error to obtain the Landsat surface reflectivity difference image at the time adjacent to the known time.
Specifically, the following formula can be adopted:
ΔF(xij，yij，λ)＝m×R(xi，yi，λ)×W(xij，yij，λ)+ΔF(c，λ) (10)
wherein, Δ F (x)ij，yijAnd lambda) is the Landsat surface reflectivity difference image of the predicted time adjacent to the known time. As defined above, in formula (10), m represents the number of Landsat pixels included in one MODIS pixel, Δ G (c, λ) is the variation of c-type feature in λ band, and W (x)ij,yijAnd λ) is the normalized weight distribution coefficient.
When the predicted time is sufficiently close to the known time, for example, when the predicted time is a time adjacent to the known time, Δ F (x) obtained as described aboveij，yijLambda) can be directly used as the prediction result of the Landsat surface reflectivity difference image at the prediction time, but when the prediction time is far away from the known time, the space prediction value obtained in the process of the first step is obtained
Figure BDA0003058552820000125
There will be a large error, so it is necessary to introduce bayesian unmixing theory to predict images that are far away from the known time.
And step S140, decomposing all MODIS surface reflectivity difference images from the known time to the prediction time according to the Bayesian unmixing theory according to the MODIS surface reflectivity difference background field and the current year-predicted MODIS surface reflectivity difference image data set, and obtaining Bayesian decomposition values of all MODIS surface reflectivity difference images from the known time to the prediction time.
Since the MODIS surface reflectivity difference image from the known time to the prediction time is also a 500m coarse spatial resolution image, a 30m fine spatial resolution image needs to be obtained through Bayesian unmixing theory decomposition, and then a relationship can be established between the fine spatial resolution image and the Landsat difference image data, so as to predict the Landsat image data.
This step S140 is based on the following two assumptions: assuming that the MODIS earth surface reflectivity difference background field obeys an average value of RpAnd the covariance matrix is Gaussian distribution of sigma p, and meanwhile, the MODIS surface reflectivity difference image data set of the current year is supposed to be predicted to follow a mean value of R0The covariance matrix is ∑0Then according to Bayesian unmixing theory, the posterior distribution of MODIS surface reflectivity difference value also obeys a mean value of mueAnd the covariance matrix is Gaussian distribution of Σ e, and a Bayesian decomposition value corresponding to each MODIS surface reflectivity difference image can be obtained by the following formula:
∑e＝[∑p-1+AT∑0-1A]-1 (11)
μe＝∑e×[AT∑0-1R0+∑p-1Rp]-1 (12)
wherein A represents the area fraction of each feature class, μeRepresenting the estimated value with the minimum variance, i.e. the pixel value after the mixed pixel decomposition, and further according to mueAnd constructing a Bayesian decomposition value corresponding to each MODIS surface reflectivity difference image.
Specifically, assuming that the bayesian decomposition value Mt1 or Mtk of the MODIS surface reflectivity difference image is an n-row m-column image, Mt1 or Mtk is an entire image composed of pixel values μ, each pixel value in the image is μ, that is, the decomposition value obtained by formula (12), Mt1 or Mtk is a set of formula (12), and the process of obtaining Mt1 or Mtk according to μ construction is as follows:
Figure BDA0003058552820000141
and S150, constructing a linear regression model to derive all the Landsat surface reflectivity difference images from the known time to the prediction time according to the predicted Landsat surface reflectivity difference images at the adjacent time to the known time and Bayesian decomposition values of all the MODIS surface reflectivity difference images from the known time to the prediction time.
The Bayesian decomposition values of all MODIS surface reflectivity difference images from the known time to the prediction time obtained according to the Bayesian unmixing theory still have spectral characteristics from MODIS data and do not have spatial detail information of Landsat data.
The invention assumes that the MODIS earth surface reflectivity difference and the Landsat earth surface reflectivity difference have the following linear relationship:
Lt1＝p×Mt1+qt1 (13)
Ltk＝p×Mtk+qtk (14)
wherein L ist1And LtkRespectively correspond to t1Time t andkthe Landsat earth surface reflectivity difference images at the moment are all fine spatial resolution images of 30 m; mt1And MtkRespectively correspond to t1Time t andkbayesian decomposition values of the MODIS surface reflectivity difference image at the moment are both fine spatial resolution images of 30m, p and q are fitting coefficients of a linear equation, k is a natural number and is less than or equal to 45.
And assuming that M and q are in a direct proportion relationship, the following steps are provided:
Figure BDA0003058552820000142
transforming equation (15) yields:
Figure BDA0003058552820000143
and, when it is predicted that there are n available landform reflectance images of known time in the year, n sets of slave t are combined1To tnBayesian decomposition of all MODIS surface reflectance difference images at the time,and weights w of different magnitudes are given according to the time distance, and the formula (16) is transformed into:
Figure BDA0003058552820000151
l in the formulas (16) and (17)tkNamely the derived Landsat surface reflectivity difference image corresponding to the prediction time.
In addition, L ist1The landform surface reflectance difference image Δ F (x) at a time adjacent to the known time predicted in step S130 is used as the difference imageij，yijλ). The invention only needs to be at a known time t1The Landsat earth surface reflectivity difference image prediction is performed once according to step S130, and then all Landsat earth surface reflectivity difference images from the known time to the prediction time are obtained as unknowns by solving equations (13) and (14) in parallel, that is, by using bayesian decomposition values.
And step S160, fusing the Landsat surface reflectivity image at the known moment and all the Landsat surface reflectivity difference images from the known moment to the prediction moment to obtain the Landsat surface reflectivity image at the prediction moment.
In step S160, image fusion may be performed according to the following formula:
Figure BDA0003058552820000152
wherein L iskLandsat surface reflectance image for time of prediction, L1Is the Landsat surface reflectance image at a known time, Lt1…LtkSequentially all Landsat surface reflectivity difference images from the known time to the predicted time, tk＜t1Indicating that the predicted time is before the known time, tk≥t1Indicating that the predicted time is equal to or after the known time.
Thus, a Landsat surface reflectance image L of the predicted time is obtainedkI.e. the final prediction result image.
In addition, regardless of whether the predicted time is close to or far from the known time, the Landsat surface reflectance difference image L at a time adjacent to the known time needs to be predicted in step S130t1Except that L is small in time spant1Landsat surface reflectivity image L capable of directly and at known time1The Landsat earth surface reflectivity image L at the predicted time can be obtained by addingk(ii) a When the time span is large, L needs to be firstly processedt1Inputting the data into a Bayesian unmixing model, and deriving all Landsat surface reflectivity difference images (L) from the known time to the prediction time by using Bayesian decomposition valuest1…Ltk) Then, the Landsat surface reflectivity difference images (L)t1…Ltk) And Landsat surface reflectance image L at a known time1Adding to obtain Landsat surface reflectivity image L at the predicted timek。
In conclusion, the collected MODIS earth surface reflectivity images are subjected to pairwise difference at adjacent moments to construct a multi-year MODIS earth surface reflectivity difference image data set, so that the time-varying information of the MODIS earth surface reflectivity images is obtained, and an MODIS earth surface reflectivity difference background field is constructed according to the multi-year average value of the MODIS earth surface reflectivity difference image data set; decomposing all MODIS surface reflectivity difference images from the known moment to the prediction moment according to a Bayesian unmixing theory, and using an MODIS surface reflectivity difference background field as prior information for restricting the decomposition value of a mixed pixel; and finally, deducing to obtain all Landsat earth surface reflectivity difference images from the known moment to the prediction moment by constructing a linear regression model, and fusing the Landsat earth surface reflectivity images at the known moment to obtain the Landsat earth surface reflectivity image at the prediction moment.
On one hand, the MODIS earth surface reflectivity difference background field is used as the pixel with poor prior information optimization quality, so that the abnormal value caused by cloud pollution can be corrected by improving the weight (sigma) of the prior information even in the cloud coverage area, the cloud pollution problem can be effectively solved, and an accurate prediction result can be obtained when the MODIS image quality is poor. On the other hand, the MODIS surface reflectivity difference image data set construction mechanism enables the changed information to be captured in time, and is different from the existing fusion model which only utilizes two MODIS images. In addition, because the invention does not use a search strategy of similar pixels in a fusion model such as STARFM, ESTARFM and the like, the Bayesian unmixing is directly carried out by taking an MODIS coarse resolution pixel as a unit, the multi-year average value of an MODIS surface reflectivity difference image data set is used as prior information for constraint, the obtained Bayesian unmixing value is closer to the actual situation, the boundary of the pixel is more uniform, the patch trace of the MODIS pixel boundary can not be reserved, and the higher image quality can be reserved.
The method for fusing the earth surface reflectivity images belongs to the same technical concept as the method for fusing the earth surface reflectivity images, and the embodiment of the invention also provides a device for fusing the earth surface reflectivity images. Fig. 2 is a block diagram of an apparatus for fusing a reflectivity image of a ground according to an embodiment of the present invention, and referring to fig. 2, an apparatus 200 for fusing a reflectivity image of a ground according to the present invention includes:
the image collecting unit 210 is used for collecting all available MODIS surface reflectivity images of the target area for many years and collecting Landsat surface reflectivity images of one or more available known moments of the target area;
the MODIS difference data set constructing unit 220 is configured to perform pairwise difference between adjacent times on the collected MODIS surface reflectivity images to construct a multi-year MODIS surface reflectivity difference image data set, and construct a MODIS surface reflectivity difference background field according to a multi-year average value of the MODIS surface reflectivity difference image data set;
a Landsat difference prediction unit 230, configured to predict, according to the Landsat surface reflectance image at the known time, the MODIS landform reflectance image corresponding to the known time, and the MODIS landform reflectance image at the time adjacent to the known time, a Landsat surface reflectance difference image at the time adjacent to the known time;
the bayesian unmixing unit 240 is configured to decompose all MODIS surface reflectivity difference images from the known time to the prediction time according to the bayesian unmixing theory according to the MODIS surface reflectivity difference background field and the current-year-predicted MODIS surface reflectivity difference image data set, and obtain bayesian decomposition values of all MODIS surface reflectivity difference images from the known time to the prediction time;
a Landsat difference derivation unit 250, configured to derive all Landsat earth surface reflectance difference images from the known time to the prediction time by constructing a linear regression model according to the predicted Landsat earth surface reflectance difference image at the time adjacent to the known time and bayesian decomposition values of all MODIS earth surface reflectance difference images from the known time to the prediction time;
and an image fusion unit 260, configured to fuse the Landsat surface reflectance image at the known time and all Landsat surface reflectance difference images from the known time to the prediction time to obtain a Landsat surface reflectance image at the prediction time.
In an embodiment of the present invention, the Landsat difference prediction unit 230 is specifically configured to:
establishing a linear regression relationship between the MODIS earth surface reflectivity image and the Landsat earth surface reflectivity image at a known moment to obtain a spatial prediction of the Landsat earth surface reflectivity at a moment adjacent to the known moment;
assuming that the change of the same type of ground objects in a short time is the same, obtaining the time prediction of the Landsat ground surface reflectivity at the time adjacent to the known time, wherein a residual exists between the time prediction and the real image;
assuming that the result of the spatial prediction is closer to the true value, and obtaining the error of the temporal prediction relative to the spatial prediction;
introducing a weight distribution coefficient to distribute the error and the residual error;
and predicting the Landsat earth surface reflectivity difference image based on the weight distribution coefficient and the residual error to obtain the Landsat earth surface reflectivity difference image at the moment adjacent to the known moment.
In an embodiment of the present invention, the bayesian unmixing unit 240 is specifically configured to:
assuming that the MODIS earth surface reflectivity difference background field obeys an average value of RpAnd the covariance matrix is Gaussian distribution of sigma p, and meanwhile, the MODIS surface reflectivity difference image data set of the current year is supposed to be predicted to follow a mean value of R0The covariance matrix is ∑0(ii) a gaussian distribution of;
according to the Bayesian unmixing theory, the posterior distribution of the MODIS surface reflectivity difference values is also obeyed a mean value of mueThe covariance matrix is a Gaussian distribution of Σ e, and μ is obtainedeThe minimum variance of the image element is the value of the image element after the mixed image element is decomposed, and then the minimum variance of the image element is calculated according to the value of mueAnd constructing a Bayesian decomposition value corresponding to each MODIS surface reflectivity difference image.
In an embodiment of the present invention, the Landsat difference deriving unit 250 is specifically configured to:
constructing a linear equation of the MODIS earth surface reflectivity difference value and the Landsat earth surface reflectivity difference value:
Lt1＝p×Mt1+qt1
Ltk＝p×Mtk+qtk
wherein L ist1And LtkRespectively correspond to t1Time t andklandsat surface reflectance difference image at time, Mt1And MtkRespectively correspond to t1Time t andkbayesian decomposition values of MODIS surface reflectivity difference images at the moment, p and q are fitting coefficients of a linear equation, k is a natural number and is less than or equal to 45;
and assuming that M and q are in a direct proportion relationship, obtaining:
Figure BDA0003058552820000191
and, when it is predicted that there are n available landform reflectance images of known time in the year, n sets of slave t are combined1To tnBayesian decomposition values of all MODIS surface reflectivity difference images at the moment are given with weights w of different sizes according to time distance, and the above formula is transformed into:
Figure BDA0003058552820000192
said LtkNamely the derived Landsat surface reflectivity difference image corresponding to the prediction time.
In an embodiment of the present invention, the image fusion unit 260 is specifically configured to:
image fusion is performed according to the following formula:
Figure BDA0003058552820000193
wherein L iskLandsat surface reflectance image for time of prediction, L1Is the Landsat surface reflectance image at a known time, Lt1…LtkSequentially all Landsat surface reflectivity difference images from the known time to the predicted time, tk＜t1Indicating that the predicted time is before the known time, tk≥t1Indicating that the predicted time is equal to or after the known time.
The method belongs to the same technical concept as the method for fusing the earth surface reflectivity images, and the embodiment of the invention also provides electronic equipment. Fig. 3 shows a block diagram of an electronic device according to an embodiment of the invention, and referring to fig. 3, at a hardware level, the electronic device of the invention comprises: a processor and a memory. Optionally also an interface module, a communication module, etc. The Memory may include a Memory, such as a Random-Access Memory (RAM), and may also include a non-volatile Memory, such as at least one disk Memory. Of course, the electronic device may also include hardware required for other services.
The processor, the interface module, the communication module, and the memory may be connected to each other via an internal bus, which may be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, an EISA (Extended Industry Standard Architecture) bus, or the like. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one double-headed arrow is shown in FIG. 3, but this does not indicate only one bus or one type of bus.
A memory for storing computer executable instructions. The memory provides computer executable instructions to the processor through the internal bus.
A processor executing computer executable instructions stored in the memory and specifically configured to perform the following operations:
collecting all available MODIS earth surface reflectivity images of the target area for many years, and collecting Landsat earth surface reflectivity images of one or more available known moments of the target area;
performing pairwise difference on the collected MODIS surface reflectivity images at adjacent moments to construct a multi-year MODIS surface reflectivity difference image data set, and constructing an MODIS surface reflectivity difference background field according to the multi-year average value of the MODIS surface reflectivity difference image data set;
predicting to obtain a Landsat earth surface reflectivity difference image at a time adjacent to the known time according to the Landsat earth surface reflectivity image at the known time, the MODIS earth surface reflectivity image corresponding to the known time and the MODIS earth surface reflectivity image at the time adjacent to the known time; and
decomposing all MODIS surface reflectivity difference images from the known moment to the prediction moment according to the MODIS surface reflectivity difference background field and the current year-predicted MODIS surface reflectivity difference image data set according to a Bayesian unmixing theory to obtain Bayesian decomposition values of all MODIS surface reflectivity difference images from the known moment to the prediction moment;
according to the Landsat surface reflectivity difference image at the adjacent moment with the known moment and the Bayesian decomposition values of all MODIS surface reflectivity difference images from the known moment to the prediction moment, all Landsat surface reflectivity difference images from the known moment to the prediction moment are obtained by constructing a linear regression model and deducing;
and fusing the Landsat surface reflectivity image at the known moment and all the Landsat surface reflectivity difference images from the known moment to the prediction moment to obtain the Landsat surface reflectivity image at the prediction moment.
The functions performed by the apparatus for fusing earth's surface reflectance images according to the embodiment of the invention shown in fig. 2 can be implemented in or by a processor. The processor may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method may be performed by integrated logic circuits of hardware in a processor or instructions in the form of software. The Processor may be a general-purpose Processor, including a Central Processing Unit (CPU), a Network Processor (NP), and the like; but also Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components. The various methods, steps and logic blocks disclosed in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in connection with the embodiments of the present invention may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software module may be located in ram, flash memory, rom, prom, or eprom, registers, etc. storage media as is well known in the art. The storage medium is located in a memory, and a processor reads information in the memory and completes the steps of the method in combination with hardware of the processor.
The electronic device may further perform the steps performed by the method for fusing the surface reflectance image in fig. 1, and implement the functions of the embodiment shown in fig. 1, which are not described herein again.
Embodiments of the present invention also provide a computer-readable storage medium storing one or more programs which, when executed by a processor, implement the aforementioned method of fusing earth surface reflectance images, and are specifically configured to perform:
collecting all available MODIS earth surface reflectivity images of the target area for many years, and collecting Landsat earth surface reflectivity images of one or more available known moments of the target area;
performing pairwise difference on the collected MODIS surface reflectivity images at adjacent moments to construct a multi-year MODIS surface reflectivity difference image data set, and constructing an MODIS surface reflectivity difference background field according to the multi-year average value of the MODIS surface reflectivity difference image data set;
predicting to obtain a Landsat earth surface reflectivity difference image at a time adjacent to the known time according to the Landsat earth surface reflectivity image at the known time, the MODIS earth surface reflectivity image corresponding to the known time and the MODIS earth surface reflectivity image at the time adjacent to the known time; and
decomposing all MODIS surface reflectivity difference images from the known moment to the prediction moment according to the MODIS surface reflectivity difference background field and the current year-predicted MODIS surface reflectivity difference image data set according to a Bayesian unmixing theory to obtain Bayesian decomposition values of all MODIS surface reflectivity difference images from the known moment to the prediction moment;
according to the Landsat surface reflectivity difference image at the adjacent moment with the known moment and the Bayesian decomposition values of all MODIS surface reflectivity difference images from the known moment to the prediction moment, all Landsat surface reflectivity difference images from the known moment to the prediction moment are obtained by constructing a linear regression model and deducing;
and fusing the Landsat surface reflectivity image at the known moment and all the Landsat surface reflectivity difference images from the known moment to the prediction moment to obtain the Landsat surface reflectivity image at the prediction moment.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
The present invention is described in terms of flowcharts and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
In a typical configuration, an electronic device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium. Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in the process, method, article, or apparatus that comprises the element.
The above are merely examples of the present invention, and are not intended to limit the present invention. Various modifications and alterations to this invention will become apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.

Claims (10)

1. A method of fusing earth surface reflectance images, comprising:
collecting all available MODIS earth surface reflectivity images of the target area for many years, and collecting Landsat earth surface reflectivity images of one or more available known moments of the target area;
performing pairwise difference on the collected MODIS surface reflectivity images at adjacent moments to construct a multi-year MODIS surface reflectivity difference image data set, and constructing an MODIS surface reflectivity difference background field according to the multi-year average value of the MODIS surface reflectivity difference image data set;
predicting to obtain a Landsat earth surface reflectivity difference image at a time adjacent to the known time according to the Landsat earth surface reflectivity image at the known time, the MODIS earth surface reflectivity image corresponding to the known time and the MODIS earth surface reflectivity image at the time adjacent to the known time; and
decomposing all MODIS surface reflectivity difference images from the known moment to the prediction moment according to the MODIS surface reflectivity difference background field and the current year-predicted MODIS surface reflectivity difference image data set according to a Bayesian unmixing theory to obtain Bayesian decomposition values of all MODIS surface reflectivity difference images from the known moment to the prediction moment;
according to the Landsat surface reflectivity difference image at the adjacent moment with the known moment and the Bayesian decomposition values of all MODIS surface reflectivity difference images from the known moment to the prediction moment, all Landsat surface reflectivity difference images from the known moment to the prediction moment are obtained by constructing a linear regression model and deducing;
and fusing the Landsat surface reflectivity image at the known moment and all the Landsat surface reflectivity difference images from the known moment to the prediction moment to obtain the Landsat surface reflectivity image at the prediction moment.
2. The method of claim 1, wherein predicting a landform surface reflectivity difference image at a time adjacent to a known time according to a landform surface reflectivity image at the known time, an MODIS surface reflectivity image corresponding to the known time, and an MODIS surface reflectivity image at a time adjacent to the known time comprises:
establishing a linear regression relationship between the MODIS earth surface reflectivity image and the Landsat earth surface reflectivity image at a known moment to obtain a spatial prediction of the Landsat earth surface reflectivity at a moment adjacent to the known moment;
assuming that the change of the same type of ground objects in a short time is the same, obtaining the time prediction of the Landsat ground surface reflectivity at the time adjacent to the known time, wherein a residual exists between the time prediction and the real image;
assuming that the result of the spatial prediction is closer to the true value, obtaining the error of the temporal prediction relative to the spatial prediction;
introducing a weight distribution coefficient to distribute the error and the residual error;
and predicting the Landsat earth surface reflectivity difference image based on the weight distribution coefficient and the residual error to obtain the Landsat earth surface reflectivity difference image at the moment adjacent to the known moment.
3. The method as claimed in claim 1, wherein decomposing all MODIS surface reflectivity difference images from the known time to the predicted time according to bayesian unmixing theory based on the MODIS surface reflectivity difference background field and the MODIS surface reflectivity difference image dataset predicted in the current year to obtain bayesian decomposition values of all MODIS surface reflectivity difference images from the known time to the predicted time comprises:
assuming that the MODIS earth surface reflectivity difference background field obeys an average value RpAnd the covariance matrix is Gaussian distribution of sigma p, and meanwhile, the MODIS surface reflectivity difference image data set of the current year is supposed to be predicted to follow a mean value of R0The covariance matrix is ∑0(ii) a gaussian distribution of;
according to the Bayesian unmixing theory, the posterior distribution of the MODIS surface reflectivity difference values is also obeyed a mean value of mueAnd the covariance matrix is Gaussian distribution of Σ e to obtain the μeThe minimum variance of (a);
according to said μeAnd constructing a Bayesian decomposition value corresponding to each MODIS surface reflectivity difference image.
4. The method of claim 1, wherein the deriving all the Landsat earth reflectivity difference images from the known time to the predicted time by constructing a linear regression model according to the predicted Landsat earth reflectivity difference images at the adjacent time to the known time and the bayesian decomposition values of all the MODIS earth reflectivity difference images from the known time to the predicted time comprises:
constructing a linear equation of the MODIS earth surface reflectivity difference value and the Landsat earth surface reflectivity difference value:
Lt1＝p×Mt1+qt1
Ltk＝p×Mtk+qtk
wherein L ist1And LtkRespectively correspond to t1Time t andklandsat surface reflectance difference image at time, Mt1And MtkRespectively correspond to t1Time t andkbayesian decomposition values of MODIS surface reflectivity difference images at the moment, p and q are fitting coefficients of a linear equation, k is a natural number and is less than or equal to 45;
and assuming that M and q are in a direct proportion relationship, obtaining:
Figure FDA0003058552810000031
and, when it is predicted that there are n available landform reflectance images of known time in the year, n sets of slave t are combined1To tnBayesian decomposition values of all MODIS surface reflectivity difference images at the moment are given with weights w of different sizes according to time distance, and the above formula is transformed into:
Figure FDA0003058552810000032
said LtkNamely the derived Landsat surface reflectivity difference image corresponding to the prediction time.
5. The method of claim 4, wherein fusing the Landsat surface reflectance image at the known time and all the Landsat surface reflectance difference images from the known time to the predicted time to obtain the Landsat surface reflectance image at the predicted time comprises:
image fusion is performed according to the following formula:
Figure FDA0003058552810000033
wherein L iskLandsat surface reflectance image for time of prediction, L1Is the Landsat surface reflectance image at a known time, Lt1...LtkSequentially all Landsat surface reflectivity difference images from the known time to the predicted time, tk＜t1Indicating that the predicted time is before the known time, tk≥t1Indicating that the predicted time is equal to or after the known time.
6. An apparatus for fusing earth surface reflectivity images, comprising:
the image collecting unit is used for collecting all available MODIS earth surface reflectivity images of the target area for many years and collecting Landsat earth surface reflectivity images of one or more available known moments of the target area;
the MODIS difference data set construction unit is used for carrying out pairwise difference at adjacent moments on the collected MODIS earth surface reflectivity images to construct a multi-year MODIS earth surface reflectivity difference image data set, and constructing an MODIS earth surface reflectivity difference background field according to the multi-year average value of the MODIS earth surface reflectivity difference image data set;
the Landsat difference value prediction unit is used for predicting to obtain a Landsat difference value image of the earth surface reflectivity of the time adjacent to the known time according to the Landsat earth surface reflectivity image of the known time, the MODIS earth surface reflectivity image corresponding to the known time and the MODIS earth surface reflectivity image of the time adjacent to the known time;
the Bayesian unmixing unit is used for decomposing all MODIS surface reflectivity difference images from the known time to the prediction time according to the MODIS surface reflectivity difference background field and the current year-predicted MODIS surface reflectivity difference image data set and obtaining Bayesian decomposition values of all MODIS surface reflectivity difference images from the known time to the prediction time;
the Landsat difference derivation unit is used for deriving and obtaining all Landsat surface reflectivity difference images from the known time to the prediction time by constructing a linear regression model according to the Landsat surface reflectivity difference images at the adjacent time with the known time and Bayesian decomposition values of all MODIS surface reflectivity difference images from the known time to the prediction time;
and the image fusion unit is used for fusing the Landsat surface reflectivity image at the known moment and all the Landsat surface reflectivity difference images from the known moment to the prediction moment to obtain the Landsat surface reflectivity image at the prediction moment.
7. The apparatus of claim 6, wherein the Landsat difference prediction unit is specifically configured to:
establishing a linear regression relationship between the MODIS earth surface reflectivity image and the Landsat earth surface reflectivity image at a known moment to obtain a spatial prediction of the Landsat earth surface reflectivity at a moment adjacent to the known moment;
assuming that the change of the same type of ground objects in a short time is the same, obtaining the time prediction of the Landsat ground surface reflectivity at the time adjacent to the known time, wherein a residual exists between the time prediction and the real image;
assuming that the result of the spatial prediction is closer to the true value, obtaining the error of the temporal prediction relative to the spatial prediction;
introducing a weight distribution coefficient to distribute the error and the residual error;
and predicting the Landsat earth surface reflectivity difference image based on the weight distribution coefficient and the residual error to obtain the Landsat earth surface reflectivity difference image at the moment adjacent to the known moment.
8. The apparatus of claim 6, wherein the Bayesian unmixing unit is specifically configured to:
assuming that the MODIS earth surface reflectivity difference background field obeys an average value RpAnd the covariance matrix is Gaussian distribution of sigma p, and meanwhile, the MODIS surface reflectivity difference image data set in the current year is supposed to be predicted to obey a mean value of R0The covariance matrix is sigma0(ii) a gaussian distribution of;
according to the Bayesian unmixing theory, the posterior distribution of the MODIS surface reflectivity difference values is also obeyed a mean value of mueAnd the covariance matrix is the Gaussian distribution of sigma e to obtain the mueThe minimum variance of (a);
according to said μeAnd constructing a Bayesian decomposition value corresponding to each MODIS surface reflectivity difference image.
9. The apparatus of claim 6, wherein the Landsat difference derivation unit is specifically configured to:
constructing a linear equation of the MODIS earth surface reflectivity difference value and the Landsat earth surface reflectivity difference value:
Lt1＝p×Mt1+qt1
Ltk＝p×Mtk+qtk
wherein L ist1And LtkRespectively correspond to t1Time t andklandsat surface reflectance difference image at time, Mt1And MtkRespectively correspond to t1Time t andkbayesian decomposition values of MODIS surface reflectivity difference images at the moment, p and q are fitting coefficients of a linear equation, k is a natural number and is less than or equal to 45;
and assuming that M and q are in a direct proportion relationship, obtaining:
Figure FDA0003058552810000061
when the Landsat surface reflectance images at n available known times in the year are predicted, the Landsat surface reflectance images are combinedn sets of slaves t1To tnBayesian decomposition values of all MODIS surface reflectivity difference images at the moment are given with weights w of different sizes according to time distance, and the above formula is transformed into:
Figure FDA0003058552810000062
said LtkNamely the derived Landsat surface reflectivity difference image corresponding to the prediction time.
10. The apparatus according to claim 9, wherein the image fusion unit is specifically configured to:
image fusion is performed according to the following formula:
Figure FDA0003058552810000063
wherein L iskLandsat surface reflectance image for time of prediction, L1Is the Landsat surface reflectance image at a known time, Lt1...LtkSequentially all Landsat surface reflectivity difference images from the known time to the predicted time, tk＜t1Indicating that the predicted time is before the known time, tk≥t1Indicating that the predicted time is equal to or after the known time.
CN202110506228.8A 2021-05-10 2021-05-10 Method and device for fusing earth surface reflectivity images Active CN113222010B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110506228.8A CN113222010B (en) 2021-05-10 2021-05-10 Method and device for fusing earth surface reflectivity images

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110506228.8A CN113222010B (en) 2021-05-10 2021-05-10 Method and device for fusing earth surface reflectivity images

Publications (2)

Publication Number Publication Date
CN113222010A true CN113222010A (en) 2021-08-06
CN113222010B CN113222010B (en) 2022-10-04

Family

ID=77094199

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110506228.8A Active CN113222010B (en) 2021-05-10 2021-05-10 Method and device for fusing earth surface reflectivity images

Country Status (1)

Country Link
CN (1) CN113222010B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108932710A (en) * 2018-07-10 2018-12-04 武汉商学院 Remote sensing Spatial-temporal Information Fusion method
CN109033543A (en) * 2018-06-29 2018-12-18 北京师范大学 A kind of heterogeneous area's vegetation fraction estimation method, device and equipment of earth's surface
CN110147832A (en) * 2019-05-08 2019-08-20 北京师范大学 The method model of high space-time timing EVI index is obtained based on bayesian theory
US20200082151A1 (en) * 2018-09-06 2020-03-12 National Central University Method of Top-of-Atmosphere Reflectance-Based Spatiotemporal Image Fusion Using Aerosol Optical Depth
CN111242022A (en) * 2020-01-10 2020-06-05 西安科技大学 High-resolution FAPAR estimation method based on low-resolution remote sensing product downscaling

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109033543A (en) * 2018-06-29 2018-12-18 北京师范大学 A kind of heterogeneous area's vegetation fraction estimation method, device and equipment of earth's surface
CN108932710A (en) * 2018-07-10 2018-12-04 武汉商学院 Remote sensing Spatial-temporal Information Fusion method
US20200082151A1 (en) * 2018-09-06 2020-03-12 National Central University Method of Top-of-Atmosphere Reflectance-Based Spatiotemporal Image Fusion Using Aerosol Optical Depth
CN110147832A (en) * 2019-05-08 2019-08-20 北京师范大学 The method model of high space-time timing EVI index is obtained based on bayesian theory
CN111242022A (en) * 2020-01-10 2020-06-05 西安科技大学 High-resolution FAPAR estimation method based on low-resolution remote sensing product downscaling

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
LEI YANG 等: "Monitoring and Risk Assessment of High-Temperature Heat Damage for Summer Maize Based on Remote Sensing Data", 《IGARSS 2020 - 2020 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENSING SYMPOSIUM》 *
SONG JINLING 等: "The method on generating LAI production by fusing BJ-1 remote sensing data and modis LAI product", 《2009 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENSING SYMPOSIUM》 *
杨磊 等: "基于遥感数据的夏玉米高温热害监测评估", 《应用气象学报》 *

Also Published As

Publication number Publication date
CN113222010B (en) 2022-10-04

Similar Documents

Publication Publication Date Title
CN110738252B (en) Space autocorrelation machine learning satellite precipitation data downscaling method and system
US11798228B2 (en) Method for updating road signs and markings on basis of monocular images
CN105354841B (en) A kind of rapid remote sensing image matching method and system
Boulch et al. Ionospheric activity prediction using convolutional recurrent neural networks
CN110428376B (en) FPGA-based linear array CCD satellite image on-satellite geometric correction method
CN109033543B (en) Method, device and equipment for estimating vegetation coverage of heterogeneous land areas
CN113870425A (en) Forest accumulation space mapping method based on random forest and multi-source remote sensing technology
CN115236655A (en) Landslide identification method, system, equipment and medium based on fully-polarized SAR
CN114037891A (en) High-resolution remote sensing image building extraction method and device based on U-shaped attention control network
Moghaddam et al. A statistical variable selection solution for RFM ill-posedness and overparameterization problems
Jiang et al. Super-resolution for terrain modeling using deep learning in high mountain Asia
CN108229426B (en) Remote sensing image change vector change detection method based on difference descriptor
CN114462486A (en) Training method of image processing model, image processing method and related device
Kosar et al. A case study comparing citizen science aurora data with global auroral boundaries derived from satellite imagery and empirical models
CN113222010B (en) Method and device for fusing earth surface reflectivity images
CN112255604A (en) Method and device for judging accuracy of radar data and computer equipment
CN115984646B (en) Remote sensing cross-satellite observation oriented distributed target detection method and device and satellite
CN111583166A (en) Image fusion network model construction and training method and device
CN116128762A (en) Satellite image preprocessing method and device, electronic equipment and storage medium
US7310440B1 (en) Replacement sensor model for optimal image exploitation
CN109919998B (en) Satellite attitude determination method and device and terminal equipment
CN117148350B (en) SAR image-based earth surface temperature space downscaling method, device and equipment
CN104463895B (en) A kind of earth's surface monitoring image processing method based on SAR
CN117609533B (en) Automatic screening method, device and equipment for synthetic aperture radar interference image pair
Bellet et al. End-to-end Learning for Land Cover Classification using Irregular and Unaligned SITS by Combining Attention-Based Interpolation with Sparse Variational Gaussian Processes

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant