CN111353937B - Super-resolution reconstruction method of remote sensing image - Google Patents
Super-resolution reconstruction method of remote sensing image Download PDFInfo
- Publication number
- CN111353937B CN111353937B CN202010129703.XA CN202010129703A CN111353937B CN 111353937 B CN111353937 B CN 111353937B CN 202010129703 A CN202010129703 A CN 202010129703A CN 111353937 B CN111353937 B CN 111353937B
- Authority
- CN
- China
- Prior art keywords
- image
- ground object
- reflectivity
- band
- remote sensing
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 30
- 238000002310 reflectometry Methods 0.000 claims abstract description 62
- 230000003595 spectral effect Effects 0.000 claims abstract description 51
- 238000009826 distribution Methods 0.000 claims abstract description 25
- 238000003384 imaging method Methods 0.000 claims abstract description 22
- 239000000126 substance Substances 0.000 claims abstract description 11
- 238000001228 spectrum Methods 0.000 claims description 18
- 230000005855 radiation Effects 0.000 claims description 15
- 239000000203 mixture Substances 0.000 claims description 14
- 238000000605 extraction Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 6
- 238000012937 correction Methods 0.000 claims description 3
- 239000013598 vector Substances 0.000 description 16
- 238000004422 calculation algorithm Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 239000011159 matrix material Substances 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000000985 reflectance spectrum Methods 0.000 description 3
- 238000002834 transmittance Methods 0.000 description 3
- 230000005457 Black-body radiation Effects 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 239000012237 artificial material Substances 0.000 description 1
- 230000003416 augmentation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000000701 chemical imaging Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 229910052500 inorganic mineral Inorganic materials 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 239000011707 mineral Substances 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000011435 rock Substances 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000002689 soil Substances 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4053—Scaling of whole images or parts thereof, e.g. expanding or contracting based on super-resolution, i.e. the output image resolution being higher than the sensor resolution
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/28—Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/194—Terrestrial scenes using hyperspectral data, i.e. more or other wavelengths than RGB
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Multimedia (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a super-resolution reconstruction method of a remote sensing image, which comprises the following steps: firstly, carrying out radiometric calibration on an input remote sensing image, and converting a gray value of the image into a spectral radiance value; then converting the spectral radiance value of each point in the image into ground object rate data through a remote sensing link imaging model; then calculating a ground object distribution model in the image, comparing the ground object reflectivity in the ground object distribution model with a standard ground object reflectivity database, matching the real ground object corresponding to each classification through an error function model, and calculating the spectral reflectivity of the to-be-supersub-band according to the matched substance and the spectral information of the to-be-supersub-band; and finally, completing super-resolution reconstruction of the super-sub-band to be processed through a remote sensing imaging link model.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to the field of image super-resolution reconstruction.
Background
With the rapid development of science and technology in recent years, the resolution of a sensor carried by a remote sensing platform is higher and higher, and the requirement of people on the spatial resolution of an image is also higher and higher. The resolution of the image represents how much information is contained, as the resolution of the image increases, the details in the image typically become more numerous, and a high resolution image facilitates target recognition and accurate interpretation.
However, due to the influence of factors such as shooting environment and hardware equipment, the resolution of the obtained image generally cannot meet the research requirement. In particular, multispectral and hyperspectral remote sensing images, because of their band limitation, cannot obtain high resolution image data meeting the needs of researchers in all visible and infrared bands. A method capable of super-resolution reconstruction of a low-resolution image of a band of interest from a high-resolution image of an existing band is now an urgent need.
Disclosure of Invention
Aiming at the technical problems, the invention aims to provide a novel super-resolution reconstruction method of a remote sensing image, which can obtain a remote sensing image of an interested wave band by utilizing a plurality of existing remote sensing images and realize super-resolution reconstruction work of the interested wave band by utilizing the intrinsic characteristics of the existing images.
In order to solve the technical problems, the invention provides a super-resolution reconstruction method of a remote sensing image, which comprises the following steps:
(1) Performing radiometric calibration on an input remote sensing image to be superseparated, and converting a gray value of the image into a spectral radiance value;
(2) Converting the spectral radiance value of each point in the image into ground object reflectivity data through a remote sensing link imaging model;
(3) According to whether the low-resolution image with input exists in the super-sub-band or not, calculating the reflectivity of the ground object in the image by using different calculation methods, comparing the reflectivity with a standard ground object reflectivity database, and matching the real ground object corresponding to each classification through an error function model;
(4) Calculating spectral reflectivity of the to-be-supersub-band by using different methods according to whether the to-be-supersub-band has an input low-resolution image or not, and calculating spectral reflectivity of the ground object according to the matched substance and the spectral information of the to-be-supersub-band;
(5) And (3) completing super-resolution reconstruction of the super-sub-band to be processed through a remote sensing imaging link model.
Preferably, in the step (1), the step of obtaining the spectral radiance value is as follows:
(1) Obtaining a camera calibration coefficient according to the existing remote sensing image file information;
(2) And performing radiation calibration processing on the remote sensing image by using the calibration coefficient, and converting the gray value of the image into a spectrum radiation brightness value.
Preferably, in the step (2), the step of calculating the reflectivity of the ground object is as follows:
(1) Constructing a remote sensing link imaging model according to a remote sensing imaging principle;
(2) Calculating specific numerical values of all parameters in the remote sensing link imaging model based on atmospheric correction;
(3) And calculating the reflectivity of the ground object by using each parameter value in the remote sensing link imaging model.
Preferably, in the step (3), the step of matching the ground object type is as follows:
(1) Establishing a standard ground object reflectivity database by using the existing digital ground object reflectivity database;
(2) Aiming at the condition that a low-resolution image is input in the super-sub-band, extracting a spectrum curve of an end member in an image scene by using an end member extraction method, comparing the spectrum curve with a standard ground object reflectivity database, determining the type of the ground object, removing the ground objects which are not in the scene, and constructing a complete dictionary;
(3) Clustering the input high-resolution images by using a Gaussian mixture model aiming at the condition that no low-resolution image is input in the super-sub-band, comparing the ground feature reflectivity of each point in the images with a built standard ground feature reflectivity database, and matching the real ground features corresponding to each classification by using an error function model.
Preferably, in the step (4), the step of calculating the spectral reflectance of the ground object to be supersub-band is as follows:
(1) Aiming at the condition that a low-resolution image is input to the super-sub-band, calculating a ground object distribution model by using a sparse representation method, and obtaining an optimal solution of a ground object distribution coefficient through continuous iterative optimization so as to calculate the spectral reflectivity of the super-sub-band;
(2) And calculating the spectral reflectivity of the ground object by using Gaussian mixture probability distribution according to the spectral information of the matched substance and the to-be-supersub-band under the condition that the to-be-supersub-band is not input with the low-resolution image.
The beneficial effects of the invention are as follows: the super-resolution reconstruction can be performed on the low-resolution image of the interested wave band according to the high-resolution image of the known wave band, so that the high-resolution image in the interested wave band is obtained.
Drawings
FIG. 1 is a schematic flow chart of the method of the present invention.
Detailed Description
The technical solution of the present invention will now be fully described with reference to fig. 1. The following description is of some, but not all embodiments of the invention. All other embodiments, based on the embodiments of the invention, which are obtained by a person skilled in the art without making any inventive effort, are within the scope of the claims of the present invention.
The super-resolution reconstruction method of the remote sensing image provided by the invention comprises the following steps of:
(1) Performing radiometric calibration on an input remote sensing image to be superseparated, and converting a gray value of the image into a spectral radiance value;
(2) Converting the spectral radiance value of each point in the image into ground object reflectivity data through a remote sensing link imaging model;
(3) According to whether the low-resolution image with input exists in the super-sub-band or not, calculating the reflectivity of the ground object in the image by using different calculation methods, comparing the reflectivity with a standard ground object reflectivity database, and matching the real ground object corresponding to each classification through an error function model;
(4) Calculating spectral reflectivity of the to-be-supersub-band by using different methods according to whether the to-be-supersub-band has an input low-resolution image or not, and calculating spectral reflectivity of the ground object according to the matched substance and the spectral information of the to-be-supersub-band;
(5) And (3) completing super-resolution reconstruction of the super-sub-band to be processed through a remote sensing imaging link model.
Preferably, in the step (1), the step of obtaining the spectral radiance value is as follows:
(1) Obtaining a camera calibration coefficient according to the existing remote sensing image file information;
(2) And performing radiation calibration processing on the remote sensing image by using the calibration coefficient, and converting the gray value of the image into a spectrum radiation brightness value.
Preferably, in the step (2), the step of calculating the reflectivity of the ground object is as follows:
(1) Constructing a remote sensing link imaging model according to a remote sensing imaging principle;
(2) Calculating specific numerical values of all parameters in the remote sensing link imaging model based on atmospheric correction;
(3) And calculating the reflectivity of the ground object by using each parameter value in the remote sensing link imaging model.
Preferably, in the step (3), the step of matching the ground object type is as follows:
(1) Establishing a standard ground object reflectivity database by using the existing digital ground object reflectivity database;
(2) Aiming at the condition that a low-resolution image is input in the super-sub-band, extracting a spectrum curve of an end member in an image scene by using an end member extraction method, comparing the spectrum curve with a standard ground object reflectivity database, determining the type of the ground object, removing the ground objects which are not in the scene, and constructing a complete dictionary;
(3) Clustering the input high-resolution images by using a Gaussian mixture model aiming at the condition that no low-resolution image is input in the super-sub-band, comparing the ground feature reflectivity of each point in the images with a built standard ground feature reflectivity database, and matching the real ground features corresponding to each classification by using an error function model.
Preferably, in the step (4), the step of calculating the spectral reflectance of the ground object to be supersub-band is as follows:
(1) Aiming at the condition that a low-resolution image is input to the super-sub-band, calculating a ground object distribution model by using a sparse representation method, and obtaining an optimal solution of a ground object distribution coefficient through continuous iterative optimization so as to calculate the spectral reflectivity of the super-sub-band; (2) And calculating the spectral reflectivity of the ground object by using Gaussian mixture probability distribution according to the spectral information of the matched substance and the to-be-supersub-band under the condition that the to-be-supersub-band is not input with the low-resolution image.
Example 1
1. Spectral radiance calculation
The radiation calibration refers to the process of converting the brightness gray value of an image into absolute radiation brightness when the spectral reflectivity or the spectral radiation brightness of the ground object needs to be calculated. For radiometric calibration of remote sensing images, the absolute calibration coefficient (which can be read from the image file information) can be used for solving the brightness value, and the formula is as follows:
wherein: QCAL is the original quantized DN value; LMIN λ A radiance value at qcal=0; LMAX λ A radiance value at qcal=qalamax; qcmax is the maximum quantized scaled pixel value; QCALMIN is the minimum quantized scaled pixel value. Qcelmin=0 in general, so the formula can be reduced to:
and (3) the following steps:
Offset＝LMIN λ
the end can be reduced to:
L λ ＝Gain·DN+Offset
wherein: l (L) λ The unit of the converted radiation brightness value is W.m -2 ·sr -1 ·μm -1 The method comprises the steps of carrying out a first treatment on the surface of the DN is satellite load observation value, namely the gray value of the recorded ground object; gain is the scaling slope in W.m -2 ·sr -1 ·μm -1 The method comprises the steps of carrying out a first treatment on the surface of the Offset is the absolute scaling factor Offset, also in units of: w.m -2 ·sr -1 ·μm -1 。
2. Ground object reflectivity calculation
Along with the change of factors such as observation wave bands, observation conditions and the like, the pixel value of each point in the image is changed, but the substance corresponding to each point is unchanged, so that a solid foundation can be provided for the subsequent super-resolution reconstruction by considering that each point in the image is restored to a real (closest to the real) ground object.
In order to reduce errors as much as possible and restore real ground features, a remote sensing link imaging model is constructed based on a remote sensing imaging principle, and the remote sensing link imaging model is shown in the following formula: each point in the image is converted into eigenvalue of ground feature-reflectivity:
in the above, L λ The radiation intensity of the lambda spectrum of the remote sensing to the ground received by the space-based sensing device is represented; e (E) s ′ λ Representing the intensity of solar illumination radiation reaching the outer layer of the atmosphere, controlled by the azimuth angle of the sun; σ' represents the solar zenith angle; r (λ) represents the spectral reflectance of a typical feature in the λ band; r is (r) d (lambda) diffuse reflectance of the ground object; τ 1 (lambda) represents the atmospheric transmittance of the sun to the ground, τ 2 (lambda) represents the atmospheric transmittance of the ground to the sensor; f represents sky shape parameter, and the value is 0-1; epsilon (lambda) is the specific spectral emissivity of the ground object; l (L) Tλ The blackbody radiation spectrum brightness with the temperature of T; e (E) dsλ The irradiation intensity of the sunlight scattered by the atmosphere is reflected for the ground surface; e (E) dελ Irradiance reflected by the earth's surface for atmospheric downlink thermal radiation; l (L) usλ The radiance of sunlight scattered by the atmosphere; l (L) uελ The brightness of the heat radiation is the upward heat radiation of the atmosphere.
On the basis of not considering the shape of the earth surface, the formulation can be simplified as:
the gray value of each point in the image is converted into the ground object reflectivity value through the remote sensing link model, and the following formula can be obtained:
x＝Tk
in the above formula, T is the relation between the spectral radiance and the reflectivity, and k is the reflectivity of the ground object.
The earth surface temperature, namely the temperature of the ground, can well reflect the balance of earth surface energy, is a key factor for researching the earth surface, adopts a single window algorithm, and has the earth surface temperature inversion formula as follows:
C＝ε·τ
D＝(1-τ)[1+τ·(1-ε)]
in the above formula, a, b is a linear regression coefficient, wherein a= -67.355351, b= 0.458606; t (T) a For the average action temperature of the atmosphere, T can be calculated by an empirical formula a =16.0110+0.92621×300; t is the bright temperature received by the sensor; l (L) λ Spectral radiance for an input band image; k (K) 1 ，K 2 Is constant and is determined before satellite transmission 1 = 774.89, unit mW/cm -2 /sr/μm；K 2 = 1321.08, unit K; epsilon is the emissivity of the earth's surface, and can be obtained by looking up a table; τ is the atmospheric transmittance and can be obtained by simulation using Modtran software.
3. Ground object type matching
3.1 construction of a Standard ground object reflectivity database (dictionary)
The standard ground object reflectivity database comprises three internationally representative digital ground object reflectivity spectrum databases, which are respectively: a USGS ground object reflectance spectrum database, a JHU ground object reflectance spectrum database, and a JPL ground object reflectance spectrum database. The wavelength ranges from near ultraviolet to long wave infrared, and the covering materials comprise thousands of common rocks, minerals, soil, vegetation, various artificial materials and the like.
And establishing a required ground object reflectivity spectrum dictionary by performing band screening and interpolation operation on the constructed ground object reflectivity database. Because the spectral resolution in the database is very high, and the spectral reflectance curve of the substance in a very short wave band range mostly shows a smooth state and little abrupt change and step change exist, the database can be supplemented by a bicubic interpolation mode, and a normalized complete dictionary is established.
3.2 end member extraction
End member extraction is the first step of calculating the hyperspectral image scene feature type and distribution thereof. By "end member" is meant that the pure data representing the idealized class sample in the hyperspectral image is often free of absolute pure pixels in the scene of the multispectral image due to factors such as insufficient spatial resolution of the spectral imaging detector and atmospheric interference, and each pixel is usually formed by linearly mixing 2 or more ground objects. In order to obtain a ground object distribution model in a scene, ground object spectrum information of the scene can be calculated by a mixed pixel decomposition method, and the ground object spectrum information represents the spectral reflectivity of each ground object in the scene. The method adopts an N-FINDR end member extraction algorithm to solve the ground object spectrum curve in the image, and comprises the following specific steps:
(1) The hyperspectral image is x= { X i I=1, 2, n, initializing end member number p, reducing the dimension of the original image to (p-1), and adding one full 1 line, resulting in an image of X' = { X i ′},i＝1,2,...,n；
(2) Initializing a sample counter j=0, and an end member counter k=0, wherein an end member matrix is E= [ E ] 1 ,e 2 ,....,e p ]；
(3) If j is greater than or equal to n, turning to the step (7), otherwise j=j+1;
(4) If k is greater than or equal to p, k=0, go to step (3), otherwise k=k+1, go to step (5);
(5) Constructing a new end member matrix E' = [ E 1 ′,e 2 ′,...,e k-1 ′,e j ′,e k+1 ′,....,e p ′]The kth vector in the end member matrix E is replaced by the jth vector with heavy X ', the convex body volumes V (E) and V (E') of E and E 'are calculated respectively, if V (E) > V (E'), the step (4) is carried out, and otherwise, the step (6) is carried out;
(6) E=e', go to step (4);
(7) And selecting p vectors corresponding to the vectors selected from the end member matrix E in the original image X as end members of the hyperspectral image.
In order to ensure the feasibility of the end-member extraction procedure, firstly, the consistency of the image size needs to be ensured, for this purpose, a bicubic interpolation mode is adopted to up-sample the low resolution image (except the wave band with the highest resolution) in the input data, the up-sample is unified to the highest resolution of the input data, and the up-sampled input image data is extracted by adopting an N-FINDR end-member extraction algorithm to extract pure pixels.
And according to the extracted end member spectrum curve, matching the end member spectrum curve with a constructed dictionary, and determining the specific ground object type. The matching condition of the scene end member and the spectrum library ground object is evaluated by adopting the related distance so as to determine the ground object type of the end member, and the method is defined as follows:
in the above, d s,t For the correlation distance between the end member vector and the corresponding spectral reflectance vector in the dictionary, x s As an end-member vector, the number of the end-member vectors,is the average value of the end member vector, x t For the corresponding spectral band spectral reflectance vector in the dictionary, < +.>Is x t Average value of (2).
Because the algorithm for calculating the feature distribution is dependent on the completeness of the database, the database needs to be supplemented and updated, if the matching error is too large during the feature distribution calculation, the end member can be considered as a new substance which does not exist in the database, the database needs to be supplemented at the moment, a new temporary database is constructed, and the end member of the non-existing substance is supplemented on the basis of the original database to serve as a new substance.
After the feature distribution is determined, dimensions of the dictionary can be reduced, features which are not in the scene are removed, a complete dictionary U is constructed, and the spectral reflectivity of the feature represented by the dictionary is introduced:
k＝U·z
in the above formula, z represents a coefficient.
Thus, the solution formula for the luminance value can be converted into:
y＝MBTUz
3.3 Gaussian mixture clustering
Aiming at the characteristics of various pixel types and complex mixed pixel components, a Gaussian mixed clustering method is adopted to cluster the images. The gaussian mixture model is a commonly used model describing a mixture density distribution, i.e. a mixture of a plurality of gaussian distributions. The Gaussian mixture model is a density estimation method of edition parameters, integrates the advantages of the parameter estimation method and the cost parameter estimation method, is not limited to a specific probability density function form, has no relation with the size of a sample set in model complexity, and comprises the following specific algorithm:
(1) Setting several categories, namely, several Gaussian distributions;
(2) For each Gaussian distribution, assigning values to the mean and variance of the Gaussian distribution at random;
(3) Calculating, for each sample, its probability under a respective gaussian distribution;
(4) For each gaussian, the contribution of each sample to the gaussian can be represented by the probability below it, with a large probability representing a large contribution and vice versa. This computes the weighted mean and variance using the contribution of the sample to the gaussian as a weight. Then replacing the original mean and variance;
(5) Repeating the steps (3) and (4) until the mean and variance of each gaussian distribution converge.
3.4 Gaussian mixture cluster matching evaluation index
And evaluating the matching condition of the spectral reflectivity of the input image and the ground object of the spectral library by adopting the related distance so as to determine the specific ground object type, wherein the definition is as follows:
in the above, d s,t For the correlation distance between the end member vector and the corresponding spectral reflectance vector in the dictionary, x s As an end-member vector, the number of the end-member vectors,is the average value of the end member vector, x t For the corresponding spectral band spectral reflectance vector in the dictionary, < +.>Is x t The smaller the correlation distance, the higher the matching accuracy is considered.
4. Sparse representation and optimization solution
Probability solution under Gaussian distribution is a complex problem of high-dimensional multivariable, and the probability solution can be carried out by adopting a sparse representation method in machine learning, and the specific representation mode is as follows:
where φ is a regularization term, λ' is a regularization coefficient, D h ,D v Is two block diagonal linear operators (each with the same block) that approximate the horizontal and vertical derivatives of the image to z.
The Lagrange method commonly used for solving the sparse dictionary optimization problem at present is adopted, and the augmentation Lagrange formula is as follows:
wherein: d, d 1 ,d 2 ,d 3 For scaling the langerhans multiplier, μ is a weight coefficient (μ > 0), the remaining parameters are derived by the ADMM solution, see formula:
v 1 ＝TUz
v 2 ＝D h z
v 3 ＝D v z
by setting an initial value (optional), and through continuous iterative optimization, converging to the minimum, the parameters can be solved, and substituted into the following formula to solve the representation coefficient z:
z＝(I+D h D h T +D v D v T ) -1 {U(v 1 +d 1 )+D h T (v 2 +d 2 )+D v T (v 3 +d 3 )}。
the previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (3)
1. The super-resolution reconstruction method of the remote sensing image is characterized by comprising the following steps of:
step 1), carrying out radiometric calibration on an input image to be superdivided, and converting a gray value of the image into a spectral radiance value;
step 2) converting the spectral radiance value of each point in the image into ground object reflectivity data through a remote sensing link imaging model;
step 3) according to whether the low-resolution image is input or not in the super-sub-band, calculating the reflectivity of the ground object in the image by using an end member extraction or Gaussian cluster calculation method, comparing the reflectivity with a standard ground object reflectivity database, and matching the real ground object corresponding to each classification through an error function model; in step 3), calculating and determining the ground object type comprises the following steps:
step 3.1, establishing a standard ground object reflectivity database by utilizing the existing digital ground object reflectivity database;
step 3.2, aiming at the condition that a low-resolution image is input in the super-sub-band to be detected, extracting a spectrum curve of an end member in an image scene by using an end member extraction method, comparing the extracted spectrum curve with a standard ground object reflectivity database, determining the type of the ground object, removing the ground object which is not in the scene, and constructing a complete dictionary;
step 3.3, clustering the input high-resolution images by using a Gaussian mixture model aiming at the condition that no low-resolution image is input in the super-sub-band, comparing the ground feature reflectivity of each point in the images with a built standard ground feature reflectivity database, and matching the real ground feature corresponding to each classification by using an error function model;
step 4) calculating the spectral reflectivity of the to-be-supersub-band by using a sparse representation or Gaussian mixture probability distribution method according to whether the to-be-supersub-band has the input low-resolution image or not; in the step 4), calculating the reflectivity of the ground object to be supersub-band comprises the following steps:
step 4.1, calculating a ground object distribution model by using a sparse representation method according to the condition that a low-resolution image is input in the to-be-supersub-band, and obtaining an optimal solution of a ground object distribution coefficient through continuous iterative optimization so as to calculate the spectral reflectivity of the to-be-supersub-band;
step 4.2, calculating the ground object spectral reflectivity by using Gaussian mixture probability distribution according to the spectral information of the matched substance and the to-be-supersub-band aiming at the condition that the to-be-supersub-band is not input with a low resolution image;
and 5) completing super-resolution reconstruction of the super-sub-band to be processed through a remote sensing imaging link model.
2. The method for super-resolution reconstruction of a remote sensing image according to claim 1, wherein in step 1), calculating the spectral radiance value comprises the steps of:
step 1.1, reading and obtaining scaling coefficients of a camera in image file information;
and 1.2, performing radiation scaling treatment on the image by using the scaling coefficient, and converting the gray value of the image into a spectral radiation brightness value.
3. The method for super-resolution reconstruction of a remote sensing image according to claim 1, wherein in the step 2), calculating the ground object reflectivity image comprises the steps of:
step 2.1, constructing a remote sensing link imaging model according to a remote sensing imaging principle;
step 2.2, calculating the values of all parameters in a remote sensing link imaging model formula based on atmospheric correction;
and 2.3, calculating the ground object reflectivity image by using the obtained parameter values.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010129703.XA CN111353937B (en) | 2020-02-28 | 2020-02-28 | Super-resolution reconstruction method of remote sensing image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010129703.XA CN111353937B (en) | 2020-02-28 | 2020-02-28 | Super-resolution reconstruction method of remote sensing image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111353937A CN111353937A (en) | 2020-06-30 |
| CN111353937B true CN111353937B (en) | 2023-09-29 |
Family
ID=71195905
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010129703.XA Active CN111353937B (en) | 2020-02-28 | 2020-02-28 | Super-resolution reconstruction method of remote sensing image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111353937B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111798378A (en) * | 2020-07-08 | 2020-10-20 | 北京航空航天大学 | Thermal infrared image super-resolution reconstruction evaluation method |
| CN111899257A (en) * | 2020-08-14 | 2020-11-06 | 哈尔滨工业大学 | Ground object spectral reflectivity image extraction method based on multi-temporal intrinsic image decomposition |
| CN112767292B (en) * | 2021-01-05 | 2022-09-16 | 同济大学 | Geographic weighting spatial hybrid decomposition method for space-time fusion |
| CN112800857A (en) * | 2021-01-07 | 2021-05-14 | 北京中云伟图科技有限公司 | Bare land rapid extraction method based on high-resolution satellite data |
| CN112967350B (en) * | 2021-03-08 | 2022-03-18 | 哈尔滨工业大学 | Hyperspectral remote sensing image eigen decomposition method and system based on sparse image coding |
| CN114841879B (en) * | 2022-04-24 | 2023-06-20 | 中国科学院空天信息创新研究院 | Water body atmosphere correction method and system |
| CN115031854A (en) * | 2022-06-22 | 2022-09-09 | 中国科学院空天信息创新研究院 | Thermal infrared hyperspectral earth surface temperature emissivity spectrum inversion method and device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107274343A (en) * | 2017-06-01 | 2017-10-20 | 清华大学 | Multi-spectral remote sensing image spectrum super-resolution method based on library of spectra under a kind of sparse framework |
| CN108830846A (en) * | 2018-06-12 | 2018-11-16 | 南京航空航天大学 | A kind of high-resolution all band Hyperspectral Remote Sensing Image emulation mode |
| CN109410165A (en) * | 2018-11-14 | 2019-03-01 | 太原理工大学 | A kind of multi-spectral remote sensing image fusion method based on classification learning |
-
2020
- 2020-02-28 CN CN202010129703.XA patent/CN111353937B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107274343A (en) * | 2017-06-01 | 2017-10-20 | 清华大学 | Multi-spectral remote sensing image spectrum super-resolution method based on library of spectra under a kind of sparse framework |
| CN108830846A (en) * | 2018-06-12 | 2018-11-16 | 南京航空航天大学 | A kind of high-resolution all band Hyperspectral Remote Sensing Image emulation mode |
| CN109410165A (en) * | 2018-11-14 | 2019-03-01 | 太原理工大学 | A kind of multi-spectral remote sensing image fusion method based on classification learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111353937A (en) | 2020-06-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111353937B (en) | Super-resolution reconstruction method of remote sensing image | |
| CN112766199B (en) | Hyperspectral image classification method based on self-adaptive multi-scale feature extraction model | |
| US20060017740A1 (en) | Diurnal variation of geo-specific terrain temperatures in real-time infrared sensor simulation | |
| CN108830846B (en) | High-resolution full-waveband hyperspectral remote sensing image simulation method | |
| CN107688003B (en) | Blade reflectivity satellite remote sensing extraction method for eliminating vegetation canopy structure and earth surface background influence | |
| Cazorla et al. | Using a sky imager for aerosol characterization | |
| CN114581349A (en) | Visible light image and infrared image fusion method based on radiation characteristic inversion | |
| CN113887124A (en) | Method for predicting photovoltaic power station cloud layer distribution state under multi-dimensional mixed weather | |
| CN109671038A (en) | One kind is based on the classified and layered relative radiometric correction method of pseudo- invariant features point | |
| Del Rocco et al. | Real-time spectral radiance estimation of hemispherical clear skies with machine learned regression models | |
| Young et al. | Development and validation of the AFIT scene and sensor emulator for testing (ASSET) | |
| CN115792908B (en) | Target detection method based on high-resolution multi-angle spaceborne SAR feature fusion | |
| CN110070513B (en) | Radiation correction method and system for remote sensing image | |
| CN117035066A (en) | Ground surface temperature downscaling method coupling geographic weighting and random forest | |
| CN109410165B (en) | Multispectral remote sensing image fusion method based on classification learning | |
| CN116628946A (en) | Space-based low-illumination low-light-level remote sensing image simulation method | |
| CN115015147A (en) | High-spatial-resolution high-spectral thermal infrared remote sensing image simulation method | |
| Rahman | Influence of atmospheric correction on the estimation of biophysical parameters of crop canopy using satellite remote sensing | |
| CN114296061A (en) | Cross calibration method based on multivariate variable detection and different radiation transmission models | |
| KR102125723B1 (en) | Method and apparatus for relative radiometric normalization of image | |
| Stow | Radiometric Correction of Remotely Sensed Data | |
| CN113592737B (en) | Remote sensing image topography correction effect evaluation method based on entropy weight method | |
| CN118154429B (en) | Remote sensing image reconstruction method | |
| Czech et al. | Estimation of daylight spectral power distribution from uncalibrated hyperspectral radiance images | |
| Xu et al. | Ill-posed surface emissivity retrieval from multi-geometry hyperspectral images using a hybrid deep neural network |
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 |