A method, system, and readable storage medium for nonlinear simulation of remote sensing images

Abstract

This invention discloses a method, system, and readable storage medium for nonlinear simulation of remote sensing images. The method includes: inputting a panchromatic image and a multispectral band image; registering the multispectral image with the panchromatic image and resampling it according to high spatial resolution bands; statistically analyzing the mean, root mean square error, and correlation coefficient matrix of the panchromatic and multispectral images; constructing an intermediate band I using the low-resolution multispectral image; constructing an image simulation operator using histogram feature matching of the remote sensing images; and calculating the feature fusion coefficient k. E and feature extraction coefficient k I The method of this invention determines the image simulation scheme and performs image simulation band by band. There is no limitation on the number of image bands, and the simulation results of remote sensing images can be predicted. This invention provides two image simulation modes: panchromatic image injection into multispectral image and multispectral image injection into panchromatic image, enabling high-fidelity simulation of the spectral and spatial information of ground features in remote sensing images.

Description

Technical Field

[0001] This invention relates to the field of remote sensing image fusion technology, and more specifically, to a method, system, and readable storage medium for nonlinear simulation of remote sensing images. Background Technology

[0002] With the development of multi-platform, multi-sensor, multi-weather, multi-temporal, and multi-resolution remote sensing technologies, the types of remote sensing images available for use in various industries are becoming increasingly diverse. These rich and varied remote sensing images not only provide a flexible space for object selection in remote sensing image application research, but also pose challenges to image application preprocessing research such as selection, synthesis, correction, and enhancement of remote sensing images.

[0003] Different remote sensing image data possess different basic image characteristics such as spatial resolution, temporal resolution, and spectral resolution, resulting in varying application performance and potential across different application fields. Traditional remote sensing image processing focuses on enhancing general image features such as color, texture, and tonal gradation of individual image types; remote sensing image fusion processing, on the other hand, emphasizes integrating the basic features of different image types into a new remote sensing image, fully exploring its comprehensive application potential and improving its application performance. Over the past two decades, remote sensing image fusion simulation technology, as a new direction in remote sensing image processing, has made significant progress and achieved a series of new results.

[0004] From the perspective of fusion principles, remote sensing image fusion methods can be divided into three categories: fusion methods based on color space theory, fusion methods based on information analysis principles, and fusion methods based on mathematical operations. Fusion methods based on color space theory apply the main color models RGB, CMYK, Lab, IHS, HSV, etc. to image fusion. Among them, the IHS model is a classic image fusion method (Te-ming TU et al., A new look at IHS-look image fusion methods, Information Fusion 2, 2001, 177-18). The color model components of this type of method are relatively independent and can be controlled separately, which can accurately describe color characteristics (Li Lin et al., Comparison of fusion methods of panchromatic and multispectral images of ZY-3 satellite, Transactions of the Chinese Society of Agricultural Engineering, 30(16), 2014, 157-165). The advantage is that it can perform image fusion quickly and simply. The disadvantage is that although the fusion result obtained by simple component substitution maintains the details of the panchromatic image well, it usually changes the spectral characteristics of the satellite remote sensing image before and after fusion. Information analysis-based fusion methods apply spatial filtering, principal component analysis (PCA), Fourier transform (FFT), wavelet transform, Gram-Schimdt transform, curvelet transform, contourlet transform, ridgelet transform, bandelet transform, wedgelet transform, and beamlet transform to image fusion. These methods inject detailed information from the panchromatic band into multispectral images (Xiao Liang et al., Progress and Challenges of Multi-Source Spatial-Spectral Remote Sensing Image Fusion Methods, Journal of Image and Graphics, 25(05), 2020). The resulting fused image has good spectral continuity, but the computation is complex and often results in the loss of some high-frequency information. Fusion methods based on mathematical operations mainly include ratio methods, difference methods, weighted superposition, multiplicative amplification, and mixed arithmetic operations. Among these methods, the classic methods are the Brovey fusion method and the CN fusion method.

[0005] Brovey fusion is a product-based fusion method proposed by American scholar Brovey, also known as color normalization transformation fusion (Vrabel J. Multispectral Imagery Band Sharpening Study. Photogrammetric Engineering & Remote Sensing, 1996, 62(9): 1075-1083). It is a typical nonlinear fusion method that multiplies three multispectral bands with a high spatial resolution panchromatic image, simplifying the image fusion operation coefficients and preserving the spectral information of the source image to a large extent. However, it can only perform image fusion on three bands. CN fusion (color normalized transformation), also known as energy subdivision transformation, uses high spatial resolution panchromatic bands to enhance the low spatial resolution multispectral bands of the input image. This method only fuses the input bands that are included in the spectral range of the fusion image bands, while other input bands are directly output without fusion processing. Brovey and CN fusion methods require that the spectral response range of the panchromatic band be consistent with or similar to that of the multispectral image. Otherwise, significant deviations in local brightness and darkness will occur in the fused image, altering the spectral characteristics of the multispectral images before and after fusion and causing spectral distortion. To address this, scholars have made various optimizations and improvements to Brovey, CN, and other fusion methods. Some scholars have introduced adaptive optimization methods to adjust the weight parameters of each multispectral band and the panchromatic band. By improving the Brovey fusion algorithm through weighted averaging, the spatial resolution characteristics of the image can be improved while preserving the spectral physical characteristics of the multispectral source image (Lin Zhilei et al., Research on ALI Image Fusion Algorithm Based on Improved Brovey Transform. Remote Sensing Technology and Application, 2020, 35(4):893-900). Some scholars have utilized the nonlinear spectral data mining characteristics of the Kernel Principal Component Transform (KPCA) algorithm to extract the three principal components KPC1, KPC2, and KPC3 with the highest information content from multispectral images. Then, the Brovey algorithm is used to perform normalization and fusion operations on the three principal components and the panchromatic band, making the spatial and spectral information in the fusion result richer (Ke Hongxia et al., A Remote Sensing Image Fusion Method Based on KPCA and Brovey Transform, Journal of Chongqing Jiaotong University (Natural Science Edition), 2019, 38(2)). Some scholars have introduced wavelet analysis to improve the shortcomings of Brovey transform, such as the large influence of noise on the fused image and the excessive retention of sporadic details in high-resolution images (Chen Sijin et al., Improving the Brovey Remote Sensing Image Fusion Method Using Wavelet Analysis, Journal of Surveying and Mapping Institute, 2004, 21(2)).

[0006] Since the emergence of fusion methods such as Brovey and CN, their improvement has been continuous and endless. Overall, current nonlinear fusion methods for remote sensing images still suffer from limitations in the number of affected bands to be fused and the unpredictability of the fusion results. Furthermore, there is still room for improvement in the fidelity of the spectral and spatial information of the fused images. Summary of the Invention

[0007] In view of the above problems, the purpose of this invention is to provide a method, system and readable storage medium for nonlinear simulation of remote sensing images.

[0008] The first aspect of this invention provides a method for nonlinear simulation of remote sensing images, the method comprising:

[0009] S1: Input panchromatic and multispectral images to obtain panchromatic band image P and multispectral band image M. i ;

[0010] Where i∈[1,n], M i Let M1, M2, ..., M n Any band in a multispectral image;

[0011] S2: Spatial registration of panchromatic and multispectral images to ensure that the geometric spatial position of the same ground feature is consistent in both panchromatic and multispectral images, and resampling of the multispectral image according to the high spatial resolution image.

[0012] S3: Calculate the mean, standard deviation, and covariance matrices of panchromatic and multispectral images.

[0013] S4: Construct intermediate band I using low spatial resolution multispectral imagery;

[0014] S5: Constructing image simulation operators using histogram feature matching of remote sensing images;

[0015] S6: Select an image simulation method and determine the feature extraction coefficients k I and feature fusion coefficient k E High-resolution multispectral image simulation calculations are performed based on remote sensing images, image simulation methods, and characteristic parameters.

[0016] S7: Synthesize and store high-resolution multispectral image simulation results.

[0017] Preferably, the intermediate band I is:

[0018] I = ω1M1 + ω2M2 + ... + ω i M i …+ω n M n

[0019] In the formula, ω1+ω2+…ωi …+ω n =1, ω i ≥0.

[0020] Preferably, the image simulation operator includes an image simulation operator M that injects multispectral band images into a panchromatic band image mode. P Image simulation operator I P And the image simulation operator P that injects panchromatic band images into multispectral band image modes. M Image simulation operator I M .

[0021] Preferably, the image simulation operator I P Image simulation operator M P Image simulation operator P M Image simulation operator I M They are respectively:

[0022]

[0023]

[0024] Where, μ M For multispectral band image M i The mean, μ P Let μ be the mean value of the panchromatic image P. I Let σ be the mean of the intermediate band I. M For multispectral band image M i The mean square error, σ I Let σ be the mean square error of the intermediate band I. P Let be the mean square error of the panchromatic image P.

[0025] Preferably, the feature extraction coefficient k I Including but not limited to the following typical solutions:

[0026] (1) When injecting multispectral band images into panchromatic band image mode:

[0027]

[0028] (2) When injecting panchromatic imagery into a multispectral imagery mode:

[0029]

[0030] Where r(P,M) P ) represents the panchromatic image P and the image analog operator M. P The correlation coefficient, r(I,M) P () is the intermediate band I and the image analog operator M P The correlation coefficient, r(P,I)P ) represents the panchromatic image P and the image analog operator I. P The correlation coefficient, r(I,I) P () is the intermediate band I and the image analog operator I P Correlation coefficient, r(M) i ,I P ) is a multispectral band image M i With image analog operator I P The correlation coefficient, r(I) P M P Image simulation operator I P With image analog operator M P Correlation coefficient, r(M) i M P ) is a multispectral band image M i With image analog operator M P Correlation coefficient, r(M) P ,I P ) is the image simulation operator M P With image analog operator I P The correlation coefficient.

[0031] Preferably, the feature fusion coefficient k E Specifically:

[0032] (1) When injecting multispectral imagery into panchromatic imagery:

[0033]

[0034] (2) When injecting panchromatic imagery into multispectral imagery:

[0035]

[0036] Where r(P,P) M ) represents the panchromatic image P and the image analog operator P. P Correlation coefficient, r(M) i P) represents the multispectral band image M. i The correlation coefficient with the panchromatic image P.

[0037] Preferably, the formula for simulating high-resolution multispectral images is:

[0038] (1) When injecting multispectral imagery into panchromatic imagery:

[0039]

[0040] (2) When injecting panchromatic imagery into multispectral imagery:

[0041]

[0042] Where, σ M For multispectral band image M i The mean square error, μ P Let P be the mean value of the panchromatic image. For image analog operator M P The mean, Image analog operator I P The mean, For image analog operator M P The mean squared error, Image analog operator I P The mean square error, σ Mf denoted as the mean square error of the high-resolution multispectral image.

[0043] Preferably, the root mean square error σ of the high-resolution multispectral image is... Mf The calculation formula is:

[0044] (1) When injecting multispectral imagery into panchromatic imagery:

[0045]

[0046] (2) When injecting panchromatic imagery into multispectral imagery:

[0047]

[0048] A second aspect of the present invention provides a remote sensing image nonlinear simulation system, including a memory and a processor. The memory includes a remote sensing image nonlinear simulation method program, which, when executed by the processor, implements the steps of the remote sensing image nonlinear simulation method.

[0049] A third aspect of the present invention provides a computer-readable storage medium including a remote sensing image nonlinear simulation method program, wherein when the remote sensing image nonlinear simulation method program is executed by a processor, it implements the steps of the remote sensing image nonlinear simulation method.

[0050] Compared with existing technologies, the beneficial effects of this invention are as follows: Based on the principle of image scale invariance, this invention constructs an image simulation operator using histogram feature matching of remote sensing images. It derives a nonlinear simulation method for remote sensing images by utilizing spatial and spectral detail differences between images at different spatial scales. This method has no limitation on the number of image bands and can predict the simulation results of remote sensing images. This method provides two image simulation modes: panchromatic band injection into multispectral bands and multispectral band injection into panchromatic bands, enabling high-fidelity simulation of the spectral and spatial information of remote sensing images.

[0051] Visual analysis of the simulated high-resolution multispectral imagery results validated the effectiveness of the proposed scheme. Compared to the original multispectral imagery, the simulated high-resolution multispectral imagery significantly enriches the spatial information of ground features, including geometric details, textures, edges, and layers. Furthermore, compared to the original multispectral imagery, the spectral characteristics and color display of ground features such as water bodies, vegetation, exposed surfaces, and buildings remain largely stable in the simulated high-resolution multispectral imagery. Attached Figure Description

[0052] Figure 1 This is a flowchart of a nonlinear simulation method for remote sensing images as described in Example 1.

[0053] Figure 2 This is a panchromatic (P) image (0.5-meter resolution).

[0054] Figure 3 This is a multispectral NRG false-color composite image (2-meter resolution).

[0055] Figure 4 This is a multispectral RGB true-color composite image (2-meter resolution).

[0056] Figure 5 This is an image of the intermediate band I (2-meter resolution).

[0057] Figure 6 This is a standard false-color composite image of the simulated multispectral band image (0.5-meter resolution).

[0058] Figure 7 This is a true-color composite image of the simulated multispectral band image (0.5-meter resolution). Detailed Implementation

[0059] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0060] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0061] Example 1

[0062] like Figure 1 As shown in the figure, this embodiment discloses a nonlinear simulation method for remote sensing images, the method comprising:

[0063] S1: Input panchromatic and multispectral images to obtain panchromatic band image P and multispectral band image M. i ;

[0064] Where i∈[1,n], M i Let M1, M2, ..., M n Any band in a multispectral image;

[0065] S2: Spatial registration of panchromatic and multispectral images to ensure that the geometric spatial position of the same ground feature is consistent in both panchromatic and multispectral images, and resampling of the multispectral image according to the high spatial resolution image.

[0066] S3: Calculate the mean, standard deviation, and covariance matrices of panchromatic and multispectral images.

[0067] S4: Construct intermediate band I using low spatial resolution multispectral imagery;

[0068] S5: Constructing image simulation operators using histogram feature matching of remote sensing images;

[0069] S6: Select an image simulation method and determine the feature extraction coefficients k I and feature fusion coefficient k E High-resolution multispectral image simulation calculations are performed based on remote sensing images, image simulation methods, and characteristic parameters.

[0070] S7: Synthesize and store high-resolution multispectral image simulation results.

[0071] It should be noted that this embodiment utilizes high-resolution panchromatic remote sensing imagery and low-resolution multispectral imagery. Based on the principle of scale invariance of remote sensing images, a fusion simulation operator is constructed using histogram feature matching to perform nonlinear simulation of high-resolution multispectral images. This overcomes the shortcomings of traditional nonlinear fusion methods, such as limitations on the number of image bands to be fused and the difficulty in predicting the fusion results. Let the remote sensing imagery have multiple low-spatial-resolution multispectral imagery M. i And a high spatial resolution panchromatic image P.

[0072] Generally, satellite remote sensing imagery consists of multiple low-resolution multispectral band images M1, M2, ... M i …、M n Given a high-resolution panchromatic image P, assume that there exists a low-resolution panchromatic band I (intermediate band) corresponding to the high-resolution panchromatic image P, and a high-resolution multispectral image Mf corresponding to the low-resolution multispectral image M.

[0073] Let μ M μ P μI μ Mf σ represents the mean values ​​of the low-resolution multispectral image, the high-resolution panchromatic image, the low-resolution panchromatic image, and the high-resolution multispectral image, respectively. M σ P σ I σ Mf These represent the mean square errors of low-resolution multispectral imagery, high-resolution panchromatic imagery, low-resolution panchromatic imagery, and high-resolution multispectral imagery, respectively. r(P,I) is the correlation coefficient between P and I, and r(Mf,M) is the mean square error of the imagery. i ) for Mf and M i The correlation coefficient.

[0074] This embodiment is based on the principle of scale invariance of remote sensing images, and according to the low spatial resolution multispectral band image M... i The spatial and spectral detail differences among high spatial resolution panchromatic image P, low spatial resolution panchromatic image I, and high spatial resolution multispectral image Mf were investigated, and a nonlinear simulation method for high-resolution multispectral images was constructed.

[0075] Furthermore, this embodiment also provides preferred feature parameters: the remote sensing image nonlinear simulation method of the present invention constructs a pair of image simulation operators M based on histogram feature matching. P and I P And a pair of feature parameters were constructed—the feature extraction coefficient is k. I The feature fusion coefficient is k E This allows for the precise injection of high-precision geometric texture information from the original panchromatic image P and rich spectral information from the original multispectral image M into the simulated image, thus achieving the advantage of high fidelity in both image spectral and spatial information.

[0076] In this embodiment, the intermediate band I is:

[0077] I = ω1M1 + ω2M2 + ... + ω i M i …+ω n M n

[0078] In the formula, ω1+ω2+…ω i …+ω n =1, ω i ≥0.

[0079] In this embodiment, the image simulation operator includes an image simulation operator M that injects multispectral band images into a panchromatic band image mode. P Image simulation operator I P And the image simulation operator P that injects panchromatic band images into multispectral band image modes.M Image simulation operator I M .

[0080] In this embodiment, the image simulation operator I P Image simulation operator M P Image simulation operator P M Image simulation operator I M They are respectively:

[0081]

[0082] Where, μ M For multispectral band image M i The mean, μ P Let μ be the mean value of the panchromatic image P. I Let σ be the mean of the intermediate band I. I Let σ be the mean square error of the intermediate band I. P Let be the mean square error of the panchromatic image P.

[0083] In this embodiment, the feature extraction coefficient k I Including but not limited to the following typical solutions:

[0084] (1) When injecting multispectral band images into panchromatic band image mode:

[0085]

[0086] (2) When injecting panchromatic imagery into a multispectral imagery mode:

[0087]

[0088] Where r(P,M) P ) represents the panchromatic image P and the image analog operator M. P The correlation coefficient, r(I,M) P () is the intermediate band I and the image analog operator M P The correlation coefficient, r(P,I) P ) represents the panchromatic image P and the image analog operator I. P The correlation coefficient, r(I,I) P () is the intermediate band I and the image analog operator I P Correlation coefficient, r(M) i ,I P ) is a multispectral band image M i With image analog operator I P The correlation coefficient, r(I) P M P Image simulation operator I P With image analog operator M PCorrelation coefficient, r(M) i N P ) is a multispectral band image M i With image analog operator M P Correlation coefficient, r(M) P ,I P ) is the image simulation operator M P With image analog operator I P The correlation coefficient.

[0089] It should be noted that in this embodiment, the multispectral band image is injected with the panchromatic band image mode feature parameter k. I The determination process is as follows:

[0090] Let the feature extraction factor

[0091] ① Determine k from r(E,P)=0 I :

[0092] From r(E,P)=0, we get: Cov(E,P)=0

[0093] Right now:

[0094] Find k I The values ​​are as follows:

[0095]

[0096] ② Determine k from r(E,I)=0 I :

[0097] From r(E,I)=0, we get: Cov(E,I)=0,

[0098] Right now:

[0099] Find k I The values ​​are as follows:

[0100]

[0101] ③From r(E,M) i ) = 0, determine k I

[0102] From r(E,M) i ) = 0, therefore: Cov(E,M) i ) = 0

[0103] Right now:

[0104] Find k I The values ​​are as follows:

[0105]

[0106] ④By r(E,I P ) = 0, determine k I :

[0107] By r(E,I P ) = 0, therefore: Cov(E,I) P ) = 0

[0108] Right now:

[0109] Find k I The values ​​are as follows:

[0110] k I =r(I P M P )

[0111] ⑤ By r(E,M) P ) = 0, determine k I

[0112] From r(E,M) P ) = 0, therefore: Cov(E,M) P ) = 0

[0113] Right now:

[0114] Find k I The values ​​are as follows:

[0115]

[0116] In this embodiment, the feature fusion coefficient k E Specifically:

[0117] (1) When injecting multispectral imagery into panchromatic imagery:

[0118]

[0119] (2) When injecting panchromatic imagery into multispectral imagery:

[0120]

[0121] Where r(P,P) M ) represents the panchromatic image P and the image analog operator P. P Correlation coefficient, r(M) i P) represents the multispectral band image M. i The correlation coefficient with the panchromatic image P.

[0122] It should be noted that in this embodiment, the feature parameter k E The determination process is as follows:

[0123] make:

[0124] From r(M) i When f) reaches its maximum value, k can be determined. E The method is as follows:

[0125]

[0126] because

[0127] The following is given k I The correlation coefficient r(M) i The maximum value analysis of f is used to determine k. E :

[0128] make:

[0129] a E =1-2k I r(M P ,I P )+k I 2

[0130] b E =2[r(P,M P )-k I r(P,I P )]

[0131] ω(k E )=1+2k E [r(P,M P )-k I r(P,I P )]+k E 2 [1-2k I r(M P ,I P )+k I 2 ]

[0132] but:

[0133] ω(k E ) = a E k E 2 +b E k E +1

[0134] make:

[0135] m E =r(M i M P )-k1r(M I ,I P )

[0136] n E =r(M I ,P)

[0137] u(k E )=r(M i ,P)+k E [r(M i M P )-k I r(M i ,I P )]

[0138] but:

[0139] u(k E ) = m E k E +n E

[0140] Therefore:

[0141] h(k I ,k E )=r(M i f)=uω -0.5

[0142] That is, to perform a given k I The correlation coefficient h(k) I ,k E )=r(M i Maximum value analysis of f):

[0143] Assuming k is given I Determine k E Make h(k) I ,k E )=r(M i f) has a maximum value.

[0144] Depend on:

[0145] h(k I ,k E )=r(M i f)=uω -0.5

[0146] We can obtain:

[0147]

[0148] make:

[0149] y(k E ) = m E ·ω-0.5u·(2a E k E +b E )

[0150] =m E ·(a E k E 2 +b E k E +1)-(m E k E +n E )·(a E k E +0.5b E )

[0151] =(0.5m) E b E -a E n E )k E +m E -0.5b E n E

[0152] a = 0.5m E b E -a E n E

[0153] b = m E -0.5b E n E

[0154] but:

[0155] y(k E )=ak E +b

[0156] y(k E )=ak E +b is k E It is a linear function.

[0157] When a = 0 and b ≠ 0, y(k) E )≡b, r(M i f) The condition for the existence of an extremum is ω = 0. According to the physical meaning of the correlation coefficient, ω = 0 is impossible, therefore... r(M i f) has no extreme values. That is, r(M) i f) is a monotonically increasing or monotonically decreasing function.

[0158] When a = 0 and b = 0, y(k) E )≡0, r(M i f) is a constant.

[0159] When a>0, y(k) E ) is an increasing function. hour, hour, hour, Therefore At that time, r(M) i f) has a local minimum value.

[0160] When a < 0, y(k) E ) is a decreasing function. hour, hour, hour, At that time, r(M) i f) has a maximum value.

[0161] Therefore, given k I The correlation coefficient r(M) i When f) has a maximum value,

[0162] Find:

[0163]

[0164] Similarly, the panchromatic band image injected into the multispectral band image mode based on histogram matching remote sensing image nonlinear fusion can be obtained as follows:

[0165]

[0166] in,

[0167]

[0168] in, These are image simulation operators P M I M The mean, These are image simulation operators P M I M The mean squared error, k I k represents the feature extraction coefficient. E is the feature fusion coefficient.

[0169] In this embodiment, the formula for simulating high-resolution multispectral images is:

[0170] (1) When injecting multispectral imagery into panchromatic imagery:

[0171]

[0172] (2) When injecting panchromatic imagery into multispectral imagery:

[0173]

[0174]

[0175] Where, σ M For multispectral band image M i The mean square error, μ P Let P be the mean value of the panchromatic image. For image analog operator M P The mean, Image analog operator I P The mean, For image analog operator M P The mean squared error, Image analog operator I P The mean square error, σ Mf denoted as the mean square error of the high-resolution multispectral image.

[0176] It should be noted that, in this embodiment, the high-resolution multispectral image simulation method is as follows:

[0177] (1) Panchromatic space information extraction factor and k1

[0178] Inferred from the scale invariance of remote sensing images: (P-μ) P )=k1(I-μ I )

[0179] make:

[0180] E1=(P-μ P )-k1(I-μ I ) = 0

[0181] Since the zero vector is independent of any non-zero vector, E1 is independent of any non-zero vector X1 (X1 has the same dimension as P), therefore:

[0182] r(E1,X1)=0

[0183] We can obtain:

[0184]

[0185] It can be derived as follows:

[0186]

[0187] Therefore:

[0188]

[0189] (2) The multispectral mean-filtered image is proportional to k2

[0190] When the mean-filtered image is proportional, generally the following applies:

[0191]

[0192]

[0193] (3) Scale invariance corollary—The relationship between high-resolution images and low-resolution product images is derived from:

[0194]

[0195] have to:

[0196]

[0197] We can obtain:

[0198]

[0199] again:

[0200]

[0201] Therefore:

[0202]

[0203] Right now:

[0204]

[0205] make:

[0206]

[0207] have to:

[0208]

[0209] m1 and m2 are any real numbers greater than zero. X1 and X2 are integers related to M. i Any remote sensing image with the same dimensions P and I.

[0210] (4) Nonlinear simulation method for high-resolution multispectral images

[0211] From the above formula, we can obtain:

[0212]

[0213] make:

[0214]

[0215] but:

[0216]

[0217] make:

[0218]

[0219] The above results were matched to the multispectral band image M. i The general scheme for nonlinear simulation of multispectral injection panchromatic mode is as follows:

[0220]

[0221] in,

[0222]

[0223] X1 and X2 are related to M i Images of P and I with the same dimension, where m1 and m2 are any real numbers greater than zero. The image analog operator M is respectively P I P The mean, The image analog operator M is respectively P I P The mean squared error of k. I k represents the feature extraction coefficient. E is the feature fusion coefficient.

[0224] As a specific embodiment, in order to achieve the purpose of nonlinear simulation of high-resolution remote sensing images, this embodiment mainly uses ENVI remote sensing image processing software, and is further described with a satellite remote sensing image with panchromatic band (P), blue band (B), green band (G), red band (R), and near-infrared (N).

[0225] 1. Input remote sensing image.

[0226] Open a WV-02 remote sensing image with panchromatic (P), blue (B), green (G), red (R), and near-infrared (N) bands. Figure 2 , 3Images 4 and 5 are respectively a panchromatic band image (0.5-meter resolution), a multispectral NRG composite image (2-meter resolution), and a multispectral RGB composite image (2-meter resolution) (the images are the result of stretching by 0.1% according to the default ENVI settings).

[0227] 2. Select the multispectral blue band (B), green band (G), red band (R), and near-infrared band (N) to construct the intermediate band I. The calculation formula is: b = fix(0.25*b1 + 0.25*b2 + 0.25*b3 + 0.25*b4), where b1, b2, b3, and b4 are the B, G, R, and N band images, respectively. Figure 5 This is an image of the intermediate band I (2-meter resolution).

[0228] 3. Statistically calculate the mean and standard deviation of image bands for panchromatic (P), intermediate (I), and multispectral (M) images. See Table 1.

[0229] Table 1. Statistical table of mean and standard deviation of each image band.

[0230]

[0231]

[0232] 4. Constructing image simulation operators

[0233] Taking the injection of multispectral band image into panchromatic band image simulation mode as an example, an image simulation operator M is constructed. P and I P Specifically, the image simulation operator M P and I P The calculation formulas are shown in the table below, where b1, b2, and b3 represent the panchromatic band image P, the intermediate band I, and the multispectral band image M, respectively. i Image simulation operator M P and I P The calculation formula is shown in Table 2.

[0234] Table 2 Image Simulation Operator M P and I P Calculation formula

[0235] NP long((b3-546.276444)*((86.753642 / 78.163084)*(b1-359.989578)+450.413964) / (b2)) RP long((b3-320.170811)*((86.753642 / 78.163084)*(b1-359.989578)+450.413964) / (b2)) GP long((b3-536.913406)*((86.753642 / 78.163084)*(b1-359.989578)+450.413964) / (b2)) BP long((b3-399.795694)*((86.753642 / 78.163084)*(b1-359.989578)+450.413964) / (b2)) IP long((b2-450.413964)*((86.753642 / 78.163084)*(b1-359.989578)+450.413964) / (b2))

[0236] 5. Combine the panchromatic image P and the multispectral image M i Fusion operator M P and I P The high spatial resolution images are resampled and synthesized into a single image file. Then, the mean μ, standard deviation σ, and other image characteristic statistical parameters of each band are calculated. The basic characteristic statistical parameters of each band image are shown in Tables 3 and 4.

[0237] Table 3. Statistical table of mean and standard deviation of each image band.

[0238]

[0239]

[0240] Table 4. Statistical table of correlation coefficients for different image bands

[0241]

[0242] 6. Multispectral image simulation calculation

[0243] This embodiment uses characteristic simulation parameters. Taking this as an example, we perform nonlinear simulation calculations for four bands of multispectral data. The specific simulation algorithm is as follows:

[0244] (1) The simulation calculation expression corresponding to the near-infrared band is fix(228.705554 / 127.984142*((b1-359.989578)+1.961354*(0.334059*(b3-(-7.578292))-0.649821*0.884653*(b2-(-1.510122))))+546.276444), where b1 is the panchromatic band image P, b2 is the simulation factor IP, and b3 is the simulation factor NP. The simulated near-infrared band (Nf) image is obtained by calculation.

[0245] (2) The simulation calculation expression corresponding to the red band is fix(95.810041 / 93.147246*((b1-359.989578)+0.972298*(0.817592*(b3-(0.959661))-0.873531*0.884653*(b2-(-1.510122))))+320.170811), where b1 is the panchromatic band image P, b2 is the simulation factor IP, and b3 is the simulation factor RP. The simulated red band (Rf) image is obtained by calculation.

[0246] (3) The simulation calculation expression corresponding to the green band is fix(91.546259 / 96.380218*((b1-359.989578)+1.072913*(0.859124*(b3-(0.610569))-0.842034*0.884653*(b2-(-1.510122))))+536.913406), where b1 is the panchromatic band image P, b2 is the simulation factor IP, and b3 is the simulation factor GP. The simulated green band (Gf) image is obtained by calculation.

[0247] (4) The simulation calculation expression for the blue band is fix(53.523395 / 104.416769*((b1-359.989578)+1.205097*(1.473825*(b3-(0.527238))-0.783518*0.884653*(b2-(-1.510122))))+399.795694), where b1 is the panchromatic band image P, b2 is the simulation factor IP, and b3 is the simulation factor BP. The simulated blue band (Bf) image is obtained by calculation.

[0248] The simulated near-infrared (Nf), red (Rf), and green (Gf) bands are combined according to the red, green, and blue channels to create a standard false-color image, as shown below. Figure 6 The simulated red band (Rf), green band (Gf), and blue band (Bf) are combined according to the red, green, and blue channels to form a true-color image, as shown below. Figure 7 .

[0249] This embodiment performs band data statistical analysis on three types of remote sensing images: the original WV-02 multispectral image, the simulated image obtained by the method described in this embodiment, and the image fused using the Gram-Schmidit method. The comparison of their image band statistical characteristic parameters is shown in Table 5. As can be seen from the data in the table, compared to the original multispectral image, the spatial resolution of the simulated image obtained using the method of this invention is improved from 2 meters to 0.5 meters, significantly enhancing the spatial accuracy of the image. The simulated image simultaneously possesses both the original multispectral band image and the panchromatic band image. Compared to the GS fused image, the information entropy of this method varies, but the correlation between bands is lower, and the data structure is significantly optimized, facilitating various image applications such as ground feature identification, interpretation, and analysis.

[0250] Table 5 Comparison of band statistical characteristic parameters of original multispectral images, GS fused images, and images simulated by this method

[0251]

[0252]

[0253] This embodiment describes a nonlinear simulation method for remote sensing images, primarily targeting images with panchromatic bands and multispectral bands such as near-infrared, red, green, and blue bands. First, based on the assumption of image scale invariance, this method utilizes the spatial and spectral detail differences between images at different spatial scales to construct a nonlinear simulation method for high-resolution multispectral images. Second, addressing the difficulty of predicting fusion results in traditional nonlinear image simulation methods, a pair of image simulation operators is constructed using histogram matching, achieving predictability of the image simulation results and a high degree of inheritance of the original image's fine spectral and spatial information. Finally, an optimized method for selecting image simulation feature parameters is provided.

[0254] This method improves the spectral preservation capability of nonlinear image simulation while maintaining the spatial details of panchromatic images. The fused image is vivid in color, rich in information, stable in spectral information, and easy to visually inspect and automatically classify. Especially in the context of the rapid development of high-resolution satellite remote sensing, it plays a significant role in promoting the application of domestically produced high-resolution imagery in various industries both domestically and internationally.

[0255] The spatial resolution of the simulated multispectral image was greatly improved, and the spatial information was greatly enriched. The geometric texture, spatial details, edge clarity and layering of ground features in the image were comprehensively improved. At the same time, the stability of the spectral characteristics and color display of ground features in the original multispectral image was maintained, which greatly improved the richness of remote sensing image information and the overall image quality.

[0256] Because multispectral remote sensing images possess rich spectral information and abundant color information for features such as water bodies, vegetation, buildings, exposed rocks, and exposed soil, with significant differences in spectral information and color display among different features, multispectral images often suffer from low spatial resolution and lack texture and detail information. This leads to inaccurate identification of feature types and locations during remote sensing analysis, affecting the application effectiveness of multispectral images. The method described in this embodiment is based on the principle of scale invariance in remote sensing images. It utilizes histogram feature matching to construct image simulation operators and employs nonlinear simulation methods to reconstruct multispectral images. This approach not only enhances the spatial information, clarity, and layering of features in multispectral images but also highly preserves the true spectral characteristics and color display of multispectral images. It does not alter the spectral characteristics and display effects of typical features such as water bodies, vegetation, soil, and buildings, effectively improving the overall quality of multispectral images and enhancing the potential for remote sensing image mapping and classification applications.

[0257] Example 2

[0258] This embodiment discloses a remote sensing image nonlinear simulation system, including a memory and a processor. The memory includes a remote sensing image nonlinear simulation method program. When the remote sensing image nonlinear simulation method program is executed by the processor, it implements the steps of a remote sensing image nonlinear simulation method as described in Embodiment 1.

[0259] Example 3

[0260] This embodiment discloses a computer-readable storage medium, which includes a remote sensing image nonlinear simulation method program. When the remote sensing image nonlinear simulation method program is executed by a processor, it implements the steps of the remote sensing image nonlinear simulation method described in Embodiment 1.

[0261] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0262] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0263] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0264] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0265] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

Claims

1. A method for non-linear simulation of remote sensing imagery, characterized in that, The method includes: S1: input panchromatic image and multispectral image, obtain panchromatic band image and multispectral band image ; in, , for , ... Any band in a multispectral image; S2: Spatial registration is performed between the panchromatic image and the multispectral image to ensure that the geometric spatial position of the same ground feature is consistent in both the panchromatic and multispectral images, and the multispectral image is resampled according to the high spatial resolution image. S3: Calculate the mean, root mean square, and covariance matrices of panchromatic and multispectral images; S4: Constructing intermediate bands using low spatial resolution multispectral images ; S5: Constructing image simulation operators using histogram feature matching of remote sensing images; S6: Select an image simulation method and determine the feature extraction coefficients. and feature fusion coefficient High-resolution multispectral image simulation calculations are performed based on remote sensing images, image simulation methods, and characteristic parameters. S7: Synthesize and store high-resolution multispectral image simulation results; The image simulation operator includes an image simulation operator for injecting multispectral band images into panchromatic band images. and image simulation operator And image simulation operators for injecting panchromatic images into multispectral images. and image simulation operator ; The image simulation operator Image simulation operator Image simulation operator Image simulation operator They are respectively: ； ； ； ； in, Multispectral band images The mean, panchromatic image The mean, Intermediate band The mean, Multispectral band images The mean squared error, Intermediate band The mean squared error, panchromatic image The mean squared error; The feature extraction coefficient Including but not limited to the following options: (1) When injecting multispectral band images into panchromatic band image mode: ； (2) When injecting panchromatic band images into multispectral band image modes: ； in, panchromatic image With image simulation operator The correlation coefficient, Intermediate band With image simulation operator The correlation coefficient, panchromatic image With image simulation operator The correlation coefficient, Intermediate band With image simulation operator The correlation coefficient, Multispectral band images With image simulation operator The correlation coefficient, Image simulation operator With image simulation operator The correlation coefficient, Multispectral band images With image simulation operator The correlation coefficient, Image simulation operator With image simulation operator The correlation coefficient; The feature fusion coefficient Specifically: (1) When injecting multispectral band image into panchromatic band image: ； (2) When injecting panchromatic band image into multispectral band image: ； in, panchromatic image With image simulation operator The correlation coefficient, Multispectral band images With panchromatic image The correlation coefficient; The formula for simulating high-resolution multispectral images is as follows: (1) When injecting multispectral band image into panchromatic band image: ； (2) When injecting panchromatic band image into multispectral band image: ； in, Multispectral band images The mean squared error, panchromatic image The mean, Image simulation operator The mean, Image simulation operator The mean, Image simulation operator The mean squared error, Image simulation operator The mean squared error, The mean square error of the high-resolution multispectral image; The mean square error of the high-resolution multispectral image The calculation formula is: (1) When injecting multispectral band image into panchromatic band image: ； (2) When injecting panchromatic band image into multispectral band image: 。 2. The nonlinear simulation method for remote sensing images according to claim 1, characterized in that, The intermediate band for: ； In the formula, , .

3. A nonlinear simulation system for remote sensing images, characterized in that, The system includes a memory and a processor. The memory includes a remote sensing image nonlinear simulation method program, which, when executed by the processor, implements the steps of a remote sensing image nonlinear simulation method as described in any one of claims 1 to 2.

4. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a remote sensing image nonlinear simulation method program, which, when executed by a processor, implements the steps of a remote sensing image nonlinear simulation method as described in any one of claims 1 to 2.

Images