Color balancing methods, systems, and products in remote sensing image mosaicking based on regional network adjustment.

Abstract

This invention discloses a color balancing method, system, and product for remote sensing image stitching based on regional network adjustment. First, the survey area is determined, and a radiometric model is established for the remote sensing images to be corrected using a second-order polynomial. Then, based on the radiometric consistency constraint of the overlapping areas of the remote sensing images, an error matrix is ​​constructed. Next, considering the overlapping areas of all remote sensing images in the survey area, a regional network adjustment model matrix is ​​established. Taking into account three quality factors—pixel count, standard deviation, and tone matching degree—the quality weight matrix is ​​solved. The least squares method is used to minimize the sum of squares of the weighted residuals to solve for the parameter vector. Finally, substituting the parameter vector, each remote sensing image to be corrected in the survey area is radiometrically corrected image by image to complete the color balancing. This invention can greatly eliminate the cumulative error of multiple remote sensing images during the color balancing process, resulting in global consistency in the color balancing of multiple remote sensing images. It is a superior color balancing method for remote sensing image stitching compared to existing methods.

Description

Technical Field

[0001] This invention belongs to the field of remote sensing image processing technology, and relates to a color balancing method, system and product in remote sensing image mosaicking, specifically a color balancing method, system and product in remote sensing image mosaicking based on regional network adjustment and polynomial radiation model. Background Technology

[0002] With the rapid development of remote sensing technology, various types of sensors have acquired a large number of high-resolution remote sensing images, which have been widely used in various fields. Meanwhile, to meet the application needs of more scenarios and create continuous remote sensing image products covering larger areas, it is usually necessary to stitch together multiple adjacent images. In remote sensing image stitching, it can be observed that, influenced by factors such as changes in lighting conditions during image imaging, differences in sensor spectral response, different shooting angles, changes in atmospheric composition, and the spatial heterogeneity of surface reflectivity, there are often certain degrees of radiometric differences between adjacent images, affecting the visual continuity and accuracy of the stitched images.

[0003] Traditional color-matching methods in remote sensing image mosaicking often employ linear adjustment models, such as mean matching, histogram matching, and regression transform, to reduce grayscale differences between images by adjusting brightness and contrast. However, these methods are poorly adaptable to complex factors such as changes in illumination, differences in surface reflectivity, and differences in sensor response, leading to color differences or halos at the mosaicking boundaries. To overcome the limitations of linear models, some studies have introduced nonlinear radiometric correction models (such as polynomial fitting and neural network models), which can more accurately describe the radiometric relationships between different images. Additionally, color space-based methods, such as HSV and Lab, can better preserve structural information, but consistency remains insufficient in complex terrain or large-scale images. There is also the adaptive block Wallis transform method, which divides the image into blocks, performs a Wallis transform on each block separately, and then uses bilinear interpolation and other methods to smooth the transition and avoid "block artifacts." Each of these methods has its own advantages and disadvantages. Linear methods are more suitable for situations with small radiation differences and stable spectral responses of ground objects, and can better preserve image details, but they are sensitive to parameter selection. Nonlinear methods can be used for situations with large radiation differences and obvious illumination distortion. They are simple and easy to implement, but may result in local detail distortion or fail to completely eliminate contrast differences at the stitching points.

[0004] Existing technologies still tend to suffer from local overcorrection or overall color inconsistency in large-scale, multi-source image stitching, lacking global constraints on the entire stitching area. Overall, existing color balancing techniques still have shortcomings in handling complex terrain scenes and maintaining global color consistency. Summary of the Invention

[0005] To address the technical problem of insufficient global color consistency caused by the cumulative error in the color balancing process of multiple remote sensing images, this invention provides a color balancing method, system, and product for remote sensing image stitching based on regional network adjustment and a polynomial radiation model, which enables global consistency in the color balancing of multiple remote sensing images.

[0006] The technical solution adopted by the method of the present invention is: a color balancing method for remote sensing image mosaicking based on regional network adjustment, comprising the following steps:

[0007] Step 1: Determine the survey area, establish a radiometric model for the remote sensing images to be corrected in the survey area, and select a reference image;

[0008] Step 2: Based on the radiometric consistency constraints of the overlapping areas of remote sensing images, construct the error equations between each pair of images in the overlapping areas and determine the design matrix. Parameter vector constant vector Error vector Construct the error matrix - = ;

[0009] Step 3: Merge the error matrices of all overlapping areas in the survey area to establish the regional network adjustment model matrix. Where A is the design matrix containing the relationship between the parameters of the second-order polynomial radiometric model and the residuals of all overlapping images to be corrected. is a parameter vector containing the parameters of the second-order polynomial radiometric model for all overlapping images to be corrected, and L is a constant vector containing the radiometric consistency target values ​​of the second-order polynomial radiometric model for all overlapping images to be corrected. It is the error vector containing the residuals of the second-order polynomial radiation model of the images to be corrected in all overlapping areas;

[0010] Step 4: Considering the three quality factors of pixel count, standard deviation, and tone matching, solve for the quality weight matrix. ;

[0011] Step 5: Minimize the sum of squares of the weighted residuals using the least squares method to solve for the parameter vector of the regional network adjustment model. ;

[0012] Step 6: For each remote sensing image to be corrected in the survey area, convert the parameter vector... The corresponding parameters are then used to correct the image radiation values ​​by substituting them into the radiation model described in step 1, thus completing the color uniformity.

[0013] Preferably, in step 1, the radiation model ;in, , , For parameters of a second-order polynomial; The original radiometric values ​​of the image; This is the radiation value after radiometric correction of the image.

[0014] Preferably, in step 2, the reference image is denoted as image 0. Based on the condition that the radiometric values ​​of the remote sensing images should meet the consistency constraint in the overlapping area after radiometric correction, the error equation is divided into the following two cases:

[0015] (1) Image to be corrected If there is an overlap with reference image 0, then the error equation is:

[0016] ;

[0017] in: , , For images Parameters of the second-order polynomial radiation model; and Corresponding to the images to be corrected The original radiometric values ​​of each pixel in the overlapping area of ​​the reference image 0; For residuals;

[0018] (2) Image to be corrected With the image to be corrected If there is an overlapping region, then the error equation is:

[0019] ;

[0020] in: , , For images Parameters of the second-order polynomial radiation model; , , For images Parameters of the second-order polynomial radiation model; and It corresponds to the image to be corrected. and The original radiance values ​​of each pixel in the overlapping region; This is the residual.

[0021] Preferably, in step 2, the reference image 0 and the image to be corrected are... Images to be corrected Among them, the image to be corrected Image to be corrected has 0 overlap with reference image. Compared with reference image 0, image to be corrected overlapping;

[0022] The design matrix is ​​then:

[0023]

[0024] in, Corresponding image In images The original radiometric values ​​of the pixels in the overlapping area with image 0; Corresponding image In images The original radiometric values ​​of the pixels in the overlapping area with image 0; Corresponding image In images With images The original radiance values ​​of the overlapping pixels; Corresponding image In images With images The original radiance values ​​of the overlapping pixels;

[0025] Parameter vector in, , , For images Parameters of the second-order polynomial radiation model; , , For images Parameters of the second-order polynomial radiation model;

[0026] constant vector ;in, Corresponding image 0 in image The original radiometric values ​​of the pixels in the overlapping area with image 0; Corresponding image 0 in image The original radiometric values ​​of the pixels in the overlapping area with image 0;

[0027] Error vector ;in, , , This is the residual.

[0028] Preferably, in step 3, if the test area is within The images total Given an overlapping region, with image 0 as the reference image, the regional network adjustment model matrix is ​​as follows: ;in: Size is ; Size is ; Size is ; Size is .

[0029] Preferably, step 4, solving the quality weight matrix W, specifically includes the following sub-steps:

[0030] Step 4.1: Count the number of pixels in each overlapping area of ​​the survey area, calculate the total number of pixels in the overlapping areas of the survey area, calculate the standard deviation of each overlapping area, and calculate the tone matching degree of each overlapping area;

[0031] Step 4.2: Calculate the image number in the survey area. Quality weight of each overlapping region ;in, For the first Number of pixels in each overlapping region; This represents the total number of pixels within the overlapping area. For the first The standard deviation of the overlapping regions; For images and In the The tonal matching degree of the overlapping areas, in which the two images and In the overlapping region The pixel values ​​are treated as two random variables. and , , They are and The marginal probability density; yes and The joint probability density;

[0032] Step 4.3: Solve for the obtained mass weight matrix ;in, For the first Quality weights for each overlapping region ; This represents the total number of overlapping regions.

[0033] Preferably, in step 5, the parameter vector of the regional network adjustment model... ;in: A is the quality weight matrix; A is the design matrix of the regional network adjustment model; This is the constant vector of the regional network adjustment model.

[0034] Preferably, in step 6, within the test area The images total There are several overlapping areas, with image 0 serving as the reference image and the remaining images as images to be corrected; images 1 to 1... Substitute the corresponding second-order polynomial radiation model parameters into the second-order polynomial radiation model. Then, it is corrected to achieve color uniformity.

[0035] The technical solution adopted by the system of this invention is: a color balancing system for remote sensing image mosaicking based on regional network adjustment, comprising:

[0036] One or more processors;

[0037] A storage device for storing one or more programs, which, when executed by one or more processors, enable the one or more processors to implement the color balancing method in remote sensing image mosaicking based on regional network adjustment.

[0038] The technical solution adopted by the product of the present invention is: a color uniform product for remote sensing image mosaic based on regional network adjustment, including computer program instructions, which, when the computer program instructions are run on a computer, cause the computer to execute the color uniform method for remote sensing image mosaic based on regional network adjustment.

[0039] Compared with the prior art, the beneficial effects of the present invention include:

[0040] (1) The present invention uses the second-order polynomial radiometric model parameters of each remote sensing image as unknown parameters and incorporates them into the global optimization framework of regional network adjustment to solve them uniformly. This results in the technical effect of radiometric correction of multiple images having global consistency and effectively eliminating the cumulative color difference caused by image-by-image transmission in traditional methods.

[0041] (2) The present invention uses the overlapping area as the observation basis to construct the radiation error equation, and combines the image quality (three quality factors such as the number of pixels, standard deviation and tone matching degree) to weight the observed values, thereby obtaining a technical effect that is highly robust to low-quality or unreliable overlapping areas and more reliable and stable color uniform results. Attached Figure Description

[0042] The technical solutions of the present invention will be further illustrated below using embodiments and specific implementation methods. In addition, some accompanying drawings are used in the description of the technical solutions. Those skilled in the art can obtain other drawings and the intent of the present invention from these drawings without any creative effort.

[0043] Figure 1 This is a flowchart of a method according to an embodiment of the present invention;

[0044] Figure 2 This is a schematic diagram of the experimental results of an embodiment of the present invention. Detailed Implementation

[0045] To facilitate understanding and implementation of the present invention by those skilled in the art, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0046] Please see Figure 1 This embodiment provides a color balancing method for remote sensing image mosaicking based on regional network adjustment, which includes the following steps:

[0047] Step 1: Determine the survey area. For the remote sensing images to be corrected in the survey area, establish a radiation model using a second-order polynomial and select a reference image.

[0048] In one implementation, a survey area is defined, which includes multiple radiometric remote sensing images to be corrected and remote sensing images that can be used as reference images, with the images overlapping each other.

[0049] In one implementation, the reference image is selected using a statistical feature-based selection method. Here, images with mean and variance close to the median of all images are used as reference images to avoid extremely bright or dark images and to ensure stability.

[0050] In one implementation, the second-order polynomial used to establish the radiation model is as follows:

[0051] ;

[0052] in, , , For parameters of a second-order polynomial; The original radiometric values ​​of the image to be corrected; This represents the radiation value of the image to be corrected after radiation correction.

[0053] After obtaining the second-order polynomial parameters of a remote sensing image, radiometric correction is performed on all radiometric values ​​of the image to be corrected using this formula.

[0054] Step 2: Based on the radiometric consistency constraints of the overlapping areas of remote sensing images, construct the error equations between each pair of images in the overlapping areas and determine the design matrix. Parameter vector constant vector Error vector Construct the error matrix - = ;

[0055] After radiometric correction, the radiometric values ​​of two adjacent remote sensing images should meet consistency constraints in the overlapping area. The reference image does not require correction by default, and the reference image and its radiometrically corrected adjacent image should meet consistency constraints in the overlapping area. Based on this condition, an error equation and error matrix are established.

[0056] In one implementation, the reference image is designated as image 0. Based on the condition that the radiometric values ​​of the remote sensing images, after radiometric correction, should meet the consistency constraint in the overlapping area, the error equation is divided into the following two cases:

[0057] (1) Image to be corrected If there is an overlapping area with reference image 0, then the image to be corrected is located in the overlapping area. The radiation values ​​at the corresponding points are corrected for second-order polynomial radiation. Equal to the radiation value at the corresponding point of the reference image 0 ,Right now:

[0058] ;

[0059] ;

[0060] in: and Corresponding to the images to be corrected The original radiometric values ​​of each pixel in the overlapping area of ​​the reference image 0; , , For images Parameters of the second-order polynomial radiation model; Corresponding image to be corrected The radiation value of the pixel in the overlapping region after second-order polynomial radiometric correction; , , For images Parameters of the second-order polynomial radiation model; For residuals;

[0061] The corresponding error equation is:

[0062] .

[0063] (2) Image to be corrected With the image to be corrected There are overlapping areas; images to be corrected. With the image to be corrected If there is an overlapping area, then the image to be corrected is located in the overlapping area. The radiation values ​​at the corresponding points are corrected for second-order polynomial radiation. Equal to the image to be corrected The radiation values ​​at the corresponding points are corrected for second-order polynomial radiation. ,Right now:

[0064] ;

[0065] ;

[0066] ;

[0067] in: and Corresponding to the images to be corrected And the original radiance values ​​of the individual pixels in the overlapping region; and Corresponding to the images to be corrected and The radiation values ​​of the pixels in the overlapping region after second-order polynomial radiometric correction; , , For images Parameters of the second-order polynomial radiation model; , , For images Parameters of the second-order polynomial radiation model; This is the residual.

[0068] The corresponding error equation is:

[0069] ;

[0070] Assume there is a reference image 0 and an image to be corrected. Images to be corrected Among them, the image to be corrected Image to be corrected has 0 overlap with reference image. Compared with reference image 0, image to be corrected Overlap. Therefore, the following system of error equations exists:

[0071]

[0072] in, Corresponding image In images The original radiometric values ​​of the pixels in the overlapping area with image 0; Corresponding image In images The original radiometric values ​​of the pixels in the overlapping area with image 0; Corresponding image In images With images The original radiance values ​​of the overlapping pixels; Corresponding image In images With images The original radiance values ​​of the overlapping pixels; , , For images Parameters of the second-order polynomial radiation model; , , For images Parameters of the second-order polynomial radiation model; , , This is the residual.

[0073] In one implementation, the coefficients of the second-order polynomial radiation model parameters are extracted from the above error equation set to form the design matrix. Construction The design matrix is ​​as follows:

[0074] ;

[0075] in, Corresponding image In images The original radiometric values ​​of the pixels in the overlapping area with image 0; Corresponding image In images The original radiometric values ​​of the pixels in the overlapping area with image 0; Corresponding image In images With images The original radiance values ​​of the overlapping pixels; Corresponding image In images With images The original radiance values ​​of the overlapping pixels.

[0076] Based on the above set of error equations, extract the parameters of the second-order polynomial radiation model to form a parameter vector. Construct... The parameter vector is as follows:

[0077] ;

[0078] in, , , For images Parameters of the second-order polynomial radiation model; , , For images The parameters of the second-order polynomial radiation model.

[0079] Based on the above system of error equations, extract the constant terms to form a constant vector. Construct... The constant vector is as follows:

[0080] ;

[0081] in, Corresponding image 0 in image The original radiometric values ​​of the pixels in the overlapping area with image 0; Corresponding image 0 in image The original radiometric value of the pixels in the overlapping area with image 0.

[0082] The residuals are extracted from the above set of error equations to form an error vector. Construct... The error vector is as follows:

[0083] ;

[0084] in, , , This is the residual.

[0085] Based on the above error equations and design matrix Parameter vector constant vector Error vector Constructing the error matrix - = ;in, To design the matrix, For parameter vectors, A constant vector, This is the error vector.

[0086] Step 3: Merge the error matrices of all overlapping areas in the survey area to establish the regional network adjustment model matrix. Where A is the design matrix containing the relationship between the parameters of the second-order polynomial radiometric model and the residuals of all overlapping images to be corrected. is a parameter vector containing the parameters of the second-order polynomial radiometric model for all overlapping images to be corrected, and L is a constant vector containing the radiometric consistency target values ​​of the second-order polynomial radiometric model for all overlapping images to be corrected. It is the error vector containing the residuals of the second-order polynomial radiation model of the images to be corrected in all overlapping areas;

[0087] In one implementation, constructing the regional network adjustment model matrix requires considering all overlapping areas within the survey area, taking into account the error matrices of all overlapping areas within the survey area to form a large matrix system.

[0088] Assuming the survey area is within The images total There are overlapping areas; image 0 is the reference image, and the image to be corrected is... Images to be corrected There are overlapping areas with reference image 0. Reference image 0 does not require radiometric correction. From image 1 to image 2... The process iterates through adjacent images that overlap with the error equations, constructing a system of error equations. This system is shown below:

[0089]

[0090] in, and Corresponding images ,image In images With images The original radiance values ​​of the overlapping pixels; , , For images Parameters of the second-order polynomial radiation model; This is the residual.

[0091] Referring to step 2, the corresponding regional network adjustment model matrix can be constructed as follows:

[0092] ;

[0093] in: Size is , Size is , Size is , Size is .

[0094] ;

[0095] ;

[0096] ;

[0097] .

[0098] For the image to be corrected and the reference image, the value corresponding to L is the radiometric value of the reference image in the overlapping area (that is, the radiometric value of the image to be corrected after correction is expected to be equal to the radiometric value of the reference image, that is, after adjustment, the radiometric values ​​in the overlapping area are consistent, and the reference image is the standard value); for the two images to be corrected, the value corresponding to L is 0 (that is, after subtracting the radiometric values ​​of the two images after correction, the expected result is 0, that is, after adjustment, the radiometric values ​​in the overlapping area are consistent).

[0099] Step 4: Considering the three quality factors of pixel count, standard deviation, and tone matching, solve for the quality weight matrix. ;

[0100] To reflect the impact of each overlapping region on the overall result during radiometric correction, weight allocation is determined based on the quality characteristics of the overlapping regions. These quality characteristics comprehensively consider three factors: the number of pixels in the overlapping region, the standard deviation, and the tone matching degree. The number of pixels characterizes the range of influence of the overlapping region, the standard deviation reflects the stability of the radiometric values ​​within that region, and the tone matching degree measures the color consistency between adjacent images. By comprehensively evaluating these multiple quality factors, a reasonable allocation of weights for overlapping regions can be achieved, thereby improving the reliability and overall consistency of the radiometric correction results.

[0101] In one implementation, solving the quality weight matrix W specifically includes the following sub-steps:

[0102] Step 4.1: Obtain basic parameters of the overlapping area. Count the number of pixels in each overlapping area of ​​the survey area. Calculate the total number of pixels in the overlapping area of ​​the survey area. Calculate the standard deviation of each overlapping region. Calculate the hue matching degree of each overlapping area. .

[0103] (1) Solve for the total number of pixels in the overlapping area The formula is as follows:

[0104] ;

[0105] in, This represents the number of overlapping areas within the survey area; For the first Number of pixels in each overlapping region; This represents the total number of pixels in all overlapping areas within the survey area.

[0106] When considering weights based on the number of pixels, the region with the larger weight contains more pixels, and thus represents the larger range of influence of the overlapping region.

[0107] (2) Calculate the standard deviation of each overlapping region. as follows:

[0108] Hypothetical Image With images Overlapping in the overlapping area Then there will be images. mean Standard deviation as follows:

[0109] ;

[0110] ;

[0111] in, For the first Number of pixels in each overlapping region For images In the overlapping region No. The radiation value of each pixel. For images In the overlapping region The average radiation value, For images In the overlapping region The standard deviation of the radiation value. Similarly, there are images. The mean and standard deviation are as follows:

[0112] ;

[0113] ;

[0114] in, For the first Number of pixels in each overlapping region For images In the overlapping region No. The radiation value of each pixel. For images In the overlapping region The average radiation value, For images In the overlapping region The standard deviation of the radiometric values. The standard deviation of the merged two images (overlap area). The standard deviation is as follows:

[0115] ;

[0116] in, For images In the overlapping region The standard deviation of the radiation value, For images In the overlapping region The standard deviation of the radiation value, Overlapping area The standard deviation.

[0117] When considering weights based on standard deviation, the weights are inversely proportional to the standard deviation. The smaller the standard deviation, the larger the weight, which indicates a higher stability of the radiation values ​​in the overlapping region.

[0118] (3) Solve for hue matching degree as follows:

[0119] Hypothetical Image With images Overlapping in the overlapping area Place the two images in the overlapping area The pixel values ​​are treated as two random variables. and Its hue matching degree is defined as:

[0120] ;

[0121] in: , They are and Marginal probability density (estimated via histogram); yes and The joint probability density (estimated by joint histogram).

[0122] When considering weights based on tone matching degree, the greater the mutual information, the more similar the overlapping areas of the two images are; the higher the tone matching degree, the higher the color consistency between adjacent images.

[0123] Step 4.2: Calculate the image number in the survey area. Quality weight of each overlapping region ;in, For the first Number of pixels in each overlapping region; This represents the total number of pixels within the overlapping area. For the first The standard deviation of the overlapping regions; For images and In the Tone matching degree of overlapping areas;

[0124] Step 4.3: Solve for the obtained mass weight matrix ;in, For the first Quality weights for each overlapping region ; This represents the total number of overlapping regions. The mass weight matrix obtained by solving is a diagonal matrix.

[0125] Step 5: Minimize the sum of squares of the weighted residuals using the least squares method to solve for the parameter vector of the regional network adjustment model. ;

[0126] In one implementation, the parameter vector is solved using the least squares indirect adjustment principle. The goal of the least squares method is to minimize the sum of squares of the weighted residuals. Then, through the normal equation, we obtain:

[0127] ;

[0128] in: A is the quality weight matrix; A is the design matrix of the regional network adjustment model; This represents the constant vector of the regional network adjustment model; This is the parameter vector of the regional network adjustment model.

[0129] Step 6: For each remote sensing image to be corrected in the survey area, convert the parameter vector... The corresponding parameters are then used to correct the image radiation values ​​by substituting them into the radiation model described in step 1, thus completing the color uniformity.

[0130] In one implementation, within the survey area The images total There are three overlapping areas, with image 0 serving as the reference image and the remaining images as images to be corrected. Reference image 0 does not require radiometric correction, while images 1 to... Perform frame-by-frame radiometric correction to achieve color uniformity, as follows:

[0131] Parameter vector of the regional network adjustment model for:

[0132] ;

[0133] Then the images to be corrected 1 to The corresponding second-order polynomial radiation model parameters are:

[0134] The parameters of the second-order polynomial radiation model for image 1 are: , , ;

[0135] The parameters of the second-order polynomial radiation model for image 2 are: , , ;

[0136] The parameters of the second-order polynomial radiation model for image 3 are: , , ; ......

[0137] image The parameters of the second-order polynomial radiation model are: , , ;

[0138] image The parameters of the second-order polynomial radiation model are: , , ; ......

[0139] image The parameters of the second-order polynomial radiation model are: , , ;

[0140] image The parameters of the second-order polynomial radiation model are: , , ;

[0141] Image 1 to be corrected The corresponding parameters of the second-order polynomial radiation model are substituted into the second-order polynomial radiation model and corrected as follows:

[0142] ;

[0143] The method of the present invention will be further illustrated below through specific experiments.

[0144] This experiment selected multiple radiometric remote sensing images of a certain area to be corrected, as well as reference images, with overlapping between them. First, a radiometric model was established for the images to be corrected using a second-order polynomial. Then, based on the radiometric consistency constraint of the overlapping areas of the remote sensing images, an error matrix was constructed. Next, considering the overlapping areas of all remote sensing images in the survey area, a regional network adjustment model matrix was established. Taking into account three quality factors—pixel count, standard deviation, and tone matching—the quality weight matrix was solved. The least squares method was used to minimize the sum of squares of the weighted residuals to solve for the parameter vector. Finally, substituting the parameter vector, radiometric correction was performed on each image to be corrected in the survey area, completing the color uniformity. See below. Figure 2 This is a schematic diagram illustrating the effect obtained in this experiment. Figure 2 As can be seen from this, the present invention can greatly eliminate the cumulative error of multiple remote sensing images in the color balancing process, so that the color balancing of multiple remote sensing images has global consistency.

[0145] It should be understood that the embodiments described above are only some, not all, of the embodiments of the present invention. Furthermore, the technical features of the various embodiments or individual embodiments provided by the present invention can be arbitrarily combined to form feasible technical solutions. Such combinations are not constrained by the order of steps and / or structural composition patterns, but must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0146] It should be understood that the above description of the preferred embodiments is quite detailed, but it should not be considered as a limitation on the scope of protection of this invention. Those skilled in the art, under the guidance of this invention, can make substitutions or modifications without departing from the scope of protection of the claims of this invention, and all such substitutions or modifications fall within the scope of protection of this invention. The scope of protection of this invention should be determined by the appended claims.

Claims

1. A method for color uniformity in remote sensing image mosaic based on network adjustment, characterized in that, The method comprises the following steps: Step 1: determining a measurement area, establishing a radiation model for remote sensing images to be corrected in the measurement area, and selecting a reference image; Step 2: According to the consistency constraint of the radiation of the overlapping area of remote sensing images, the error equation between two images in the overlapping area is constructed, and the design matrix is determined , parameter vector , constant vector , error vector , error matrix is constructed - = ; Wherein, the reference image is denoted as image 0, according to the consistency constraint condition that the remote sensing image radiation value should satisfy after radiation correction in the overlapping area, the error equation is divided into the following two cases: (1) image to be corrected With reference to the image 0 overlap area, the error equation is: ； wherein: , , are second order polynomial radiometric model parameters of the image ; and respectively correspond to the original radiometric values of the image to be corrected and of the reference image 0 at the respective pixels of the overlap area; is the residual; (2) image to be corrected with the image to be corrected If there is an overlapping region, the error equation is: ； wherein: , , are second order polynomial radiometric model parameters of the image ; , , are second order polynomial radiometric model parameters of the image ; and are the respective original radiometric values of the image and in the overlapping area pixels; is the residual; for the reference image 0, the image to be corrected , the image to be corrected , wherein the image to be corrected overlaps the reference image 0, the image to be corrected overlaps the reference image 0, the image to be corrected ; The design matrix is: wherein, corresponding image in image the raw radiance values of the pixels in the overlap region with image 0; corresponding image in image the raw radiance values of the pixels in the overlap region with image 0; corresponding image in image the raw radiance values of the pixels in the overlap region with image ; corresponding image in image the raw radiance values of the pixels in the overlap region with image ; Parameter vector wherein, , , are second order polynomial radiance model parameters of the image ; , , are second order polynomial radiance model parameters of the image ; Constant vector ; wherein, corresponding to image 0 in image the raw radiance value of the pixel in the overlap region with image 0; corresponding to image 0 in image the raw radiance value of the pixel in the overlap region with image 0; error vector ; wherein, , , is a residual; Step 3: merging all the error matrices of the overlapping areas in the survey area to establish a regional network adjustment model matrix ; wherein A is a design matrix containing the relationship between the second-order polynomial radiation model parameters of the images to be corrected in all the overlapping areas and the residuals, is a parameter vector containing the second-order polynomial radiation model parameters of the images to be corrected in all the overlapping areas, L is a constant vector containing the radiation consistency target values of the second-order polynomial radiation model of the images to be corrected in all the overlapping areas, is an error vector containing the residuals of the second-order polynomial radiation model of the images to be corrected in all the overlapping areas; Step 4: Considering the three quality factors of the number of pixels, standard deviation and hue matching degree, the quality weight matrix is solved ; Step 5: Solve the parameter vector of the network adjustment model by minimizing the sum of squares of weighted residuals using least squares ; Step 6: for each remote sensing image to be corrected in the survey area, the parameter vector corresponding to the parameter is brought into the radiation model described in Step 1 to correct the image radiation value, and the color is completed.

2. The method according to claim 1, wherein the method is characterized in that: In step 1, the radiation model ; wherein, , , is a second order polynomial parameter; is an original radiation value of the image; is a radiation value of the image after radiation correction.

3. The method according to claim 1, wherein the method is characterized by, In step 3, if the measurement area is The image common The image 0 is the reference image, then the area network adjustment model matrix ; wherein: The size is ; The size is ; The size is ; The size is .

4. The method according to claim 1, wherein the method is characterized by, In step 4, the solving of the quality weight matrix W specifically comprises the following sub-steps: Step 4.1: counting the number of pixels in each overlapping area in the measurement area, calculating the total number of pixels in the overlapping area in the measurement area, calculating the standard deviation of each overlapping area, and calculating the color matching degree of each overlapping area; Step 4.2: Calculate the quality weight of the image overlap area in the measurement area Step 4.3: Calculate the quality weight of the image overlap area in the measurement area ; wherein, is the number of pixels in the overlap area; is the total number of pixels in the overlap area; is the standard deviation of the overlap area; is the standard deviation of the image and is the hue matching degree of the image and in the overlap area, wherein the pixel values of the two images and in the overlap area are considered as two random variables and , , , are the marginal probability densities of and , is the joint probability density of and . Step 4.3: Solving the resulting mass weight matrix ; where, is the mass weight of the jth overlapping region, ; is the total number of overlapping regions.​ 5. The method of claim 1, wherein the method further comprises: In step 5, the parameter vector of the network adjustment model ; wherein: is a quality weight matrix; A is a design matrix of the network adjustment model; is a constant vector of the network adjustment model.

6. The method according to any one of claims 1-5, wherein, In step 6, the measurement area is divided into The image common to the range of images The image 0 is the reference image, and the remaining images are the images to be corrected; the images to be corrected 1 to The corresponding second-order polynomial radiation model parameters are respectively brought into the second-order polynomial radiation model , and correction is performed to complete the color uniformity.

7. A color uniformity system in a remote sensing image stitching based on a network adjustment, characterized in that, Comprise: One or more processors; A storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the color uniformization method in the regional net adjustment based remote sensing image splicing as claimed in any one of claims 1 to 6.

8. A color-uniform product in remote sensing image stitching based on network adjustment, comprising computer program instructions, characterized in that: When the computer program instructions run on the computer, the computer executes the color uniformization method in the regional net adjustment based remote sensing image splicing as claimed in any one of claims 1 to 6.

Images