Satellite image radiation quality evaluation method, device, and storage medium

Abstract

The present invention provides a method, computer device, and computer-readable storage medium for evaluating the radiation quality of satellite images. The method comprises: Step A, performing a contour extraction operation on the satellite image to be evaluated; Step B, obtaining N evaluation indicators for the satellite image to be evaluated, wherein the N evaluation indicators are divided into the following three categories: grayscale statistical feature indicators, texture statistical feature indicators, and radiation accuracy feature indicators; Step C, constructing a quality evaluation regression model and training the quality evaluation regression model using samples from a training sample set; and Step D, substituting the N evaluation indicators for the satellite image to be evaluated into the trained quality evaluation regression model to obtain a comprehensive evaluation index value for the satellite image to be evaluated. The present invention covers a wide range of data and features, enabling comprehensive evaluation of satellite images, greatly improving the objectivity and accuracy of the evaluation.

Description

Technical Field

[0001] The present invention relates to the field of satellite image quality evaluation, and in particular to a method, a computer device, and a computer-readable storage medium for evaluating satellite image radiation quality. Background Art

[0002] With so many types of remote sensing satellite imagery available, researchers face a challenge in selecting the appropriate data for specific applications. To assess the quality of remote sensing satellite data, researchers typically use subjective or objective evaluation methods, employing three to five image quality metrics, to analyze the quality of different types of remote sensing satellite imagery.

[0003] In the process of realizing the present invention, the applicant discovered that the accuracy of traditional satellite image radiation quality evaluation technology is low and the effect is not ideal. Summary of the Invention

[0004] (1) Technical issues to be solved

[0005] The present invention aims to at least partially solve at least one of the above technical problems.

[0006] (2) Technical solution

[0007] To achieve the above objectives, according to one aspect of the present invention, a method for evaluating the radiometric quality of satellite images is provided, comprising:

[0008] Step A, performing a contour extraction operation on the satellite image to be evaluated, removing the outer edge portion that has no analytical value, and retaining the inner portion that has analytical value;

[0009] Step B: obtaining N evaluation indicators of the satellite image to be evaluated, wherein the N evaluation indicators are divided into the following three categories: grayscale statistical feature indicators, texture statistical feature indicators, and radiation accuracy feature indicators;

[0010] Step C: constructing a quality evaluation regression model, and training the quality evaluation regression model using samples in the training sample set, wherein the quality evaluation regression model is a regression model based on the XGBoost algorithm;

[0011] In step D, the N evaluation indicators of the satellite image to be evaluated are substituted into the trained quality evaluation regression model to obtain the comprehensive evaluation indicator value of the satellite image to be evaluated.

[0012] In some embodiments of the present invention, after step D, the further step includes: adding the N evaluation indicators and the corresponding comprehensive evaluation indicator values ​​of the satellite image to be evaluated to one sample in the training sample set; in step C, the training sample set includes the following three types of training samples:

[0013] The first type of training samples has an index set consisting of various evaluation indices calculated based on satellite images of satellites passing through ground calibration sites, and its true value is the comprehensive evaluation index value of satellite images obtained by measuring with ground equipment.

[0014] The second type of training samples, whose index set is the various evaluation indicators of the previous satellite images of the source satellite of the satellite image to be evaluated, and its true value is the artificial comprehensive evaluation index value of the previous satellite images;

[0015] The third type of training samples, whose index set is the various evaluation indicators of the previous satellite images of the source satellite of the satellite image to be evaluated, and the true value of the sample is the automatic comprehensive evaluation index value obtained by step D of the previous satellite images;

[0016] Among them, at the initial stage of training, the sum of the number of training samples of the first category and the second category accounts for 100% of the total number of samples; in the mature stage of training, the proportion of training samples of the first category in the total number of samples is less than 30%; the number of samples of the second category accounts for less than 30% of the total number of samples; the number of samples of the third category accounts for more than 40% of the total number of samples, and the proportion of samples of the third category in the total number of samples gradually increases with the gradual accumulation of satellite images.

[0017] In some embodiments of the present invention, step C comprises:

[0018] Sub-step C1: Construct a quality evaluation regression model L = A·X, where:

[0019] L is the comprehensive evaluation index value, and its value range is 0%-100%;

[0020] X is a column vector consisting of N evaluation indicators, and A is a row vector consisting of the deviation impact factors corresponding to each evaluation indicator. During the training process, [L, X] is the input content of the quality evaluation regression model, and A is the content to be obtained through training;

[0021] Sub-step C2: Initialize A in the quality evaluation regression model;

[0022] In sub-step C3, the quality evaluation regression model is trained iteratively using the samples in the training sample set and the XGBoost algorithm until the prediction accuracy requirement is met.

[0023] In some embodiments of the present invention, in sub-step C1:

[0024] A=(h1 ... h n1 w1 ... w n2 f1 ... f n3 )

[0025] Among them: H1 is the first grayscale statistical characteristic index, h1 is its corresponding deviation influence factor; H n1 is the n1th grayscale statistical characteristic index, h n1 Its corresponding deviation influence factor; W1 is the first texture statistical feature index, w1 is its corresponding deviation influence factor; W n2 is the n2th texture statistical feature index, w n2 Its corresponding deviation impact factor; F1 is the first radiation accuracy characteristic index, f1 is its corresponding deviation impact factor; F n3 is the n3th radiation accuracy characteristic index, f n3 is the corresponding deviation influence factor; n1, n2, and n3 are the number of grayscale statistical feature indicators, texture statistical feature indicators, and radiation accuracy feature indicators, respectively. n1+n2+n3=N, where N is the total number of evaluation indicators.

[0026] In some embodiments of the present invention, in sub-step C2, for each evaluation indicator, its initial corresponding deviation impact factor is: if not set by the user, it takes the default value; if set by the user, it takes the user-set value.

[0027] In some embodiments of the present invention, in sub-step C3, the learning rate of the XGBoost algorithm is 0.1, the total number of trees is 500, and other parameters are set to default values. The maximum depth of the tree, the minimum leaf split value, the sample sampling, and the attribute column sampling parameters are optimized using the grid search GridSearchCV function in the Sciki-learn package to optimize the XGBoost parameters.

[0028] In some embodiments of the present invention, before step B, the following step is further included: step A, performing a contour extraction operation on the satellite image to be evaluated, removing the outer edge portion that has no analytical value, and retaining the inner portion that has analytical value.

[0029] In some embodiments of the present invention, grayscale statistical feature indicators include: radiation mean, minimum value, maximum value, peak value, median value, brightness range value, variance, standard deviation and signal-to-noise ratio; texture statistical feature indicators include: information entropy, angular second moment, detail energy, edge energy, local stability, information capacity, contrast and clarity; radiation accuracy feature indicators include: skewness, steepness, edge radiation distortion, gain adjustment distortion and power spectrum.

[0030] In some embodiments of the present invention, among the grayscale statistical feature indicators:

[0031] The radiation mean M is:

[0032] Minimum value M min For: Mmin =MIN(g(x,y));

[0033] Maximum value M max For: M max =MAX(g(x,y));

[0034] The peak value is the highest point on the histogram curve corresponding to the satellite image to be evaluated;

[0035] The median is the median of the frequency distribution of the histogram curve corresponding to the satellite image to be evaluated;

[0036] The brightness range value of each band is the difference between the maximum grayscale value and the minimum grayscale value in the band;

[0037] The variance is the average squared difference between the grayscale values ​​of all pixels and the mean, and its square root is the standard deviation;

[0038] The signal-to-noise ratio was calculated using the local mean and local standard deviation method;

[0039] Where S is the image area, g(x,y) is the grayscale value corresponding to the position (x,y) in the satellite image to be evaluated;

[0040] In some embodiments of the present invention, among the texture statistical feature indicators:

[0041] Information entropy is: Where i is the gray level, L is the total number of gray levels, b i The gray value of level i, p(b i ) is b i Probability of occurrence;

[0042] The angular second moment is: in, is the square of the pixel value of the gray-level co-occurrence matrix;

[0043] The detail energy is: in Regional variance, expressed as: f(x+i,y+j) is the variance of the pixel with coordinates (x+i,y+j); m f (x,y) represents the average grayscale value of the region, the average value of the sum of the variances of all (2M+1)*(2M+1) regions in the image:

[0044] The edge energy is: m, n are the pixel values ​​of the length and width of the image respectively; two normalized edge operators E1 and E2 of 45 degrees and 135 degrees are used to perform convolution calculation on the satellite image f(x,y) to eliminate the blocking effect, and the image e(x,y) is obtained after adding them together, where e(x,y)=E1(f(x,y))+E2(f(x,y)).

[0045] Local stability is: Where L is the total number of gray levels, is the gray-level co-occurrence matrix pixel value;

[0046] The information capacity is: Among them, Norm log (G1, G2) = log[Num(G1, G2)] / log[max{Num(G1, G2)}] is called the logarithmic peak normalized histogram; max{Num(G1, G2)} is the frequency peak of the two-dimensional histogram; ω is the cumulative constraint domain;

[0047] The contrast ratio is: Where |ij|=n, is the normalized gray-level co-occurrence matrix pixel value;

[0048] The clarity is: in: Where i and j cannot be equal to 0 at the same time, m and n are the number of rows and columns of the satellite image to be evaluated, and f(x, y) is the satellite image to be evaluated;

[0049] In some embodiments of the present invention, in the radiation accuracy characteristic index,

[0050] The skewness reflects the degree to which the shape of the image histogram distribution deviates from the symmetrical shape around the mean. A positive skewness indicates that the distribution of the asymmetric side tends to be more positive, and a negative skewness indicates that the distribution of the asymmetric side tends to be more negative. The skewness is: Where i is the gray level, m is the arithmetic mean, b i The gray value of level i, p(b i ) is b i The probability of occurrence, d refers to the standard deviation of the image;

[0051] The steepness is:

[0052] Edge radiation distortion: R is the mean vector of the row vector, that is: R = [mean(p(1,j)),mean(p(2,j))…mean(p(i,j))] i＝1～N,j＝1～M, where M, N are the rows and columns of the satellite image to be evaluated, Mean is the mean, Std is the standard deviation, and p(i,j) is the grayscale value of the satellite image to be evaluated in row i and column j;

[0053] The gain adjustment distortion is: L is the mean vector of the column vector, that is: L = [mean(p(i,1)),mean(p(i,2))…mean(p(i,j))] i＝1～N,j＝1～M ;

[0054] The power spectrum comprehensively reflects the intensity of each spatial frequency component of the image, specifically: PSD = log2Sum, |F(u,v)| 2 =R 2 (u,v)+I 2 (u,v); R(u,v) and I(u,v) are the real and imaginary parts of F(u,v) after Fourier transform.

[0055] To achieve the above purpose, according to a second aspect of the present invention, a computer device for evaluating satellite image radiation quality is provided, comprising:

[0056] An application plug-in layer storing computer plug-in programs;

[0057] The basic platform layer provides an interface for calling computer plug-in programs stored in the application plug-in layer;

[0058] The business subsystem is configured to execute the above method when calling the computer plug-in program stored in the application plug-in layer through the basic platform layer.

[0059] In some embodiments of the present invention, the basic platform layer automatically calls the computer plug-in program of the application plug-in layer in the form of an XML order to achieve automated radiation quality evaluation.

[0060] In some embodiments of the present invention, the basic platform layer provides one or more of the following interface groups: component communication interface, COM compatible interface, general UI interface, CPU+GPU acceleration interface.

[0061] In order to achieve the above purpose, according to a third aspect of the present invention, a computer-readable storage medium is further provided, wherein the computer-readable storage medium stores computer instructions, and when the computer instructions are executed, the above method is implemented.

[0062] (3) Beneficial effects

[0063] It can be seen from the above technical solution that the present invention has at least one of the following beneficial effects:

[0064] (1) Comprehensively utilize grayscale statistical feature indicators, texture statistical feature indicators, and radiation accuracy feature indicators to evaluate the radiation quality of satellite images. The data and features cover a wide range and can evaluate satellite images in an all-round way, greatly improving the objectivity and accuracy of the evaluation.

[0065] (2) Using a regression model based on the XGBoost algorithm and large sample training, combined with deep learning training to obtain total parameters, the trained parameters are more objective. Moreover, as the number of image dataset samples gradually increases, the frequency and quality of model training gradually increase, and the subsequent evaluation accuracy and objectivity will become higher and higher, with greater room for improvement, which is conducive to meeting the increasingly high actual needs of scientific research.

[0066] (3) At the initial stage of training, the first and second types of training samples are used for training; then the third type of training samples are gradually added. The first type of training samples, whose index set is the various evaluation indicators calculated based on the satellite images of the satellite passing through the ground calibration field, and whose sample true values ​​are the comprehensive evaluation index values ​​of the satellite images obtained by measuring using ground equipment. The second type of training samples, whose index set is the various evaluation indicators of the previous satellite images of the source satellite of the satellite image to be evaluated, and whose sample true values ​​are the manual comprehensive evaluation index values ​​of the previous satellite images. The third type of training samples, whose index set is the various evaluation indicators of the previous satellite images of the source satellite of the satellite image to be evaluated, and whose sample true values ​​are the automatic comprehensive evaluation index values ​​of the previous satellite images obtained through step D.

[0067] Through the above methods, on the one hand, the stability and reliability of the data in the initial stage of training are guaranteed, especially the introduction of satellite image-related samples based on the satellite transit ground calibration field, which can ensure that the training of the regression model based on the XGBoost algorithm is carried out in the correct training direction and avoid training going astray, thereby greatly improving the effect of the initial training and preventing the occurrence of training errors; on the other hand, in the mature stage, the third type of training samples gradually increases, while the proportion of the first and second types of training samples gradually decreases, and the total sample size continues to increase, ensuring that the evaluation effect is more and more objective and accurate.

[0068] (4) Before the formal evaluation, the satellite image to be evaluated is subjected to contour extraction operation to remove the outer edge parts that have no analytical value and retain the inner parts that have analytical value, which can reduce its grayscale and reduce the amount of calculation.

[0069] (5) Grayscale statistical characteristic indicators include: radiation mean, minimum, maximum, peak, median, brightness range, variance, standard deviation and signal-to-noise ratio; texture statistical characteristic indicators include: information entropy, angular second moment, detail energy, edge energy, local stability, information capacity, contrast and clarity; radiation accuracy characteristic indicators include: skewness, steepness, edge radiation distortion, gain adjustment distortion and power spectrum.

[0070] The above characteristic indicators cover all angles of the satellite image to be evaluated, are more comprehensive and do not miss any details, can ensure the evaluation of any type of satellite image, and ensure the stability of the evaluation. Compared with these existing patents, the radiation evaluation of this patent covers a wider range and also improves the accuracy of the evaluation results.

[0071] (6) The entire radiation evaluation process is fully automated and does not require human intervention, which greatly improves the degree of automation. It adopts a platform + plug-in deployment model, which can automatically input and output an image throughout the process, improving convenience.

[0072] (7) Although the evaluation process does not require human intervention, professionals can exert influence during the training phase, such as adding new training data, so that the trained evaluation model can more objectively reflect the true quality of the image to be evaluated, and achieve mutual complementation and mutual promotion of subjective and objective evaluation results.

[0073] In summary, the present invention aims to achieve the following objectives by providing a satellite image radiation quality evaluation method: ① constructing a mathematical model that meets the requirements of optical satellite image evaluation; ② improving the accuracy, objectivity and reliability of satellite image radiation quality evaluation; ③ reducing manual intervention and realizing the automation of the satellite image radiation quality evaluation process; ④ achieving mutual complementation and mutual promotion of subjective and objective evaluation results, and providing technical support for expanding the application of optical remote sensing satellite imagery in the remote sensing field. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 This is a flowchart of a satellite image radiation quality assessment method according to an embodiment of the present invention.

[0075] Figure 2 for Figure 1 Schematic diagram of the method for evaluating the radiometric quality of satellite images.

[0076] Figure 3 This is a structural diagram of a computer device for evaluating satellite image radiation quality according to an embodiment of the present invention. DETAILED DESCRIPTION

[0077] The present invention aims to achieve the following objectives by providing a satellite image radiation quality evaluation method: 1. construct a mathematical model that meets the requirements of optical satellite image evaluation; 2. improve the accuracy, objectivity and reliability of satellite image radiation quality evaluation; 3. reduce manual intervention and realize the automation of the satellite image radiation quality evaluation process; 4. achieve mutual complementation and mutual promotion of subjective and objective evaluation results, and provide technical support for expanding the application of optical remote sensing satellite imagery in the remote sensing field.

[0078] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings.

[0079] In an exemplary embodiment of the present invention, a method for evaluating the radiometric quality of satellite images is provided.

[0080] Figure 1 This is a flowchart of a satellite image radiation quality assessment method according to an embodiment of the present invention. Figure 2 for Figure 1 The diagram below shows the method for evaluating the radiometric quality of satellite images. Figure 1 and Figure 2 As shown, this embodiment includes:

[0081] Step A, performing a contour extraction operation on the satellite image to be evaluated, removing the outer edge portion that has no analytical value, and retaining the inner portion that has analytical value;

[0082] Step B: obtaining N evaluation indicators of the satellite image to be evaluated, wherein the N evaluation indicators include the following three types of evaluation indicators: grayscale statistical feature indicators, texture statistical feature indicators, and radiation accuracy feature indicators;

[0083] Step C: constructing a quality evaluation regression model, and training the quality evaluation regression model using samples in the training sample set, wherein the quality evaluation regression model is a regression model based on the XGBoost algorithm;

[0084] Step D: Substitute the N evaluation indicators of the satellite image to be evaluated into the trained quality evaluation regression model to obtain the comprehensive evaluation indicator value of the satellite image to be evaluated.

[0085] It can be seen that this embodiment can greatly improve the accuracy and reliability of the evaluation through the regression model based on the XGBoost algorithm and the comprehensive statistical feature indicators that do not miss any details.

[0086] The following describes each step in detail.

[0087] In step A, the satellite image to be evaluated is subjected to contour extraction, removing the outer edge portions that are not valuable for analysis and retaining the inner portions that are valuable for analysis. This processing in step A can reduce the grayscale of the satellite image to be evaluated, thereby reducing the amount of computation required.

[0088] In step B, N evaluation indicators of the satellite image to be evaluated are obtained. The N evaluation indicators include the following three types of evaluation indicators: grayscale statistical feature indicators, texture statistical feature indicators, and radiation accuracy feature indicators. Figure 2 As shown in the figure, after the satellite image data to be evaluated is input into the computer system for contour extraction, the radiation index will be automatically calculated according to the preset formula to obtain various indicators. Among them, the radiation quality evaluation indicators mainly include three categories:

[0089] (1) Grayscale statistical characteristic indicators, including radiation mean, minimum, maximum, peak, median, brightness range, variance, standard deviation and signal-to-noise ratio;

[0090] (2) Texture statistical feature indicators, including information entropy, angular second moment, detail energy, edge energy, local stability, information capacity, contrast and clarity;

[0091] (3) Radiation accuracy characteristic indicators, including skewness, steepness, edge radiation distortion, gain adjustment distortion and power spectrum.

[0092] The above characteristic indicators cover all angles of the satellite image to be evaluated, are more comprehensive and do not miss any details, can ensure the evaluation of any type of satellite image, and ensure the stability of the evaluation. Compared with these existing patents, the radiation evaluation of this patent covers a wider range and also improves the accuracy of the evaluation results.

[0093] Of course, those skilled in the art can, in combination with the above description, add or delete certain indicators from the various evaluation indicators to implement the present invention, but the specific effects may vary slightly. However, as long as the evaluation indicators include the three categories: grayscale statistical feature indicators, texture feature statistical indicators, and radiometric accuracy feature indicators, they are all within the scope of protection claimed by the present invention.

[0094] 1. Grayscale statistical characteristic indicators:

[0095] Grayscale statistical characteristic indicators include radiation mean, minimum, maximum, peak, median, brightness range, variance, standard deviation and signal-to-noise ratio. The calculation is described as follows:

[0096] (1) Radiation mean

[0097] The mean radiation value is the average value of the brightness of the entire image.

[0098]

[0099] Where S is the image area and g(x,y) is the grayscale value corresponding to the position (x,y).

[0100] (2) Minimum value

[0101] The minimum value is the minimum brightness of the entire image. min :

[0102] M min =MIN(g(x,y))

[0103] Where g(x,y) is the grayscale value corresponding to the position (x,y).

[0104] (3) Maximum value

[0105] The maximum value is the maximum value of the brightness of the entire image. max

[0106] M max =MAX(g(x,y))

[0107] (4) Peak

[0108] The peak is the brightness value that appears most frequently, that is, the highest point on the histogram curve. If an image containing several regions has several peaks, it means that the spectral resolution and signal-to-noise ratio of the remote sensor are relatively high.

[0109] (5) Median

[0110] The median is the value in the middle of a frequency distribution where the area to the left of the distribution is equal to the area to the right of the distribution.

[0111] (6) Brightness range

[0112] The brightness range value of each band is the difference between the maximum grayscale value and the minimum grayscale value in the band.

[0113] (7) Variance

[0114] Variance is the average squared difference between the grayscale values ​​of all pixels and the mean, with the square root of this difference being the standard deviation. Variance reflects the amount of information in an image, that is, the degree of data dispersion. The greater the deviation and the more discrete the data, the more it reflects the differences between different ground targets, indicating a richer image. This directly impacts the information content of the resulting color composite image. Its calculation formula is:

[0115]

[0116] where u is the mean of p(I,j).

[0117] (8) Covariance

[0118] Covariance is the average of the product of the difference between the pixel brightness values ​​of two bands in an image and their band means, reflecting the correlation between the bands. The correlation coefficient is generally between -1 and +1. If the correlation coefficient is greater than 0, it means that the two bands are positively correlated, that is, an increase in the brightness value of one band will cause an increase in the brightness value of the other band. Conversely, if the correlation coefficient is less than 0, the two bands are negatively correlated.

[0119] For two images, the mean is Where S is the image area.

[0120]

[0121] or

[0122]

[0123] Where σ 2 fg Represents the covariance of two adjacent images, σ 2 ff Represents the variance of the reference image f(x,y).

[0124] (9) Signal-to-noise ratio:

[0125] The signal-to-noise ratio generally refers to the ratio of the instrument signal to the noise signal strength, which is calculated using the local mean and local standard deviation method, the signal-to-noise ratio PSNR.

[0126]

[0127] Among them, M is the image mean, Std local is the local standard deviation of the image, is the sum of the local standard deviations that satisfy the maximum majority condition, and max(n) is the number of local standard deviations that satisfy the maximum majority condition.

[0128] 2. Texture statistical feature indicators

[0129] Texture statistical feature indicators, including information entropy, angular second moment, detail energy, edge energy, local stability, information capacity, contrast and clarity, are calculated as follows:

[0130] (1) Information entropy

[0131] Information entropy mainly reflects the degree of detail of the ground object information contained in the image, and is generally expressed by entropy. Entropy is a measure of the richness of image information from the perspective of information theory. By comparing the information entropy of images, the ability to express details of the image can be compared. At present, the application range of Shannon information entropy is relatively wide. This is a method of calculating remote sensing images as discrete memoryless information sources, that is, it is assumed that each point on the image exists independently. If the influence of image texture is considered, the remote sensing image can be regarded as a discrete memory information source, and the Markov entropy is calculated to express the amount of information. The calculation method of information entropy is:

[0132]

[0133] Where i is the gray level, L is the total number of gray levels, b i The gray value of level i, p(b i ) is b i Probability of occurrence.

[0134] (2) Angular second moment

[0135] The angular second-order moment is the sum of the squares of the pixel values ​​in the grayscale co-occurrence matrix, also known as energy. It is a measure of the uniformity of the grayscale distribution of an image and is mainly used to observe the coarseness and directional characteristics of the image texture. From the overall image perspective, the larger the angular second-order moment value, the coarser the texture, and vice versa. The calculation method for the angular second-order moment is:

[0136]

[0137] in, is the square of the pixel value of the gray-level co-occurrence matrix.

[0138] (3) Detail energy:

[0139] Detail energy and edge energy describe the details and edge shape characteristics of an image from the perspective of high-frequency components in the image frequency domain. Detail energy evaluates the quality of an image from a local perspective. The calculation method for detail energy is:

[0140]

[0141] in Regional variance, expressed as:

[0142]

[0143] m f (x,y) represents the average gray value of the region,

[0144]

[0145] Detail energy is the average of the sum of the variances of all (2M+1)*(2M+1) regions in an image, reflecting the richness of detail information in the image. Larger values ​​indicate finer textures, richer information, and clearer images. Generally, M is 1 or 2.

[0146] (4) Edge energy:

[0147] Edges are important information about the shape features of an image. Edges are high-frequency information in an image, but unlike noise signals, they are directional and can be extracted using anisotropic filters.

[0148] Edges are important information about image shape and detail. Edges are high-frequency information, but unlike noise signals, they are directional and can be extracted using anisotropic filters. Convolution is performed on the image using two normalized edge operators, E1 and E2, at 45 and 135 degrees, respectively, to eliminate blocking artifacts. The sum of these two values ​​yields the image's e(x, y), which is:

[0149] e(x,y)=E1(f(x,y))+E2(f(x,y))

[0150]

[0151] The edge energy is defined as

[0152]

[0153] The edge energy of an image indicates the richness and clarity of the edges in the image. The larger the edge energy value, the higher the image quality.

[0154] (5) Local stationarity

[0155] Local stability is used to measure the local changes in image texture and reflect the homogeneity of the image. The best effect is achieved when the value is 1. The calculation formula is:

[0156]

[0157] Where L is the total number of gray levels, is the gray-level co-occurrence matrix pixel value.

[0158] (6) Information capacity

[0159] The information capacity is defined based on a two-dimensional histogram to compare different distributions. The information capacity is:

[0160]

[0161] Among them, Norm log(G1,G2)=log[Num(G1,G2)] / log[max{Num(G1,G2)}] is called the logarithmic peak normalized histogram; max{Num(G1,G2)} is the frequency peak of the two-dimensional histogram; ω is the cumulative constraint domain, which is a measure within the histogram definition domain. It should take a sufficiently large area, and the correlation of the histogram in this area is the largest, so that the information capacity has macro-statistical significance and real information significance.

[0162] (7) Contrast

[0163] Image contrast reflects the degree of clarity of the target in the image compared to the background. The greater the contrast, the more obvious the target information in the image, so it is crucial for evaluating the quality of remote sensors. The greater the image contrast, the clearer the image texture. The better the reflection of the ground object, the better the image quality. The calculation method of image contrast is:

[0164]

[0165] Where |ij|=n, is the normalized gray-level co-occurrence matrix pixel value.

[0166] (8) Clarity

[0167] Clarity is also called average gradient. The calculation method of point sharpness algorithm is:

[0168]

[0169] Wherein, df / dx is the grayscale change rate perpendicular to the edge, and f(b)-f(a) is the total contrast in the direction perpendicular to the edge.

[0170] By discretizing it into a two-dimensional image and taking the square root, the formula for calculating the clarity can be obtained as follows:

[0171]

[0172] in:

[0173]

[0174] Where i and j cannot be equal to 0 at the same time, m and n are the number of rows and columns of the image, and f(x, y) is the satellite image to be evaluated.

[0175] Image clarity is the arithmetic difference between each pixel of the image and its surrounding pixels multiplied by The sum of the normalized values. m, n are the number of rows and columns of the image. Since a 3*3 scanning frame is used, the starting column, row and ending column, row are not included in the calculation.

[0176] 3. Radiation accuracy characteristic indicators

[0177] The radiation accuracy characteristic indicators include skewness, steepness, edge radiation distortion, gain adjustment distortion and power spectrum. The calculation is described as follows:

[0178] (1) Skewness:

[0179] An image's grayscale histogram is a statistical representation of the grayscale distribution of an image. The grayscale values ​​reflect the amount of radiation emitted by the object, while the distribution range of the histogram reflects the richness of the image information. The grayscale histogram morphology of an image can be analyzed and evaluated, with skewness and steepness being two important indicators.

[0180] Skewness reflects the degree to which the shape of the image histogram distribution deviates from the symmetrical shape around the mean. Positive skewness means that the distribution of asymmetric sides tends to be more positive, while negative skewness means that the distribution of asymmetric sides tends to be more negative. Its calculation formula is:

[0181]

[0182] Where i is the gray level, b i The gray value of level i, m is the arithmetic mean, p(b i ) is b i The probability of occurrence, d refers to the standard deviation of the image.

[0183] (2) Steepness (slope):

[0184] Steepness expresses whether the distribution shape of the image histogram is concentrated near the average value or extends to the edge. The higher the steepness, the narrower the dynamic range of the image grayscale, which means the image is more prominent. Its calculation formula is:

[0185]

[0186] (3) Edge radiation distortion:

[0187] Edge radiation distortion reflects the degree of radiation inhomogeneity, and its calculation method is to find the ratio of the mean and variance of the row direction mean vector.

[0188]

[0189] R is the mean vector of the row vector, that is:

[0190] R=[mean(p(1,j)),mean(p(2,j))…mean(p(i,j))] i＝1～N,j＝1～M ,

[0191] Where M and N are the rows and columns of the image, Mean is the mean, Std is the standard deviation, and p(i,j) is the grayscale value of the satellite image to be evaluated in row i and column j;

[0192] (4) Gain adjustment distortion:

[0193] The gain adjustment distortion reflects the degree of radiation inhomogeneity and is calculated by finding the ratio of the mean and variance of the column-wise mean vector.

[0194]

[0195] L is the mean vector of the column vector, that is:

[0196] L=[mean(p(i,1)),mean(p(i,2))…mean(p(i,j))] i＝1～N,j＝1～M ,

[0197] Among them, M and N are the rows and columns of the image, Mean is the mean, and Var is the variance.

[0198] (5) Power spectrum:

[0199] The image power spectrum comprehensively reflects the intensity of each spatial frequency component of the image. The digital image f(x,y) is transformed by Fourier transform:

[0200]

[0201] The frequency spectrum of the image is: |F(u,v)| 2

[0202]

[0203] |F(u,v)| 2 =R 2 (u,v)+I 2 (u,v)

[0204] The normalized power spectrum of the image is: PSD = log2 Sum; R(u,v) and I(u,v) are the real and imaginary parts of F(u,v) after Fourier transform.

[0205] In step C of this embodiment, a quality evaluation regression model is constructed and trained using samples in the training sample set, wherein the quality evaluation regression model is a regression model based on the XGBoost algorithm. This step C further includes:

[0206] Sub-step C1: Construct a quality evaluation regression model L = A·X, where L is the comprehensive evaluation index value, ranging from 0% to 100%; X is a column vector consisting of N evaluation indicators, and A is a row vector consisting of the deviation impact factors corresponding to each evaluation indicator, expressed as follows:

[0207] A=(h1 ... hn1 w1 ... w n2 f1 ... f n3 )

[0208] Among them: H1 is the first grayscale statistical characteristic index, h1 is its corresponding deviation influence factor; H n1 is the n1th grayscale statistical characteristic index, h n1 Its corresponding deviation influence factor; W1 is the first texture statistical feature index, w1 is its corresponding deviation influence factor; W n2 is the n2th texture statistical feature index, w n2 Its corresponding deviation impact factor; F1 is the first radiation accuracy characteristic index, f1 is its corresponding deviation impact factor; F n3 is the n3th radiation accuracy characteristic index, f n3 is the corresponding deviation influence factor; n1, n2, and n3 are the number of grayscale statistical feature indicators, texture statistical feature indicators, and radiation accuracy feature indicators, respectively, n1+n2+n3=N; N is the total number of evaluation indicators.

[0209] During the training process, [L, X] is the input content of the quality evaluation regression model, and A is the content that needs to be obtained through training.

[0210] Sub-step C2: Initialize A in the quality evaluation regression model;

[0211] In sub-step C2, for each evaluation indicator, its initial corresponding deviation impact factor is: if not set by the user, it takes the default value; if set by the user, it takes the user-set value. The default value can be set to 1 / N for each parameter's deviation impact factor.

[0212] In sub-step C3, the quality evaluation regression model is trained iteratively using the samples in the training sample set and the XGBoost algorithm until the prediction accuracy requirement is met.

[0213] In the sub-step C3, the learning rate of the XGBoost algorithm is 0.1, the total number of trees is 500, and other parameters are set to default values. The maximum depth of the tree, the minimum leaf split value, the sample sampling, and the attribute column sampling parameters are optimized using the grid search GridSearchCV function in the Sciki-learn package to optimize the XGBoost parameters.

[0214] 1. Training sample set

[0215] Regarding the training sample set, it contains the following three types of training samples:

[0216] The first type of training samples has an index set consisting of various evaluation indices calculated based on satellite images of satellites passing through ground calibration sites, and its true value is the comprehensive evaluation index value of satellite images obtained by measuring with ground equipment.

[0217] The second type of training samples, whose index set is the various evaluation indexes of the previous satellite images of the source satellite of the satellite image to be evaluated, and whose sample true value is the artificial comprehensive evaluation index value of the previous satellite images;

[0218] The third type of training samples, whose index set is the various evaluation indicators of the previous satellite images of the source satellite of the satellite image to be evaluated, and the sample true value is the automatic comprehensive evaluation index value obtained by step D of the previous satellite image.

[0219] Among them, at the initial stage of training, the sum of the number of training samples of the first category and the second category accounts for 100% of the total number of samples; in the mature stage of training, the proportion of training samples of the first category in the total number of samples is less than 30%; the number of samples of the second category accounts for less than 30% of the total number of samples; the number of samples of the third category accounts for more than 40% of the total number of samples, and the proportion of samples of the third category in the total number of samples gradually increases with the gradual accumulation of satellite images.

[0220] Through the above methods, on the one hand, the stability and reliability of the initial training data are guaranteed; on the other hand, in the mature stage, the third type of training samples gradually increases, while the proportion of the first and second types of training samples gradually decreases, and the total sample size continues to increase, ensuring that the evaluation effect is getting better and better.

[0221] 2. Use XGBoost algorithm to iterate the training of quality evaluation regression model

[0222] XGBoost is an ensemble learning method that achieves a task through the collaboration of multiple weak learners, often achieving better generalization performance than a single learner. Boosting is a type of ensemble learning algorithm that boosts weak learners to strong ones. Examples include Adaboost, GBDT, and XGBoost. XGBoost, or eXtreme Gradient Boosting, implements a generalized tree boosting algorithm.

[0223] This algorithm can solve both classification and regression problems. The XGBoost algorithm has attracted widespread attention due to its high prediction accuracy and exceptional efficiency. This algorithm imposes a penalty on the number of leaf nodes in each tree, limiting the growth of leaf nodes and making each tree weaker. It also introduces a learning rate to further reduce the influence of each tree. This approach comes at the cost of a larger number of trees, but the results are well worth it.

[0224] The XGBoost algorithm works as follows: First, the training set and the true sample values ​​are used to train the first tree. This tree is then used to predict the training set and obtain the predicted value for each sample. The predicted value is then subtracted from the true value to obtain the residual. Next, the second tree is trained, using the residual as the standard value. Finally, the residuals obtained from training the two trees are used to train the third tree, and so on. The objective function of this model is defined as follows:

[0225]

[0226] The absolute deviation Dif and relative deviation PDif formulas are defined as follows:

[0227] Dif=Ref pre -Ref sim

[0228]

[0229] The process of using the improved XGBoost algorithm for satellite image quality assessment model regression training is as follows:

[0230] Step 1: The key influencing factors of deviation obtained based on the deviation influencing factor analysis are used as the input of the model;

[0231] Step 2: By combining the key factors affecting the deviation in different ways, different models are constructed for training and prediction.

[0232] Step 3: Construct sample training and test sets

[0233] Step 4: Use the XGBoost model to build a deep learning training model and initialize the model parameters. The objective function of the XGBoost regression model is:

[0234]

[0235] Step 5: Continue iterating until the prediction accuracy requirement is met and the training ends;

[0236] Step 6: By adjusting the parameters, output the predicted value of the model, and compare and analyze the original simulation value with the model predicted value, evaluate the prediction effect, and obtain the influencing factor of the final deviation, thereby obtaining the various parameters of the quality evaluation model.

[0237] When calculating the comprehensive evaluation index value, this embodiment uses the XGBoost model and deep learning method to train a more objective sample set, so that the total parameters of the quality evaluation are more objective, and then avoid human subjective factors, thereby improving the evaluation accuracy. Specifically, the regression model based on the XGBoost algorithm and large sample training are used, and the total parameters are obtained by combining deep learning training. Therefore, the trained parameter effect is more objective and has room for improvement. As the data samples of the image data set gradually increase, after more training, some better results will be obtained, that is, the evaluation results will become more and more objective and accurate.

[0238] In step D of this embodiment, the N evaluation indicators of the satellite image to be evaluated are substituted into the trained quality evaluation regression model to obtain the comprehensive evaluation indicator value of the satellite image to be evaluated.

[0239] It should be noted that, in response to the previous sub-step C3, step D also includes: adding the N evaluation indicators and the corresponding comprehensive evaluation indicator values ​​of the satellite image to be evaluated as one sample in the training sample set.

[0240] What needs to be specially explained about this embodiment is that the entire radiation evaluation process is fully automated and does not require human intervention, which reduces manual workload, greatly improves the degree of automation, and enhances the objectivity of the evaluation.

[0241] This concludes the introduction to satellite image radiation quality evaluation in this embodiment.

[0242] In another exemplary embodiment of the present invention, a computer device for evaluating the radiation quality of satellite images is also provided. Figure 3 This is a structural diagram of a computer device for evaluating satellite image radiation quality according to an embodiment of the present invention. Figure 3 The computer device for evaluating satellite image radiation quality in this embodiment includes:

[0243] The application plug-in layer stores computer plug-in programs, which in this example are programs for calculating various radiation evaluation indicators, XGBoost programs, deep learning training programs, etc.

[0244] A basic platform layer, which provides an interface for calling computer plug-in programs stored in the application plug-in layer;

[0245] A business subsystem configured to execute the method described in the above embodiment when calling the computer plug-in program stored in the application plug-in layer through the basic platform layer;

[0246] Among them, the basic platform layer automatically calls the computer plug-in program of the application plug-in layer in the form of XML orders to realize automated radiation quality evaluation; the basic platform layer provides one or more of the following interface groups: component communication interface, COM compatible interface, universal UI interface, CPU+GPU acceleration interface.

[0247] Specifically, the deployment scheme of the platform and plug-in in this embodiment is:

[0248] Basic platform layer: Based on the client-side infrastructure framework, common modules, module packaging standards, and module configuration standards provided by the application platform, business algorithm engineering is implemented. The application platform provides the business subsystem with a client-side infrastructure framework and common modules. When engineering plug-in business algorithms, the platform's basic program provides component communication interfaces, COM-compatible interfaces, universal UI interfaces, CPU+GPU acceleration interfaces, and other algorithm engineering interfaces to effectively improve algorithm operation efficiency.

[0249] Application Plug-in Layer: Plug-in programs run on the platform layer and are automatically invoked by the platform via XML orders to implement automated radiation quality assessment. This independent quality assessment software is the core component of the entire subsystem, enabling radiation quality monitoring of various data. Data assessment software operates automatically, enabling automated quality assessments based on configuration options. More complex analysis algorithms can be run on background computing nodes to improve analysis efficiency.

[0250] Those skilled in the art should understand that this embodiment adopts a platform + plug-in-based deployment mode, which can automatically input and output an image and calculate radiation quality indicators throughout the entire process, thereby improving convenience.

[0251] In another exemplary embodiment of the present invention, a computer-readable storage medium is provided, wherein computer instructions are stored in the computer-readable storage medium, and when the computer instructions are executed, the method described above is implemented.

[0252] So far, the embodiments of the present invention have been introduced.

[0253] So far, the embodiments of the present invention have been described in detail with reference to the accompanying drawings. Based on the above description, those skilled in the art should have a clear understanding of the present invention.

[0254] In summary, the present invention calculates comprehensive evaluation indicators with more comprehensive metrics, comprehensively considering various perspectives, and achieving a wider coverage of radiation evaluation, thereby improving the accuracy of the evaluation results. Furthermore, the use of deep learning methods to train a more objective sample set eliminates subjective factors, making the overall quality evaluation parameters more objective. This invention improves the automation and convenience of satellite imagery evaluation, provides technical support for expanding the application of optical remote sensing satellite imagery in the remote sensing field, and has considerable promotional value.

[0255] It should be noted that for certain implementations, if they are not the key content of the present invention and are well known to ordinary technicians in the relevant technical field, they are not described in detail in the drawings or the main text of the specification. In this case, they can be understood by referring to relevant existing technologies.

[0256] Furthermore, the above embodiments are provided solely for the purpose of enabling the present invention to satisfy legal requirements. The present invention may be implemented in many different forms and should not be construed as limited to the embodiments set forth herein. Furthermore, the above definitions of the various elements and methods are not limited to the specific structures, shapes, or methods described in the embodiments, and those skilled in the art may readily modify or substitute them.

[0257] Furthermore, unless specifically described or required to occur sequentially, the order of the steps is not limited to the order listed above and may be varied or rearranged based on desired design requirements. Furthermore, the above embodiments may be mixed and matched with each other or with other embodiments based on design and reliability considerations. That is, the technical features of different embodiments may be freely combined to form more embodiments.

[0258] The algorithms and displays provided herein are not related to any particular computer, virtual system, or other inherent devices. Various general-purpose systems may also be used together with the teachings herein. Based on the above description, it is apparent that the structure required for constructing such systems is suitable. In addition, the present invention is not directed to any specific programming language. It should be understood that various programming languages ​​can be utilized to implement the present invention, and the description of specific languages ​​above is intended to disclose the preferred embodiment of the present invention.

[0259] The present invention can be implemented by means of hardware including several different elements and by means of a suitably programmed computer. The various component embodiments of the present invention can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art will appreciate that a microprocessor or digital signal processor (DSP) can be used in practice to implement some or all of the functions of some or all of the components in the relevant equipment according to an embodiment of the present invention. The present invention can also be implemented as a device or apparatus program (e.g., a computer program and a computer program product) for executing part or all of the methods described herein. Such a program implementing the present invention can be stored on a computer-readable medium, or can have the form of one or more signals. Such a signal can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0260] All modules in the embodiments of the present invention may be hardware structures. Physical implementations of the hardware structures include but are not limited to physical devices. Physical devices include but are not limited to transistors, memristors, and DNA computers.

[0261] Those skilled in the art will appreciate that the modules in the devices in the embodiments can be adaptively changed and set in one or more devices different from the embodiments. The modules or units or components in the embodiments can be combined into one module or unit or component, and in addition they can be divided into multiple sub-modules or sub-units or sub-components. Except that at least some of such features and (or) processes or units are mutually exclusive, all features of the invention in this specification (including claims, abstracts and drawings) and all processes or units of any method or device invented in this way can be combined in any combination. Unless otherwise expressly stated, each feature of the invention in this specification (including the accompanying claims, abstracts and drawings) can be replaced by an alternative feature that provides the same, equivalent or similar purpose. Moreover, in a unit claim that lists several devices, several of these devices can be embodied by the same hardware item.

[0262] Similarly, it should be understood that in order to streamline the present invention and aid understanding of one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this inventive approach should not be interpreted as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as reflected in the claims, various inventive aspects lie in less than all the features of the preceding individual embodiments. Accordingly, the claims that follow the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of the present invention.

[0263] The specific embodiments described above provide a detailed description of the objectives, technical solutions and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for evaluating satellite image radiation quality, characterized in that: include: Step A: Perform contour extraction on the satellite image to be evaluated, remove the outer edge parts that have no analytical value, and retain the inner parts that have analytical value; Step B: obtaining N evaluation indicators of the satellite image to be evaluated, wherein the N evaluation indicators are divided into the following three categories: grayscale statistical feature indicators, texture statistical feature indicators, and radiation accuracy feature indicators; Step C: constructing a quality evaluation regression model, and training the quality evaluation regression model using samples in the training sample set, wherein the quality evaluation regression model is a regression model based on the XGBoost algorithm; Step D: Substituting the N evaluation indicators of the satellite image to be evaluated into the trained quality evaluation regression model to obtain the comprehensive evaluation indicator value of the satellite image to be evaluated; adding the N evaluation indicators of the satellite image to be evaluated and the corresponding comprehensive evaluation indicator value to one sample in the training sample set; Wherein, in step C, the training sample set includes the following three types of training samples: The first type of training samples has an index set consisting of various evaluation indices calculated based on satellite images of satellites passing through ground calibration sites, and its true value is the comprehensive evaluation index value of satellite images obtained by measuring with ground equipment. The second type of training samples, whose index set is the various evaluation indexes of the previous satellite images of the source satellite of the satellite image to be evaluated, and whose sample true value is the artificial comprehensive evaluation index value of the previous satellite images; The third type of training samples, whose index set is the various evaluation indicators of the previous satellite images of the source satellite of the satellite image to be evaluated, and the sample true value is the automatic comprehensive evaluation index value obtained by step D of the previous satellite image; Among them, at the initial stage of training, the sum of the number of training samples of the first category and the second category accounts for 100% of the total number of samples; in the mature stage of training, the proportion of training samples of the first category in the total number of samples is less than 30%; the number of samples of the second category accounts for less than 30% of the total number of samples; the number of samples of the third category accounts for more than 40% of the total number of samples, and the proportion of samples of the third category in the total number of samples gradually increases with the gradual accumulation of satellite images.

2. The method according to claim 1, characterized in that The step C comprises: Sub-step C1: Construct a quality evaluation regression model L = A·X, where: L is the comprehensive evaluation index value, and its value range is 0%-100%; X is a column vector consisting of N evaluation indicators, and A is a row vector consisting of the deviation impact factors corresponding to each evaluation indicator. During the training process, [L, X] is the input content of the quality evaluation regression model, and A is the content to be obtained through training; Sub-step C2: Initialize A in the quality evaluation regression model; In sub-step C3, the quality evaluation regression model is trained iteratively using the samples in the training sample set and the XGBoost algorithm until the prediction accuracy requirement is met.

3. The method according to claim 2, characterized in that In the sub-step C1: Among them: H1 is the first grayscale statistical characteristic index, h1 is its corresponding deviation influence factor; H n1 is the n1th grayscale statistical characteristic index, h n1 Its corresponding deviation influence factor; W1 is the first texture statistical feature index, w1 is its corresponding deviation influence factor; W n2 is the n2th texture statistical feature index, w n2 Its corresponding deviation impact factor; F1 is the first radiation accuracy characteristic index, f1 is its corresponding deviation impact factor; F n3 is the n3th radiation accuracy characteristic index, f n3 is the corresponding deviation influence factor; n1, n2, and n3 are the number of grayscale statistical feature indicators, texture statistical feature indicators, and radiation accuracy feature indicators, respectively. n1+n2+n3=N, where N is the total number of evaluation indicators.

4. The method according to claim 3, characterized in that In the sub-step C2, for each evaluation indicator, its initial corresponding deviation impact factor is: if the user does not set it, it takes the default value; if the user sets it, it takes the user set value; and / or In the sub-step C3, the learning rate of the XGBoost algorithm is 0.1, the total number of trees is 500, and other parameters are set to default values. The maximum depth of the tree, the minimum leaf split value, the sample sampling, and the attribute column sampling parameters are optimized using the grid search GridSearchCV function in the Sciki-learn package to optimize the XGBoost parameters.

5. The method according to claim 1, wherein: The grayscale statistical characteristic indicators include: radiation mean, minimum value, maximum value, peak value, median value, brightness range value, variance, standard deviation and signal-to-noise ratio; The texture statistical feature indicators include: information entropy, angular second moment, detail energy, edge energy, local stability, information capacity, contrast and clarity; The radiation accuracy characteristic indicators include: skewness, steepness, edge radiation distortion, gain adjustment distortion and power spectrum.

6. The method according to claim 5, characterized in that: Among the grayscale statistical characteristic indicators: The radiation mean M is: Minimum value M min For: M min =MIN(g(x,y)); Maximum value M max For: M max =MAX(g(x,y)); The peak value is the highest point on the histogram curve corresponding to the satellite image to be evaluated; The median is the median of the frequency distribution of the histogram curve corresponding to the satellite image to be evaluated; The brightness range value of each band is the difference between the maximum grayscale value and the minimum grayscale value in the band; The variance is the average squared difference between the grayscale values ​​of all pixels and the mean, and its square root is the standard deviation; The signal-to-noise ratio was calculated using the local mean and local standard deviation method; Where S is the image area, g(x,y) is the grayscale value corresponding to the position (x,y) in the satellite image to be evaluated; Among the texture statistical feature indicators: Information entropy is: Where i is the gray level, L is the total number of gray levels, b i The gray value of level i, p(b i ) is b i Probability of occurrence; The angular second moment is: in, is the square of the pixel value of the gray-level co-occurrence matrix; The detail energy is: in Regional variance, expressed as: f(x+i,y+j) is the variance of the pixel with coordinates (x+i,y+j); m f (x,y) represents the average grayscale value of the region, the average value of the sum of the variances of all (2M+1)*(2M+1) regions in the image: The edge energy is: m, n are the pixel values ​​of the length and width of the image respectively; two normalized edge operators E1 and E2 of 45 degrees and 135 degrees are used to perform convolution calculation on the satellite image f(x,y) to eliminate the blocking effect, and the image e(x,y) is obtained after adding them together, where e(x,y)=E1(f(x,y))+E2(f(x,y)). Local stability is: Where L is the total number of gray levels, is the gray-level co-occurrence matrix pixel value; The information capacity is: Among them, Norm log (G1, G2) = log[Num(G1, G2)] / log[max{Num(G1, G2)}] is called the logarithmic peak normalized histogram; max{Num(G1, G2)} is the frequency peak of the two-dimensional histogram; ω is the cumulative constraint domain; The contrast ratio is: Where |ij|=n, is the normalized gray-level co-occurrence matrix pixel value; The clarity is: in: Where i and j cannot be equal to 0 at the same time, m and n are the number of rows and columns of the satellite image to be evaluated, and f(x, y) is the satellite image to be evaluated; Among the radiation accuracy characteristic indicators, The skewness reflects the degree to which the shape of the image histogram distribution deviates from the symmetrical shape around the mean. A positive skewness indicates that the distribution of the asymmetric side tends to be more positive, and a negative skewness indicates that the distribution of the asymmetric side tends to be more negative. The skewness is: Where i is the gray level, m is the arithmetic mean, b i The gray value of level i, p(b i ) is b i The probability of occurrence, d refers to the standard deviation of the image; The steepness is: Edge radiation distortion: R is the mean vector of the row vector, that is: R = [mean(p(1,j)),mean(p(2,j))…mean(p(i,j))] i＝1～N,j＝1～M , where M, N are the rows and columns of the satellite image to be evaluated, Mean is the mean, Std is the standard deviation, and p(i,j) is the grayscale value of the satellite image to be evaluated in row i and column j; The gain adjustment distortion is: L is the mean vector of the column vector, that is: L = [mean(p(i,1)),mean(p(i,2))…mean(p(i,j))] i＝1～N,j＝1～M ; The power spectrum comprehensively reflects the intensity of each spatial frequency component of the image, specifically: PSD = log2Sum, |F(u,v)| 2 =R 2 (u,v)+I 2 (u,v); R(u,v) and I(u,v) are the real and imaginary parts of F(u,v) after Fourier transform.

7. A computer device for evaluating satellite image radiation quality, characterized in that: include: An application plug-in layer storing computer plug-in programs; A basic platform layer, which provides an interface for calling computer plug-in programs stored in the application plug-in layer; The business subsystem is configured to execute the method according to any one of claims 1 to 6 when calling the computer plug-in program stored in the application plug-in layer through the basic platform layer.

8. The computer device according to claim 7, wherein: The basic platform layer automatically calls the computer plug-in program of the application plug-in layer in the form of an XML order to realize automated radiation quality evaluation; and / or The basic platform layer provides one or more of the following interface groups: component communication interface, COM compatible interface, general UI interface, CPU+GPU acceleration interface.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, which implement the method according to any one of claims 1 to 6 when executed.

Images