An optoelectronic simulation image processing method and system based on spearman rank correlation coefficient

Abstract

This invention provides a photoelectric simulation image processing method and system based on Spearman's rank correlation coefficient. Both the method and system utilize the Spearman correlation coefficient, extracting image feature points using the Harris feature point detection algorithm to perform a non-parametric measurement of the image's rank correlation, accurately expressing the similarity between the simulated and real images. The simulation model parameters are then modified based on the Spearman correlation coefficient until it approaches an ideal value, at which point further modifications are stopped. This invention inputs a measured image into the simulation model to obtain a simulated image. The Spearman rank correlation coefficient is obtained by comparing the measured and simulated images. Based on this coefficient, the relevant parameters of the simulation model are optimized to obtain an improved simulation model. This method eliminates the need to include the global pixels of both the measured and simulated images in the improvement process, accurately evaluating the similarity between the simulated and measured images with minimal computational cost.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a photoelectric simulation image processing method and system based on Spearman's rank correlation coefficient. Background Technology

[0002] The statements in this section are merely background information relating to this disclosure and do not necessarily constitute prior art.

[0003] Simulated images can be created by mathematically modeling real-world imaging processes. This process involves scene geometry, radiometry, thermal analysis, and sensor characteristics. Simulated images can help researchers analyze the physical properties of different targets under simulated conditions, or conduct preliminary evaluations of sensor systems before manufacturing. They can also be used to validate or optimize algorithm models related to target detection and recognition, thereby significantly reducing R&D costs and improving performance.

[0004] The credibility of a simulation image refers to the accuracy with which the model reproduces the real world. Accuracy verification focuses on evaluating the accuracy of the reconstructed imaging model. One important method for evaluating the accuracy of a reconstructed imaging model is to compare the simulation image with real-world images. The comparison results help improve the simulation model and optimize relevant parameters.

[0005] Therefore, accurately evaluating the similarity between simulated images and measured images has become a key and challenging issue in the field of optoelectronic image simulation. Summary of the Invention

[0006] This invention provides a photoelectric simulation image processing method and system based on Spearman's rank correlation coefficient. The method inputs a measured image into a simulation model to obtain a simulated image. By comparing the measured image and the simulated image, the Spearman's rank correlation coefficient is obtained. Based on the Spearman's rank correlation coefficient, the relevant parameters of the simulation model are optimized to obtain an improved simulation model. This method does not require incorporating the global pixels of the measured image and the simulated image into the improvement of the simulation model. It accurately evaluates the similarity between the simulated image and the measured image with minimal computational load.

[0007] The technical solution for achieving the objective of this invention is as follows:

[0008] On one hand, the present invention provides a photoelectric simulation image processing method based on Spearman's rank correlation coefficient, comprising:

[0009] The first simulated image is obtained by inputting the measured image into the simulation model;

[0010] Extract a series of target feature points from the measured image and the first simulated image;

[0011] Extract the data values ​​of each of the target feature points, and calculate the first Spearman rank correlation coefficient based on the data values;

[0012] The relevant parameters of the simulation model were optimized based on the first Spearman rank correlation coefficient to obtain the improved simulation model;

[0013] The measured image is input into the improved simulation model to obtain the second simulation image. The second Spearman rank correlation coefficient is obtained based on the measured image and the second simulation image. The improved simulation model is then optimized again to obtain a further improved simulation model.

[0014] Repeat the above process until the Nth Spearman rank correlation coefficient approaches the ideal value.

[0015] Based on one aspect, as a further limitation of the embodiments of the present invention, the step of inputting the measured image into the simulation model to obtain the first simulation image includes:

[0016] A simulation model is obtained by mathematically modeling the imaging link in the real world.

[0017] The measured image is input into the simulation model, and the simulation model outputs the first simulation image corresponding to the measured image.

[0018] Based on one aspect, as a further limitation of the embodiments of the present invention, the extraction of a series of target feature points from the measured image and the first simulated image includes:

[0019] Read the measured image and the first simulated image;

[0020] The measured image and the first simulated image are processed such that the pixels of the measured image and the first simulated image represent the same unit;

[0021] The Harris feature point detection method was used to extract a series of target feature points of interest from the measured image and the first simulated image, respectively.

[0022] Based on one aspect, as a further limitation of the embodiments of the present invention, the Harris feature point detection method includes:

[0023] The local window is used to move simultaneously in the horizontal and vertical directions of the measured image or the first simulated image;

[0024] Calculate the pixel value change E(x,y) within the local window. If the pixel value change E(x,y) is less than the preset change, continue to move the local window in the horizontal and vertical directions. If the pixel value change E(x,y) is greater than the preset change, calculate the corner response function R of the region where this local window is located.

[0025] Thresholding is applied to the corner response function R, and the pixels corresponding to the local maxima of the corner response function R are extracted as target feature points. The local window is then moved further in the horizontal and vertical directions.

[0026] Based on one aspect, as a further limitation of the embodiments of the present invention, the extraction of data values ​​for each of the target feature points includes:

[0027] A series of target feature points are extracted by simultaneously moving the measured image or the first simulated image in both the horizontal and vertical directions using a local window.

[0028] Extract the data value corresponding to the pixel of each target feature point one by one;

[0029] The data values ​​corresponding to a series of target feature points are summarized.

[0030] Based on one aspect, as a further limitation of the embodiments of the present invention, the calculation of the first Spearman rank correlation coefficient based on the data values ​​includes:

[0031] The measured image is represented as X. i The measured images X are arranged in order according to their data values. i A series of target feature points;

[0032] Assign a grade value R(x) to each of the series of target feature points. i Thus, we obtain R(x). i =1,2,3,…,n;

[0033] The first simulated image is represented as Y i The first simulation image Y is arranged in order according to its data values. i A series of target feature points;

[0034] A series of target feature points in the measured image correspond one-to-one with a series of target feature points in the first simulated image, and a level value R(y) is assigned to a series of target feature points in the first simulated image. i That is, the first simulated image Y i Sort value among all values;

[0035] Calculate the target feature point level value R(x) of the measured image. i The target feature point level value R(y) in the first simulated image. i Square of the difference

[0036] The first Spearman rank correlation coefficient is calculated using the following formula;

[0037]

[0038] In the formula, ρ1 is the first Spearman rank correlation coefficient. Let R(x) be the value of R(x). i With R(y) i The square of the difference, where n is the number of target feature points in the series.

[0039] Based on one aspect, as a further limitation of the embodiments of the present invention, the second Spearman rank correlation coefficient is obtained based on the measured image and the second simulated image, including:

[0040] The improved simulation model outputs a second simulated image corresponding to the input measured image;

[0041] Read the measured image and the second simulated image;

[0042] The measured image and the second simulated image are processed such that the pixels of the measured image and the second simulated image represent the same unit;

[0043] The Harris feature point detection method was used to extract a series of target feature points of interest from the measured image and the second simulated image, respectively.

[0044] A series of target feature points are extracted by simultaneously moving the measured image or the second simulated image in both the horizontal and vertical directions using a local window;

[0045] Extract the data value corresponding to the pixel of each target feature point one by one;

[0046] Summarize the data values ​​corresponding to a series of target feature points;

[0047] The measured image is represented as X. i The measured images X are arranged in order according to their data values. i A series of target feature points;

[0048] Assign a grade value R(x) to each of the series of target feature points. i Thus, we obtain R(x). i =1,2,3,…,n;

[0049] The second simulated image is represented as Z. i Arrange the second simulation image Z in sequence according to the data values ​​of the second simulation image. i A series of target feature points;

[0050] A series of target feature points in the measured image correspond one-to-one with a series of target feature points in the second simulated image, and a level value R(Z) is assigned to a series of target feature points in the second simulated image.i That is, the second simulated image Z i Sort value among all values;

[0051] Calculate the target feature point level value R(x) of the measured image. i The target feature point level value R(Z) in the second simulated image. i Square of the difference

[0052] The second Spearman rank correlation coefficient is calculated using the following formula;

[0053]

[0054] In the formula, ρ2 is the second Spearman rank correlation coefficient. Let R(x) be the value of R(x). i With R(Z) i The square of the difference, where n is the number of target feature points in the series.

[0055] Based on one aspect, as a further limitation of the embodiments of the present invention, the above process is repeated until the Nth Spearman rank correlation coefficient approaches the ideal value, including:

[0056] The first Spearman correlation coefficient to the Nth Spearman rank correlation coefficient all indicate the direction of the correlation between X and Y;

[0057] If X increases and Y tends to increase, then the Spearman correlation coefficient is positive; if X increases and Y tends to decrease, then the Spearman correlation coefficient is negative; if the Spearman correlation coefficient is 0, it indicates that Y has no trend when X increases.

[0058] If X and Y are positively correlated, the Spearman correlation coefficient is 1; if X and Y are negatively correlated, the Spearman correlation coefficient is -1.

[0059] On the other hand, the present invention provides a photoelectric simulation image processing system based on Spearman's rank correlation coefficient, comprising:

[0060] The input unit inputs the measured image into the simulation model, and the simulation model outputs the first simulation image; the input unit inputs the measured image into the improved simulation model, and the improved simulation model outputs the second simulation image; the input unit inputs the measured image into the further improved simulation model, and the further improved simulation model outputs the third simulation image.

[0061] The processing unit processes the measured image and the first simulated image to obtain a first Spearman rank correlation coefficient; the processing unit processes the measured image and the second simulated image to obtain a second Spearman rank correlation coefficient; the processing unit processes the measured image and the third simulated image to obtain a third Spearman rank correlation coefficient.

[0062] An adjustment unit optimizes the relevant parameters of the simulation model based on a first Spearman rank correlation coefficient to obtain an improved simulation model; the adjustment unit further optimizes the relevant parameters of the improved simulation model based on a second Spearman rank correlation coefficient to obtain a further improved simulation model.

[0063] In another respect, as a further limitation of the embodiments of the present invention, the processing unit uses the following formula to obtain the Spearman rank correlation coefficient;

[0064]

[0065] In the formula, ρ is the Spearman rank correlation coefficient. Let R(x) be the value of R(x). i With R(Z) i The square of the difference, where n is the number of target feature points in the series, R(x) i R(y) represents the target feature point level value in the measured image. i Let R(Z) be the target feature point level value of the first simulated image. i The value represents the target feature point level in the second simulated image.

[0066] Compared with the prior art, the beneficial effects of the present invention are:

[0067] 1. This invention inputs the measured image into the simulation model to obtain the simulation image, and obtains the Spearman rank correlation coefficient by comparing the measured image and the simulation image. Based on the Spearman rank correlation coefficient, the relevant parameters of the simulation model are optimized to obtain an improved simulation model. It does not require the global pixels of the measured image and the simulation image to be included in the improvement of the simulation model. It accurately evaluates the similarity between the simulation image and the measured image with minimal computation.

[0068] 2. This invention extracts image feature points using the Harris feature point detection algorithm and performs a non-parametric measurement of the image rank correlation, which can accurately express the similarity between the simulated image and the real image. Then, the parameters of the simulation model are modified according to the Spearman correlation coefficient until the Spearman rank correlation coefficient approaches the ideal value and the simulation model is no longer modified. Attached Figure Description

[0069] Figure 1A flowchart of a photoelectric simulation image processing method based on Spearman's rank correlation coefficient provided by the present invention;

[0070] Figure 2 A schematic diagram of a photoelectric simulation image processing system based on Spearman's rank correlation coefficient provided for this invention;

[0071] Figure 3 This is a comparison illustration between the measured image and the first simulated image provided in an embodiment of the present invention;

[0072] Figure 4 This is an illustration of a series of target feature points extracted from the measured image and the first simulated image, provided for an embodiment of the present invention.

[0073] Figure 5 Provided for embodiments of the present invention Figure 4 A sorting table that sorts a series of target feature points according to their pixel values;

[0074] Figure 6 This is a comparison chart of the error results of feature point level values ​​between the measured image and the first simulated image provided in this embodiment of the invention;

[0075] Figure 7 This is an illustration of a series of target feature points extracted from the measured image and the second simulated image, provided for an embodiment of the present invention.

[0076] Figure 8 Provided for embodiments of the present invention Figure 7 A sorting table that sorts a series of target feature points according to their pixel values;

[0077] Figure 9 This is a comparison of the error results of the feature point level values ​​between the measured image and the second simulated image provided in this embodiment of the invention. Detailed Implementation

[0078] The present invention will now be described in detail with reference to the embodiments shown in the accompanying drawings. However, it should be noted that these embodiments are not intended to limit the present invention. Equivalent changes or substitutions in function, method, or structure made by those skilled in the art based on these embodiments are all within the scope of protection of the present invention.

[0079] Example 1:

[0080] Please see Figure 1 , Figure 1 This invention provides a method for photoelectric simulation image processing based on Spearman's rank correlation coefficient, comprising:

[0081] Step 1: Input the measured image into the simulation model to obtain the first simulation image.

[0082] The above process of inputting the measured image into the simulation model to obtain the first simulation image includes:

[0083] A simulation model is obtained by mathematically modeling the imaging link in the real world.

[0084] The measured image is input into the simulation model, and the simulation model outputs the first simulation image corresponding to the measured image.

[0085] Please see Figure 3 , Figure 3 This is a comparison illustration between the measured image and the first simulated image provided in an embodiment of the present invention; Figure 3 (a) is the measured image. Figure 3 (b) The first simulation image generated in the simulation model for the scene as a reference to the measured image. Figure 3 (a) and Figure 3 (b) shows a water inlet on the top of the cube box, through which water can be poured into the box in actual use. Figure 3 (b) The white frame above the water inlet is a simulated light source.

[0086] Step 2: Extract a series of target feature points from the measured image and the first simulation image.

[0087] The above-mentioned extraction of a series of target feature points from the measured image and the first simulated image includes:

[0088] Read the measured image and the first simulated image;

[0089] Process the measured image and the first simulated image so that the pixels in the measured image and the first simulated image represent the same unit;

[0090] The Harris feature point detection method was used to extract a series of target feature points of interest from the measured image and the first simulation image, respectively.

[0091] The aforementioned Harris feature point detection methods include:

[0092] The local window is used to move simultaneously in the horizontal and vertical directions on either the measured image or the first simulated image;

[0093] Calculate the pixel value change E(x,y) within the local window. If the pixel value change E(x,y) is less than the preset change, continue to move the local window in the horizontal and vertical directions. If the pixel value change E(x,y) is greater than the preset change, calculate the corner response function R of the region where this local window is located.

[0094] Thresholding is applied to the corner response function R, and the pixels corresponding to the local maxima of the corner response function R are extracted as target feature points. The local window is then moved further in the horizontal and vertical directions.

[0095] Please see Figure 4 , Figure 4 This illustration shows a series of target feature points extracted from the measured image and the first simulated image, as provided in an embodiment of the present invention. Figure 4 (a) Illustration of the extraction of 8 target feature points from the measured image. Figure 4 (b) Illustration of the eight target feature points extracted from the first simulation image. Figure 3 (a) and Figure 4 (a) The four corners of the measured image are not clear, so a portion of the measured image is cropped and used to generate the first simulated image in the simulation model. Figure 3 (b) and Figure 4 (b) The four corners of the first simulation image are relatively clear. Figure 4 (a) and Figure 4 (b) The eight target feature points extracted are located at: background, water inlet, top surface of box, left surface of box (center), left surface of box (corner), right surface of box (center), right surface of box (corner) and edge of box.

[0096] Step 3: Extract the data values ​​of each target feature point and calculate the first Spearman rank correlation coefficient based on the data values.

[0097] The data values ​​extracted for each target feature point mentioned above include:

[0098] A series of target feature points are extracted by moving a local window simultaneously in the horizontal and vertical directions in the measured image or the first simulated image.

[0099] Extract the data value corresponding to the pixel of each target feature point one by one;

[0100] The data values ​​corresponding to a series of target feature points are summarized.

[0101] The first Spearman rank correlation coefficient calculated based on the data values ​​above includes:

[0102] The measured image is represented as X i Arrange the measured images X in order according to their data values. i A series of target feature points;

[0103] Assign a grade value R(x) to each of the series of target feature points. i Thus, we obtain R(x). i =1,2,3,…,n;

[0104] The first simulated image is represented as Y.i Arrange the first simulation image Y in sequence according to the data values ​​of the first simulation image. i A series of target feature points;

[0105] A series of target feature points in the measured image correspond one-to-one with a series of target feature points in the first simulated image, and a level value R(y) is assigned to the series of target feature points in the first simulated image. i That is, the first simulated image Y i Sort value among all values;

[0106] Calculate the target feature point ranking value R(x) in the measured image. i The target feature point level value R(y) in the first simulated image. i Square of the difference

[0107] The first Spearman rank correlation coefficient is calculated using the following formula;

[0108]

[0109] In the formula, ρ1 is the first Spearman rank correlation coefficient. Let R(x) be the value of R(x). i With R(y) i The square of the difference, where n is the number of target feature points in the series.

[0110] Please see Figure 5 , Figure 5 Provided for embodiments of the present invention Figure 4 A sorted table of target feature points based on their pixel values. Figure 5 The background feature point level of the measured image is 8, the water inlet feature point level of the measured image is 2, the top surface feature point level of the box in the measured image is 1, the left surface (center) feature point level of the box in the measured image is 5, the left surface (corner) feature point level of the box in the measured image is 6, the right surface (center) feature point level of the box in the measured image is 4, the right surface (corner) feature point level of the box in the measured image is 7, and the edge feature point level of the box in the measured image is 3. Figure 5The background feature point level of the first simulated image is 2, the water inlet feature point level of the first simulated image is 3, the top surface feature point level of the box in the first simulated image is 1, the left surface (center) feature point level of the box in the first simulated image is 6, the left surface (corner) feature point level of the box in the first simulated image is 8, the right surface (center) feature point level of the box in the first simulated image is 4, the right surface (corner) feature point level of the box in the first simulated image is 7, and the edge feature point level of the box in the first simulated image is 5. The error is calculated as follows: the error of the target feature point level value between the measured image and the target feature point level value of the first simulation image is 8-2=6, the error of the water inlet feature point level value is |2-3|=1, the error of the top surface feature point level value of the box is 1-1=0, the error of the left surface (center) feature point level value of the box is 5-6=1, the error of the left surface (corner) feature point level value of the box is |6-8|=2, the error of the right surface (center) feature point level value of the box is 4-4=0, the error of the right surface (corner) feature point level value of the box is 7-7=0, and the error of the edge feature point level value of the box is |3-5|=2.

[0111] Substitute the above error into the formula The first Spearman rank correlation coefficient was obtained as 0.4524.

[0112] Step 4: Optimize the relevant parameters of the simulation model based on the first Spearman rank correlation coefficient to obtain the improved simulation model.

[0113] In practical applications, the relevant parameters of the simulation model are optimized based on the first Spearman rank correlation coefficient of 0.4524. The relevant parameters of the simulation model include: object geometric information and relative position; material parameters: reflectivity, emissivity, heat capacity, etc.; scene parameters: temperature, light source, etc.; physical model: reflection model, shading model, atmospheric absorption, etc.

[0114] Step 5: Input the measured image into the improved simulation model to obtain the second simulation image. Based on the measured image and the second simulation image, obtain the second Spearman rank correlation coefficient. Optimize and improve the simulation model again to obtain a further improved simulation model.

[0115] The second Spearman rank correlation coefficient obtained above based on the measured image and the second simulated image includes:

[0116] The improved simulation model outputs a second simulated image corresponding to the input measured image;

[0117] Read the measured image and the second simulated image;

[0118] The measured image and the second simulated image are processed so that the pixels in the measured image and the second simulated image represent the same unit;

[0119] The Harris feature point detection method was used to extract a series of target feature points of interest from the measured image and the second simulated image, respectively.

[0120] A series of target feature points are extracted by moving a local window simultaneously in the horizontal and vertical directions on either the measured image or the second simulated image.

[0121] Extract the data value corresponding to the pixel of each target feature point one by one;

[0122] Summarize the data values ​​corresponding to a series of target feature points;

[0123] The measured image is represented as X i Arrange the measured images X in order according to their data values. i A series of target feature points;

[0124] Assign a grade value R(x) to each of the series of target feature points. i Thus, we obtain R(x). i =1,2,3,…,n;

[0125] The second simulated image is represented as Z. i Arrange the second simulation image Z in order according to the data values ​​of the second simulation image. i A series of target feature points;

[0126] A series of target feature points in the measured image correspond one-to-one with a series of target feature points in the second simulated image, and a level value R(Z) is assigned to the series of target feature points in the second simulated image. i That is, the second simulated image Z i Sort value among all values;

[0127] Calculate the target feature point ranking value R(x) in the measured image. i The target feature point level value R(Z) in the second simulated image. i Square of the difference

[0128] The second Spearman rank correlation coefficient is calculated using the following formula;

[0129]

[0130] In the formula, ρ2 is the second Spearman rank correlation coefficient. Let R(x) be the value of R(x). i With R(Z) i The square of the difference, where n is the number of target feature points in the series.

[0131] Please see Figure 7 , Figure 8 and Figure 9 , Figure 7 This is an illustration of a series of target feature points extracted from the measured image and the second simulated image, provided for an embodiment of the present invention. Figure 8 Provided for embodiments of the present invention Figure 7 A sorting table that sorts a series of target feature points according to their pixel values; Figure 9 This is a comparison result of the feature point level value error between the measured image and the second simulated image provided in this embodiment of the invention; Figure 8 Substituting the mean error into the formula The second Spearman rank correlation coefficient was obtained as 0.9048, while the ideal value for the Spearman rank correlation coefficient is 1.

[0132] Step 6: Repeat the above process until the Nth Spearman rank correlation coefficient approaches the ideal value.

[0133] Repeat the above process until the Nth Spearman rank correlation coefficient approaches the ideal value, including:

[0134] The first Spearman correlation coefficient to the Nth Spearman rank correlation coefficient all indicate the direction of the correlation between X and Y;

[0135] If X increases and Y tends to increase, then the Spearman correlation coefficient is positive; if X increases and Y tends to decrease, then the Spearman correlation coefficient is negative; if the Spearman correlation coefficient is 0, it indicates that Y has no trend when X increases.

[0136] If X and Y are positively correlated, the Spearman correlation coefficient is 1; if X and Y are negatively correlated, the Spearman correlation coefficient is -1.

[0137] This invention uses the Spearman correlation coefficient, a statistical indicator, and extracts image feature points through the Harris feature point detection algorithm. It performs a non-parametric measurement of the hierarchical correlation of images, which can accurately express the similarity between simulated images and real images. It has the advantages of being less affected by outliers and not being affected by unstructured image distortion, thus providing a reference for the verification of simulation models.

[0138] Example 2:

[0139] Based on the photoelectric simulation image processing method based on Spearman's rank correlation coefficient disclosed in Example 1, please refer to... Figure 2 , Figure 2 This invention provides a schematic diagram of a photoelectric simulation image processing system based on Spearman's rank correlation coefficient, comprising:

[0140] The input unit inputs the measured image into the simulation model, and the simulation model outputs the first simulation image; the input unit inputs the measured image into the improved simulation model, and the improved simulation model outputs the second simulation image; the input unit inputs the measured image into the improved simulation model again, and the improved simulation model again outputs the third simulation image.

[0141] The processing unit processes the measured image and the first simulated image to obtain the first Spearman rank correlation coefficient; the processing unit processes the measured image and the second simulated image to obtain the second Spearman rank correlation coefficient; the processing unit processes the measured image and the third simulated image to obtain the third Spearman rank correlation coefficient.

[0142] The adjustment unit optimizes the relevant parameters of the simulation model based on the first Spearman rank correlation coefficient to obtain an improved simulation model; the adjustment unit optimizes the relevant parameters of the improved simulation model based on the second Spearman rank correlation coefficient to obtain a further improved simulation model.

[0143] In practical applications, the above processing unit uses the following formula to obtain the Spearman rank correlation coefficient;

[0144]

[0145] In the formula, ρ is the Spearman rank correlation coefficient. Let R(x) be the value of R(x). i With R(Z) i The square of the difference, where n is the number of target feature points in the series, R(x) i R(y) represents the measured feature point level value of the target in the image. i Let R(Z) be the target feature point level value in the first simulated image. i The value of the target feature point level in the second simulated image.

[0146] The above processing unit processes the measured and simulated images as follows: read the two images respectively; convert the two images to the same unit, if they are the same, no conversion is needed; use the Harris feature point detection algorithm to extract a series of target feature points or feature regions of interest from the two images respectively; extract the values ​​of feature points in the image one by one, and if it is a feature region, calculate the average value; calculate the Spearman correlation coefficient based on the extracted data.

[0147] The core of the Harris algorithm in this embodiment of the invention is to use a local window to move on the image and determine corner points based on the changes in grayscale. If there is a large change in the grayscale value within the window, then there is a corner point in the area where the window is located. The Harris algorithm includes the following three steps: (1) When the window moves simultaneously in the horizontal and vertical directions, calculate the change in pixel value E(x,y) within the window; (2) Calculate a corner response function R for each window; (3) Perform thresholding on the function. If R>threshold, it means that the window corresponds to a corner feature.

[0148] In this embodiment of the invention, if there are a total of n marked feature points, (1) arrange the real image data X in order. i Each is assigned a grade value R(x). i =1,2,3…,n; (2) Arrange the simulated image data Y i This ensures that each feature point corresponds one-to-one and is assigned a level value R(y). i That is, Y i (3) Calculate R(x) among all values. i With R(y) i Square of the difference The correlation coefficient of the grades is calculated according to the following formula:

[0149]

[0150] The Spearman correlation coefficient indicates the direction of the correlation between X and Y. If Y tends to increase as X increases, the Spearman correlation coefficient is positive. If Y tends to decrease as X increases, the Spearman correlation coefficient is negative. A Spearman correlation coefficient of 0 indicates that Y shows no trend as X increases. The Spearman correlation coefficient is 1 when X and Y are positively correlated and -1 when they are negatively correlated.

[0151] This invention employs the Spearman correlation coefficient to measure a finite number of feature points in an image, offering the advantage of being less affected by outliers. Because it measures only the monotonic relationship between variables rather than a strictly linear one, it has better applicability to specific scenarios compared to other correlation coefficients. Compared to histogram-based comparison methods, this method reflects the image structure, not just grayscale distribution information. Compared to structural similarity methods, this method is more adaptable to unstructured image distortions. Furthermore, since this technique uses a finite number of labeled data points for calculation, the computational load is extremely small, making it suitable for hypothesis testing in large-scale, specific scenarios.

[0152] The detailed descriptions listed above are merely specific descriptions of feasible embodiments of the present invention, and are not intended to limit the scope of protection of the present invention. All equivalent embodiments or modifications made without departing from the spirit of the present invention should be included within the scope of protection of the present invention.

[0153] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0154] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A photoelectric simulation image processing method based on Spearman's rank correlation coefficient, characterized in that, include: The first simulated image is obtained by inputting the measured image into the simulation model; Extract a series of target feature points from the measured image and the first simulated image; Extract the data values ​​of each of the target feature points, and calculate the first Spearman rank correlation coefficient based on the data values; The relevant parameters of the simulation model were optimized based on the first Spearman rank correlation coefficient to obtain the improved simulation model; The measured image is input into the improved simulation model to obtain the second simulation image. The second Spearman rank correlation coefficient is obtained based on the measured image and the second simulation image. The improved simulation model is then optimized again to obtain a further improved simulation model. Repeat the above process until the Nth Spearman rank correlation coefficient approaches the ideal value; The extraction of a series of target feature points from the measured image and the first simulated image includes: Read the measured image and the first simulated image; The measured image and the first simulated image are processed such that the pixels of the measured image and the first simulated image represent the same unit; The Harris feature point detection method was used to extract a series of target feature points of interest from the measured image and the first simulated image, respectively. The extraction of data values ​​for each of the target feature points includes: A series of target feature points are extracted by simultaneously moving the measured image or the first simulated image in both the horizontal and vertical directions using a local window. Extract the data value corresponding to the pixel of each target feature point one by one; The data values ​​corresponding to a series of target feature points are summarized.

2. The photoelectric simulation image processing method based on Spearman's rank correlation coefficient according to claim 1, characterized in that, The step of inputting the measured image into the simulation model to obtain the first simulation image includes: A simulation model is obtained by mathematically modeling the imaging link in the real world. The measured image is input into the simulation model, and the simulation model outputs the first simulation image corresponding to the measured image.

3. The photoelectric simulation image processing method based on Spearman's rank correlation coefficient according to claim 1, characterized in that, Harris feature point detection methods include: The local window is used to move simultaneously in the horizontal and vertical directions of the measured image or the first simulated image; Calculate the change in pixel values ​​within a local window If the pixel value changes If the change is less than a preset amount, the local window continues to move horizontally and vertically. If the pixel value change... If the change exceeds the preset amount, the corner response function R of the area where this local window is located is calculated. Thresholding is applied to the corner response function R, and the pixels corresponding to the local maxima of the corner response function R are extracted as target feature points. The local window is then moved further in the horizontal and vertical directions.

4. The photoelectric simulation image processing method based on Spearman's rank correlation coefficient according to claim 1, characterized in that, The calculation of the first Spearman rank correlation coefficient based on the data values ​​includes: The measured image is represented as The measured images are arranged in order according to their data values. A series of target feature points; Assign a level value to each of the target feature points. ,get ; The first simulated image is represented as The first simulation image is arranged in order according to its data values. A series of target feature points; A series of target feature points in the measured image correspond one-to-one with a series of target feature points in the first simulated image, and a level value is assigned to a series of target feature points in the first simulated image. That is, the first simulated image Sort value among all values; Calculate the target feature point level value of the measured image. Compared with the target feature point level value of the first simulated image Square of the difference ; The first Spearman rank correlation coefficient is calculated using the following formula; In the formula, The first Spearman rank correlation coefficient, for and The square of the difference, This represents the number of target feature points in a series.

5. The photoelectric simulation image processing method based on Spearman's rank correlation coefficient according to claim 1, characterized in that, The second Spearman rank correlation coefficient is obtained based on the measured image and the second simulated image, including: The improved simulation model outputs a second simulated image corresponding to the input measured image; Read the measured image and the second simulated image; The measured image and the second simulated image are processed such that the pixels of the measured image and the second simulated image represent the same unit; The Harris feature point detection method was used to extract a series of target feature points of interest from the measured image and the second simulated image, respectively. A series of target feature points are extracted by simultaneously moving the measured image or the second simulated image in both the horizontal and vertical directions using a local window; Extract the data value corresponding to the pixel of each target feature point one by one; Summarize the data values ​​corresponding to a series of target feature points; The measured image is represented as The measured images are arranged in order according to their data values. A series of target feature points; Assign a level value to each of the target feature points. ,get ; The second simulated image is represented as Arrange the second simulation image in order according to its data values. A series of target feature points; A series of target feature points in the measured image correspond one-to-one with a series of target feature points in the second simulated image, and a level value is assigned to a series of target feature points in the second simulated image. That is, the second simulated image Sort value among all values; Calculate the target feature point level value of the measured image. Compared with the target feature point level value of the second simulated image Square of the difference ; The second Spearman rank correlation coefficient is calculated using the following formula; In the formula, The second Spearman rank correlation coefficient, for and The square of the difference, This represents the number of target feature points in a series.

6. The photoelectric simulation image processing method based on Spearman's rank correlation coefficient according to claim 1, characterized in that, Repeat the above process until the Nth Spearman rank correlation coefficient approaches the ideal value, including: The first Spearman rank correlation coefficient to the Nth Spearman rank correlation coefficient all indicate the direction of correlation between X and Y; like Increase, If the trend is towards increasing, then the Spearman rank correlation coefficient is positive; if Increase, If the Spearman rank correlation coefficient tends to decrease, then it is negative; if the Spearman rank correlation coefficient is 0, then it indicates that when... When increasing There is no tendency; like and If there is a positive correlation, then the Spearman rank correlation coefficient is 1. and If there is a negative correlation, then the Spearman rank correlation coefficient is: .

7. A photoelectric simulation image processing system based on Spearman's rank correlation coefficient, characterized in that, include: The input unit inputs the measured image into the simulation model, and the simulation model outputs the first simulation image. The input unit inputs the measured image into the improved simulation model, and the improved simulation model outputs a second simulation image; the input unit inputs the measured image into the improved simulation model again, and the improved simulation model again outputs a third simulation image; The processing unit processes the measured image and the first simulated image to obtain a first Spearman rank correlation coefficient; the processing unit processes the measured image and the second simulated image to obtain a second Spearman rank correlation coefficient; the processing unit processes the measured image and the third simulated image to obtain a third Spearman rank correlation coefficient. An adjustment unit optimizes the relevant parameters of the simulation model based on a first Spearman rank correlation coefficient to obtain an improved simulation model; the adjustment unit further optimizes the relevant parameters of the improved simulation model based on a second Spearman rank correlation coefficient to obtain a further improved simulation model. The processing unit is specifically used to read the measured image and the first simulated image; process the measured image and the first simulated image so that the pixels of the measured image and the first simulated image represent the same unit; and use the Harris feature point detection method to extract a series of target feature points of interest from the measured image and the first simulated image respectively. The processing unit is specifically used to extract a series of target feature points by simultaneously moving a local window in the measured image or the first simulation image in both the horizontal and vertical directions; and to extract the data value corresponding to the pixel of each target feature point one by one. The data values ​​corresponding to a series of target feature points are summarized.

8. The photoelectric simulation image processing system based on Spearman's rank correlation coefficient according to claim 7, characterized in that, The processing unit uses the following formula to obtain the Spearman rank correlation coefficient; In the formula, The Spearman rank correlation coefficient. for and The square of the difference, The number of a series of target feature points. The measured image target feature point level value, The target feature point level value of the first simulated image. The value represents the target feature point level in the second simulated image.

Images