Method for designing a surgical illumination light source based on a color commercial imaging device

Abstract

The present application relates to a surgical illumination light source design method based on color commercial imaging equipment. The spectral reflectance reconstruction stage: extracting the characteristic color and corresponding spectral reflectance from the biological tissue multispectral image; the extracted biological tissue characteristic color and spectral reflectance data are established as a lookup table from the imaging equipment response value to the spectral reflectance, and the lookup table is used to generate the spectral reflectance image of the biological tissue according to the imaging equipment response value interpolation. The surgical illumination design stage: extracting the characteristic spectral reflectance sample from the reflectance image in the above stage; three optimal wave bands are selected by using the spectral wave band selection method. Three primary color LED light sources are selected according to the three optimal wave bands to combine a new light source to improve the visual clarity. The present application generates a multispectral image quickly with the help of color commercial imaging equipment, and analyzes the differences between different biological tissues from the spectral level to design a surgical LED light source that can improve the visual clarity of the biological tissue surface.

Description

Technical Field

[0001] This invention belongs to the field of biomedical photonics technology and relates to a surgical illumination source design method, specifically a surgical illumination source design method based on a color commercial imaging device. Background Technology

[0002] In the medical field, the quality of surgical lighting is crucial for the smooth execution of surgical procedures. High-quality lighting can effectively improve the clarity of biological tissues, enabling surgeons to accurately distinguish different tissue textures and greatly improving surgical efficiency. Surgical lighting requires strict quality control and typically uses thermal radiation sources, which inevitably leads to increased ambient temperature during surgery when the lights are on. Furthermore, existing surgical lighting sources only provide good visibility and are not designed based on the spectral characteristics of biological tissues themselves. To address these issues, some researchers have proposed adjusting the LED light source.

[0003] LED light sources possess excellent characteristics such as energy efficiency and high adjustability. Based on these characteristics, different LEDs can be combined to achieve optimal lighting performance. This process requires the use of spectral information from the surface of biological tissue, and capturing multispectral images is an effective way to obtain this information. However, in existing methods, capturing multispectral images of biological tissue requires the use of a multispectral imaging system (MSIS), but such devices consist of complex optics, which makes the imaging process time-consuming and expensive.

[0004] In order to obtain LED surgical lighting sources with optimal visual clarity, a portable and rapid method is needed to acquire spectral reflectance information of biological tissues, so as to analyze the differences between different tissues at the spectral level, design LED surgical lighting sources that can enhance visual clarity, and achieve real-time control. Summary of the Invention

[0005] The first objective of this invention is to address the problems of time-consuming and complex acquisition of biological tissue spectral reflectance information in existing surgical light source optimization methods and the inability of existing surgical lighting sources to provide the best visual clarity. This invention proposes a surgical lighting source design method based on a commercial color imaging device. This method enables the rapid acquisition of spectral reflectance images of biological tissues using only a commercial color imaging device, without relying on complex and expensive multispectral imaging equipment. Furthermore, it allows for the analysis of differences between different tissues at the spectral level to design surgical lighting that improves the clarity of biological tissue surfaces.

[0006] This invention discloses a surgical lighting design method based on a color commercial imaging device, comprising a spectral reflectance reconstruction stage and a surgical lighting design stage;

[0007] The spectral reflectance reconstruction stage is used to reconstruct the spectral reflectance image of biological tissue based on a commercial color imaging device; the commercial color imaging device is a color camera, preferably a commercial color digital camera; specifically:

[0008] Step 1.1: Use a multispectral imaging system (MSIS) to acquire spectral reflectance images of biological tissues, and use formula (1) to map them to the spectral sensitivity function space of a commercial camera to obtain the RGB value corresponding to each pixel.

[0009] D = MR (1)

[0010] Where D is a 3xn real matrix representing the set of RGB space pixels, where each column represents a 3-channel camera RGB value; M is a 3xt real matrix representing the transformation matrix from t spectral channels to RGB channels; and R is a txn real matrix, which is the vectorized representation of spectral reflectance.

[0011] Step 1.2: Based on the RGB values ​​of commercial cameras, use a clustering algorithm to cluster the RGB values ​​of pixels to obtain a dataset of RGB characteristic colors and spectral reflectance of biological tissues, and divide it into training samples and test samples;

[0012] Preferably, the clustering algorithm chosen is either K-means clustering based on Euclidean distance or Mean shift clustering. The spectral reflectance image can be converted to the CIELAB color space before clustering the LAB values.

[0013] Step 1.3: Based on the lattice regression method, a lookup table (LUT) is established to map the RGB values ​​of the training samples to metamerism black, using the RGB values ​​and corresponding spectral reflectance. The specific process is as follows:

[0014] Step 1.3.1: Decompose the spectral reflectance of the training samples into metamerism b using formulas (2), (3), and (4).

[0015] A = L.*S (2)

[0016] R = A (A T A) -1 A T (3)

[0017] b=(IR)r (4)

[0018] Where L is a vector with t rows and 1 column, which is the vector representation of the spectral power distribution of the light source; S is a matrix with t rows and 3 columns, which is the matrix representation of the spectral sensitivity function of the three-channel camera; * indicates multiplication of each row; A is a matrix with t rows and 3 columns. Then the R matrix can be calculated from the A matrix using formula (3).

[0019] Step 1.3.2: Determine the range based on the maximum and minimum values ​​of the three dimensions of RGB of the training sample, and then manually divide each dimension by average. A total of m vertices can be obtained, which are called lattice points.

[0020] Step 1.3.3: Suppose there are n training sample pairs {d} i ,b i}, i=1,…,n. d represents the RGB response vector of a commercial camera, which is 3 rows and n columns; b represents the vector representation of its corresponding metameris black, which is t rows and n columns.

[0021] Therefore, for each training sample, the camera response d i A set of weights {w} can be determined i,j Each training sample corresponding to metamerism can be linearly represented by equation (5):

[0022]

[0023] Where {s j}, j=1,…,m represents the metamerism black corresponding to the m lattice points in the lookup table.

[0024] Step 1.3.4: The regression error between the real training samples and the interpolated estimated training samples can be written as shown in equation (6), and a second-order difference smoothing constraint is added. The entire metamerism black reconstruction model is as follows:

[0025]

[0026] Where s h ,s j ,s l This represents the metamerism black corresponding to three adjacent lattice points in dimension k in the lookup table. The main function of the regularization parameter α (>0) is to balance the solution accuracy and smoothness, and it needs to be set manually.

[0027] Step 1.3.5: For the RGB response of the test camera that needs to be reconstructed, the metamerism black lookup table established by the model in step 1.3.4 can be used to interpolate and reconstruct the metamerism black through equation (7).

[0028]

[0029] Where b' represents the metamerism black reconstructed from the test sample, a j This represents the interpolation weighting coefficients for reconstructing RGB values ​​in RGB space and the nearest 8 lattice points. This indicates the metameritic black corresponding to the lattice point in the metameritic black lookup table established in step 1.3.4.

[0030] Step 1.4: Calculate the basic stimulus spectrum of the test sample using formula (8).

[0031] r'=A(A T A) -1 d test (8)

[0032] Where d test represents the RGB value vector matrix of the test sample. r' represents the basic stimulus spectrum of the test sample.

[0033] Step 1.5: Use formula (9) to combine metamerism black and basic stimulus spectrum to reconstruct spectral reflectance.

[0034]

[0035] in This represents the reconstructed spectral reflectance.

[0036] The surgical lighting design stage utilizes the spectral reflectance obtained from the spectral reflectance reconstruction stage to process and obtain a new lighting source; specifically:

[0037] Step 2.1: Use a clustering algorithm to cluster the spectral reflectance images of biological tissues obtained in the spectral reflectance reconstruction stage, and extract n feature spectral reflectance samples.

[0038] As preferred, the clustering algorithms selected are the K-means clustering algorithm and the Mean shift clustering algorithm, which are based on Euclidean distance.

[0039] Step 2.2: Divide the wavelength range of spectral reflectance into 3 regions. These 3 regions mainly represent the wavelength ranges of red, green and blue. The specific wavelength ranges of the three regions can be selected according to the actual situation.

[0040] Preferably, within the visible light range of 380nm-780nm, the blue region band can be selected from 380nm-500nm, the green band from 500nm-600nm, and the red band from 600nm-780nm.

[0041] Step 2.3: Using the spectral band selection algorithm, calculate the information content of each band of the n feature spectral reflectance extracted in Step 2.2, and sort the bands in each region according to the amount of information. Select the three optimal bands for each region based on the band with the highest information content.

[0042] As a preferred method, variance, standard deviation, mean deviation, and information entropy are selected for calculating the information content of each band.

[0043] Step 2.4: Based on the three bands selected in Step 2.3, select LED light sources with center wavelengths corresponding to the three bands. If no LED light source with the peak wavelength is available, select three LED light sources that contain the three bands respectively.

[0044] Step 2.5: Using formula (10), calculate the tristimulus values ​​of the four light source colors for the three LED light sources, CIED65 light source and standard chromaticity observer function obtained in step 2.4.

[0045] p=kQl (10)

[0046]

[0047] Where p represents the tristimulus value vector, which is a 3x1 vector. k is the scaling factor, a scalar. Q is the matrix representation of the three channels of the standard chromaticity observer function, which is a 3xn matrix. l is the vector representation of the spectral power distribution of the light source, which is an nx1 vector. q represents the standard chromaticity observer function. The vector representation of a channel is a 1-row, n-column vector.

[0048] Preferably, the CIE standard observer function can be selected from CIE1931 2°, CIE1964 10°, CIE2006 2°, or CIE2006 10° standard observers.

[0049] Step 2.6: Based on the tristimulus values ​​of the four light source colors obtained in Step 2.5, convert them to CIE chromaticity diagram coordinate representation.

[0050] As a preferred option, the CIE chromaticity diagram can be either the CIE1931 xy chromaticity diagram or the CIE1964 xy chromaticity diagram.

[0051] Step 2.7: Using the chromaticity coordinates of the four light source colors calculated in Step 2.6, determine whether the chromaticity coordinates of the CIED65 light source color are contained within the triangle formed by the chromaticity coordinates of the three LED light source colors.

[0052] As a preferred method, the area method or the vector cross product method can be used to determine whether a point exists in the triangle formed by the three vertices.

[0053] Step 2.8: If the determination in Step 2.7 is yes, then proceed to Step 2.9. If the determination in Step 2.7 is no, then exclude the band with the least information content among the three bands, and select a second-best band from the region to which the excluded band belongs, repeating Steps 2.4-2.8. The second-best band is the band with the fourth-highest information content.

[0054] Step 2.9: The three LED light sources obtained from the final iteration above can be combined to fit white light. Using formula (12) with the tristimulus values ​​of the CIED65 light source color as a reference, the intensity coefficients of the three LED light sources are calculated.

[0055] β=Q + p d65 (12)

[0056] Where β is the intensity coefficient of the three LED light sources, which is a 3x1 vector. Q represents the tristimulus value matrix of the three LED light source colors, which is a 3x3 vector, with each column representing a tristimulus value vector of one LED light source color. p d65 This is a vector of tristimulus values ​​for the CIED65 light source color, arranged in 3 rows and 1 column. '+' indicates taking the pseudo-inverse of the matrix.

[0057] Step 3.0: Combine the intensity coefficient vector β obtained in step 2.8 into a new light source using formula (13).

[0058] l=Sβ (13)

[0059] Where l is a vector representation of the spectral power distribution of the new light source, which is a vector with t rows and 1 column. S is a vector representation of the spectral power distribution of the three LED light sources, which is a vector with t rows and 3 columns.

[0060] A second object of the present invention is to provide an electronic device including a processor and a memory, the memory storing machine-executable instructions executable by the processor, the processor executing the machine-executable instructions to implement the method.

[0061] A third object of the present invention is to provide a machine-readable storage medium storing machine-executable instructions that, when invoked and executed by a processor, cause the processor to implement the method described herein.

[0062] The beneficial effects of this invention are:

[0063] It is possible to acquire multispectral images of biological tissues using a single commercial digital camera without relying on multispectral imaging equipment. Furthermore, based on the spectral characteristics of biological tissues, a surgical illumination source with improved visual clarity is designed using LED light sources of three primary colors. Attached Figure Description

[0064] Figure 1 This is a flowchart of the surgical lighting design method based on RGB images from a color commercial camera in this invention.

[0065] Figure 2This is a schematic diagram of the process of creating a lookup table (LUT) in this invention. Detailed Implementation

[0066] To more clearly demonstrate the technical implementation of the present invention, the invention will be further described in detail below with reference to the accompanying drawings and examples.

[0067] according to Figure 1 This invention provides a surgical illumination source design method based on a color commercial imaging device, mainly comprising two stages: spectral reflectance reconstruction and surgical illumination source design. The first stage includes: extracting feature colors and their corresponding spectral reflectance from acquired multispectral images of biological tissue using a clustering algorithm, and generating a spectral reflectance database; decomposing the spectral reflectance of training samples into metameritic black; establishing a lookup table (LUT) from camera response values ​​(RGB) to metameritic black using the extracted biological tissue feature colors and metameritic black data; and using this lookup table to interpolate the RGB values ​​of the camera to generate the metameritic black of the biological tissue. Finally, combining this with the corresponding basic stimulus spectrum, the final spectral reflectance is generated. The second stage includes: extracting feature spectral reflectance samples from the reflectance image obtained in the first stage using a clustering method; and selecting three primary color LED light sources using a multispectral band selection method for linear combination to optimize a new light source, thereby increasing the color difference among all feature samples under this light source illumination, thus improving organ visibility during surgery.

[0068] This invention discloses a surgical lighting design method based on RGB images from a color commercial camera, comprising a spectral reflectance reconstruction stage and a surgical lighting design stage;

[0069] The spectral reflectance reconstruction stage is used to reconstruct the spectral reflectance image of biological tissue (taking pork tissue as an example) based on the RGB response values ​​of a commercial camera; specifically:

[0070] Step 1.1: Use a multispectral imaging system (MSIS) to acquire spectral reflectance images of pork tissue, and use formula (1) to map them to the spectral sensitivity function space of a commercial camera to obtain the RGB value corresponding to each pixel.

[0071] D = MR (1)

[0072] Where D is a 3x3 real matrix representing the set of pixels in RGB space, with each column representing a 3-channel camera RGB value; M is a 3x3 real matrix representing the transformation matrix from t spectral channels to RGB channels; R is a tx3 real matrix representing the vectorized representation of spectral reflectance. Here, t is typically 31, because the spectral range and resolution are 400-700nm, and spectral intervals of 10nm are sufficient to generate effective color information.

[0073] Step 1.2: Based on the RGB values ​​of commercial cameras, use a clustering algorithm to cluster the RGB values ​​of pixels in multiple images of pork tissue spectral reflectance to obtain a dataset of RGB feature colors and spectral reflectance of pork tissue, and divide it into training samples and test samples.

[0074] The clustering algorithm chosen is either K-means clustering or Mean shift clustering, based on Euclidean distance. The spectral reflectance image can be converted to the CIELAB color space before clustering the LAB values.

[0075] Step 1.3: Based on the lattice regression method, a lookup table (LUT) is established to map the RGB values ​​of the training samples to metamerism black, using the RGB values ​​and corresponding spectral reflectance. The specific process is as follows:

[0076] Step 1.3.1: Decompose the spectral reflectance of the training samples into metamerism b using formulas (2), (3), and (4).

[0077] A = L.*S (2)

[0078] R = A(A T A) -1 A T (3)

[0079] b=(IR)r (4)

[0080] Where L is a vector with t rows and 1 column, which is the vector representation of the spectral power distribution of the light source; S is a matrix with t rows and 3 columns, which is the matrix representation of the spectral sensitivity function of the three-channel camera; * indicates multiplication of each row; A is a matrix with t rows and 3 columns. Then the R matrix can be calculated from the A matrix using formula (3).

[0081] Step 1.3.2: Determine the range based on the maximum and minimum values ​​of the RGB dimensions of the training samples, and then manually average each dimension to obtain a total of m vertices (called lattice points, e.g., ...). Figure 2 (Right) Evenly distributed points.

[0082] Step 1.3.3: Suppose there are n training sample pairs {d} i ,bi}, i=1,…,n. d represents the RGB response vector of a commercial camera, which is 3 rows and n columns; b represents the vector representation of its corresponding metameris black, which is t rows and n columns.

[0083] Therefore, for each training sample, the camera response d i A set of weights {w} can be determined i,j Each training sample corresponding to metamerism can be linearly represented by equation (5):

[0084]

[0085] Where {s j}, j=1,…,m represents the metamerism black corresponding to the m lattice points in the lookup table.

[0086] Step 1.3.4: The regression error between the real training samples and the interpolated estimated training samples can be written as shown in equation (9), and a second-order difference smoothing constraint is added. The entire metamerism black reconstruction model is as follows:

[0087]

[0088] Where s h ,s j ,s l This represents the metamerism black corresponding to three adjacent lattice points in dimension k in the lookup table. The main function of the regularization parameter α (>0) is to balance the solution accuracy and smoothness, and it needs to be set manually.

[0089] Step 1.3.5: For the RGB response of the test camera that needs to be reconstructed, the metamerism black lookup table established by the model in step 1.3.4 can be used to interpolate and reconstruct the metamerism black through equation (7).

[0090]

[0091] Where b' represents the metamerism black reconstructed from the test sample, a j This represents the interpolation weighting coefficients for reconstructing RGB values ​​in RGB space and the nearest 8 lattice points. This indicates the metameritic black corresponding to the lattice point in the metameritic black lookup table established in step 1.3.4.

[0092] Step 1.4: Calculate the basic stimulus spectrum of the test sample using formula (8).

[0093] r'=A(A T A) -1 d test (8)

[0094] Where d testrepresents the RGB value vector matrix of the test sample. r' represents the basic stimulus spectrum of the test sample.

[0095] Step 1.5: Use formula (9) to combine metamerism black and basic stimulus spectrum to reconstruct spectral reflectance.

[0096]

[0097] in This represents the reconstructed spectral reflectance.

[0098] The surgical lighting design stage utilizes the spectral reflectance obtained from the spectral reflectance reconstruction stage to process and obtain a new lighting source; specifically:

[0099] Step 2.1: Use a clustering algorithm to cluster the spectral reflectance image of pork tissue obtained in the spectral reflectance reconstruction stage, and extract n (e.g., 50 samples) feature spectral reflectance samples.

[0100] The clustering algorithms selected are K-means clustering based on Euclidean distance and Mean shift clustering. Step 2.2: Divide the wavelength range of spectral reflectance into three regions, which mainly represent the wavelength ranges of red, green, and blue. For example, the blue region is selected from the wavelength range of 400-500nm, the green region from the wavelength range of 500-600nm, and the red region from the wavelength range of 600-700nm.

[0101] Within the visible light range of 380nm-780nm, the blue band has a selectable range of 380nm-500nm, the green band has a selectable range of 500nm-600nm, and the red band has a selectable range of 600nm-780nm.

[0102] Step 2.3: Using a spectral band selection algorithm, calculate the information content of each band of the 50 feature spectral reflectance extracted in Step 2.2, and sort the bands in each region according to the amount of information. Select the optimal band for each region based on the band with the highest information content.

[0103] The method for calculating the information content of each band is to select variance, standard deviation, and information entropy.

[0104] Step 2.4: Based on the three bands selected in Step 2.3, select LED light sources with center wavelengths corresponding to the three bands. If no LED light source with the peak wavelength is available, select three LED light sources that contain the three bands respectively.

[0105] Step 2.5: Using formula (10), calculate the tristimulus values ​​of the four light source colors for the three LED light sources, CIED65 light source and standard chromaticity observer function obtained in step 2.4.

[0106] t=kQs (10)

[0107]

[0108] Where t represents the tristimulus value vector, which is a 3x1 vector. k is the scaling factor, a scalar. Q is the matrix representation of the three channels of the standard chromaticity observer function, which is a 3xn matrix. s is the vector representation of the spectral power distribution of the light source, which is an nx1 vector. q represents the standard chromaticity observer function. The vector representation of a channel is a 1-row, n-column vector.

[0109] Preferably, the CIE standard observer function can select CIE193 12°, CIE196 4 10°, CIE3 200 62°, and CIE200 6 10° standard observers.

[0110] Step 2.6: Based on the tristimulus values ​​of the four light source colors obtained in Step 2.5, convert them to CIE chromaticity diagram coordinate representation.

[0111] The CIE chromaticity diagrams available are CIE1931 xy chromaticity diagram and CIE1964 xy chromaticity diagram.

[0112] Step 2.7: Using the coordinates of the four chromaticity diagrams calculated in Step 2.6, determine whether the chromaticity coordinates of the CIED65 light source color are contained within the triangle formed by the chromaticity coordinates of the three LED light source colors.

[0113] To determine whether a point exists within a triangle formed by its three vertices, the area method and the vector cross product method can be used.

[0114] Step 2.8: If the determination in Step 2.7 is yes, then proceed to Step 2.9. If the determination in Step 2.7 is no, then exclude the band with the least information among the three bands, and select a second-best band from the region to which the excluded band belongs, and repeat Steps 2.4-2.8.

[0115] Step 2.9: The three LED light sources obtained from the final iteration above can be combined to fit white light. Using formula (12) with the tristimulus values ​​of the CIED65 light source color as a reference, the intensity coefficients of the three LED light sources are calculated.

[0116] β=Q + x d65 (12)

[0117] Where β is the intensity coefficient of the three LED light sources, which is a 3x1 vector. Q represents the tristimulus value matrix of the three LED light source colors, which is a 3x3 vector, with each column representing a tristimulus value vector of one LED light source color. X d65 This is a vector of tristimulus values ​​for the CIED65 light source color, arranged in 3 rows and 1 column. '+' indicates taking the pseudo-inverse of the matrix.

[0118] Step 3.0: Combine the intensity coefficient vector β obtained in step 2.8 into a new light source using formula (13).

[0119] l=Sβ (13)

[0120] Where l is a vector representation of the spectral power distribution of the new light source, which is a vector with t rows and 1 column. S is a vector representation of the spectral power distribution of the three LED light sources, which is a vector with t rows and 3 columns.

Claims

1. A method for designing surgical illumination sources based on color commercial imaging equipment, characterized in that... This includes the spectral reflectance reconstruction stage and the surgical lighting design stage; The spectral reflectance reconstruction stage is used to reconstruct the spectral reflectance image of biological tissue based on the response value of commercial imaging equipment. Specifically: Step 1.1: Use the multispectral imaging system MSIS to acquire spectral reflectance images of biological tissues, and use formula (1) to map them to the spectral sensitivity function space of a commercial camera to obtain the RGB value corresponding to each pixel. (1) in It is a 3xn real matrix representing the set of pixels in RGB space, where each column represents a 3-channel RGB value from the camera; It is a 3x3 real matrix, representing the transformation matrix from t spectral channels to RGB channels; It is a t-row, n-column real matrix, which is the vectorized representation of spectral reflectance; Step 1.2: Based on the RGB values ​​of commercial cameras, use a clustering algorithm to cluster the RGB values ​​of pixels to obtain a dataset of RGB characteristic colors and spectral reflectance of biological tissues, and divide it into training samples and test samples; Step 1.3: Based on the lattice regression method, a lookup table (LUT) is established to map the RGB values ​​of the training samples to metamerism black, using the RGB values ​​and corresponding spectral reflectance. The specific process is as follows: Step 1.3.1: Decompose the spectral reflectance of the training samples into metamerism using formulas (2), (3), and (4). ; (2) (3) (4) in Let be a vector of row t and column 1, which is a vector representation of the spectral power distribution of the light source. It is a t-row, 3-column matrix, which is the matrix representation of the spectral sensitivity function of the three-channel camera; This indicates multiplication by each row; Let t be a matrix with 3 rows and t columns; then A matrix can be derived from... The matrix is ​​calculated using formula (3); Step 1.3.2: Determine the range based on the maximum and minimum values ​​of the three dimensions of RGB of the training samples, and then manually average each dimension to obtain a total of m vertices, called lattice points; Step 1.3.3: There are a training sample pairs , , This represents the RGB response vector of a commercial camera, arranged in 3 rows and 'a' columns. The vector representation of its corresponding metamerism black is t rows and a columns; Camera response for each training sample Determine a set of weights Each training sample corresponds to metamerism black, which is linearly represented by equation (5): ， (5) in , This represents the metamerism black corresponding to the m lattice points in the lookup table; Step 1.3.4: The regression error between the real training samples and the interpolated estimated training samples can be written as shown in equation (6), and a second-order difference smoothing constraint is added. The entire metamerism black reconstruction model is as follows: (6) in This represents the metamerism black corresponding to three adjacent lattice points in dimension k in the lookup table; Represents the regular expression parameter. ; Step 1.3.5: For the RGB response of the test camera that needs to be reconstructed, the metamerism black lookup table established by the model in step 1.3.4 can be used to interpolate and reconstruct the metamerism black through equation (7); (7) in This represents the metamerism black reconstructed from the test sample. This represents the interpolation weighting coefficients for reconstructing RGB values ​​and the nearest 8 lattice points in RGB space. This represents the metamerism corresponding to the lattice point in the metamerism lookup table established in step 1.3.4; Step 1.4: Calculate the basic stimulus spectrum of the test sample using formula (8); (8) in This represents the RGB value vector matrix of the test sample. This represents the basic stimulus spectrum of the test sample; Step 1.5: Use formula (9) to combine metamerism black and the basic stimulus spectrum to reconstruct the spectral reflectance; (9) in This represents the reconstructed spectral reflectance; The surgical lighting design stage is used to obtain the spectral reflectance of biological tissues from the spectral reflectance reconstruction stage, and then process it to obtain a new lighting source.

2. The method according to claim 1, characterized in that... The surgical lighting design phase specifically includes: Step 2.1: Use a clustering algorithm to cluster the spectral reflectance images of biological tissues obtained in the spectral reflectance reconstruction stage, and extract... One characteristic spectral reflectance sample; Step 2.2: Divide the wavelength range of spectral reflectance into 3 regions. These 3 regions mainly represent the wavelength ranges of red, green and blue. The specific wavelength ranges of the three regions can be selected according to the actual situation. Step 2.3: Use the spectral band selection algorithm to calculate the values ​​extracted in Step 2.

2. The information content of each band of the characteristic spectral reflectance; sort the bands of each region according to the information content, and select the top three bands with the largest information content; Step 2.4: Based on the three bands selected in Step 2.3, select LED light sources with center wavelengths of the three bands respectively. If there is no LED light source with the center wavelength, select three LED light sources that contain the three bands respectively. Step 2.5: Using formula (10), calculate the tristimulus values ​​of the four light source colors for the three LED light sources, CIED65 light source and standard chromaticity observer function obtained in step 2.4 respectively; (10) (11) Where p represents the tristimulus value vector, which is a 3-row, 1-column vector; k is the proportionality coefficient, which is a scalar; It is a matrix representation of the three channels of the standard chromaticity observer function, which is a 3xn matrix; It is a vector representation of the spectral power distribution of the light source, which is an n-row, 1-column vector; Represents the standard chromaticity observer function The vector representation of a channel is a vector with 1 row and n columns; Step 2.6: Based on the tristimulus values ​​of the four light source colors obtained in Step 2.5, convert them to CIE chromaticity diagram coordinate representation; Step 2.7: Using the coordinates of the four chromaticity diagrams calculated in Step 2.6, determine whether the chromaticity coordinates of the CIED65 light source color are contained within the triangle formed by the chromaticity coordinates of the three LED light source colors; Step 2.8: If the determination in step 2.7 is yes, then continue to step 2.9; if the determination in step 2.7 is no, then exclude the band with the least information among the three bands, and select a second-best band from the region to which the excluded band belongs, and repeat steps 2.4-2.

8. Step 2.9: Use the three LED light sources finally determined after the iterations in Steps 2.7-2.8 to combine them to fit white light; use formula (12) to calculate the intensity coefficients of the three LED light sources based on the tristimulus values ​​of the CIED65 light source color; (12) in Let be the intensity coefficients of the three LED light sources, which are a 3x1 vector; The matrix represents the tristimulus values ​​of the three LED light source colors. It is a 3x3 vector, with each column representing a tristimulus value vector of one LED light source color. This is a vector of tristimulus values ​​for the CIED65 light source color, which is a 3x1 vector; '+' indicates taking the pseudo-inverse of the matrix; Step 3.0: Use formula (13) to combine them into a new light source; (13) in It is a vector representation of the spectral power distribution of the new light source, which is a vector with t rows and 1 column; Let be a vector representation of the spectral power distribution of the three LED light sources, which is a vector with t rows and 3 columns.

3. The method according to claim 2, characterized in that... In steps 1.2 and 2.1, the clustering algorithm selected is either the K-means clustering algorithm based on Euclidean distance or the Mean shift clustering algorithm.

4. The method according to claim 2, characterized in that... In step 2.2, within the visible light range of 380nm-780nm, the blue region band selection range is 380nm-500nm, the green band selection range is 500nm-600nm, and the red band selection range is 600nm-780nm.

5. The method according to claim 2, characterized in that... In step 2.3, the method for calculating the information content of each band is to select variance, standard deviation, mean deviation, and information entropy.

6. The method according to claim 2, characterized in that... In step 2.5, the CIE standard observer function is selected from CIE193 12°, CIE1964 10°, CIE2006 2°, or CIE2006 10° standard observers.

7. The method according to claim 2, characterized in that... In step 2.6, the CIE chromaticity diagram can be selected from the CIE1931 xy chromaticity diagram or the CIE1964 xy chromaticity diagram.

8. The method according to claim 2, characterized in that... In step 2.7, the method to determine whether a point exists in the triangle formed by the three vertices is the area method or the vector cross product method.

9. An electronic device, characterized in that, The method includes a processor and a memory, the memory storing machine-executable instructions that can be executed by the processor, the processor executing the machine-executable instructions to implement the method of any one of claims 1-8.

10. A machine-readable storage medium, characterized in that, The machine-readable storage medium stores machine-executable instructions that, when invoked and executed by a processor, cause the processor to perform the method described in any one of claims 1-8.

Images