Fluorescence hyperspectral unmixing method, device and system in membrane pollution forming process

Abstract

A fluorescence hyperspectral unmixing method, device and system in a membrane pollution formation process, the method comprising: acquiring a fluorescence hyperspectral vector of a target membrane within a predetermined time period and a predetermined spectral range, using a strong robustness minimum volume unmixing analysis method to perform transmission projection on the fluorescence hyperspectral vector, and performing target function soft constraint unmixing calculation to obtain a target function soft constraint unmixing result; the method comprises the following steps of: obtaining end member spectral data according to a preset time period, determining a membrane pollution forming process of a corresponding pollutant in the preset time period according to the end member spectral data to obtain a plurality of end member spectral data, and respectively executing the following operations on each piece of end member spectral data: determining the membrane pollution forming process of the corresponding pollutant in the preset time period according to the end member spectral data, by using the transmission projection method, the influence of a pixel-related scale factor is eliminated, the direction of an original spectral vector is maintained, angular displacement in the projection process is avoided, the projection effect is improved, the calculation process is high in robustness, good in flexibility, high in unmixing efficiency and more accurate in unmixing result, and the formation process of membrane pollution can be presented more truly.

Description

Technical Field

[0001] This article relates to the field of fluorescence hyperspectral detection technology, and in particular to a fluorescence hyperspectral unmixing method, device and system for a membrane contamination formation process. Background Art

[0002] At present, when using hyperspectral for membrane pollution analysis, the minimum variance spectrum analysis (MVSA) algorithm is mainly used to achieve spectral unmixing. The minimum variance spectrum analysis uses orthogonal projection when obtaining spectral vector data, which changes the direction of the spectral vector and introduces angular displacement, resulting in large errors in the obtained spectral vector data; at the same time, the minimum variance spectrum analysis algorithm model needs to strictly meet the hard constraint of "abundance coefficient is non-negative" when calculating the simplex. However, in actual hyperspectral images, due to the presence of noise and outliers, the simplex expands in space, and the position of the vertices of the simplex in the signal subspace is very different from the actual vertex position, resulting in large unmixing errors, irregular unmixing waveforms, and low robustness. The hard constraints in the calculation process also make the algorithm inflexible and slow to unmix. Therefore, there is an urgent need for a membrane pollution hyperspectral analysis method with strong robustness, good accuracy, and high unmixing efficiency. Summary of the invention

[0003] The present application provides a method, device and system for unmixing fluorescence hyperspectral in the process of membrane pollution formation, which improves the projection effect, has strong robustness, good flexibility, high unmixing efficiency, more accurate unmixing results, and can more realistically present the process of membrane pollution formation.

[0004] On the one hand, the embodiment of the present application provides a method for unmixing fluorescence hyperspectral in a membrane pollution formation process, comprising: Acquire a fluorescence hyperspectral vector of a target film within a predetermined time period and a predetermined spectral range; A strong robust minimum volume unmixing analysis method is used to perform transmission projection on the fluorescence hyperspectral vector, and an objective function soft constraint unmixing calculation is performed to obtain end-member spectral data, wherein the end-member spectral data is multiple, and each end-member spectral data corresponds to a pollutant; The following operations are performed on each end member spectrum data respectively: a membrane pollution formation process of the corresponding pollutant in the predetermined time period is determined according to the end member spectrum data.

[0005] On the other hand, an embodiment of the present application also provides a fluorescence hyperspectral unmixing device for a film contamination formation process, comprising: a field programmable gate array, the field programmable gate array comprising a processor and a memory; the memory is used to store a fluorescence hyperspectral unmixing program for a film contamination formation process; the processor is used to read the fluorescence hyperspectral unmixing program for a film contamination formation process, and perform the fluorescence hyperspectral unmixing method for a film contamination formation process as described in the above embodiment.

[0006] On the other hand, the embodiment of the present application further provides a fluorescence hyperspectral unmixing system for a membrane pollution formation process, comprising: a fluorescence hyperspectral imager and a fluorescence hyperspectral unmixing device for a membrane pollution formation process as described in the above embodiment; The fluorescence hyperspectral imager is used to collect the fluorescence hyperspectral image continuous analog quantity of the target film in a predetermined time period and a predetermined spectral range, and send the fluorescence hyperspectral image continuous analog quantity to the fluorescence hyperspectral unmixing device of the film pollution formation process; The device for unmixing fluorescence hyperspectral during the formation of membrane pollution is used to carry out the method for unmixing fluorescence hyperspectral during the formation of membrane pollution as described in the above embodiment.

[0007] Compared with the related art, the fluorescence hyperspectral unmixing method, device and system for the membrane pollution formation process in the embodiment of the present application not only eliminates the influence of pixel-related proportional factors by the transmission projection method, but also maintains the direction of the original spectral vector, avoids angular displacement in the projection process, and significantly improves the projection effect. In addition, a highly robust minimum volume unmixing algorithm based on soft constraints is adopted, and the calculation process has strong robustness, good flexibility, high unmixing efficiency, and more accurate unmixing results, which can more realistically present the membrane pollution formation process.

[0008] Other features and advantages of the present application will be described in the following description, and partly become apparent from the description, or be understood by implementing the present application. Other advantages of the present application can be realized and obtained by the schemes described in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The accompanying drawings are used to provide an understanding of the technical solution of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the technical solution of the present application and do not constitute a limitation on the technical solution of the present application.

[0010] Figure 1 This is a flow chart of a fluorescence hyperspectral unmixing method for a membrane pollution formation process according to an embodiment of the present application; Figure 2 It is a flow chart of a fluorescence hyperspectral unmixing method for a membrane pollution formation process specifically exemplified in an embodiment of the present application; Figure 3Schematic diagram of a fluorescence hyperspectral unmixing device in a membrane pollution formation process according to an embodiment of the present application; Figure 4 It is a schematic diagram of a fluorescence hyperspectral unmixing system in the membrane pollution formation process of an embodiment of the present application. DETAILED DESCRIPTION

[0011] The present application describes multiple embodiments, but the description is exemplary rather than restrictive, and it is obvious to those skilled in the art that there may be more embodiments and implementations within the scope of the embodiments described in the present application. Although many possible feature combinations are shown in the drawings and discussed in the specific embodiments, many other combinations of the disclosed features are also possible. Unless specifically limited, any feature or element of any embodiment may be used in combination with any other feature or element in any other embodiment, or may replace any other feature or element in any other embodiment.

[0012] The present application includes and contemplates combinations of features and elements known to those of ordinary skill in the art. The embodiments, features and elements disclosed in the present application may also be combined with any conventional features or elements to form a unique invention scheme defined by the claims. Any features or elements of any embodiment may also be combined with features or elements from other invention schemes to form another unique invention scheme defined by the claims. Therefore, it should be understood that any feature shown and / or discussed in the present application may be implemented individually or in any appropriate combination. Therefore, except for the limitations made according to the attached claims and their equivalents, the embodiments are not subject to other restrictions. In addition, various modifications and changes may be made within the scope of protection of the attached claims.

[0013] In addition, when describing representative embodiments, the specification may have presented the method and / or process as a specific sequence of steps. However, to the extent that the method or process does not rely on the specific order of the steps described herein, the method or process should not be limited to the steps of the specific order described. As will be understood by those of ordinary skill in the art, other sequences of steps are also possible. Therefore, the specific sequence of the steps set forth in the specification should not be interpreted as a limitation to the claims. In addition, the claims for the method and / or process should not be limited to the steps of performing them in the order written, and those skilled in the art can easily understand that these sequences can be changed and still remain within the spirit and scope of the embodiments of the present application.

[0014] Membrane separation technology is a technology that uses semipermeable membranes to separate and purify substances. It is widely used in current sewage treatment plants and is an indispensable technical means in the field of modern water purification. In the process of water purification, various pollutants in the water have physical, chemical or mechanical effects on the membrane, which in turn causes the adsorption and deposition of various pollutants on the membrane surface and in the membrane pores, causing the membrane pore size to become smaller or blocked, reducing the membrane flux, causing membrane pollution, and then reducing the separation and purification capacity of the membrane, shortening the service life of the membrane. The formation process of membrane pollution is related to the properties of the membrane itself, the characteristics of sewage, and the interaction between the membrane and sewage, such as the hydrophilicity, pore size, charge, surface roughness of the membrane, and the pollutant type, temperature, pH value, flow rate, pressure and other factors of the sewage. Studying the formation process of membrane pollution is of great significance for improving membrane performance, optimizing membrane separation and filtration processes, reducing membrane cleaning frequency, and reducing sewage treatment costs.

[0015] The present application embodiment provides a method for unmixing fluorescence hyperspectral in a membrane pollution formation process, including S100-S300, such as Figure 1 As shown: S100: Acquire a fluorescence hyperspectral vector of a target film within a predetermined time period and a predetermined spectral range; S200: using a strong robust minimum volume unmixing analysis method to perform transmission projection on the fluorescence hyperspectral vector, and performing objective function soft constraint unmixing calculation to obtain end-member spectral data, wherein the end-member spectral data is multiple, and each end-member spectral data corresponds to a pollutant; S300: performing the following operations on each end member spectral data respectively: determining the membrane contamination formation process of the corresponding pollutant in the predetermined time period according to the end member spectral data.

[0016] In this embodiment, in step S100, a fluorescence hyperspectral imager can be used to obtain fluorescence hyperspectral data in a predetermined spectral range of 440 nm to 719 nm. The resolution of the fluorescence hyperspectral data can be 1.088 nm. The fluorescence hyperspectral data contains a total of (719-440) / 1.088=256 bands. The fluorescence hyperspectral data of these 256 bands are processed to obtain a fluorescence hyperspectral vector.

[0017] In this embodiment, the fluorescence hyperspectral data is continuous-time analog data collected in a predetermined time period, and has continuity in time and space. Moreover, compared with ordinary hyperspectral data, fluorescence hyperspectral data has high sensitivity and strong selectivity in pollutant detection, can detect low concentrations of pollutants, and can effectively distinguish different pollutants. The fluorescence spectrum also provides information on the internal molecular structure of pollutants, helps to identify specific functional groups of pollutants, and provides rich spectral information. Therefore, the space-time data of the target membrane pollution process can be obtained based on the fluorescence hyperspectral data.

[0018] In this embodiment, the target membrane may include a microfiltration membrane, an ultrafiltration membrane, a nanofiltration membrane or a reverse osmosis membrane; the pollutants may include but are not limited to suspended particles, colloidal particles, bacteria, viruses, heavy metal ions, soluble inorganic salts and natural organic matter.

[0019] In this embodiment, the fluorescence hyperspectral unmixing method for the membrane fouling formation process can be applied to various indoor water treatment fields, such as municipal sewage treatment, industrial wastewater treatment, seawater desalination, and drinking water preparation.

[0020] In this embodiment, the robust minimum volume unmixing analysis method (RMVAU for short) is an unmixing algorithm based on convex geometry, which aims to reduce the impact of noise and outliers on the unmixing results, and robustly estimates the components of the mixed signal by minimizing the volume of the mixing model. The algorithm is executed using a field programmable gate array (FPGA), and the fluorescence hyperspectral vector is transmitted and projected, and the objective function soft-constrained unmixing calculation is performed to obtain end-member spectral data corresponding to the pollutants one by one.

[0021] In this embodiment, step S300 can track the end-member spectral data at multiple time points in a predetermined time period, and then determine the membrane pollution formation process of the pollutant corresponding to the end-member spectral data in the predetermined time period. The fully constrained least squares method (FCLS, full name Full-Component Least Squares) can be used to invert the end-member spectral data to estimate the abundance of the pollutants and obtain an abundance distribution map of the pollutants. The abundance distribution map is further rendered in RGB color to obtain a distribution map of the membrane pollution formation process.

[0022] The fluorescence hyperspectral unmixing method for the membrane pollution formation process of this embodiment adopts a transmission projection method, which not only eliminates the influence of pixel-related proportional factors, but also maintains the direction of the original spectral vector, avoids angular displacement in the projection process, and significantly improves the projection effect. In addition, a strong robust minimum volume unmixing algorithm based on soft constraints is adopted, and the calculation process has strong robustness, good flexibility, high unmixing efficiency, and more accurate unmixing results, which can more realistically present the membrane pollution formation process.

[0023] In an exemplary embodiment, step S200 may include steps S210-S240: S210: Obtaining a low-dimensional linear subspace vector according to the fluorescence hyperspectral vector; S220: Performing orthogonal calculation on the low-dimensional linear subspace vector to obtain an orthogonal matrix; S230: performing transmission projection calculation on the orthogonal matrix to obtain spectral vector data that eliminates the influence of pixel-related scale factors; S240: Perform objective function soft-constrained unmixing calculation on the spectral vector data to obtain the endmember spectral data.

[0024] In this embodiment, in step S220, the low-dimensional linear subspace vector is subjected to orthogonal calculation, which can be achieved by singular value decomposition, and the low-dimensional linear subspace vector is converted into a diagonal matrix by singular value decomposition; it can also be achieved by Gram-Schmidt orthogonalization, and the orthogonal calculation can simplify the subsequent calculation steps.

[0025] In this embodiment, due to the problem of spectral variability, it is necessary to eliminate the influence of spectral variability on subsequent unmixing calculations. Spectral variability refers to the influence on the spectrum caused by changes in illumination and target film surface morphology. Spectral variability is ultimately manifested as a pixel-related proportional factor that affects the abundance vector. In the current related technology, orthogonal projection is generally used to solve this problem, but orthogonal projection has the following problems: orthogonal projection will change the direction of the original spectral vector, and orthogonal projection will cause an angular displacement between the original spectral vector and the corresponding projection vector, and this angular displacement will cause a large error in the projection vector. Due to the above problems with orthogonal projection, step S230 of this embodiment adopts a perspective projection method. Perspective projection will not change the direction of the original spectral vector, and will not introduce an angular displacement between the original spectral vector and the corresponding projection vector. After simulation calculation, the average angular displacement of the orthogonal projection is 7.45 degrees, while the average angular displacement of the perspective projection is approximately 0 degrees. The spectral vector data obtained by the perspective projection in this embodiment not only eliminates the influence of the pixel-related proportional factor, but also overcomes the projection defects caused by the orthogonal projection, and significantly improves the projection effect.

[0026] In an exemplary embodiment, step S240 may include steps S241-S242: S241: using a minimum volume simplex analysis method to perform vertex component analysis on the spectral vector data to obtain initial end-member spectral data; S242: Using a combined method of upper bound minimization and sequential quadratic programming to perform soft-constrained convex optimization calculation of an objective function on the initial end-member spectral data to obtain the end-member spectral data.

[0027] In this embodiment, in the current related technology, when the MVSA algorithm is used, the algorithm model is strictly constrained by the minimum volume idea, requiring all spectral vectors to be enclosed in a simplex, and at the same time ensuring that the volume of the simplex is minimized. From the perspective of constraints, it is necessary to strictly meet the hard constraint of "abundance coefficient is non-negative". However, in actual hyperspectral images, there is a high probability that there are noise and outliers in the spectral vectors, which is manifested as some pixels being far away from the simplex in the subspace. If the constraint of "abundance coefficient is non-negative" of each pixel is strictly met at this time, in order to ensure that the simplex can surround all abnormal points, the simplex will expand in space, and finally the position of the vertex of the simplex in the signal subspace will be far from the actual vertex position. In order to solve this problem, in this embodiment, the unmixing calculation of step S242 can adopt the upper bound minimization and sequential quadratic programming combined method (abbreviated as MM-SQP, full name Majorization-Minimization and Sequence Quadratic Programming). Specifically, a soft constraint that penalizes the objective function when the abundance coefficient is negative is used in the calculation process, which enhances the robustness of the algorithm in the presence of outliers or large noise, and maintains the minimum volume of the singlet as much as possible.

[0028] In this embodiment, the initial endmember spectral data obtained in step S241 is an initial unmixing result obtained by using a minimum volume simplex analysis method based on convex geometry, and is a result of vertex component analysis (VCA for short, full name Vertex Component Analysis) based on a pure pixel algorithm.

[0029] In this embodiment, after step S241, an endmember matrix may be constructed according to the initial endmember spectral data, wherein each column represents an initial endmember data point and different rows represent different features or attributes.

[0030] In this embodiment, a soft constraint term is introduced into the objective function in step S242, as shown in formula (1): (1) In formula (1), represents the constraint value, Representative end member matrix No. OK, Representative end member matrix No. List, Represents the end member matrix Line Elements of a column; represents the constraint function, Represents the parameter of the constraint function, i.e., the abundance coefficient; this soft constraint term assigns 0 to non-negative abundance coefficients and assigns a negative value to negative abundance coefficients. In this way, the abundance coefficients from outliers or noise have negative slack. This soft constraint is used to replace the hard constraint "abundance coefficient is non-negative" in the existing MVSA algorithm. After introducing this soft constraint, the sequential quadratic programming problem is solved as shown in formula (2): (2) In formula (2), represents the end member matrix after the iteration is completed, represents the sequential quadratic programming function, is the objective function after adding the above soft constraints, represents the output terminal matrix of each iteration, represents the characteristic matrix of the endmember matrix, represents the eigenvectors of the endmember matrix, Represents the objective function The gradient vector of Represents the objective function The Hessian matrix of the initial endmember spectral data is calculated through soft constraints to achieve soft convex optimization of the objective function. The endmember spectral data obtained are local optimal solution data. The calculation process has good flexibility and strong robustness, which helps to obtain the optimal calculation results faster.

[0031] In this embodiment, there are multiple end-member spectral data obtained in step S242, and each end-member spectral data corresponds to a pollutant. After executing step S242, step S300 is executed to determine the membrane pollution formation process of the corresponding pollutant in the predetermined time period according to each end-member spectral data. Step S300 performs inversion based on the end-member spectral data to estimate the abundance information of the pollutant corresponding to the end-member spectral data. The fully constrained least squares method (FCLS, full name Full-Component Least Squares) can be used to implement abundance estimation to obtain an abundance distribution map of the pollutant. The abundance distribution map is rendered in RGB color to obtain a distribution map of the membrane pollution formation process.

[0032] In an exemplary embodiment, step S210 may include steps S211-S213: S211: De-noising the fluorescence hyperspectral vector to obtain a first observed spectral vector after de-noising; S212: performing signal-to-noise ratio analysis on the first observed spectrum vector to obtain a noise correlation matrix; S213: Using the first observed spectrum vector and the noise correlation matrix to perform minimum error signal subspace identification calculation to obtain a low-dimensional linear subspace vector.

[0033] In this embodiment, since the number of end-member spectral data obtained in step S200 is usually much smaller than the number of bands of the fluorescence hyperspectral data obtained in step S100 when performing fluorescence hyperspectral unmixing, the minimum error signal subspace identification calculation in step S213 can adopt the minimum error hyperspectral signal identification algorithm (HySime, full name hyperspectral signal identification by minimum error). HySime is based on the principle of minimizing error and is an unsupervised, fully automatic algorithm based on feature decomposition and independent of any tuning parameters. By selecting the most representative partial spectrum, the first observed spectral vector can be reduced in dimension, which is a preliminary estimate of the number of end-member spectral data. It can accurately reduce the dimensionality of the high-dimensional first observed spectral vector to obtain a low-dimensional linear subspace vector.

[0034] In an exemplary embodiment, step S211 may include steps S2111-S2112: S2111: performing band selection within a predetermined spectral range, and using the fluorescence hyperspectral vector in the selected band as a band selection observation spectral vector; S2112: Perform block matching stereo denoising on the band-selected observed spectrum vector to obtain the first observed spectrum vector.

[0035] In this embodiment, step S2111 is performed to select representative characteristic bands, which reduces the subsequent calculation amount on the one hand, and eliminates the bands with poor quality on the other hand. The band selected observation spectrum vector obtained after band selection still retains the physical meaning of the band, and also removes the band with low signal-to-noise ratio, and finally obtains the band selected observation spectrum vector with significantly improved signal-to-noise ratio.

[0036] In this embodiment, step S2111 can perform band selection by using the variance threshold method or the principal component analysis method. When the variance threshold method is used to select the band, for the visualization image of the observed spectral vector of each band, a uniform image area is selected, and the mean and standard deviation of the observed spectral vector data in the uniform image area are calculated, and the ratio of the mean to the standard deviation is used as the signal-to-noise ratio of the observed spectral vector of the band.

[0037] In an exemplary real-time method, after the signal-to-noise ratios of the observed spectral vectors of all bands are obtained, the observed spectral vectors with a signal-to-noise ratio less than the threshold value are removed according to a preset signal-to-noise ratio threshold value.

[0038] In another exemplary real-time method, after obtaining the signal-to-noise ratio of the observed spectral vectors of all bands, the observed spectral vectors that are not within the signal-to-noise ratio threshold range are removed according to a preset signal-to-noise ratio threshold range; for example, according to the signal-to-noise ratio threshold range, the first five and last twenty observed spectral vectors in the signal-to-noise ratio sorting are removed, because the first five data with too high signal-to-noise ratio may be outliers and have a negative impact on subsequent calculations, or may over-rely on these five data in subsequent calculations, causing overfitting or data imbalance; the last twenty data with too low signal-to-noise ratio introduce a lot of noise and should be removed.

[0039] In this embodiment, step S2112 can perform block matching stereo denoising (BM3D for short, full name Block-Matching 3D Filtering) on ​​the band selected observed spectral vector after the band selection in step S2111 to obtain the first observed spectral vector. The BM3D algorithm combines the spatial context information and three-dimensional changes between local blocks of the image, and can effectively reduce the additive Gaussian white noise in the observed spectral vector by utilizing the similarity of local blocks, while considering the retention of rare pixels. After denoising, the image scatter points corresponding to the observed spectral vector are denser, reflecting a lower variance.

[0040] In an exemplary embodiment, step S212 may include steps S2121-S2122: S2121: performing noise estimation on the first observed spectrum vector to obtain a noise estimation matrix; S2122: Calculate the correlation matrix of the noise estimation matrix to obtain the noise correlation matrix.

[0041] In this embodiment, step S2121 can use a multivariate regression signal estimation method to perform noise estimation on the first observed spectral vector to obtain a noise estimation matrix. The noise estimation matrix can be used to estimate and describe noise statistical characteristics to help understand and process the noise components in the first observed spectral vector.

[0042] In this embodiment, the noise correlation matrix in step S2122 can provide correlation information between different parts within the noise signal, which helps to understand the spatial and temporal structure of the noise.

[0043] In an exemplary embodiment, step S213 may include steps S2131-S2134: S2131: performing multivariate regression signal estimation on the first observed spectrum vector to obtain a signal correlation matrix; S2132: performing a minimization projection error power calculation on the signal correlation matrix to obtain a projection error power matrix; S2133: Calculate the noise power of the noise correlation matrix to obtain a noise power matrix; S2134: Add the projection error power matrix and the noise power matrix, and use the added matrix as the low-dimensional linear subspace vector.

[0044] In this embodiment, in step S2131, a signal correlation matrix can be obtained by constructing a multivariate regression model. The signal correlation matrix can be used to analyze the correlation between the first observed spectral vectors of different bands, perform characteristic analysis of different bands, and provide information support for the identification of different pollutants.

[0045] In this embodiment, the signal correlation matrix is ​​minimized by calculating the projection error power through step S2132, and the signal correlation matrix can be eigenvalue decomposition to obtain eigenvalues ​​and their corresponding eigenvectors, so as to better understand the signal structure and information of the first observed spectral vector. By removing the eigenvalues ​​that contribute less to the overall information and retaining the most important eigenvalues, a projection matrix is ​​constructed based on the retained eigenvalues, the first observed spectral vector is projected into a new low-dimensional subspace, and the error power is calculated during the projection process to minimize the error power, thereby achieving error minimization and dimensionality reduction of the first observed spectral vector.

[0046] In this embodiment, in step S2133, the power spectral density of the noise signal can be calculated according to the noise correlation matrix. The power spectral density can represent the energy distribution of the noise signal in the frequency domain. A noise power matrix can be constructed according to the power spectral density. The noise power matrix can show the power level of the noise at different frequencies.

[0047] In this embodiment, in step S2134, the projection error power matrix and the noise power matrix are respectively increasing functions and decreasing functions of the dimension of the low-dimensional linear subspace vector. If the dimension of the low-dimensional linear subspace vector is overestimated, the noise power matrix dominates, and if the dimension of the low-dimensional linear subspace vector is underestimated, the projection error power matrix dominates.

[0048] In an exemplary embodiment, step S100 may include steps S110-S120: S110: Acquire a continuous analog value of the fluorescence hyperspectral image of the target film within the predetermined time period and within the predetermined spectral range; S120: Performing analog-to-digital conversion on the continuous analog quantity of the fluorescence hyperspectral image to obtain an observed spectrum vector as the fluorescence hyperspectral vector, wherein the observed spectrum vector is a visualized digital quantity.

[0049] In this embodiment, step S110 can use a fluorescence hyperspectral imager to collect continuous analog quantities of fluorescence hyperspectral images within a predetermined spectral range. The predetermined spectral range is a fluorescence spectrum with a wavelength of 440nm to 719nm. The fluorescence hyperspectral imager can be an instrument currently available on the market, or a fluorescence light sheet hyperspectral microscope imager built by the laboratory can be used.

[0050] In this embodiment, step S120 can perform analog-to-digital conversion through a field programmable gate array FPGA processor, convert the continuous analog quantity of the fluorescence hyperspectral image into a visualized fluorescence hyperspectral vector in real time, and save the fluorescence hyperspectral vector in the memory of the field programmable gate array FPGA.

[0051] The following is a detailed description of the above-mentioned fluorescence hyperspectral unmixing method for the film pollution formation process using a specific example. The example includes the following steps: Figure 2 As shown: S1: Obtain the continuous analog quantity of the fluorescence hyperspectral image of the target membrane in a predetermined time period, with a wavelength range of 440nm to 719nm and a resolution of 1.088nm; S2: The variance threshold method is used to select the continuous analog quantity of the fluorescence hyperspectral image in each band, and the 5 bands with the highest signal-to-noise ratio and the 20 bands with the lowest signal-to-noise ratio are removed to obtain the band-selected observation spectrum vector; S3: performing BM3D denoising on the band-selected observed spectrum vector to obtain a first observed spectrum vector; S4: performing noise estimation on the first observed spectrum vector to obtain a noise estimation matrix; S5: Calculate the correlation matrix of the noise estimation matrix to obtain the noise correlation matrix; S6: Perform multivariate regression signal estimation on the first observed spectral vector to obtain a signal correlation matrix; S7: Minimize the projection error power calculation on the signal correlation matrix to obtain a projection error power matrix; S8: Calculate the noise power of the noise correlation matrix to obtain a noise power matrix; S9: Add the projection error power matrix and the noise power matrix to obtain a low-dimensional linear subspace vector; S10: perform orthogonal calculation on the low-dimensional linear subspace vector to obtain an orthogonal matrix; S11: performing transmission projection calculation on the orthogonal matrix to obtain spectral vector data that eliminates the influence of pixel-related scale factors; S12: using the minimum volume simplex analysis method, performing vertex component analysis on the spectral vector data to obtain initial end-member spectral data; S13: using the upper bound minimization and sequential quadratic programming combined method to perform soft-constrained convex optimization calculation of the objective function on the initial endmember spectral data to obtain the endmember spectral data; S14: using the FCLS algorithm to perform abundance inversion and estimation on the end-member spectral data respectively, and obtaining the abundance distribution map of each pollutant; S15: Perform RGB color rendering on the abundance distribution map of each pollutant to obtain a distribution map of the membrane pollution formation process.

[0052] The present application also provides a fluorescence hyperspectral unmixing device for a film pollution formation process, including: a field programmable gate array, such as Figure 3 As shown; The field programmable gate array includes a processor and a memory; The memory is used to store the fluorescence hyperspectral unmixing program of the film pollution formation process; The processor is used to read the film pollution formation process fluorescence hyperspectral unmixing program and perform the film pollution formation process fluorescence hyperspectral unmixing method as described in the above embodiment.

[0053] In this embodiment, a field programmable gate array FPGA is used to perform a fluorescence hyperspectral unmixing method for the membrane pollution formation process. FPGA has high computing efficiency and good real-time performance, and can meet the requirements of fluorescence hyperspectral unmixing for the membrane pollution formation process in the above-mentioned embodiment.

[0054] The present application also provides a fluorescence hyperspectral unmixing system for a membrane pollution formation process, including: a fluorescence hyperspectral imager and the fluorescence hyperspectral unmixing device for a membrane pollution formation process described in the above embodiment, such as Figure 4 As shown; The fluorescence hyperspectral imager is used to collect the fluorescence hyperspectral image continuous analog quantity of the target film in a predetermined time period and a predetermined spectral range, and send the fluorescence hyperspectral image continuous analog quantity to the fluorescence hyperspectral unmixing device of the film pollution formation process; The device for unmixing fluorescence hyperspectral information during the formation of membrane pollution is used to carry out the method for unmixing fluorescence hyperspectral information during the formation of membrane pollution as described in the above embodiment.

[0055] In this embodiment, the fluorescence hyperspectral imager can collect continuous analog quantities of fluorescence hyperspectral images with a wavelength of 440nm to 719nm and a resolution of 1.088nm. Instruments currently available on the market can be used, or a fluorescence light sheet hyperspectral microscope imager built by the laboratory can be used.

[0056] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In hardware implementations, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all components may be implemented as software executed by a processor, such as a digital signal processor or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or transient medium). As is known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

Claims

1. A fluorescence hyperspectral unmixing method for a membrane fouling formation process, characterized in that: include: Acquire a fluorescence hyperspectral vector of a target film within a predetermined time period and a predetermined spectral range; A strong robust minimum volume unmixing analysis method is used to perform transmission projection on the fluorescence hyperspectral vector, and an objective function soft constraint unmixing calculation is performed to obtain end-member spectral data, wherein the end-member spectral data is multiple, and each end-member spectral data corresponds to a pollutant; The following operations are performed on each end member spectrum data respectively: a membrane pollution formation process of the corresponding pollutant in the predetermined time period is determined according to the end member spectrum data.

2. The method for unmixing fluorescence hyperspectral in the membrane pollution formation process according to claim 1, characterized in that: The strong robust minimum volume unmixing analysis method is used to perform transmission projection on the fluorescence hyperspectral vector, and to perform soft constraint unmixing calculation of the objective function to obtain end member spectral data, including: Obtaining a low-dimensional linear subspace vector according to the fluorescence hyperspectral vector; Performing orthogonal calculation on the low-dimensional linear subspace vector to obtain an orthogonal matrix; Perform transmission projection calculation on the orthogonal matrix to obtain spectral vector data that eliminates the influence of pixel-related scale factors; The objective function soft-constrained unmixing calculation is performed on the spectral vector data to obtain the endmember spectral data.

3. The method for unmixing fluorescence hyperspectral in the membrane pollution formation process according to claim 2, characterized in that: The step of performing objective function soft constraint unmixing calculation on the spectral vector data to obtain the endmember spectral data includes: Using a minimum volume simplex analysis method, performing vertex component analysis on the spectral vector data to obtain initial end-member spectral data; The initial end-member spectral data are subjected to objective function soft-constrained convex optimization calculation by adopting the upper bound minimization and sequential quadratic programming combined method to obtain the end-member spectral data.

4. The method for unmixing fluorescence hyperspectral in the membrane pollution formation process according to claim 2, characterized in that: The step of obtaining a low-dimensional linear subspace vector according to the fluorescence hyperspectral vector includes: De-noising the fluorescence hyperspectral vector to obtain a first observed spectral vector after de-noising; Performing a signal-to-noise ratio analysis on the first observed spectral vector to obtain a noise correlation matrix; The first observed spectrum vector and the noise correlation matrix are used to perform minimum error signal subspace identification calculation to obtain a low-dimensional linear subspace vector.

5. The method for unmixing fluorescence hyperspectral in the membrane pollution formation process according to claim 4, characterized in that: The step of denoising the fluorescence hyperspectral vector to obtain a denoised first observed spectral vector includes: Perform band selection within a predetermined spectral range, and use the fluorescence hyperspectral vector in the selected band as a band selection observation spectral vector; Block matching stereo denoising is performed on the band-selected observed spectral vector to obtain the first observed spectral vector.

6. The method for unmixing fluorescence hyperspectral in the membrane pollution formation process according to claim 4, characterized in that: The performing signal-to-noise ratio analysis on the first observed spectrum vector to obtain a noise correlation matrix includes: Performing noise estimation on the first observed spectrum vector to obtain a noise estimation matrix; A correlation matrix of the noise estimation matrix is ​​calculated to obtain the noise correlation matrix.

7. The method for unmixing fluorescence hyperspectral in the membrane pollution formation process according to claim 4, characterized in that: The method of using the first observed spectrum vector and the noise correlation matrix to perform minimum error signal subspace identification calculation to obtain a low-dimensional linear subspace vector includes: Performing multivariate regression signal estimation on the first observed spectral vector to obtain a signal correlation matrix; Minimizing the projection error power calculation on the signal correlation matrix to obtain a projection error power matrix; Performing noise power calculation on the noise correlation matrix to obtain a noise power matrix; The projection error power matrix and the noise power matrix are added, and the resultant matrix is ​​used as the low-dimensional linear subspace vector.

8. The method for unmixing fluorescence hyperspectral in the membrane pollution formation process according to claim 1, characterized in that: The step of obtaining the fluorescence hyperspectral vector of the target film within a predetermined time period and a predetermined spectral range comprises: Acquire a continuous analog value of the fluorescence hyperspectral image of the target film within the predetermined time period and within the predetermined spectral range; The continuous analog quantity of the fluorescence hyperspectral image is converted into a digital value to obtain a visualized digital quantity, and the visualized digital quantity is used as the fluorescence hyperspectral vector.

9. A fluorescence hyperspectral unmixing device for a membrane fouling formation process, characterized in that: Includes: Field Programmable Gate Array; The field programmable gate array includes a processor and a memory; The memory is used to store the fluorescence hyperspectral unmixing program of the film pollution formation process; The processor is used to read the film pollution formation process fluorescence hyperspectral unmixing program and perform the film pollution formation process fluorescence hyperspectral unmixing method as described in any one of claims 1-8.

10. A fluorescence hyperspectral unmixing system for membrane fouling formation process, characterized in that: include: A fluorescence hyperspectral imager and a fluorescence hyperspectral unmixing device for the membrane pollution formation process as claimed in claim 9; The fluorescence hyperspectral imager is used to collect the fluorescence hyperspectral image continuous analog quantity of the target film in a predetermined time period and a predetermined spectral range, and send the fluorescence hyperspectral image continuous analog quantity to the fluorescence hyperspectral unmixing device of the film pollution formation process; The membrane pollution formation process fluorescence hyperspectral unmixing device is used to perform the membrane pollution formation process fluorescence hyperspectral unmixing method as described in any one of claims 1-8.

Images