Fluorescence hyperspectral three-dimensional online in-situ imaging analysis device and analysis method suitable for membrane filtration
Through the fluorescence hyperspectral three-dimensional online in-situ imaging analysis device, the complexity and dynamics of the membrane pollution process are solved, and the rapid detection and visualization of membrane pollutants is realized, providing a non-invasive analysis platform for direct quantification Distribution and changes of pollutants.
Patent Information
- Application Number
- CN202211217680.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-09-30
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2042-09-30
AI Technical Summary
The prior art is difficult to effectively explore the complexity and dynamic process of membrane pollution, especially in situ visualization and quantification of the distribution of pollutants during membrane filtration, which limits the in-depth study of the membrane pollution mechanism.
Using fluorescence hyperspectral three-dimensional online in-situ imaging analysis device, the membrane filtration process is performed through the hyperspectral three-dimensional imaging mechanism, combined with the collaborative control mechanism to achieve rapid detection and demix of pollutant components and spatial distribution, and output color pollutant distribution images .
The label-free autofluorescence distinguishes the components of membrane contaminants and quickly detects its 2/3-dimensional spatial distribution on/in-film, directly visualizes and quantifies the formation and development of the pollution process, providing a non-invasive membrane contamination imaging platform.
Smart Images

Figure CN115656038B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of detection technology and relates to a fluorescence hyperspectral three-dimensional online in-situ imaging analysis device and an analysis method suitable for membrane filtration. Background Art
[0002] Membrane technology has developed rapidly in recent years and has been widely used in various environmental applications such as drinking water / wastewater treatment and seawater desalination due to its advantages in screening and energy efficiency, simplicity and manufacturing scalability. However, all membrane technologies face the inevitable phenomenon of membrane fouling, which will reduce product quantity and quality, increase energy consumption, and ultimately shorten membrane life. Therefore, understanding the mechanism of membrane fouling is crucial to improving the sustainability of membrane technology.
[0003] The formation and development of membrane fouling is a function of filtration time and is affected by many factors. It is a complex dynamic process. In this process, several fouling behaviors, including pore shrinkage / blocking, filter cake layer, concentration polarization, etc., are usually involved and play different roles in different fouling stages. In addition to the complexity of fouling behavior, the coupled effects of mechanical and chemical kinetics can also significantly affect the fouling process. For example, membrane-pollutant interaction and pollutant-pollutant interaction make different contributions in the initial or subsequent fouling stages. Due to the complexity of the structure and type of pollutants, it is increasingly important to characterize membrane fouling. In order to explore the characteristics and formation mechanism of membrane fouling, many detection methods have been applied, such as scanning electron microscopy, confocal scanning laser microscopy (CLSM), electrical impedance spectroscopy, etc. to explore the formation of membrane fouling layer, surface morphology during membrane fouling, and the rate of change of membrane pores; at the same time, since impurities with multiple chemical structures such as NOM and EPS have characteristic absorption or emission characteristics, molecular spectroscopy methods such as UV-visible spectroscopy, fluorescence spectroscopy, infrared spectroscopy, and Raman spectroscopy are ideal tools for identifying the chemical structure of these impurities. With the rapid development of spectral instruments and chemometric methods for spectral data processing, spectral methods are one of the most effective platform tools for characterizing membrane fouling. However, these technologies generally have problems such as shallow observation depth, complex characterization process, and inability to observe in situ, which limits further research on the mechanism of membrane fouling.
[0004] With the development of molecular spectroscopy instruments and microscopic imaging technology, the microscopic spectral imaging technology developed by combining the two has attracted more and more attention. It can not only identify specific pollutants but also characterize the morphological information of membrane pollution and is gradually becoming the most effective platform tool for characterizing the membrane pollution process. At the same time, since traditional detection methods require various methods of processing of membrane samples to cooperate with their imaging platforms, it is difficult to achieve in situ imaging. Summary of the invention
[0005] The technical problem to be solved by the present invention is that, in view of the defects of the prior art, a fluorescent hyperspectral three-dimensional online in-situ imaging analysis device suitable for membrane filtration is provided, which can realize label-free spontaneous fluorescence to distinguish the components of membrane pollutants and quickly detect their 2 / 3-dimensional spatial distribution on / in the membrane, thereby realizing direct visualization and quantification of the formation and development of the pollution process during the membrane filtration process.
[0006] The technical solution adopted by the present invention to solve its technical problem is:
[0007] A fluorescence hyperspectral three-dimensional online in-situ imaging and analysis device suitable for membrane filtration, comprising a membrane filtration mechanism for performing membrane filtration treatment on sewage through a filter membrane, a hyperspectral three-dimensional imaging mechanism for performing three-dimensional spectral imaging on the filter membrane in the membrane filtration mechanism and obtaining the composition and spatial distribution information of pollutants, a hyperspectral three-dimensional imaging online analysis mechanism for performing online unmixing of the collected three-dimensional spectral image and outputting a color pollutant distribution image, and a coordinated control mechanism for controlling the above mechanisms;
[0008] The hyperspectral three-dimensional imaging mechanism includes a light beam illumination optical path component for providing an excitation light source, a hyperspectral imaging component, and a scanning-anti-scanning coupling component; the scanning-anti-scanning coupling component is arranged corresponding to the filter membrane in the membrane filtering mechanism.
[0009] Furthermore, in the fluorescent hyperspectral three-dimensional online in-situ imaging analysis device suitable for membrane filtration, the membrane filtration mechanism preferably includes a sewage pool component, a sample pool component, and a filter pool component arranged in sequence, and the three are connected by pipelines, wherein a filter membrane is provided in the sample pool component.
[0010] Furthermore, in the fluorescent hyperspectral three-dimensional online in-situ imaging analysis device suitable for membrane filtration, the sewage pool component preferably includes a sewage pool for holding sewage and a sewage conveying component for conveying sewage to the sample pool component; the sample pool component includes a sample pool with a closed hollow structure and a three-dimensional mobile platform for driving the sample pool to perform three-dimensional movement, a filter membrane and a filter membrane placement rack for placing the filter membrane are provided in the sample pool, and at least one light-transmitting window corresponding to the scanning-anti-scanning coupling component is provided on the sample pool corresponding to the filter membrane; the filter pool component includes a filter pool for holding filtered water and a weighing component for weighing the filtrate.
[0011] Furthermore, in the fluorescence hyperspectral three-dimensional online in-situ imaging analysis device suitable for membrane filtration, the light beam illumination optical path component preferably includes a laser light source, a filter, a reflector group and a beam expansion system for adjusting the light beam size.
[0012] Furthermore, in the fluorescent hyperspectral three-dimensional online in-situ imaging analysis device suitable for membrane filtration, the hyperspectral imaging component is preferably composed of an imaging lens group and a hyperspectral imager.
[0013] Furthermore, in the fluorescence hyperspectral three-dimensional online in-situ imaging analysis device suitable for membrane filtration, the scanning-anti-scanning coupling component preferably includes a scanning galvanometer and a scanning lens group, an anti-scanning galvanometer and an anti-scanning lens group, each of which is respectively arranged corresponding to the filter membrane, wherein the scanning lens group is connected to a displacement member for adjusting the position of the anti-scanning lens group.
[0014] Furthermore, in the fluorescence hyperspectral three-dimensional online in-situ imaging analysis device suitable for membrane filtration, the hyperspectral three-dimensional imaging online analysis mechanism preferably comprises:
[0015] An image preprocessing module for background removal, spectral denoising and band signal-to-noise ratio analysis of the collected hyperspectral data;
[0016] A hyperspectral unmixing module that decomposes the hyperspectral data preprocessed by the image preprocessing module through endmember number estimation, initial value estimation and iterative decomposition, and online identifies the composition of pollutants and their three-dimensional spatial distribution;
[0017] A multi-dimensional image comprehensive processing module for image pseudo-color display, image threshold segmentation and image area calculation.
[0018] Furthermore, in the fluorescence hyperspectral three-dimensional online in-situ imaging analysis device suitable for membrane filtration, the collaborative control mechanism preferably includes a main controller, a membrane filtration mechanism control unit, an imaging control unit, and an online analysis control unit.
[0019] A fluorescence hyperspectral three-dimensional online in-situ imaging analysis method suitable for membrane filtration, comprising:
[0020] S1, controlling the hyperspectral three-dimensional imaging mechanism to image the filter membrane of the sample pool component through the collaborative control mechanism to obtain a three-dimensional spectral imaging corresponding to the filter membrane and containing pollutant information;
[0021] S2. Controlling the high-light three-dimensional imaging online analysis mechanism through the collaborative control mechanism to perform image analysis on the three-dimensional spectral imaging to extract the pollutant components corresponding to the filtering model, and generating a color pollutant distribution image containing pollutant type and distribution information for each pollutant.
[0022] Furthermore, in the fluorescence hyperspectral three-dimensional online in-situ imaging analysis method, preferably the method also includes:
[0023] S3. The use status of the filter membrane is monitored by the cooperative control mechanism, and during the use of the filter membrane, a plurality of color pollutant distribution images corresponding to each pollutant are sequentially acquired at preset time intervals and a change process image of the pollutant is generated.
[0024] Furthermore, in the fluorescence hyperspectral three-dimensional online in-situ imaging analysis method, preferably in step S2, the high-light three-dimensional imaging online analysis mechanism performs image analysis on the three-dimensional spectral imaging to extract the pollutant components corresponding to the filter module; comprising:
[0025] S21, performing background removal, spectral stack denoising, deep learning image enhancement and band signal-to-noise ratio analysis on the three-dimensional spectral imaging in sequence to obtain hyperspectral data containing wavelength information of each pollutant;
[0026] S22, decomposing the hyperspectral data based on the number of categories of pollutants to be extracted to obtain wavelength information having a number equal to the number of categories;
[0027] S23, judging the wavelength information, and obtaining the category information of the pollutants according to the wavelength information when the wavelength information is legal; when the wavelength information is illegal, modifying the number of categories of the pollutants to be extracted, and executing step S23.
[0028] The present invention uses a membrane filtration mechanism to fully display the membrane filtration treatment process of sewage, uses a hyperspectral three-dimensional imaging mechanism to perform three-dimensional spectral imaging on the filter membrane in the membrane filtration mechanism and obtains the composition and spatial distribution information of pollutants, and then uses a hyperspectral three-dimensional imaging online analysis mechanism to perform online unmixing of the collected three-dimensional spectral image and output a color pollutant distribution image, so as to perform in-situ pollution characterization in a label-free manner. The formation and development of the pollution process in the membrane filtration process is realized by directly visualizing and quantifying it.
[0029] The present invention establishes a non-invasive membrane contamination imaging platform based on extended light sheet microscopy and combined with hyperspectral fluorescence analysis technology. In combination with a specially designed membrane filtration system, a series of two-dimensional or three-dimensional hyperspectral images are obtained using different solutions as a function of filtration time. Hyperspectral light sheet fluorescence imaging (HS-LSFM) is used to characterize the dynamic contamination process during membrane filtration, achieving label-free autofluorescence to distinguish the components of membrane contaminants and quickly detecting their 2 / 3-dimensional spatial distribution on / inside the membrane. Line scanning illumination and imaging on / inside the membrane are completed by a scanning / reverse scanning combination, that is, the scanning end provides moving line illumination, and the reverse scanning end moves synchronously to reflect the signal light into the spectral imager to complete the imaging of the membrane. In addition to being able to complete the detection of spatial distribution, it can also complete the distinction between different substances. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:
[0031] Figure 1 It is a block diagram of the relationship between the various parts of implementation 1 of the present invention;
[0032] Figure 2 Schematic diagram of the optical path relationship of each part of Example 1 of the present invention;
[0033] Figure 3 is a block diagram of the relationship between the various parts of the membrane filtration mechanism of Example 1 of the present invention;
[0034] Figure 4 Schematic diagram of the positional relationship between the scanning-anti-scanning coupling component and the sample pool component of Example 1 of the present invention;
[0035] Figure 5 is a schematic diagram of the appearance structure of a sample pool assembly according to Example 1 of the present invention;
[0036] Figure 6 It is a schematic diagram of the structure of the sample cell assembly of Example 1 of the present invention after the cover body is removed. DETAILED DESCRIPTION
[0037] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.
[0038] Embodiment 1, as Figure 1-6 As shown, a fluorescence hyperspectral three-dimensional online in-situ imaging and analysis device suitable for membrane filtration includes a membrane filtration mechanism 100 for membrane filtration treatment of sewage, a hyperspectral three-dimensional imaging mechanism 200 for performing three-dimensional spectral imaging of the filter membrane in the membrane filtration mechanism and obtaining the composition and spatial distribution information of pollutants, a hyperspectral three-dimensional imaging online analysis mechanism 300 for online unmixing of the collected three-dimensional spectral image and outputting a color pollutant distribution image, and a collaborative control mechanism 400 for controlling the above-mentioned mechanisms; the hyperspectral three-dimensional imaging mechanism 200 includes a light beam illumination optical path component 210 for providing an excitation light source, a hyperspectral imaging component 230, and a scanning-anti-scanning coupling component 220; the scanning-anti-scanning coupling component 220 is arranged corresponding to the filter membrane 500 in the membrane filtration mechanism 100.
[0039] The present invention mainly includes four parts: a membrane filtration mechanism 100, a hyperspectral three-dimensional imaging mechanism 200, a hyperspectral three-dimensional imaging online analysis mechanism 300, and a collaborative control mechanism 400. Each part has its own interdependent functions. The following is a detailed description of the four parts:
[0040] The main function of the membrane filtration mechanism 100 is to achieve continuous filtration, including a sewage pool component 110, a sample pool component 120, and a filter pool component 130, which are arranged in sequence, and are connected by a pipeline 101, wherein a filter membrane 500 is provided in the sample pool component 120. Through the continuous filtration process, pollutants are gradually accumulated on the filter membrane 500, and online real-time three-dimensional spectral imaging is performed by the hyperspectral three-dimensional imaging mechanism 200, and then online unmixing and output of a color pollutant distribution image are performed by the hyperspectral three-dimensional imaging online analysis mechanism 300, so as to achieve label-free autofluorescence to distinguish the components of membrane pollutants, and quickly detect their 2 / 3-dimensional spatial distribution on / in the membrane, so as to achieve direct visualization and quantification of the formation and development of the pollution process during the membrane filtration process.
[0041] Among them, Figure 3 As shown, the sewage pool component 110 includes a sewage pool 111 for holding sewage and a sewage conveying member 112 for conveying sewage to the sample pool component 120. The water in the sewage pool is sewage before filtration. The sewage conveying member provides continuous sewage to the entire system, which satisfies the entire process of accumulating pollutants in the filter membrane 500 to form membrane pollution, and forms a basis for accurately analyzing the membrane pollution process. The sewage conveying member 112 can adopt a variety of implementation methods, such as: a delivery pump, a pressure tank, etc. The present invention preferably adopts a pressure tank, which pressurizes the sewage pool 111 and conveys the sewage to the sample pool component 120 through the pipeline 101. The sewage pool 111 can adopt all kinds of structures suitable for the sewage and storage of the present invention, which are not limited by the present invention. The pipeline 101 is also provided with a pressure gauge 113 and a recorder 114, and the recorder 114 can be a paperless recorder.
[0042] Among them, Figure 5-6 As shown, the sample pool assembly 120 includes a sample pool 121 with a closed hollow structure and a three-dimensional moving platform 129 that drives the sample pool 121 to perform three-dimensional movement. Specifically, a filter membrane placement rack 122 for placing a filter membrane 500 is provided in the sample pool 121, and at least one light-transmitting window 123 corresponding to the scanning-anti-scanning coupling assembly 220 is provided on the sample pool 121 corresponding to the filter membrane 500. Specifically, the sample pool 121 is a closed structure, including a cover body and a pool body, which are sealed and connected. The filter membrane placement rack 122 is sealed with the inner wall of the sample pool 121, and the filter membrane placement rack 122 divides the inner cavity of the sample pool 121 into a sewage cavity and a post-filter cavity, and the sewage cavity and the post-filter cavity are respectively connected to the sewage pool assembly 110 and the filter pool assembly 130 through pipelines 101.
[0043] The filter membrane rack 122 can be arranged horizontally, vertically or tilted. Preferably, the top surface of the filter membrane rack 122 is arranged tilted so that the filter membrane 500 is tilted to adapt to the optical excitation and detection of the scanning-anti-scanning coupling component 220. The horizontal tilt angle α of the filter membrane rack 122 can be 90°>α>0°, preferably α is 20-60°. A water outlet 124 is arranged on the bottom or side wall of the sample pool 121 corresponding to the post-filtration cavity. The water outlet 124 collects filtered water and supplies it to the filter pool component 130 through the pipeline 101; a water inlet 125 is arranged on the side wall or top of the sewage cavity, which is connected to the sewage pool component 110 through the pipeline 101.
[0044] Two light-transmitting windows are provided on one side of the sewage chamber of the sample pool 121, respectively provided on the top surface and the side surface of the sample pool 121. The light-transmitting windows on the top surface and the side surface may also be formed into a whole, and the light-transmitting windows 123 correspond to the inclined filter membrane 500. The light-transmitting windows 123 are made of highly transparent quartz glass and are optical windows for providing the scanning-anti-scanning coupling assembly 220.
[0045] The three-dimensional moving platform 129 is used to adjust the position of the sample pool assembly 120, and is disposed below the sample pool assembly 120 to adjust the sample pool assembly 120 in three directions: up and down, left and right, and front and back. Preferably, the three-dimensional moving platform 129 is a three-axis platform for fixing the sample pool 121, and can move in the three directions of X-axis, Y-axis, and Z-axis respectively, so as to drive the sample pool 121 to move and adapt to imaging.
[0046] Specifically, the three-dimensional mobile platform 129 includes a vertical Z-axis guide rail, on which are provided an X-axis guide rail and a Y-axis guide rail that slide vertically with each other, thereby realizing three-axis structural movement. Its specific structure can adopt an existing three-axis platform.
[0047] The pipeline 101 is used to transport sewage or filtered water, and its material, size and other parameters and structure are selected according to actual needs. The present invention is not limited to this, and the present invention only needs to be applicable.
[0048] The filter tank assembly 130 includes a filter tank 131 and a weighing element 132 for weighing the filtrate. Specifically, an analytical balance is used for weighing, a weighing tank is placed on the analytical balance, the weighing tank is connected to the filter tank, and the filtrate in the filter tank is passed into the weighing tank for weighing. The purpose of weighing is to obtain real-time membrane flux data and monitor the use status of the filter membrane.
[0049] like Figure 1-2 As shown, the hyperspectral three-dimensional imaging mechanism 200 uses the principle of light sheet-hyperspectral fluorescence imaging to perform rapid three-dimensional spectral imaging on the filter membrane 500 of the sample pool component 120 to obtain the composition and spatial distribution information of the pollutants. The light beam illumination optical path component 210 is used to provide an excitation light source, such as Figure 2 As shown, it includes a laser light source 211, a filter 212, a reflector group 213 and a beam expansion system 214 for adjusting the size of the light beam. Among them, the laser light source 211 generates illumination excitation light, the filter 212 filters non-excitation band light, the reflector group 213 is used to adjust the laser position, and the beam expansion system 214 is used to adjust the size of the light beam. The laser light source 211 can adopt a laser light source compatible with the present invention, which is not limited here. In other structures, the filter 212, the reflector group 213 and the beam system 214 are optical structures, and specifically, products with existing structures can be adopted. The number of reflectors in the reflector group 213 is set according to actual needs, which is not limited by the present invention.
[0050] like Figure 2 As shown, the scanning-anti-scanning coupling component 220 is used to coordinate the light beam illumination optical path component 210 and the hyperspectral imaging component 230 to jointly complete hyperspectral three-dimensional imaging. After complex coordinated control, the scanning-anti-scanning coupling component 220 can complete line scanning imaging with accuracy to the micron level. The scanning-anti-scanning coupling component 220 includes two parts. The first part is a scanning galvanometer 221 and a scanning lens group 222 set corresponding to the filter membrane 500. The function of the first part is: the scanning lens group 222 further shapes the light beam into a qualified scanning light beam to form a light beam illumination, and the scanning galvanometer 221 changes the illumination light following the change of the angle of the reflector in the galvanometer to control the light beam to complete the plane line scanning of the filter membrane 500; the second part is an anti-scanning galvanometer 223 and an anti-scanning lens group 224 also set corresponding to the filter membrane 500. The anti-scanning lens group 224 is connected to a displacement member (not shown in the figure) for adjusting the position of the scanning-anti-scanning lens group 224. The function of the second part is: the anti-scanning lens group 224 images a column of fluorescence excited on the pollutant sample of the filter membrane 500 onto the anti-scanning galvanometer 223, and drives the anti-scanning galvanometer 223 and the anti-scanning lens group 224 to move synchronously with the scanning galvanometer 221 through the displacement member, so as to reflect the line scanning images at different spatial positions into the slit 2321 in the hyperspectral imaging component 230. The displacement member may specifically adopt a one-dimensional adjustment platform, such as a linear motion track, a cylinder linear motion, etc. It only needs to realize the linear motion of the anti-scanning galvanometer 223 and the anti-scanning lens group 224, and the present invention does not limit this.
[0051] like Figure 4 As shown, the scanning galvanometer 221 and the scanning lens group 222 of the first part correspond to the filter membrane 500 from the light-transmitting window 123 of the sample pool 121, and the reverse scanning galvanometer 223 and the reverse scanning lens group 223 of the second part also correspond to the filter membrane 500 from the light-transmitting window 123 of the sample pool 121 in another direction. The scanning directions of the first part and the second part form an angle of 90° relative to the filter membrane 500.
[0052] The number, size, position and structure of the galvanometer in the scanning galvanometer 221 and the lenses in the scanning lens group 222 can be based on the existing technology according to actual needs, and the present invention does not limit them.
[0053] Similarly, the number, size, position and structure of the anti-scanning galvanometer 223 and the anti-scanning lens group 224, the galvanometer in the anti-scanning galvanometer 223, and the lens in the anti-scanning lens group 224 can be based on actual needs using existing technologies, and the present invention does not limit them.
[0054] like Figure 2 As shown, a hyperspectral imaging component 230 is provided behind the anti-scanning galvanometer 223 and the anti-scanning lens group 224 , and the hyperspectral imaging component 230 is composed of an imaging lens group 231 and a hyperspectral imager 232 .
[0055] In the hyperspectral imaging component 230, the imaging lens group 231 includes a scanning lens and a sleeve lens. The scanning lens and the sleeve lens form a 4f imaging system, which collects the reverse scanning imaging beam completed by the scanning-reverse scanning coupling component 220 into the hyperspectral imager 232. The hyperspectral imager is composed of a slit 2321, a grating 2322, a lens 2323 and a CCD camera 2324. The slit 2321 filters out a column on the image plane to ensure the spectral resolution formed by the line scan. The grating 2322 performs grating spectroscopy on this column of spatial fluorescence, and the light intensity of this column is distributed in the space thereafter according to the spectrum, and is projected into the CCD camera 2324 through the imaging lens group 2323 to form a spectral distribution surface corresponding to the spatial influence of this column. The CCD camera 2324 collects the spectral distribution surface to obtain spectral spatial information.
[0056] like Figure 1 As shown, the hyperspectral three-dimensional imaging online analysis mechanism 300 includes an image preprocessing module 310, a hyperspectral unmixing module 320, and a multi-dimensional image comprehensive processing module 330. The hyperspectral three-dimensional imaging online analysis mechanism 300 performs online unmixing and visualization operations on the collected three-dimensional spectral image, and outputs an intuitive and clear color pollutant distribution image. The hyperspectral three-dimensional imaging online analysis mechanism 300 is a packaged program written based on the MABLAB app, which can complete the processing of the collected hyperspectral three-dimensional membrane pollution image online in real time.
[0057] The image preprocessing module 310 is used to perform background removal, spectral denoising, deep learning image enhancement, and band signal-to-noise ratio analysis on the collected hyperspectral data.
[0058] Among them, background removal is that after the image preprocessing module 310 reads the collected high-spectral three-dimensional film pollution raw data stack, it identifies the data by calculating the image peak signal-to-noise ratio PSSN and classifies it into background data and signal data according to the PSSN value distribution interval. After screening out the background data, the signal data is subtracted from the background data to obtain pure signal data without background data;
[0059] Spectral denoising uses two denoising algorithms for spectral stack denoising: principal component analysis (PCA) and minimum noise separation (MNF). Denoising improves the signal-to-noise ratio, and PCA denoising is preferred. The principal component analysis method decomposes the original data into feature space by calculating the covariance matrix of the data stack and solving the characteristic matrix, and then removes the dimensions irrelevant to the main signal to complete the extraction and denoising of the main data. The minimum noise separation method is similar to the principal component analysis method, but uses the characteristic matrix of the covariance matrix arranged by the signal-to-noise ratio to separate the noise. In the application scenario of this patent, the principal component analysis algorithm is more suitable for membrane filtration pollution data;
[0060] Deep learning signal enhancement adopts the U-net convolution deep learning architecture. By collecting high / medium / low signal-to-noise ratio images of this system in advance and training the deep learning network, a U-net neural network with improved signal-to-noise ratio of data of this system is obtained. The main structure of this convolutional neural network can be divided into downsampling convolution part, upsampling convolution part and jump connection part. After the training data is input, the features are extracted by downsampling convolution kernel compression, and then the image size is restored by upsampling, and the correctness of the restoration is maintained through the jump connection layer to obtain the output data. The output data is compared with the reference data to obtain the deviation value and back-propagation is performed to correct the parameters of each layer of the network, and repeated training is performed to obtain a stable network. After the above data is pre-processed and input into the neural network trained with a large amount of data, the signal-to-noise ratio can be greatly improved.
[0061] Band signal-to-noise ratio analysis uses the variance method to calculate the signal-to-noise ratio of each channel (third dimension) of three-dimensional data:
[0062]
[0063] After calculating the signal-to-noise ratio of each wavelength channel, the band with the lowest signal-to-noise ratio is screened out for data deletion. This operation can improve the accuracy of subsequent hyperspectral decomposition.
[0064] The hyperspectral unmixing module 320 is the core functional module of the present invention. It decomposes the complex hyperspectral data collected by the hyperspectral three-dimensional imaging mechanism 200 in detail, decomposes the mixed spectrum into independent components through mathematical algorithms, restores the material composition in the membrane contamination process, and provides the spatial distribution of each material composition. Specifically, it is realized through three parts: end member number estimation, initial value estimation and iterative decomposition.
[0065] In the process of estimating the number of endmembers, the HySime algorithm that minimizes the sum of the projection errors of the observed data and noise can be used to make a preliminary estimate of the number of endmembers of the data spectrum, that is, to determine the number of pollutant types. Users can also manually enter the number of endmembers. The purpose of this process is to obtain an estimated type of pollutant data so that hyperspectral decomposition can be performed based on the estimated number to obtain the corresponding number of wavelength information.
[0066] In the initial value estimation process, in order to improve the accuracy and speed of spectral decomposition, the initial value estimation part uses vertex component analysis (VCA) or PPI algorithm to perform preliminary end member decomposition to provide initial values for the subsequent formal decomposition. It can be understood that the initial value estimation process is the process of obtaining a number of pollutant types that is closer to the actual situation.
[0067] In the VCA algorithm flow, the following steps are performed in sequence: Input the preprocessed hyperspectral image matrix V∈R n×m and the number of endmembers r; generate a random Gaussian vector w with zero mean, initialize the endmember 0 matrix W; project the pixel onto the hyperplane r T u=1 to obtain the simplex Sp; take the pixel i corresponding to the endpoint of the orthogonal projection of the pixel in the simplex as the first end member to be extracted; For n=1:(r-1) project the pixel point to the orthogonal direction of the subspace formed by the first end member; take the pixel corresponding to the projection endpoint as the i+1 end member. Confirm the number of end members according to this process. The number of end members obtained is the number of pollutant types.
[0068] In the PPI algorithm flow, the following processes are performed in sequence: Input the preprocessed hyperspectral image matrix V∈R n×m and the number of end members r; use MNF to reduce the dimension of data and whiten the noise; randomly generate M random vectors (the default value is 10000); initialize the m-dimensional zero matrix q=q1,q2,…,q m ; For i = 1: M projects the pixel and records the pixel number P corresponding to the projection endpoint (extreme value) min , P max ; Find and update the PPI of the corresponding pixel number, Arrange the pixel spectra in descending order according to the PPI count, and take the first r pixel spectra in the ordered set as end members. The number of end members obtained is the number of pollutant types.
[0069] The iterative part is divided into geometric fast decomposition and synchronous precise decomposition: the geometric fast decomposition occupies less memory and is fast, which is suitable for in-situ online monitoring. Specifically, the minimum volume constrained vertex component analysis (MVSA) algorithm is used for spectral decomposition, and the fully constrained least squares (FCLS) algorithm is configured for abundance inversion. In the specific decomposition process, the following steps are performed in sequence: input the preprocessed hyperspectral image matrix V∈R n×m and the number of endmembers r; use BM3D to denoise the data; solve the V eigenvector matrix by singular value decomposition and perform dimensionality reduction; project the pixels and record the pixel numbers corresponding to the projection endpoints (extreme values); perform MM optimization-quadratic programming on the projection: perform FCLS abundance inversion on the endmember spectrum, output the endmember matrix and abundance image, and display the accuracy evaluation index; calculate SAD and root mean square error.
[0070] The synchronous precise decomposition module has high accuracy, and adopts non-negative matrix decomposition and minimum volume constrained non-negative matrix decomposition to synchronously decompose and obtain decomposed spectrum and abundance matrix. The present invention can identify substances in the membrane pollution process online and reproduce the three-dimensional spatial distribution through the hyperspectral decomposition part. The following steps are performed in sequence during the specific decomposition process: input the pre-processed hyperspectral image matrix, the number of end members r, the maximum number of iterations T, and the convergence error ζ; initialize the random end member matrix W and the abundance matrix H, i=0; in ||f (t-1) -f t ||<ζ, i≦T, update W and H alternately according to the following formula;
[0071] W←W.*(VH T ). / (WHH T ); H←H.*(W T V). / (W T WH)(Non-negative Matrix Factorization)
[0072]
[0073] (Minimum Volume Constrained Non-Negative Matrix Factorization)
[0074] Calculate SAD and root mean square error, and reconstruct to obtain abundance image; output end element matrix and abundance image, and display accuracy evaluation index.
[0075] The multi-dimensional integrated image processing module 330 provides further display and analysis of the data, including image pseudo-color display, image threshold segmentation, and image area calculation. The image pseudo-color display further displays the decomposed spectrum and abundance matrix obtained by the hyperspectral decomposition: the decomposed spectrum will be directly displayed in the form of a curve in the coordinate axis, so that the user can intuitively judge the main pollutant components of the current membrane pollution; the abundance matrix corresponding to different substances will be given different colors (pseudo-color) to display in color RGB format, so that the user can intuitively observe the spatial distribution of pollutants. Image threshold segmentation is a more detailed data processing for further membrane pollution analysis. It has a variety of built-in threshold segmentation algorithms, and uses the Triangle threshold segmentation algorithm to locate the original membrane surface. The subsequent membrane pollution image is completed by the ImageJ registration plug-in to complete the registration of the initial membrane surface position. After the positioning is completed, the Triangle threshold segmentation algorithm is used to perform threshold segmentation on the membrane pollution data, and the effective membrane pollution area is determined according to the initial membrane surface position. Image area calculation uses the positioning points generated by threshold segmentation to perform binarization operations on the image and obtain quantitative data information statistically.
[0076] like Figure 1 As shown, the collaborative control mechanism 400 includes a membrane filtration control unit 420, an imaging control unit 430, an online analysis control unit 440, and a main controller 410 for receiving signals from the above control units and issuing control instructions.
[0077] The membrane filtration control unit 410 includes an air supply controller, a pressure control valve, a pressure transmitter or a pressure gauge, a recorder, a weighing controller, and a three-dimensional movement controller. The control unit is arranged in the membrane filtration mechanism, wherein the air supply controller is arranged on the air pressure tank, and is used to control the air pressure tank, such as opening and closing. The pressure control valve is arranged in the pipeline 101 of the membrane filtration mechanism 100, and is used to adjust the pressure and flow of the pipeline 101. The pressure transmitter collects the pressure or flow information of the pipeline 101, and the recorder is used to record the pressure or flow information. The weighing controller controls the weighing of the filtered filtrate. Specifically, the rear end of the filter pool is connected to the weighing pool, and the weighing pool is placed on the weighing piece. The weighing piece is connected to the main controller through the weighing controller (RS232 serial port), and a communication reading is performed every certain time, such as 20s or 30s, to record the current water weight. The three-dimensional movement controller is used to control the three-dimensional movement of the sample pool 121 in the sample pool assembly 120, and specifically controls the action of the three-dimensional movement platform 129 to move up and down, left and right, and forward and backward.
[0078] The equipment is turned on, and the main controller 410 sends a command to the air supply controller in the membrane filtration control unit 420. The air pressure tank is turned on to transport the sewage to the filter membrane 500 in the sample pool assembly 120 for filtration. The pressure transmitter or pressure gauge transmits the pressure information of the pipeline 101 to the main controller 410. The main controller 410 controls the air pressure tank or pressure control valve to control the pressure and flow of the pipeline 101. At the same time, the main controller 410 controls the weighing filter to complete the weighing of the filtered filtrate and feeds back the weighing information to the main controller. After the detection is completed, the main controller 410 sends a command to the air supply controller in the membrane filtration control unit 420, and the air pressure tank is turned off.
[0079] The imaging control unit 430 controls the scanning galvanometer 221, the anti-scanning galvanometer 223, the displacement element, and the hyperspectral imager controller in the scanning-anti-scanning coupling component 220. The main controller 410 sends a movement instruction to control the scanning galvanometer and the anti-scanning galvanometer to move simultaneously, and the scanning galvanometer 221 and the anti-scanning galvanometer 223 work, and control the displacement element to adjust. The main controller 410 sends instructions to control the hyperspectral imager controller to perform imaging. The main controller 410 sends a signal to the scanning galvanometer 221 and the anti-scanning galvanometer 223, and controls the angle of the internal reflector through voltage. The scanning galvanometer 221 controls the illumination beam to move along the direction perpendicular to the beam propagation to complete the plane scanning illumination, and the anti-scanning galvanometer 223 moves synchronously to reflect the beam into the spectral imager to complete the two-dimensional spectral imaging.
[0080] The main controller 410 sends signals to the three-dimensional moving platform 129 below the sample pool 121 and the displacement components in the anti-scanning lens group, and moves synchronously to complete the three-dimensional spectral imaging.
[0081] The specific devices of the above collaborative control mechanism and the specific implementation of the control signal flow can adopt the existing technology and will not be described in detail here.
[0082] The collaborative control system 400 is developed based on LABVIEW software, and programming is used to realize the comprehensive control of the optical imaging module and the membrane filtration module. Human-computer interaction is realized by writing LABVIEW programs. The overall state machine architecture is adopted, and the process processing of hardware commands is realized through operations such as sequential structures and trigger events. The integrated control of hardware or lower computers is realized by combining serial communication, calling drivers, subVIs, etc. The software includes debugging mode and acquisition mode. In debugging mode, the software scans and reverses the galvanometer in the coupling component 220, and the three-dimensional displacement platform of the filter imaging device component for imaging debugging; in acquisition mode, the software coordinates and controls the two-dimensional galvanometer, the displacement stage and the hyperspectral imager to perform rapid membrane pollution hyperspectral three-dimensional imaging of multiple optical selection surfaces.
[0083] Embodiment 2, in the fluorescent hyperspectral three-dimensional online in-situ imaging analysis method suitable for membrane filtration of the present application, includes the following steps: S1, controlling the hyperspectral three-dimensional imaging mechanism to image the filter membrane of the sample pool assembly through a collaborative control mechanism to obtain a three-dimensional spectral imaging corresponding to the filter membrane and containing pollutant information. That is, based on the above-mentioned analysis device, the positional relationship between the membrane filtration mechanism and the hyperspectral three-dimensional imaging mechanism is controlled by the collaborative control mechanism, so that the hyperspectral three-dimensional imaging mechanism can image a specific section of the filter membrane of the sample pool assembly in the membrane filtration mechanism to obtain the corresponding three-dimensional spectral imaging. The three-dimensional spectral imaging information contains pollutant information on the filter membrane. The coordination process between the various mechanisms can refer to the above description.
[0084] S2. The collaborative control mechanism controls the high-light three-dimensional imaging online analysis mechanism to perform image analysis on the three-dimensional spectral imaging to extract the pollutant components corresponding to the filter module, and generate a color pollutant distribution image containing the pollutant type and distribution information for each pollutant. Specifically, after obtaining the three-dimensional spectral image corresponding to the cross-section of the filter membrane, the high-light three-dimensional imaging online analysis mechanism performs image analysis on the three-dimensional spectral imaging to extract the pollutant components corresponding to the cross-section of the filter membrane, and generates a corresponding color pollutant distribution image for each pollutant type and distribution information based on the pollutant components. For example, the color pollutant distribution image that can be generated can include the type information of the pollutant and its proportion information in all pollutant components.
[0085] Optionally, in one embodiment, the fluorescence hyperspectral three-dimensional online in-situ imaging analysis method of the present invention further includes: S3, monitoring the use status of the filter membrane through the collaborative control mechanism, and in the use process of the filter membrane, sequentially obtaining a number of the color pollutant distribution images corresponding to each pollutant at preset time intervals and generating a change process image of the pollutant. That is, monitoring the use status of the filter membrane, in the normal use process of the filter membrane, recording the color pollutant distribution images corresponding to the preset section of the filter membrane at each preset time point in time sequence, and sequentially recording the change process of the same pollutant based on the time sequence. The change process can be reflected in generating a series of color pollutant distribution images of the pollutant, and arranging them in time sequence. Based on this process, the intuitive record of the pollutant formation process in the same section of the filter membrane can be quickly realized. For multiple pollutants, it can record each pollutant separately to form an image matrix of the color pollutant distribution images of multiple pollutants, so that the distribution change law of each pollutant in the entire pollution process of the filter membrane can be intuitively obtained.
[0086] Optionally, in the fluorescence hyperspectral three-dimensional online in-situ imaging analysis method of the present invention, in the step S2, the high-light three-dimensional imaging online analysis mechanism performs image analysis on the three-dimensional spectral imaging to extract the pollutant components corresponding to the filter module; including: S21, the three-dimensional spectral imaging is sequentially subjected to background removal, spectral stack denoising, deep learning image enhancement and band signal-to-noise ratio analysis to obtain hyperspectral data containing wavelength information of each pollutant. After obtaining pure signal data without background data, the signal data is subjected to spectral stack denoising based on the spectral denoising algorithm, and the denoising process can adopt principal component analysis (PCA) or minimum noise separation (MNF). After spectral denoising, deep learning image enhancement processing is performed to improve the signal-to-noise ratio of the obtained spectrum, so as to facilitate the subsequent band signal-to-noise ratio analysis process, and it is easier to remove invalid signals, that is, band data with low signal-to-noise ratio, so as to retain useful spectral signals to the maximum extent and ensure the correctness of spectral decomposition. The specific processing process of each stage can be obtained through the working process of the above-mentioned image preprocessing module 310, which will not be repeated here.
[0087] S22. Decompose the hyperspectral data based on the number of categories of pollutants to be extracted to obtain wavelength information of the number of categories. Specifically, after obtaining the hyperspectral data, the wavelength corresponding to the spectrum can be extracted through the hyperspectral unmixing module 320, wherein the type and number information of pollutants corresponding to the filter membrane can be preset based on the use scenario of the filter membrane, and the hyperspectral data is decomposed according to the number of categories to obtain wavelength information of the number equal to the number of categories. The simple understanding is that when the number of possible pollutants is set, the number of groups of wavelength information can be decomposed from the hyperspectral data. The specific decomposition process can refer to the working process of the hyperspectral unmixing module 320 mentioned above, which will not be repeated here.
[0088] S23, judging the wavelength information, and obtaining the category information of the pollutants according to the wavelength information when the wavelength information is legal. When the wavelength information is illegal, modifying the number of categories of the pollutants to be extracted, and executing step S23. Specifically, after obtaining several groups of wavelength information, the wavelength information obtained is judged to confirm whether the spectral graph shape corresponding to the obtained single wavelength is correct, that is, whether it conforms to the general rules of the fluorescent spectrum group. If the spectral graph shape obtained conforms to the general rules of the fluorescent spectrum group, it is determined that the wavelength information obtained is legal. At this time, it can be determined that the types and quantities of the pollutants obtained are accurate, and pollutant analysis can be performed. If the spectral graph shape obtained does not conform to the general rules of the fluorescent spectrum group, it means that the judgment of the types and quantities of pollutants at this time is abnormal. At this time, it is necessary to correct the preset number of pollutant types and perform high-spectral decomposition again based on the corrected number of pollutant types. Until the obtained spectral graph shape is legal.
[0089] Based on the above process, the transformation process of pollutants in the process of pollutant deposition in a specific area of the filter membrane can be intuitively obtained. By performing the above analysis process on multiple areas of the filter membrane, the transformation process of pollutants in the entire filter membrane can be obtained, and then the pollution process of the filter membrane during use can be obtained.
Claims
1. A fluorescence hyperspectral three-dimensional online in-situ imaging analysis device suitable for membrane filtration, characterized in that: It includes a membrane filtration mechanism for performing membrane filtration treatment on sewage through a filter membrane, a high-spectral three-dimensional imaging mechanism for performing three-dimensional spectral imaging on the filter membrane in the membrane filtration mechanism and obtaining the composition and spatial distribution information of pollutants, a high-spectral three-dimensional imaging online analysis mechanism for online unmixing of the collected three-dimensional spectral image and outputting a color pollutant distribution image, and a coordinated control mechanism for controlling the above mechanisms; The hyperspectral three-dimensional imaging mechanism includes a light beam illumination optical path component for providing an excitation light source, a hyperspectral imaging component, and a scanning-anti-scanning coupling component; the scanning-anti-scanning coupling component is arranged corresponding to the filter membrane in the membrane filtering mechanism.
2. The fluorescence hyperspectral three-dimensional online in-situ imaging analysis device suitable for membrane filtration according to claim 1, characterized in that: The membrane filtration mechanism comprises a sewage pool component, a sample pool component and a filter pool component which are arranged in sequence and are connected by pipelines, wherein a filter membrane is arranged in the sample pool component.
3. The fluorescence hyperspectral three-dimensional online in-situ imaging analysis device suitable for membrane filtration according to claim 2, characterized in that: The sewage pool component includes a sewage pool for holding sewage and a sewage conveying component for conveying sewage to the sample pool component; the sample pool component includes a sample pool with a closed hollow structure and a three-dimensional moving platform for driving the sample pool to perform three-dimensional movement, a filter membrane and a filter membrane placement rack for placing the filter membrane are provided in the sample pool, and at least one light-transmitting window corresponding to the scanning-reverse scanning coupling component is provided on the sample pool corresponding to the filter membrane; the filter pool component includes a filter pool for holding filtered water and a weighing component for weighing the filtrate.
4. The fluorescence hyperspectral three-dimensional online in-situ imaging analysis device suitable for membrane filtration according to claim 1, characterized in that: The light beam illumination optical path component comprises a laser light source, a filter, a reflector group and a beam expansion system for adjusting the size of the light beam.
5. The fluorescence hyperspectral three-dimensional online in-situ imaging analysis device suitable for membrane filtration according to claim 1, characterized in that: The hyperspectral imaging component consists of an imaging lens group and a hyperspectral imager.
6. The fluorescence hyperspectral three-dimensional online in-situ imaging analysis device suitable for membrane filtration according to claim 1, characterized in that: The scanning-anti-scanning coupling assembly comprises a scanning galvanometer and a scanning lens group, an anti-scanning galvanometer and an anti-scanning lens group respectively arranged corresponding to the filter membrane, wherein the scanning lens group is connected to a displacement member for adjusting the position of the anti-scanning lens group.
7. The fluorescence hyperspectral three-dimensional online in-situ imaging analysis device suitable for membrane filtration according to claim 1, characterized in that: The hyperspectral three-dimensional imaging online analysis mechanism comprises: Image preprocessing module for background removal, spectral denoising and band signal-to-noise ratio analysis of the collected hyperspectral data; A hyperspectral unmixing module that decomposes the hyperspectral data preprocessed by the image preprocessing module through endmember number estimation, initial value estimation and iterative decomposition, and online identifies the composition of pollutants and their three-dimensional spatial distribution; A multi-dimensional image comprehensive processing module for image pseudo-color display, image threshold segmentation and image area calculation.
8. A fluorescence hyperspectral three-dimensional online in-situ imaging analysis method suitable for membrane filtration, characterized in that: Applied to the fluorescence hyperspectral three-dimensional online in-situ imaging analysis device according to any one of claims 1 to 7, the method comprises: S1, controlling the hyperspectral three-dimensional imaging mechanism to image the filter membrane of the sample pool component through the collaborative control mechanism to obtain a three-dimensional spectral imaging corresponding to the filter membrane and containing pollutant information; S2. Controlling the high-light three-dimensional imaging online analysis mechanism through the collaborative control mechanism to perform image analysis on the three-dimensional spectral imaging to extract the pollutant components corresponding to the filter membrane, and generating a color pollutant distribution image containing pollutant type and distribution information for each pollutant.
9. The fluorescence hyperspectral three-dimensional online in-situ imaging analysis method according to claim 8, characterized in that: The method further comprises: S3. The use status of the filter membrane is monitored by the cooperative control mechanism, and during the use of the filter membrane, a plurality of color pollutant distribution images corresponding to each pollutant are sequentially acquired at preset time intervals and a change process image of the pollutant is generated.
10. The fluorescence hyperspectral three-dimensional online in-situ imaging analysis method according to claim 9, characterized in that: In the step S2, the high light three-dimensional imaging online analysis mechanism performs image analysis on the three-dimensional spectral imaging to extract the pollutant components corresponding to the filter module; include: S21, performing background removal, spectral stack denoising, deep learning image enhancement and band signal-to-noise ratio analysis on the three-dimensional spectral imaging in sequence to obtain hyperspectral data containing wavelength information of each pollutant; S22, decomposing the hyperspectral data based on the number of categories of pollutants to be extracted to obtain wavelength information having a number equal to the number of categories; S23, judging the wavelength information, and obtaining the category information of the pollutants according to the wavelength information when the wavelength information is legal; when the wavelength information is illegal, modifying the number of categories of the pollutants to be extracted, and executing step S23.
Citation Information
Patent Citations
Large-scale visual membrane pollution in-situ online monitoring system suitable for membrane filtration
CN113176236A
System and Method for the Assessment of Biological Particles in Exhaled Air
US20120252058A1
