Hyperspectral data analysis method based on plasma resonance metal film substrate

Abstract

The invention discloses a hyperspectral data analysis method based on a plasma resonance metal film substrate. According to the metal nano-film substrate, scattering and signal absorption of a sample to be detected in a visible spectrum range can be remarkably enhanced through a plasma resonance effect, and meanwhile, a reflection spectrum from the substrate is effectively inhibited. The enhancement mechanism not only can realize a high contrast effect similar to dark field imaging in a reflection mode, but also does not need to use expensive optical components, so that the equipment cost can be effectively reduced. In addition, the hyperspectral data is efficiently processed and analyzed by combining a deep learning technology, and high-sensitivity and high-resolution microscopic hyperspectral imaging analysis is realized. According to the method disclosed by the invention, the spectral characteristic information of the sample can be remarkably enhanced under the condition of relatively low equipment cost, and the spectral information can be accurately and efficiently interpreted. The method is suitable for the fields of biomedicine, material science and environmental monitoring.

Description

Technical Field

[0001] The present invention relates to the field of spectral imaging technology, and particularly to a hyperspectral data analysis method based on a plasma resonance metal thin film substrate. Background Art

[0002] Microscopic hyperspectral imaging (Micro-HSI) is an emerging microscale target detection and analysis technology that combines a hyperspectral imaging system with an optical microscope to achieve detailed characterization of microscopic samples.

[0003] However, most current microscopic hyperspectral imaging technologies use bright-field imaging modes (such as transmission or reflection modes), which often generate strong background signals during sample detection, masking the spectral features of semi-transparent samples (such as microplastics or biological tissues). To improve the signal-to-noise ratio and imaging contrast of samples, researchers have developed dark-field hyperspectral imaging technology. This technology detects the scattered light of the sample and compares it with a dark background, significantly reducing background interference and improving the visibility of target signals. However, due to the weak scattered signals, dark-field hyperspectral imaging usually requires expensive high-performance light sources and detection equipment, which greatly restricts the cost and popularization of the system.

[0004] Currently, there is an urgent need for a more cost-effective optical strategy to further improve the performance of microscopic hyperspectral imaging technology while ensuring strong signals and high contrast.

[0005] Therefore, the prior art still needs to be improved and developed. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a hyperspectral data analysis method based on a plasma resonance metal thin film substrate in view of the above-mentioned defects of the prior art, aiming to solve the problem that it is difficult to balance the equipment cost and the performance of microscopic hyperspectral imaging in the prior art.

[0007] The technical solution adopted by the present invention to solve the problem is as follows:

[0008] In a first aspect, an embodiment of the present invention provides a hyperspectral data analysis method based on a plasma resonance metal thin film substrate, the method comprising:

[0009] Preparing a metal nano-film substrate with plasma resonance effect in advance, and loading a sample to be measured on the metal nano-film substrate;

[0010] Performing reflection spectrum testing on the sample to be measured through a reflection-type bright-field microscopic hyperspectral imaging system to obtain hyperspectral data;

[0011] Feature extraction is performed on the hyperspectral data through a deep neural network or a support vector machine, and analysis information of the sample to be measured is obtained based on the extracted feature data.

[0012] In one embodiment, the metal used for the metal nanofilmsubstrate includes one of gold, silver, and copper; the substrate used for the metal nanofilmsubstrate is a silicon substrate; preparing a metal nanofilmsubstrate with a surface plasmon resonance effect in advance includes:

[0013] Depositing a metal on the cleaned silicon substrate by a preset coating method to obtain the metal nanofilmsubstrate.

[0014] In one embodiment, the nanostructure morphology corresponding to the metal nanofilmsubstrate includes one of a nano-column array film, a nano-hole array film, and a planar film.

[0015] In one embodiment, loading the sample to be measured on the metal nanofilmsubstrate includes:

[0016] Performing a dispersion treatment on the sample to be measured to uniformly disperse the sample to be measured, where the dispersion treatment includes ultrasonic treatment;

[0017] Uniformly depositing the pretreated sample to be measured on the metal nanofilmsubstrate.

[0018] In one embodiment, the reflection bright-field microscopic hyperspectral imaging system includes a light source, an illumination system, a microscope, and a hyperspectral imaging device; the microscope is used to magnify the sample to be measured by a preset multiple; the microscopic hyperspectral imaging system is used to acquire hyperspectral data of the sample to be measured at a spatial resolution of preset pixels, where each pixel includes a preset number of discrete wavelengths.

[0019] In one embodiment, before performing feature extraction on the hyperspectral data through a deep neural network or a support vector machine and obtaining analysis information of the sample to be measured based on the extracted feature data, it further includes:

[0020] Performing preprocessing on the hyperspectral data, where the preprocessing includes at least one of noise filtering, spectral correction, smoothing processing, and normalization processing.

[0021] In one embodiment, the deep neural network includes a channel prior convolutional attention layer; the deep neural network is used for:

[0022] Through the channel prior convolutional attention layer, spectral features are extracted from the hyperspectral data based on channel attention, and joint features are extracted from the spectral features based on spatial attention. The joint features are used to reflect the spectral information and spatial information of the hyperspectral data.

[0023] Analytical information of the sample to be measured is obtained according to the joint features.

[0024] In a second aspect, an embodiment of the present invention further provides a hyperspectral data analysis device based on surface plasmon resonance and a metal thin film substrate. The device includes:

[0025] A placement module for loading a sample to be measured onto a pre-prepared metal nanofilm substrate with surface plasmon resonance effect;

[0026] An imaging module for performing a reflectance spectrum test on the sample to be measured through a reflection bright-field microscopic hyperspectral imaging system to obtain hyperspectral data;

[0027] An analysis module for extracting features from the hyperspectral data through a deep neural network or a support vector machine, and obtaining analytical information of the sample to be measured according to the extracted feature data.

[0028] In a third aspect, an embodiment of the present invention further provides a terminal, which includes a memory and more than one processor; the memory stores more than one program; the program includes instructions for executing the hyperspectral data analysis method based on a surface plasmon resonance metal thin film substrate as described in any one of the above; the processor is used to execute the program.

[0029] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, on which multiple instructions are stored, and characterized in that the instructions are adapted to be loaded and executed by a processor to implement the steps of the hyperspectral data analysis method based on a surface plasmon resonance metal thin film substrate as described in any one of the above.

[0030] Advantages of the present invention: The metal nanofilmsubstrate in the embodiments of the present invention can significantly enhance the scattering and absorption signals of a sample to be measured in the visible spectral range through the localized surface plasmon resonance (LSPR) effect, while effectively suppressing the reflection spectrum from the substrate. This enhancement mechanism can not only achieve a high-contrast effect similar to dark-field imaging in the reflection mode, but also does not require the use of expensive optical components, thus effectively reducing the equipment cost. In addition, the present invention also combines deep learning technology to efficiently process and analyze hyperspectral data, realizing high-sensitivity and high-resolution microscopic hyperspectral imaging analysis. Through the method of the present invention, it is possible to significantly enhance the spectral feature information of a sample at a relatively low equipment cost, and accurately and efficiently interpret the spectral information. It is applicable to the fields of biomedicine, materials science, environmental monitoring, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0032] Figure 1 It is a schematic flowchart of a hyperspectral data analysis method based on a plasmon resonance metal film substrate provided by an embodiment of the present invention.

[0033] Figure 2 It is a complete flowchart block diagram of obtaining hyperspectral data and spectral analysis data of a sample to be measured provided by an embodiment of the present invention.

[0034] Figure 3 It is a schematic structural diagram of a deep neural network (CNN-CPCA model) provided by an embodiment of the present invention.

[0035] Figure 4 It is a schematic principle diagram of a channel prior convolutional attention layer (CPCA module, including a channel attention mechanism and a spatial attention mechanism) provided by an embodiment of the present invention.

[0036] Figure 5 It is a comparative analysis diagram of the spectrum of the original light source of the Micro-HSI system, the average reflection spectrum of the Ag nanoarray substrate, and the average reflection spectrum of the Si substrate provided by an embodiment of the present invention.

[0037] Figure 6 It is the average reflection spectrum of six microplastic solution samples on the Si substrate in Comparative Example 1 provided by an embodiment of the present invention.

[0038] Figure 7It is the average reflection spectrum of six microplastic solution samples on the Ag nanoarray substrate in Example 1 provided by the embodiments of the present invention.

[0039] Figure 8 It is the confusion matrix of the CNN-CPCA model based on the Ag nanoarray substrate in Example 1 provided by the embodiments of the present invention when applied to the Micro-HSI dataset.

[0040] Figure 9 It is the average reflection spectrum of the PS microplastic solid sample on the Si substrate in Example 2 provided by the embodiments of the present invention.

[0041] Figure 10 It is the average reflection spectrum of the PS microplastic solid sample on the Ag nanoarray substrate in Example 2 provided by the embodiments of the present invention.

[0042] Figure 11 It is the confusion matrix of the CNN-CPCA model based on the Si substrate in Comparative Example 1 when applied to the Micro-HSI dataset.

[0043] Figure 12 It is a high-magnification SEM image of the Ag nanoarray substrate provided by the embodiments of the present invention.

[0044] Figure 13 It is a schematic diagram of the modules of the hyperspectral data analysis device based on surface plasmon resonance and a metal thin film substrate provided by the embodiments of the present invention.

[0045] Figure 14 It is a schematic block diagram of the terminal provided by the embodiments of the present invention. Detailed implementation manners

[0046] The present invention discloses a hyperspectral data analysis method based on a surface plasmon resonance metal thin film substrate. To make the objectives, technical solutions and effects of the present invention clearer and more definite, the following further describes the present invention in detail with reference to the accompanying drawings and by way of examples. It should be understood that the specific examples described herein are only used to explain the present invention and are not used to limit the present invention.

[0047] Those skilled in the art can understand that, unless specifically stated otherwise, the singular forms "a", "an", "the" and "said" used herein may also include the plural forms. It should be further understood that the term "comprising" used in the description of the present invention means the presence of the described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or their groups. It should be understood that when we say that an element is "connected" or "coupled" to another element, it can be directly connected or coupled to other elements, or there may also be intermediate elements. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The phrase "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.

[0048] Those skilled in the art can understand that, unless otherwise defined, all terms used herein (including technical terms and scientific terms) have the same meaning as the general understanding of those of ordinary skill in the art to which the present invention pertains. It should also be understood that terms such as those defined in a general dictionary should be understood to have a meaning consistent with the meaning in the context of the prior art, and will not be interpreted with an idealized or overly formal meaning unless specifically defined as here.

[0049] Aiming at the above-mentioned defects of the prior art, the present invention provides a hyperspectral data analysis method based on a plasma resonance metal thin film substrate, and the method includes: pre-preparing a metal nanometer thin film substrate with a plasma resonance effect, and loading a sample to be measured on the metal nanometer thin film substrate; performing a reflection spectrum test on the sample to be measured through a reflection-type bright-field microscopic hyperspectral imaging system to obtain hyperspectral data; extracting features from the hyperspectral data through a deep neural network or a support vector machine, and obtaining analysis information of the sample to be measured according to the extracted feature data. The metal nanometer thin film substrate in the present invention can significantly enhance the scattering and absorption signals of the sample to be measured in the visible spectral range through the localized surface plasmon resonance (LSPR) effect, and at the same time effectively suppress the reflection spectrum from the substrate. This enhancement mechanism can not only achieve a high-contrast effect similar to dark-field imaging in the reflection mode, but also does not require the use of expensive optical components, so the equipment cost can be effectively reduced. In addition, the present invention also combines deep learning technology to efficiently process and analyze hyperspectral data, realizing hypersensitive and high-resolution microscopic hyperspectral imaging analysis. Through the method of the present invention, it is possible to significantly enhance the spectral feature information of the sample at a relatively low equipment cost, and accurately and efficiently interpret the spectral information. It is applicable to fields such as biomedicine, materials science and environmental monitoring.

[0050] As Figure 1 shown, the method includes:

[0051] Step S100: Prepare a metal nano-film substrate with plasma resonance effect in advance, and load the sample to be measured on the metal nano-film substrate.

[0052] Specifically, the plasma resonance effect is highly sensitive to the refractive index change at the interface between the sample and the nanoarray. Different samples will exhibit different spectral responses due to their unique refractive index characteristics. In this embodiment, it is necessary to prepare a metal nano-film substrate with plasma resonance effect in advance to provide a special platform for subsequent detection of the sample to be measured. In practical applications, when the sample to be measured is placed on the metal nano-film substrate, due to the characteristic of increased electromagnetic field brought by the plasma resonance effect, the scattering and absorption signals of the sample to be measured in the visible spectral range can be enhanced, and at the same time, the reflection spectrum from the substrate can be effectively suppressed, so as to achieve a high-contrast effect similar to dark-field imaging in the reflection mode, thereby improving the detection sensitivity.

[0053] In one implementation, the metal used for the metal nano-film substrate includes one of gold, silver, and copper; the substrate used for the metal nano-film substrate is a silicon substrate; preparing a metal nano-film substrate with plasma resonance effect in advance includes:

[0054] Deposit the metal on the cleaned silicon substrate through a preset coating method to obtain the metal nano-film substrate.

[0055] Specifically, the metal nano-film can be selected as gold, or silver, or copper. The substrate can be selected as a silicon (Si) substrate. The preparation method of the metal nano-film substrate is to deposit the selected metal on the selected substrate through a specific coating method, including but not limited to the combination of thermal evaporation coating and magnetron sputtering coating with annealing treatment process. In order to better enhance the spectral signal of the sample to be measured, it is necessary to thoroughly clean the surface of the substrate before coating to ensure the removal of any residual contaminants.

[0056] For example, as Figure 2 shown, when the substrate is selected as a Si substrate and the metal nano-film is selected as an engineered Ag nanoarray film, the scattering and absorption signals of the sample in the visible spectral range (such as 600–900 nm) can be significantly enhanced through the LSPR effect, and at the same time, the reflection spectrum from the substrate can be effectively suppressed.

[0057] In one implementation, the nano-structural morphology corresponding to the metal nano-film substrate includes one of a nano-column array film, a nano-hole array film, and a planar film.

[0058] Specifically, the nanostructure morphology of the metal nanothin film substrate is selected from a nanocolumn array film, or a nanopore array film, or a planar film. In addition, the metal nanothin film substrate also has adjustable nanostructure sizes, where the nanocolumn size is between 10 nm and 500 nm, and the film thickness is between 10 nm and 500 nm, to meet the specific spectral enhancement requirements of different samples.

[0059] In one implementation, loading the sample to be measured onto the metal nanothin film substrate includes:

[0060] Performing a dispersion treatment on the sample to be measured to uniformly disperse the sample to be measured, where the dispersion treatment includes ultrasonic treatment;

[0061] Uniformly depositing the pretreated sample to be measured onto the metal nanothin film substrate.

[0062] Specifically, before loading the sample to be measured onto the metal nanothin film substrate in this embodiment, it is also necessary to perform a dispersion treatment on the sample to be measured. For example, ultrasonic treatment is performed on the sample to be measured to achieve uniform dispersion of the sample to be measured, and then the uniformly dispersed sample to be measured is deposited onto the metal nanothin film substrate, thereby realizing uniform loading of the sample to be measured, ensuring close contact between the sample and the surface of the metal nanothin film, and thus making full use of the local surface plasmon resonance effect to enhance the spectral signal of the sample to be measured.

[0063] For example, assume that the sample to be measured is a microplastic solution. The microplastic aqueous solution is ultrasonically treated to ensure uniform dispersion. Then, a micropipette is used to precisely deposit the microplastic solution onto the prepared metal nanothin film substrate, and a cover glass is covered to achieve uniform distribution of the solution on the metal nanothin film substrate, and subsequent analysis of the solution sample is performed. Or a micropipette is used to precisely deposit the microplastic solution onto the prepared metal nanothin film substrate, and then it is dried to obtain a dried sample, and subsequent analysis of the dried solid sample is performed.

[0064] Step S200: Performing a reflection spectrum test on the sample to be measured through a reflection bright-field microscopic hyperspectral imaging system to obtain hyperspectral data.

[0065] Specifically, in the reflection bright-field microscopic hyperspectral imaging system, the light emitted by the light source is irradiated onto the sample to be measured through a microscope, and the light reflected by the sample to be measured then enters the hyperspectral imaging device for imaging. In this embodiment, the reflection spectrum test is performed on the sample to be measured through such a microscopic hyperspectral imaging system to obtain detailed reflection spectrum information, that is, to obtain the hyperspectral data of the sample to be measured. The hyperspectral data can reflect the composition or structural characteristics of the sample to be measured.

[0066] In one implementation, the microscopic hyperspectral imaging system of the reflective bright field includes a light source, an illumination system, a microscope, and a hyperspectral imaging device; the microscope is used to magnify the sample to be measured by a preset magnification; the microscopic hyperspectral imaging system is used to obtain hyperspectral data of the sample to be measured with a spatial resolution of preset pixels, where each pixel includes a preset number of discrete wavelengths.

[0067] Specifically, the components of the microscopic hyperspectral imaging system of the reflective bright field include a light source, an illumination system, a microscope, and a hyperspectral imaging device, which can efficiently capture the reflected spectral data of the sample to be measured in the visible light range of 600 nm to 900 nm. For example, a microscopic hyperspectral imaging system including a microscope and an SM320 hyperspectral imaging system is used, with a magnification of 200 times, a spectral range of the system covering 400 - 1000 nanometers, a spectral resolution of 2 nanometers, and a detailed spectral profile of the sample to be measured (such as microplastic particles) is captured. Hyperspectral data is obtained with a spatial resolution of 660×360 pixels, where each pixel includes 300 discrete wavelengths.

[0068] Step S300: Extract features from the hyperspectral data through a deep neural network or a support vector machine, and obtain analysis information of the sample to be measured according to the extracted feature data.

[0069] Specifically, the deep neural network includes but is not limited to using a convolutional neural network (CNN), or a transformer, or a hybrid neural network structure. Before actual application, the deep neural network needs to go through three stages of model training, verification, and testing to ensure the accuracy and reliability of feature extraction. In this embodiment, through a deep neural network or a support vector machine, adaptive spectral feature extraction of different types of samples to be measured can be realized, and different analysis tasks including but not limited to classification and regression can be completed, which is applicable to a variety of application fields.

[0070] In one implementation, before extracting features from the hyperspectral data through a deep neural network or a support vector machine and obtaining analysis information of the sample to be measured according to the extracted feature data, it further includes:

[0071] Preprocess the hyperspectral data, and the preprocessing includes at least one of noise filtering, spectral correction, smoothing processing, and normalization processing.

[0072] Specifically, in order to ensure the accuracy of the analysis results, before inputting the hyperspectral data into a deep neural network or a support vector machine, one or more preprocessings such as noise filtering, spectral correction, smoothing processing, and normalization processing need to be performed on the hyperspectral data to improve the reliability of the data. Then, a deep neural network or a support vector machine is used to perform feature learning and extraction on the preprocessed hyperspectral data.

[0073] In one implementation, the smoothing processing specifically includes the following steps:

[0074] Fit a low-order polynomial (such as a second-order or third-order polynomial) within the window, and replace the central point value of the original data with the central value of the fitted polynomial; the original data is the original hyperspectral data;

[0075] The window moves point by point, and the steps of fitting a low-order polynomial within the window and replacing the central point value of the original data with the central value of the fitted polynomial are repeated.

[0076] In this embodiment, the data is smoothed using a sliding window, which can effectively reduce high-frequency noise while retaining spectral features such as peaks and trends.

[0077] In one implementation, the normalization processing specifically includes the following steps:

[0078] By subtracting the overall mean from each value of the original data and dividing by the standard deviation, the original data is normalized to zero mean and unit variance.

[0079] In this embodiment, through normalization processing, the influence of data dimension and amplitude can be eliminated, enabling the hyperspectral data to be compared under a unified scale.

[0080] In one implementation, the deep neural network includes a channel prior convolutional attention layer; the deep neural network is used for:

[0081] Through the channel prior convolutional attention layer, based on channel attention, feature extraction is performed on the hyperspectral data to obtain spectral features; and based on spatial attention, feature extraction is performed on the spectral features to obtain joint features, where the joint features are used to reflect the spectral information and spatial information of the hyperspectral data;

[0082] Obtain the analysis information of the sample to be measured according to the joint features.

[0083] Specifically, the input data of the deep neural network in this embodiment is hyperspectral data to be analyzed, and the hyperspectral data can be in the form of an image; the output data is the analysis information of the sample to be measured, such as the classification category. In addition to the common modules, this deep neural network particularly introduces a channel prior convolutional attention layer. The channel prior convolutional attention layer first extracts spectral features through a channel attention, and then continues to extract the features obtained by the previous channel attention mechanism through spatial attention. Finally, the final features obtained through the previous operations are put into a fully connected layer for classification and prediction. In this embodiment, the obtained final features are defined as the joint features of hyperspectrum and spatial spectrum, and these joint features can reflect the spectral features and spatial features of the hyperspectral data. This embodiment can be applied to related research on samples such as microplastics, cells or nanomaterials in fields including but not limited to biomedicine, materials science, environmental monitoring, etc.

[0084] In one implementation, as Figure 3 shown, the structure of the deep neural network specifically includes: a three-dimensional convolutional layer, a three-dimensional pooling layer, a two-dimensional convolutional layer, a two-dimensional pooling layer, a channel prior convolutional attention layer, and a fully connected layer connected in sequence.

[0085] Specifically, the deep neural network in this embodiment can extract the spatial feature information of hyperspectral data through a three-dimensional convolutional network and a spatial attention mechanism, and can extract the spectral feature information of hyperspectral data through a channel attention mechanism (as Figure 4 shown).

[0086] The advantages of the present invention are as follows:

[0087] 1. By introducing a metal nanofilmsubstrate with surface plasmon resonance effect, the present invention can enhance the scattering and absorption signals in the visible light range of 600–900 nm, thus greatly suppressing the background reflection signal of the substrate, making the reflection spectral signal in this band mainly come from the sample, thereby improving the detection sensitivity and resolution of the microscopic hyperspectral imaging system, and solving the problem of strong interference of background reflected light in traditional bright-field hyperspectral imaging.

[0088] 2. The present invention combines deep neural network technology to process hyperspectral data. Through feature learning and extraction, it effectively analyzes the complex information in high-dimensional spectral data, overcoming the limitations of low efficiency and insufficient accuracy of traditional data analysis methods. At the same time, deep neural network technology can realize adaptive hyperspectral-spatial spectrum joint feature extraction and complete different tasks including classification and regression.

[0089] 3. The present invention combines a reflective bright-field microscopic hyperspectral imaging system with an engineered metal nanofilms substrate, avoiding the dependence on expensive light sources and high-sensitivity detectors in traditional dark-field hyperspectral imaging, effectively reducing the manufacturing and use costs of the system hardware, and ensuring high efficiency performance at the same time.

[0090] Taking the Ag nanoarray film substrate and microplastic particles as an example, the Ag nanoarray film substrate can significantly enhance the spectral characteristic information of microplastic particles. Combining with a deep neural network model for data analysis can further improve the sensitivity and accuracy of detection. This method is applicable to the efficient detection and classification of microplastic particles in complex environments, and has the characteristics of economic scalability. It can quickly, simply, and at low cost classify microplastics in aqueous solutions, and the classified microplastics have a high accuracy rate, and there are few problems such as other types of microplastics still mixed in the classified microplastics.

[0091] The effectiveness of the method of the present invention is verified through two examples and a comparative example as follows:

[0092] I. Example 1: Enhancing the spectral characteristics of the microscopic hyperspectral imaging technology of microplastic solutions through an Ag nanoarray film substrate, and combining with a deep learning algorithm to achieve high-sensitivity classification and detection of microplastic solutions. The specific steps are as follows:

[0093] (1) Substrate preparation: Immerse the silicon substrate in freshly prepared sulfuric acid-hydrogen peroxide solution (98% H 2 SO 4 :30% H 2 O 2 =7:3) for 10 minutes to thoroughly clean the surface. Rinse three times with deionized water and ethanol respectively, and each rinse is accompanied by 10 minutes of ultrasonic cleaning to ensure the removal of any residual contaminants.

[0094] (2) Silver film deposition: Fix the cleaned silicon wafer in a thermal evaporation system, and place an ultra-pure silver target material in the evaporation boat. Deposit silver on the cleaned silicon substrate through thermal evaporation to form a film about 10 nanometers thick. After deposition, take out the silver-coated silicon substrate and perform an annealing treatment in a muffle furnace at 310 °C for 47 minutes to promote the formation of well-defined silver nanoarrays on the substrate.

[0095] (3) Microplastic solution preparation: Ultrasonically treat the microplastic aqueous solution to ensure uniform dispersion. Use a micropipette to precisely deposit 2.5 microliters of the microplastic solution onto the prepared substrate. Cover with a cover glass of 0.6 cm × 0.6 cm to achieve uniform distribution of the solution on the substrate surface.

[0096] (4) Microscopic Hyperspectral Imaging: A customized microscopic hyperspectral imaging system was used, including a microscope and an SM320 hyperspectral imaging system, with a magnification of 200 times. The system covers a spectral range of 400 - 1000 nm, with a spectral resolution of 2 nm, capturing the detailed spectral profiles of microplastic particles. Hyperspectral data was acquired with a spatial resolution of 660×360 pixels, and each pixel contains 300 discrete wavelengths.

[0097] (5) Data Processing and Classification: The samples were divided into six categories: PS (polystyrene), PLA (polylactic acid), PMMA (polymethyl methacrylate), PS+PLA, PS+PMMA, and PS+PLA+PMMA. The hyperspectral data was segmented into regions of 15×15 pixels, and each region was used as an individual training sample, resulting in a total of 19008 training samples. The dataset was evenly divided into a training set and a test set, each accounting for 50%. The hyperspectral data was processed with Savitzky-Golay smoothing and Z-score normalization. Classification was performed using a Convolutional Attention Hybrid Neural Network (CNN-CPCA) model or a Support Vector Machine (SVM): First, the data was preprocessed, then the preprocessed data was separately input into the (CNN-CPCA) model or the Support Vector Machine for data feature extraction, and finally, the extracted data was input into a fully connected network for classification prediction. In practical applications, all models were developed and simulated using PyTorch on a computer equipped with a GTX4090 GPU.

[0098] (6) Spectral Comparative Analysis in Result Analysis: The spectra of the original light source of the Micro-HSI system, the reflection spectra of the Ag nanorod array substrate, and the pure Si substrate were compared ( Figure 5 ). In a microscopic environment, a planar substrate usually results in very weak scattering. Under the strong incident light inherent in Micro-HSI, the reflection spectrum of the Si substrate is very close to the original light source spectrum, indicating that the Si substrate exhibits minimal scattering and absorption, and the reflection spectrum is almost the same as the incident light. In contrast, the Ag nanorod array substrate shows a significantly different reflection spectrum under the same strong incident light. The reflection band of the Ag nanorod array substrate is mainly concentrated in the wavelength range of 400 - 600 nm, and the reflectivity decreases significantly in the wavelength range of 500 - 750 nm. This behavior is attributed to the LSPR effect of the Ag nanorod array, which enhances scattering and absorption in the visible spectrum while effectively suppressing the substrate reflection in the wavelength range of 500 - 750 nm.

[0099] Micro-HSI Inspection of Microplastic Aqueous Solution Samples in Result Analysis: Further Micro-HSI inspections were carried out on microplastic aqueous solution samples that are usually translucent and difficult to distinguish under a traditional microscope. The average reflection spectra of the six microplastic solution samples on the Si substrate and the Ag nanorod array substrate are respectivelyFigure 6 and Figure 7 are shown in. These graphs compare the reflection spectra of six different microplastic samples on Si substrates and Ag nanoarray substrates. On the Si substrate, the reflection spectra of all six samples ( Figure 6 ) are very close to the spectrum of the pure Si substrate, indicating that the strong substrate reflection overwhelms the subtle reflection signals of the microplastics, making it difficult to distinguish between different microplastic types.

[0100] In contrast, the reflection spectra of the six samples on the Ag nanoarray substrate ( Figure 7 ) are significantly different from the spectrum of the pure Ag substrate. The reflection bands of these six samples are mainly concentrated in the range of 400 - 700 nm, which is significantly different from the 400 - 580 nm reflection band of the pure Ag nanoarray substrate. More importantly, compared with the almost identical reflection spectra of different samples observed on the Si substrate, each sample shows unique spectral features due to the modulation caused by LSPR on the Ag nanoarray substrate. Figure 7 clearly demonstrates the more obvious spectral curve differences of the Ag nanoarray substrate in enhancing the reflection signals of microplastic samples compared with the traditional Si substrate ( Figure 6 ). In addition, the LSPR effect of the Ag nanoarray not only suppresses the unwanted substrate reflection, but also shows high sensitivity to the different refractive indices of different samples, providing a solid foundation for high-contrast, high-precision detection and classification in Micro-HSI applications.

[0101] Result analysis - Classification performance comparison: Classification performance of the Ag nanoarray substrate: The confusion matrix of the Ag nanoarray substrate ( Figure 8 ) shows significantly superior classification performance, with an average accuracy rate reaching 98.65%. Most samples are correctly classified, and very few misclassifications are observed in all six categories.

[0102] Result analysis - Other performance indicators of the Ag nanoarray substrate: Recall: 98.58%; F1 score: 98.47%; Precision: 98.65%.

[0103] (7) Summary of Example 1: Example 1 verifies the effectiveness of the method of the present invention. The Ag nanoarray thin film substrate significantly enhances the spectral feature information of microplastic particles, and the classification analysis combined with the deep neural network model further improves the sensitivity and accuracy of detection. The method of the present invention is applicable to the efficient detection and classification of microplastic particles in complex environments, and has significant technical advantages and broad application prospects.

[0104] II. Example 2: Enhancing the spectral characteristics of the microscopic hyperspectral imaging technology for microplastic solutions through an Ag nanoarray thin film substrate, and combining with a deep learning algorithm to achieve high-sensitivity classification and detection of microplastic solutions. This example analyzes by measuring the dried solid sample, and specifically includes the following steps:

[0105] (1) Substrate preparation: Immerse the silicon substrate in a freshly prepared sulfuric acid-hydrogen peroxide solution (98% H 2 SO 4 :30% H 2 O 2 = 7:3) for 10 minutes to thoroughly clean the surface. Rinse three times with deionized water and ethanol respectively, and each rinse is accompanied by 10 minutes of ultrasonic cleaning to ensure the removal of any residual contaminants.

[0106] (2) Silver film deposition: In a thermal evaporation system, fix the clean silicon wafer and place an ultra-pure silver target material in the evaporation boat. Deposit silver on the clean silicon substrate by thermal evaporation to form a film about 10 nm thick. After deposition, take out the silver-coated silicon substrate and perform an annealing treatment in a muffle furnace at 310 °C for 47 minutes to promote the formation of well-defined silver nanoarrays on the substrate.

[0107] (3) Microplastic sample preparation: Ultrasonically treat the aqueous microplastic solution to ensure uniform dispersion. Use a micropipette to precisely deposit 2.5 μL of the microplastic solution onto the prepared substrate, and then dry it to obtain the dried sample.

[0108] (4) Microscopic hyperspectral imaging: Use a customized microscopic hyperspectral imaging system, including a microscope and an SM320 hyperspectral imaging system, with a magnification of 200 times.

[0109] (5) Data processing and classification: Divide the samples into six categories: PS (polystyrene), PET (polyethylene terephthalate), PMMA (polymethyl methacrylate), PS+PET, PS+PMMA, and PS+PET+PMMA. Divide the hyperspectral data into regions of 15×15 pixels, and each region is used as an individual training sample, obtaining a total of 19,008 training samples. Divide the dataset evenly into a training set and a test set, each accounting for 50%. Perform Savitzky-Golay smoothing and Z-score normalization on the hyperspectral data. Use a convolutional attention hybrid neural network (CNN-CPCA) model and a support vector machine (SVM) for classification, and all models are developed and simulated on a computer equipped with a GTX4090 GPU using PyTorch.

[0110] (6) Result analysis: Micro-HIS test of microplastic solid samples: The average reflection spectra of PS microplastic solid samples on the Si substrate and the Ag nanoarray substrate are shown respectively in Figure 9 and Figure 10 . These charts compare the reflection spectra of PS microplastic samples on the Si substrate and the Ag nanoarray substrate. On the Si substrate, the reflection spectra of PS in the regions with fewer and more samples ( Figure 9 ) are very close to the spectrum of the pure Si substrate, indicating that the strong substrate reflection overwhelms the subtle reflection signal of the microplastics, resulting in a poor signal-to-noise ratio for different microplastics. In contrast, when comparing the reflection spectra of the dried solid samples of PS microplastics on the Ag nanoarray substrate ( Figure 10 ), it is found that there are obvious differences between the reflection spectra of PS in the regions with fewer (point 2) and more (point 3) samples and the reflection spectrum of the substrate. The peaks that appear separately at 600 - 700 are the reflection spectrum signals of the samples, and the substrate reflection spectrum does not overlap with the sample spectrum, which significantly increases the signal-to-noise ratio of the sample spectrum. At the same time, the reflection spectrum peak (point 3) in the region with more samples is more obvious.

[0111] III. Comparative Example 1: This comparative example is a comparative example of Example 1. The only difference is that the substrate of the microplastic solution is replaced with an ordinary silicon wafer substrate.

[0112] Classification performance of the Si substrate: The overall classification accuracy shown by the confusion matrix of the Si substrate ( Figure 11 ) is only 88.70%. Quantitative analysis shows that the classification error rate of the Si substrate is about 11.30%, while the Ag nanoarray substrate achieves a significantly lower error rate, only 1.35%. Compared with the Si substrate, the error rate of the Ag nanoarray substrate is reduced by 8 times, which highlights the significant improvement in classification accuracy of the Ag nanoarray substrate.

[0113] Performance indicators of the Si substrate: Recall: 88.70%. F1 score: 88.71%. Precision: 89.02%.

[0114] The Ag nanoarray substrate is significantly superior to the Si substrate in all performance indicators, indicating that the Ag nanoarray substrate has obvious advantages in high-contrast, high-precision detection and classification in Micro-HSI applications. Figure 12 A high-magnification SEM image of the Ag nanoarray substrate is shown.

[0115] Based on the above embodiments, the present invention also provides a hyperspectral data analysis device based on surface plasmon resonance and a metal thin film substrate, as shown in Figure 13 . The device includes:

[0116] A placement module 01 for loading a sample to be measured onto a metal nano-film substrate with surface plasmon resonance effect prepared in advance.

[0117] An imaging module 02 for performing reflectance spectrum testing on the sample to be measured through a reflection bright-field microscopic hyperspectral imaging system to obtain hyperspectral data.

[0118] An analysis module 03 for extracting features from the hyperspectral data through a deep neural network or a support vector machine, and obtaining analysis information of the sample to be measured according to the extracted feature data.

[0119] Based on the above embodiments, the present invention further provides a terminal, the principle block diagram of which can be as Figure 14 shown. The terminal includes a processor, a memory, a network interface, and a display screen connected through a system bus. Among them, the processor of the terminal is used to provide computing and control capabilities. The memory of the terminal includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program stored in the non-volatile storage medium to run. The network interface of the terminal is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it realizes a hyperspectral data analysis method based on a surface plasmon resonance metal film substrate. The display screen of the terminal can be a liquid crystal display screen or an electronic ink display screen.

[0120] Those skilled in the art can understand that Figure 14 the principle block diagram shown in

[0121] merely shows the block diagram of some structures related to the solution of the present invention, and does not constitute a limitation on the terminal to which the solution of the present invention is applied. The specific terminal may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0122] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided by the present invention can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or an external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0123] In summary, the present invention discloses a hyperspectral data analysis method based on a plasma resonance metal thin film substrate. The method includes: pre-preparing a metal nanometer thin film substrate with a plasma resonance effect, and loading a sample to be measured on the metal nanometer thin film substrate; performing a reflection spectrum test on the sample to be measured through a reflection-type bright-field microscopic hyperspectral imaging system to obtain hyperspectral data; extracting features from the hyperspectral data through a deep neural network or a support vector machine, and obtaining analysis information of the sample to be measured according to the extracted feature data. The metal nanometer thin film substrate in the present invention can significantly enhance the scattering and absorption signals of the sample to be measured in the visible spectral range through the localized surface plasmon resonance (LSPR) effect, and at the same time effectively suppress the reflection spectrum from the substrate. This enhancement mechanism can not only achieve a high-contrast effect similar to dark-field imaging in the reflection mode, but also does not require the use of expensive optical components, so the equipment cost can be effectively reduced. In addition, the present invention also combines deep learning technology to efficiently process and analyze hyperspectral data, realizing high-sensitivity and high-resolution microscopic hyperspectral imaging analysis. Through the method of the present invention, it is possible to significantly enhance the spectral feature information of the sample at a relatively low equipment cost, and accurately and efficiently interpret the spectral information. It is applicable to fields such as biomedicine, materials science, and environmental monitoring.

[0124] It should be understood that the application of the present invention is not limited to the above examples. For those of ordinary skill in the art, improvements or modifications can be made according to the above description, and all such improvements and modifications shall fall within the protection scope of the appended claims of the present invention.

Claims

1. A hyperspectral data analysis method based on a plasma resonance metal film substrate, characterized in that: The method comprises: Preparing a metal nanofilm substrate with a plasma resonance effect in advance, and loading a sample to be tested on the metal nanofilm substrate; Performing a reflectance spectrum test on the sample to be tested by a reflective bright field microscopic hyperspectral imaging system to obtain hyperspectral data; The hyperspectral data is subjected to feature extraction through a deep neural network or a support vector machine, and analysis information of the sample to be tested is obtained based on the extracted feature data.

2. The method for analyzing hyperspectral data based on a plasma resonance metal thin film substrate according to claim 1, characterized in that: The metal used in the metal nano film substrate includes one of gold, silver and copper; the substrate used in the metal nano film substrate is a silicon substrate; A metal nano film substrate having a plasma resonance effect is prepared in advance, comprising: The metal is deposited on the cleaned silicon substrate by a preset coating method to obtain the metal nano film substrate.

3. The method for analyzing hyperspectral data based on a plasma resonance metal thin film substrate according to claim 1, characterized in that: The nanostructure morphology corresponding to the metal nanofilm substrate includes one of a nanocolumn array film, a nanohole array film, and a planar film.

4. The method for analyzing hyperspectral data based on a plasma resonance metal thin film substrate according to claim 1, characterized in that: The method comprises: loading a sample to be tested on the metal nano film substrate; Performing a dispersion treatment on the sample to be tested so that the sample to be tested is evenly dispersed, wherein the dispersion treatment includes ultrasonic treatment; The pre-treated sample to be tested is uniformly deposited on the metal nano film substrate.

5. The method for analyzing hyperspectral data based on a plasma resonance metal thin film substrate according to claim 1, characterized in that: The reflective bright field microscopic hyperspectral imaging system includes a light source, an illumination system, a microscope, and a hyperspectral imaging device; the microscope is used to magnify the sample to be tested by a preset multiple; the microscopic hyperspectral imaging system is used to obtain hyperspectral data of the sample to be tested with a spatial resolution of preset pixels, wherein each pixel contains a preset number of discrete wavelengths.

6. The method for analyzing hyperspectral data based on a plasma resonance metal thin film substrate according to claim 1, characterized in that: Extracting features from the hyperspectral data by using a deep neural network or a support vector machine, and obtaining analysis information of the sample to be tested according to the extracted feature data, wherein the method also includes: The hyperspectral data is preprocessed, and the preprocessing includes at least one processing method of noise filtering, spectrum correction, smoothing, and normalization.

7. The method for analyzing hyperspectral data based on a plasma resonance metal thin film substrate according to claim 1, characterized in that: The deep neural network includes a channel prior convolutional attention layer; the deep neural network is used to: Through the channel prior convolutional attention layer, feature extraction is performed on the hyperspectral data based on channel attention to obtain spectral features; and extracting features from the spectral features based on spatial attention to obtain joint features, wherein the joint features are used to reflect spectral information and spatial information of the hyperspectral data; The analysis information of the sample to be tested is obtained according to the joint feature.

8. A hyperspectral data analysis device based on plasma resonance and metal thin film substrate, characterized in that: The device comprises: A placement module is used to load the sample to be tested on a pre-prepared metal nano film substrate with plasma resonance effect; An imaging module is used to perform a reflectance spectrum test on the sample to be tested through a reflective bright field microscopic hyperspectral imaging system to obtain hyperspectral data; The analysis module is used to extract features from the hyperspectral data through a deep neural network or a support vector machine, and obtain analysis information of the sample to be tested based on the extracted feature data.

9. A terminal, characterized in that: The terminal includes a memory and one or more processors; the memory stores one or more programs; the programs contain instructions for executing the hyperspectral data analysis method based on a plasma resonance metal film substrate as described in any one of claims 1-7; and the processor is used to execute the programs.

10. A computer-readable storage medium having a plurality of instructions stored thereon, characterized in that: The instructions are suitable for being loaded and executed by a processor to implement the steps of the hyperspectral data analysis method based on a plasma resonance metal thin film substrate as described in any one of claims 1 to 7.

Images