A method for pre-simulating key parameters of spectral imaging chips for specific applications

Abstract

The present invention provides a key parameter pre-simulation method for a spectral imaging chip for a specific application. Based on the hyperspectral imaging data of a specific application target, M spectral segments of the hyperspectral imaging data are sorted from high to low using the influence on the classification and recognition accuracy as a weight; a classifier is selected for classification and recognition, and the classification and recognition accuracy of the first N spectral segments is obtained. The minimum number of spectral segments N with a classification and recognition accuracy greater than a set value is the number of characteristic spectral segments of the spectral imaging chip, and the N spectral segments are the central wavelengths of the spectral segments; a multispectral imaging model of the spectral imaging chip is established with a spectral resolution f n As a variable, the spectral information output by the multispectral imaging model is verified using a classifier to obtain f n The corresponding classification recognition accuracy, fitting f n The relationship curve between the classification and recognition accuracy is used to obtain the required spectral imaging chip resolution. This method can simulate and calculate the key parameters of multispectral imaging chips and guide the design and development of spectral imaging chips.

Description

Technical Field

[0001] The present invention belongs to the technical field of spectral imaging, and in particular relates to a key parameter pre-simulation method of a spectral imaging chip for specific applications. Background Art

[0002] Passive spectral imaging technology is a non-destructive detection technology that relies on the target's reflection of environmental electromagnetic waves to obtain the target's two-dimensional spatial information and spectral characteristics. It is widely used in satellite remote sensing, laboratory food safety, industrial production line quality monitoring and other fields. It has broad application prospects in industrial sorting, drone-borne environmental remote sensing, smart agriculture, intelligent security and other fields. Traditional hyperspectral imaging systems are bulky and costly. They have high resolution but usually use linear scanning to collect data at a low acquisition rate, which greatly limits their application in people's daily production and life. Driven by the strong national and social demand for non-contact detection, more miniaturized spectral detection equipment has emerged, especially the emergence of spectral imaging chips based on interference thin film spectroscopy technology, which has provided a new technical approach for the promotion of consumer-level applications of spectral imaging technology.

[0003] Interference-type spectral imaging spectroscopic thin films can be directly integrated onto CMOS sensors using mature and advanced semiconductor processes, forming an integrated spectral imaging detection chip. This compact design, high optical coupling efficiency, and low design and redesign costs offer broad application prospects in areas such as military counter-camouflage, remote sensing monitoring, industrial sorting, and product grading. Spectral imaging chips can be customized to meet the detection needs of diverse applications, effectively avoiding redundant information and reducing the processing requirements of high-dimensional data on classification and recognition algorithms. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for pre-simulating the key parameters of spectral imaging chips for specific applications, and to provide a universal design method for the customization process of spectral imaging chips for applications in specific fields. According to this method, the simulation of the key parameters of spectral imaging chips required for applications in a specific field can be achieved, thereby guiding the design and manufacture of customized spectral imaging chips in this field.

[0005] To achieve the above objectives, the present invention provides a method for pre-simulating key parameters of a spectral imaging chip for a specific application, comprising the following steps:

[0006] Acquire hyperspectral imaging data for specific application targets;

[0007] Hyperspectral imaging data consists of M spectral segments. The influence of each spectral segment on classification and recognition accuracy is used as the weight. Feature selection and feature extraction methods are used to sort the M spectral segments from high to low according to the weight.

[0008] Select a classifier based on the properties of a specific application target;

[0009] The classifier adds one spectrum segment in turn according to the order of M spectrum segments for classification and recognition, and obtains the classification and recognition accuracy X of the first N spectrum segments. N , N=1,…,M, when X N ≥X, X is the classification accuracy setting value, the minimum number of spectral segments N is the number of characteristic spectral segments of the spectral imaging chip, and N spectral segments are the central wavelengths of the spectral segments corresponding to the spectral imaging chip;

[0010] Establish a multispectral imaging model for spectral imaging chips;

[0011] With spectral resolution f n As a variable, the multispectral imaging model is input, and the spectral information output by the multispectral imaging model is verified using a classifier to obtain f n The corresponding classification recognition accuracy X n ; Obtain f by interpolation fitting n With X n The relationship curve of the classification recognition accuracy preset value X corresponds to the f n This is the resolution of the spectral imaging chip.

[0012] Furthermore, the feature selection and feature extraction method adopts the Fisher criterion, and the calculation criterion function is

[0013]

[0014] Among them, S w is the gap between sample classes, S b is the intra-class dispersion;

[0015] The spectrum segments are sorted from small to large according to the values ​​calculated by the criterion function.

[0016] Furthermore, the classifier also includes randomly extracting some samples from the hyperspectral imaging data as training sets and test sets before classification and identification, each sample has complete spectral information, training the classifier with the training set, and then classifying and identifying the test set.

[0017] Furthermore, the multispectral imaging model of the spectral imaging chip is H(S,f n ,N,λ1......λ N ,R), where S represents the real spectral data of the target, f n is the spectral resolution of the spectral imaging chip, N is the number of characteristic spectral bands, λ N is the central wavelength of the Nth characteristic spectrum band, and R is the sensor response function.

[0018] Furthermore, the classification recognition accuracy and spectral resolution are fitted into a curve using a cubic spline interpolation method.

[0019] Furthermore, the spectral imaging chip uses the classifier to perform classification and identification.

[0020] The present invention also provides a spectral imaging chip, wherein key parameters of the spectral imaging chip are obtained by using the above-mentioned key parameter pre-simulation method for the spectral imaging chip for specific applications.

[0021] Furthermore, the spectral imaging chip includes a pixel photosensitive unit and a spectral filtering unit, the spectral filtering unit is integrally grown on the pixel photosensitive unit, and each spectral filtering structure of the spectral filtering unit corresponds one-to-one to each pixel of the pixel photosensitive unit; the spectral filtering unit includes several periods, each period includes 2*2 spectral filtering structures, and each period includes 3 FP cavity structures with different central wavelengths and 1 fully transparent spectrum segment.

[0022] Furthermore, the FP cavity structure includes Bragg reflectors at both ends formed by alternating arrangement of high and low refractive index materials, and a SiO2 light-transmitting layer in the middle.

[0023] Furthermore, a matching layer is provided between the FP cavity structure and the pixel photosensitive unit, the matching layer is integrally deposited and grown on the pixel photosensitive unit, and the FP cavity structure is integrally deposited and grown on the matching layer.

[0024] Furthermore, a transition layer and a cutoff filter membrane are provided on the FP cavity structure. The transition layer is integrally deposited and grown on the FP cavity structure, and the cutoff filter membrane is integrally deposited and grown on the transition layer.

[0025] Furthermore, the cutoff filter membrane includes at least one layer. When the cutoff filter membrane has multiple layers, the multiple layers of cutoff filter membranes cut off different interference bands respectively. The bottom layer of the cutoff filter membrane is integrally deposited and grown on the transition layer, and the multiple layers of the cutoff filter membranes are bonded and fixed to each other.

[0026] Beneficial effects of the present invention:

[0027] This invention provides a method for pre-simulating the key parameters of spectral imaging chips for specific applications. Using hyperspectral data from a specific application, this method simulates and calculates key parameters such as the number of spectral segments, central wavelength, and spectral resolution of a multispectral imaging chip required for classification and identification of that specific application, thereby guiding the design and development of spectral imaging chips. Furthermore, the resulting spectral imaging chip can directly employ a pre-designed classification algorithm for that specific application, eliminating the need for secondary classification and identification algorithm design based on hardware data. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings are included to provide a further understanding of the embodiments of the present invention, constitute a part of the specification, illustrate the embodiments of the present invention, and together with the description, explain the principles of the present invention. Obviously, the drawings described below are only some embodiments of the present invention, and those skilled in the art can derive other drawings based on these drawings without inventive effort.

[0029] Figure 1 A flowchart of a method for pre-simulating key parameters of a spectral imaging chip for specific applications provided by a specific embodiment of the present invention;

[0030] Figure 2 A schematic diagram of the structure of a spectral imaging chip provided in a specific embodiment of the present invention;

[0031] Figure 3 These are transmittance simulation curves for different spectral bands provided by a specific embodiment of the present invention, where (a) is the transmittance for spectral band 1, (b) is the transmittance for spectral band 2, and (c) is the transmittance for spectral band 3. DETAILED DESCRIPTION

[0032] Below specific embodiments of the present invention are described in detail. In the following description, for the purpose of explanation and not limitation, specific details are set forth to help fully understand the present invention. However, it will be apparent to those skilled in the art that other embodiments that have departed from these specific details can also be used to practice the present invention.

[0033] It should be noted that, in order to avoid obscuring the present invention due to unnecessary details, the accompanying drawings only show the device structure and / or processing steps closely related to the solutions of the present invention, while omitting other details that are not closely related to the present invention.

[0034] This invention customizes the design of spectral imaging chips for specific applications, achieving the highest accuracy in classification and recognition of target scenarios while simultaneously removing redundant information and reducing the complexity of high-dimensional data processing. For customized snapshot spectral imaging chip products, it is necessary to pre-analyze the key parameters required for hardware product design, including but not limited to the number of spectral segments, central wavelength, and spectral resolution.

[0035] The present invention provides a method for pre-simulating key parameters of a spectral imaging chip for a specific application. The specific process is as follows: Figure 1 As shown, the following steps are included.

[0036] 1. Acquire hyperspectral imaging data for specific application targets and perform data preprocessing.

[0037] 2. Hyperspectral imaging data consists of M spectral segments. The influence of each spectral segment on classification and recognition accuracy is used as the weight. Feature selection and feature extraction methods are used to sort the M spectral segments from high to low according to the weight.

[0038] 3. Select a classifier based on the attributes of the specific application target;

[0039] 4. Randomly extract some samples from the hyperspectral imaging data as training sets and test sets. Each sample has complete spectral information. Use the training set to train the classifier, and then use the trained classifier to classify and identify the test set. The classification and identification process adds one spectral segment in sequence according to the order of M spectral segments, and obtains the classification and identification accuracy X of the first N spectral segments. N (N=1,…,M), when X N When ≥X, X is the classification accuracy setting value, the minimum number of spectral segments N is the number of characteristic spectral segments of the spectral imaging chip, and the first N spectral segments are the central wavelengths of the spectral segments corresponding to the spectral imaging chip.

[0040] 5. Establish the multi-spectral imaging model H(S,f n ,N,λ1......λ N , R), where S represents the real spectral data of the target (the real spectral data of the target is continuous spectral data, and the present invention takes the target hyperspectral imaging data as the real spectral data of the target), f n is the spectral resolution of the spectral imaging chip, N is the number of characteristic spectral bands, λ N is the central wavelength of the Nth characteristic spectrum band, and R is the sensor response function.

[0041] 6. With spectral resolution f n As a variable, the multispectral imaging model is input, and the spectral information output by the multispectral imaging model is verified using a classifier to obtain f n The corresponding classification recognition accuracy Xn; f is obtained by interpolation fitting n The relationship curve with Xn, the f corresponding to the preset value X of classification recognition accuracy in the curve n This is the resolution of the spectral imaging chip.

[0042] This determined the number of characteristic spectral bands, central wavelength, spectral resolution and classifier used in the spectral imaging chip.

[0043] The above-mentioned key parameter pre-simulation method of the spectral imaging chip for specific applications mainly completes the following two aspects.

[0044] 1. Determination of characteristic spectrum bands of spectral imaging chips for specific applications

[0045] Based on the hyperspectral imaging data of specific application targets, a classification and recognition algorithm is designed. The specific process mainly includes five steps: data preprocessing, feature selection and feature extraction, classifier design, classification and recognition, and post-processing. After feature selection and feature extraction, the original data M spectral segments are sorted from high to low according to the weight, and the first N (number of spectral segments M≥N) spectral segments are extracted as feature spectral segments. The classifier is selected and the target test set is classified and recognized to obtain the classification and recognition accuracy X. N , if the classification recognition accuracy requirement X is met (generally, X N ≥X), the classifier meets the classification and recognition requirements. At this time, the minimum N and its corresponding spectrum (λ1, ..., λ N ) refers to the number of spectral bands of the spectral imaging chip and its corresponding central wavelength for specific application targets.

[0046] 2. Calculation of spectral resolution parameters for spectral imaging chips for specific applications

[0047] According to the selected characteristic spectrum N, a multispectral imaging model H(S, f n , N, λ1,…,λ N , R), where S represents the real spectral data of the target. In the experiment, hyperspectral imaging data is used instead of real spectral data; f n is the spectral resolution of the spectral imaging chip; N is the number of characteristic spectral segments (N≥1), λ N is the central wavelength corresponding to the Nth spectral band; R is the response characteristic function of the spectral imaging chip sensor, including but not limited to Gaussian response.

[0048] The target hyperspectral imaging data is taken as the real spectral data S, the sensor response characteristic function R is selected as the Gaussian function, and N and its corresponding spectral bands (λ1, ..., λ N ), with a resolution of f n (n is a positive integer) is the only variable, simulating the target response H of the multispectral imaging chip. In this case, H is only f n Function. Convert the training set hyperspectral imaging data S into multispectral imaging chip response H, and use the trained classifier for verification to obtain the resolution f n Classification accuracy X n . According to the hardware design capability of the spectral imaging chip, the f n Value, get the corresponding classification recognition accuracy X n Fitting f n With X n Correlation curve, when X n =X, f nThat is, to design the optimal resolution solution for the multispectral imaging chip for this typical application, which can minimize the hardware development cost and complexity while meeting the classification and recognition accuracy requirements.

[0049] The technical solution of the present invention is described in detail below with reference to specific embodiments and drawings.

[0050] Taking the customization of spectral imaging chips for food classification and identification as an example, traditional RGB cameras primarily identify food based on three-channel color information and image topography, resulting in very limited detection capabilities. For example, there is a high rate of misclassification of foods within the same family. By designing a spectral imaging chip that incorporates multi-channel spectral information, the chip can identify the characteristic spectra of different samples and accurately identify their types.

[0051] 1. Collect hyperspectral imaging data of three foods to be identified from the same family as samples, select characteristic spectrum segments according to the Fisher criterion, and calculate the inter-class gap S based on the hyperspectral imaging data in the sample set. w Maximum, simultaneous intra-class dispersion S b The minimum spectrum segment, the calculation criterion function is

[0052]

[0053] Sort the spectral segments according to the calculated values ​​of the criterion function from small to large (the smaller the calculated value of the criterion function, the smaller the intra-class dispersion in the spectral segment, the largest the inter-class dispersion, and the spectral segment is most beneficial for inter-class classification and recognition).

[0054] In addition, feature extraction and feature selection methods such as LDA, PCA and wavelet transform can also be used to assign a weight to each spectral segment in terms of its impact on classification and recognition accuracy, and sort the spectral segments according to the impact weight.

[0055] 2. Select a classifier and sort the spectral segments by their influence weights, then take 1 spectral segment, 2 spectral segments, 3 spectral segments, etc. for classification and recognition, and obtain the classification and recognition accuracy X corresponding to the first N spectral segments. N (N=1,…,M), when X N When ≥X, the minimum number of spectral segments N is the number of spectral segments N.

[0056] Conventional classifiers such as the minimum distance method, maximum likelihood method, Mahalanobis distance method and other supervised learning methods can be selected according to the complexity of the application target. If the target attribute characteristics are simple, the minimum distance method can be used. If the target attribute characteristics are complex, the Mahalanobis distance method can be used.

[0057] After selecting the classifier, some data are randomly extracted from the hyperspectral imaging data as a training set, the classifier is trained with the training set, and the trained classifier is used for classification and recognition.

[0058] In the embodiment, the minimum distance method is selected to classify and identify the mixed sample, and the image after identification is post-processed to obtain the classification and identification results. Ultimately, the food can be accurately distinguished through the three characteristic spectrum segments.

[0059] 3. Based on the above classification experiments, it can be concluded that the customized spectral imaging chip must contain at least three spectral segments, with peak positions λ1, λ2 and λ3 respectively. Establish the spectral line model H(S,f n ,3,λ1,λ2,λ3,R), where f n represents different spectral resolutions, S represents the sample hyperspectral imaging data, and the response model R selects the Gaussian function.

[0060] 4. Select a group of mixed samples as follows, containing 34 different foods, and change f n The minimum distance method is used to classify the mixed samples according to the simulated multispectral imaging chip response H. The statistical classification and recognition results are shown in Table 1 below.

[0061] Table 1 Classification and recognition results at different spectral resolutions

[0062]

[0063] 5. Based on the relationship between classification accuracy and spectral resolution, use the cubic spline interpolation method to fit the correlation curve. Based on the classification accuracy requirements, determine the spectral resolution requirements for the characteristic spectrum of the customized spectral imaging chip. If the recognition accuracy is required to be no less than 99%, the spectral resolution should not exceed 25nm. Based on the number of spectral segments, the central wavelength of the spectral segments, and the spectral resolution, the customized chip optical design can be carried out.

[0064] 6. The customized chip design in this embodiment is as follows Figure 2 As shown in the figure, a spectral filter unit with a 2×2 spectral filter structure as a period is set on the CMOS pixel photosensitive unit. Each period includes spectral segments 1, 2, and 3 with central wavelengths λ1, λ2, and λ3 respectively, as well as a fully transparent spectral segment. The half-wave width of the spectral filter structure of the three spectral segments is no more than 25nm. There is no film structure on the surface of the fully transparent spectral segment. The introduction of the fully transparent spectral segment can increase the amount of light transmitted and improve the signal-to-noise ratio. The simulation results of the spectral filtering of spectral segments 1 to 3 are shown in Figure 1. Figure 3 shown.

[0065] Based on the above spectral imaging chip design results, a spectral filter unit with a 2×2 spectral filter structure as a period is integrally grown on the CMOS pixel photosensitive unit. The three spectral bands in the spectral filter unit are realized by FP cavity narrow-band filter films with three central wavelengths.

[0066] Each FP cavity structure consists of Bragg reflectors made of alternating high and low refractive index materials at both ends, and a central SiO2 light-transmitting layer. The FP cavity narrowband filter structure utilizes a mature growth process and exhibits excellent filtering performance.

[0067] To improve the transmittance of the center wavelength of the spectral imaging chip, a matching layer is placed between the FP cavity structure and the pixel photosensitive unit. The matching layer is deposited and grown integrally on the pixel photosensitive unit, and the FP cavity structure is deposited and grown integrally on the matching layer. Due to the significant refractive index difference between the FP cavity narrowband filter film material and the pixel photosensitive unit material of the spectral imaging chip during growth, direct growth results in a refractive index mismatch, a decrease in center wavelength transmittance, and low quantum efficiency of the spectral imaging system, affecting imaging performance. Therefore, by placing a matching layer between the pixel photosensitive unit and the narrowband filter film, the refractive index mismatch and decreased center wavelength transmittance issues can be effectively overcome, effectively improving the center wavelength transmittance of the spectral imaging chip structure. The matching layer can bridge the optical admittance between the pixel photosensitive unit and the film structure grown externally to improve the peak transmittance of the center wavelength.

[0068] The matching layer is prepared by depositing a layer of low-refractive-index material or a layer of high-refractive-index material and low-refractive-index material. The high-refractive-index material is at least one of Ta2O5, Ti3O5, TiO2, Si3N4, or Nb2O5, and the low-refractive-index material is at least one of SiO2, MgF2, and Al2O3.

[0069] It should be noted that a matching layer can also be deposited and grown integrally on the FP cavity structure. The externally grown matching layer is used to improve the peak transmittance of the central wavelength of the spectral imaging chip structure. A matching layer is set on the substrate side and the air side respectively. The addition of two matching layers not only improves the transmittance of the filter but also changes the half-wave width. Adding a matching layer on the substrate side can reduce the half-wave width, while adding it on the air side can increase the half-wave width. In order to make the bandwidth with the matching layer close to the bandwidth without the matching layer, the matching layer can be added to the substrate side and the air side at the same time, thereby ensuring that the half-wave width remains unchanged. When two matching layers are used, the film system structure of the substrate side matching layer includes HL or LHL, and the corresponding film system structure of the air side matching layer is LHL or LH, where H is a high refractive index material and L is a low refractive index material.

[0070] To reduce interference bands, a transition layer and a cutoff filter film can be added to the FP cavity structure. The transition layer is deposited and grown integrally on the FP cavity structure, and the cutoff filter film is deposited and grown integrally on the transition layer. By depositing the transition layer and cutoff filter film integrally on the FP cavity narrowband filter film, there is no gap between the cutoff filter film, transition layer, and narrowband filter film, resulting in high spectral transmittance and reduced energy loss. The integrated manufacturing process is unaffected by external environmental contamination, resulting in improved robustness, higher manufacturing efficiency, and greater integration. The introduction of the transition layer also effectively improves the peak transmittance of the spectral imaging chip structure.

[0071] The transition layer is used to transition between the FP cavity structure and the cutoff filter film, and is prepared by deposition of a low-refractive-index material. The low-refractive-index material includes at least one of SiO2, MgF2 and Al2O3.

[0072] Furthermore, to reduce the range of interference bands, the cutoff filter membrane can be provided in multiple layers. When the cutoff filter membrane has multiple layers, each of the multiple layers blocks different interference bands, and the multiple layers are bonded together. The first layer of the cutoff filter membrane is deposited integrally on the transition layer, and the subsequent layers of the cutoff filter membrane are bonded to the previous layer. This effectively simplifies the manufacturing process and widens the cutoff range of the interference band.

[0073] The cutoff filter membranes are fabricated by alternately depositing high-refractive-index materials and low-refractive-index materials. The high-refractive-index material of the cutoff filter membranes includes at least one of Ta2O5, Ti3O5, TiO2, Si3N4, or Nb2O5, and the low-refractive-index material of the cutoff filter membranes includes at least one of SiO2, MgF2, and Al2O3.

[0074] Spectral imaging chips for specific applications require the design of key spectral imaging chip parameters and the required classification and recognition algorithms. This invention performs feature selection and extraction, as well as classification and recognition algorithm design, based on the actual application environment and requirements. Furthermore, typical spectral imaging chip parameters are designed based on this designed algorithm. The resulting spectral imaging chip can directly use the designed classification algorithm for classification and recognition in this specific application, eliminating the need for secondary classification and recognition algorithm design based on hardware data.

[0075] Features described and / or illustrated above for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or used in place of features in other embodiments.

[0076] It should be emphasized that the term "include / comprises" when used herein refers to the presence of features, integers, steps or components, but does not exclude the presence or addition of one or more other features, integers, steps, components or combinations thereof.

[0077] The many features and advantages of these embodiments are apparent from this detailed description, and thus, the appended claims are intended to cover all such features and advantages of these embodiments that fall within the true spirit and scope thereof. Furthermore, since numerous modifications and changes will readily occur to those skilled in the art, it is not intended that the embodiments of the invention be limited to the exact construction and operation illustrated and described, but rather that all suitable modifications and equivalents be covered within the scope thereof.

[0078] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

[0079] Parts of the present invention that are not described in detail are well known to those skilled in the art.

Claims

1. A method for pre-simulating key parameters of a spectral imaging chip for a specific application, characterized in that: The steps include: Acquire hyperspectral imaging data for specific application targets; Hyperspectral imaging data consists of M spectral segments. The influence of each spectral segment on classification and recognition accuracy is used as the weight. Feature selection and feature extraction methods are used to sort the M spectral segments from high to low according to the weight. Select a classifier based on the properties of a specific application target; The classifier adds one spectrum segment in turn according to the order of M spectrum segments for classification and recognition, and obtains the classification and recognition accuracy X of the first N spectrum segments. N , N=1,…,M, when X N ≥X, X is the classification accuracy setting value, the minimum number of spectral segments N is the number of characteristic spectral segments of the spectral imaging chip, and N spectral segments are the central wavelengths of the spectral segments corresponding to the spectral imaging chip; Establish a multi-spectral imaging model of the spectral imaging chip, the multi-spectral imaging model of the spectral imaging chip is H(S,f n ,N,λ1......λ N ,R), where S represents the real spectral data of the target, f n is the spectral resolution of the spectral imaging chip, N is the number of characteristic spectral bands, λ N is the central wavelength of the Nth characteristic spectrum band, and R is the sensor response function; With spectral resolution f n As a variable, the multispectral imaging model is input, and the spectral information output by the multispectral imaging model is verified using a classifier to obtain f n The corresponding classification recognition accuracy X n ; Obtain f by interpolation fitting n With X n The relationship curve of the classification recognition accuracy preset value X corresponds to the f n This is the resolution of the spectral imaging chip.

2. The key parameter pre-simulation method according to claim 1, characterized in that: The feature selection and feature extraction method adopts Fisher criterion, and the calculation criterion function is: Among them, S w is the gap between sample classes, S b is the intra-class dispersion; The spectrum segments are sorted from small to large according to the values ​​calculated by the criterion function.

3. The key parameter pre-simulation method according to claim 1, characterized in that: The classifier also includes, before classification and identification, randomly extracting some samples from the hyperspectral imaging data as a training set and a test set, each sample having complete spectral information, training the classifier with the training set, and then classifying and identifying the test set.

4. The key parameter pre-simulation method according to claim 1, characterized in that: The classification recognition accuracy and spectral resolution are fitted into a curve using a cubic spline interpolation method.

5. The key parameter pre-simulation method according to claim 1, characterized in that: The spectral imaging chip uses the classifier to perform classification and identification.

6. A spectral imaging chip, characterized in that: The key parameters of the spectral imaging chip are obtained by using the key parameter pre-simulation method of the spectral imaging chip for specific applications according to any one of claims 1 to 5.

7. The spectral imaging chip according to claim 6, wherein: The spectral imaging chip includes a pixel photosensitive unit and a spectral filtering unit. The spectral filtering unit is integrally grown on the pixel photosensitive unit, and each spectral filtering structure of the spectral filtering unit corresponds one-to-one to each pixel of the pixel photosensitive unit; the spectral filtering unit includes several periods, each period includes 2*2 spectral filtering structures, and each period includes 3 FP cavity structures with different central wavelengths and 1 fully transparent spectrum segment.

8. The spectral imaging chip according to claim 7, wherein: The FP cavity structure includes Bragg reflectors at both ends formed by alternating arrangement of high and low refractive index materials, and a SiO2 light-transmitting layer in the middle.

9. The spectral imaging chip according to claim 8, characterized in that: A matching layer is provided between the FP cavity structure and the pixel photosensitive unit. The matching layer is integrally deposited and grown on the pixel photosensitive unit, and the FP cavity structure is integrally deposited and grown on the matching layer.

10. The spectral imaging chip according to claim 8, characterized in that: A transition layer and a cutoff filter membrane are provided on the FP cavity structure. The transition layer is deposited and grown in an integral manner on the FP cavity structure, and the cutoff filter membrane is deposited and grown in an integral manner on the transition layer.

11. The spectral imaging chip according to claim 10, characterized in that: The cutoff filter membrane includes at least one layer. When the cutoff filter membrane has multiple layers, the multiple layers of cutoff filter membranes cut off different interference bands respectively. The bottom layer of the cutoff filter membrane is integrally deposited and grown on the transition layer. The multiple layers of the cutoff filter membranes are bonded and fixed to each other.

Images