A Method for Constructing a Standard Spectral Library of Phytoplankton Based on Non-Uniform Interpolation
By constructing a standardized spectral library of phytoplankton based on non-uniform interpolation, the problems of accuracy and measurement precision in spectral classification of phytoplankton were solved, enabling rapid and efficient analysis for marine environmental monitoring.
Patent Information
- Application Number
- CN202410936114.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-12
- Publication Date
- 2025-10-28
- Estimated Expiration
- 2044-07-12
AI Technical Summary
Existing spectral classification methods for phytoplankton struggle to achieve accurate classification and quantitative analysis when faced with high similarity and instability in live fluorescence, especially in marine environmental monitoring where they suffer from problems such as large instrument size, high power consumption, and long processing time.
A standardized spectral library for phytoplankton was constructed using a non-uniform interpolation method. By acquiring the fluorescence spectra of phytoplankton under different environmental conditions, calculating the mean fluorescence spectrum and relative standard deviation, and performing non-uniform interpolation, a standardized spectral library covering most spectral regions was established. The library was then analyzed using a non-negative weighted least squares multiple linear regression method.
It significantly improves the accuracy and precision of algal classification, provides a technical method for rapidly assessing chlorophyll content and community structure of phytoplankton, and lays the foundation for monitoring marine aquatic ecological environment.
Smart Images
Figure CN118888013B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of marine aquatic ecological environment monitoring technology, specifically to a method for constructing a standardized spectral library of phytoplankton based on non-uniform interpolation. Background Technology
[0002] Accurate classification and measurement of marine planktonic diatoms and dinoflagellates are of great significance for monitoring marine aquatic ecosystems. Currently, the main detection methods for classifying marine diatoms and dinoflagellates include molecular probe methods, microscopic examination, machine learning-based image recognition methods, and fluorescence spectroscopy. While molecular probe methods, microscopic examination, and machine learning-based image recognition methods can generally provide identification and analysis at the genus and species level, they cannot yet fully meet the needs of rapid, real-time on-site monitoring. Fluorescence spectroscopy, which measures the fluorescence spectrum of living algae to invert the concentration of different algal species, has advantages such as fast measurement speed, no pretreatment required, and no need for cell disruption. Therefore, it has seen significant development in the rapid on-site monitoring of planktonic algae in large natural water areas such as freshwater lakes, reservoirs, and oceans.
[0003] Based on the needs of scientific research and industry applications, a series of commercial instruments for measuring algae using in vivo fluorescence methods have been developed both domestically and internationally. Among these, advanced foreign algal fluorescence analysis instruments, based on the principle of excitation fluorescence spectroscopy, can achieve quantitative analysis and measurement of algal chlorophyll in four spectral groups, but cannot distinguish diatoms. With further research, quantitative analysis methods for phytoplankton based on three-dimensional fluorescence spectroscopy can distinguish phytoplankton at a finer level, but suffer from problems such as large instrument size, high power consumption, and long processing time. Discrete three-dimensional fluorescence spectroscopy selectively measures fluorescence information at specific excitation and emission wavelengths based on the fluorescence spectral characteristics of the analyte, using optimized results to replace continuous three-dimensional fluorescence spectroscopy measurements. This solves the problem of high operating conditions required by continuous three-dimensional fluorescence spectrometers, effectively meeting the needs of online and in-situ monitoring of phytoplankton. However, the instability of in vivo fluorescence spectroscopy and the high similarity in pigment composition between marine diatoms and dinoflagellates mean that while reducing redundant spectral points, discrete three-dimensional fluorescence spectroscopy may reduce the differences between spectral groups of different algae, thus affecting the accuracy of marine diatom classification and measurement. To address this challenge, researchers have explored various methods to reduce the negative impact of high-similarity spectra on classification measurement results. For example, Su Rongguo et al. studied the characteristic fluorescence spectra of different marine diatoms and dinoflagellates, and improved the accuracy of identifying single algae by combining principal component analysis (PCA), Fisher's discriminant method, and Db7 wavelet analysis. However, constructing standardized spectral libraries applicable to different algae is also a key approach to overcoming the instability of live fluorescence spectra and achieving accurate classification and quantitative analysis of phytoplankton. Currently, the mainstream method for constructing standardized spectral libraries uses the concentration-normalized spectral average as the standardized spectrum for that algal phylum. While this method can achieve classification of different algal phyla to some extent, it cannot fully characterize the instability of live fluorescence in phytoplankton under different habitat conditions, thus affecting the accurate classification and measurement of phytoplankton.
[0004] In summary, developing a standardized spectral library construction method for phytoplankton suitable for discrete three-dimensional fluorescence spectroscopy is of great practical significance. This invention focuses on common diatoms and dinoflagellates. Based on obtaining the three-dimensional fluorescence spectra of living algae, it analyzes the three-dimensional fluorescence spectral characteristics of diatoms under different habitat conditions and studies a method for constructing a standardized spectral library for phytoplankton based on non-uniform interpolation. This provides scientific methodological support for the accurate quantification of phytoplankton and the monitoring and protection of aquatic ecological environments. Summary of the Invention
[0005] To address the aforementioned technical problems, this invention provides a method for constructing a standardized spectral library of phytoplankton based on non-uniform interpolation.
[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:
[0007] A method for constructing a canonical spectral library of phytoplankton based on non-uniform interpolation includes the following steps:
[0008] Step 1: Obtain the in vivo fluorescence spectra of different species of phytoplankton under different environmental conditions and of a certain category, and measure the standard concentration of chlorophyll a of the phytoplankton according to the national standard method.
[0009] Step 2: Divide the obtained in vivo fluorescence spectra of specific phyla of phytoplankton by their respective chlorophyll a standard concentration to obtain the concentration-normalized spectra of specific phyla of phytoplankton (hereinafter referred to as the standard spectrum).
[0010] Step 3: Based on the standard spectra of a specific phylum of phytoplankton, calculate the mean fluorescence spectrum M and the corresponding relative standard deviation S of that specific phylum of phytoplankton.
[0011] Step 4: The upper limit of the interpolation range for the standard spectral library of specific phyla of phytoplankton is M+2*S, and the lower limit is M-2*S.
[0012] Step 5: Perform non-uniform interpolation within the upper and lower limits of the standard spectral library interpolation range to obtain the standard spectral library of specific phyla of phytoplankton.
[0013] Step six: Repeat steps one through five to obtain standard spectral libraries for other specific phyla of phytoplankton.
[0014] Furthermore, step five specifically includes:
[0015] The i-th interpolated spectrum in the standard spectral library i for:
[0016] l i =M-2*S*(1-2k′);
[0017] Where i = 1, 2, ..., k; k represents the number of interpolated spectra, k > 1; k′ represents the interpolation coefficients:
[0018]
[0019]
[0020] Z α This represents the α quantile of the standard normal distribution; when α = 0, the quantile is represented by Z. 0. Approximately, when α = 1, the quantile is represented by Z. 0.99 approximate.
[0021] Furthermore, the number of interpolated spectra in the standard spectral library, k, is set to 10.
[0022] Furthermore, it also includes methods for resolving standardized spectral libraries, specifically including:
[0023] By combining the non-negative weighted least squares multiple linear regression method, the normalized spectral library is traversed, and the least squares results are output to realize the parsing of the normalized spectral library.
[0024] Compared with the prior art, the beneficial technical effects of the present invention are:
[0025] This invention significantly improves the accuracy and precision of algae classification, which exhibit high similarity and instability in live fluorescence. It not only provides an effective technical method for the rapid assessment of chlorophyll and community structure of phytoplankton, but also lays the foundation for future marine aquatic ecological environment monitoring and construction.
[0026] The standardized spectral library obtained by this invention is not a common single spectral line, but covers the elastic region generated in most spectral areas; this standardized spectral library construction method can solve the following problems: 1) the problem of misidentification of algal species caused by normalized spectral distortion of algae of the same phylum; 2) the problem of quantitative error in pigment concentration caused by differences in fluorescence efficiency. Attached Figure Description
[0027] Figure 1 The diagram shows the non-uniform interpolation of fluorescence spectra of phytoplankton; where (a) is a schematic diagram for dinoflagellates and (b) is a schematic diagram for diatoms.
[0028] Figure 2 Normalized fluorescence spectra of phytoplankton concentration under different habitat conditions; where (a) is the emission spectrum of diatoms at 450 nm and (b) is the emission spectrum of dinoflagellates at 450 nm.
[0029] Figure 3 The diagram shows the interpolation range of diatoms and dinoflagellates at different standard deviations; where (a) is the emission spectrum of diatoms at 450 nm and (b) is the emission spectrum of dinoflagellates at 450 nm.
[0030] Figure 4 The average relative error and computation time are given for different numbers of interpolations. Detailed Implementation
[0031] A preferred embodiment of the present invention will now be described in detail with reference to the accompanying drawings.
[0032] The present invention provides a method for constructing a canonical spectral library of phytoplankton based on non-uniform interpolation, comprising the following steps:
[0033] Step 1: Obtain the in vivo fluorescence spectra of different species of phytoplankton under different environmental conditions and of a certain category, and measure the standard concentration of chlorophyll a of the phytoplankton according to the national standard method.
[0034] Step 2: Divide the obtained in vivo fluorescence spectra of specific phyla of phytoplankton by their respective chlorophyll a standard concentration to obtain the concentration-normalized spectra of specific phyla of phytoplankton (hereinafter referred to as the standard spectrum).
[0035] Step 3: Based on the standard spectra of a specific phylum of phytoplankton, calculate the mean fluorescence spectrum M and the corresponding relative standard deviation S of that specific phylum of phytoplankton.
[0036] Step 4: The upper limit of the interpolation range for the standard spectral library of specific phyla of phytoplankton is M+2*S, and the lower limit is M-2*S.
[0037] Step 5: Perform non-uniform interpolation within the upper and lower limits of the standard spectral library interpolation range to obtain the standard spectral library of specific phyla of phytoplankton.
[0038] Step six: Repeat steps one through five to obtain standard spectral libraries for other specific phyla of phytoplankton.
[0039] The interpolated spectrum is represented as:
[0040] l i =M-2*S*(1-2k′);
[0041] Among them, l i The i-th interpolated spectrum is represented by k, where i = 1, 2, ..., k; k (k > 1) represents the number of interpolated spectra; M represents the normalized average value of a certain phylum of phytoplankton; S represents the relative standard deviation of the normalized spectrum of a certain phylum of phytoplankton; and k′ represents the interpolation coefficient.
[0042]
[0043]
[0044] Z α This represents the α quantile of the standard normal distribution; when α = 0, the quantile is represented by Z. 0. Approximately, when α = 1, the quantile is represented by Z. 0.99 approximate.
[0045] Step six: In order to verify the effectiveness of the standard spectral library construction method in this invention, a multiple linear regression analysis method is also combined to traverse the interpolated standard spectral library and output the least squares result.
[0046] Step one specifically includes:
[0047] Common diatoms and dinoflagellates (Synedra acusvar. Her, diatoms; Nitzschia, Hni, diatoms; Navicula, Hna, diatoms; Glenodinium, Kgd, dinoflagellates; Peridinium, Kpe, dinoflagellates) from coastal and freshwater areas of China were selected and cultured according to GB / T21805-2008 "Chemicals - Algal Growth Inhibition Tests". Appropriate amounts of stock solution were taken at different growth stages, diluted at a specific volume ratio, and the three-dimensional fluorescence spectra of the live algae were measured using a fluorescence spectrophotometer (F7000, Hitachi). Simultaneously, the standard chlorophyll a concentration of the algal samples was measured according to national standard methods.
[0048] Step two specifically includes:
[0049] By dividing the three-dimensional fluorescence spectrum of the living phytoplankton by the concentration of chlorophyll a obtained using a standard measurement method, the normalized three-dimensional fluorescence spectrum of chlorophyll a concentration of algae under different habitat conditions (hereinafter referred to as the standard spectrum) was obtained. Figure 2 The values represent the emission spectra of diatoms and dinoflagellates at 450 nm under different habitat conditions.
[0050] Steps three, four, and five specifically include:
[0051] Within a specific range of normalized spectral variations at algal concentrations, a certain number of non-uniform interpolations are performed to expand several three-dimensional fluorescence spectral standardization libraries based on normal distribution characteristics. The key to this method lies in accurately studying the interpolation range and optimal number of interpolations for the standardization library to ensure the scientific validity and practicality of the constructed algal fluorescence standardization library.
[0052] Optimal Interpolation Range: Studies have shown that algal species, growth cycles, and other factors can lead to instability in algal fluorescence spectra. The contribution of different habitat conditions to this instability can be reflected in the standard deviation. To obtain a suitable interpolation range for a standardized spectral library, this invention, referencing statistical principles, introduces different multiples of the standard deviation of the algal concentration-normalized spectrum and attempts to establish different difference ranges, such as... Figure 3 As shown in the figure, M represents the normalized spectral mean of the concentration of a specific algal phylum, and S represents the normalized spectral standard deviation of the concentration of a specific algal phylum.
[0053] Figure 3This indicates that the average value of the concentration-normalized spectrum generally lies in the center of the spectral variation range for a specific algal phylum, characterizing the spectral features of that phytoplankton phylum to some extent, but failing to reflect the fluorescence instability of the algae. The range M±S (±1 standard deviation) includes a portion of the measured concentration-normalized spectra, with a large number of measured spectra outside this range. The range M±2*S contains most of the normalized spectral variation region for a specific algal phylum. While the range M±3*S includes the normalized spectra of all algae, there is significant range redundancy, and the upper and lower limits of the interpolation range are severely distorted. This situation may introduce new analytical errors when resolving phytoplankton fluorescence spectra.
[0054] To accurately study the optimal interpolation range of the standard spectral library, this invention defines the parameters spectral range coverage C and correlation coefficient ρ to compare the merits of interpolation ranges with different multiples of standard deviation.
[0055]
[0056]
[0057] Where: C N The number of spectra representing the concentration-normalized spectra of a specific algal phylum that fall entirely within the given upper and lower limits; N represents the total number of concentration-normalized spectra of a specific algal phylum; x i represents the intensity of the spectrum of a specific algal phylum to be compared at a certain wavelength point; x represents the average intensity of the spectrum of a specific algal phylum to be compared; X represents the intensity of the spectral point of the normalized mean concentration spectrum of a specific algal phylum at a certain wavelength point; X represents the average intensity of the normalized mean concentration spectrum of a specific algal phylum.
[0058] The closer C is to 1, the higher the spectral coverage within a given range. The closer ρ is to 1, the higher the correlation between the upper and lower limits and the mean, and the lower the degree of distortion in the interpolated spectrum. The calculated coverage C and correlation coefficient ρ values for specific algal phyla within different given interpolation upper and lower limits are shown in Table 1.
[0059]
[0060] Table 1
[0061] Table 1 shows that, compared to M±S, the coverage rate C of diatoms and dinoflagellates in the M±2*S range increased to approximately 88%, an increase of about 60%. Compared to M±2*S, the coverage rate in the M±3*S range increased by 2%–10%, but the correlation coefficient ρ decreased from approximately 0.98 to approximately 0.6. This indicates that expanding the interpolation spectral range can significantly improve spectral coverage, but an excessively wide interpolation range may cause significant distortion of the interpolated spectrum. Overall, the M±2*S range achieves an average coverage rate of 88% and an average correlation coefficient of 0.98, which effectively balances the similarity and instability characteristics of the live fluorescence spectra of a given algal phylum, making it the optimal interpolation range for the standard spectral library.
[0062] Optimal Interpolation Number Study: Three-dimensional fluorescence spectra of diatoms and dinoflagellates at known concentrations obtained experimentally were used as the test set. Non-uniform interpolation was performed within the range of M±2*S using interpolation numbers of 1 (concentration-normalized spectral mean, hereinafter referred to as the average method), 5, 10, 20, 30, 40, 50, and 60. Combined with a non-negative weighted least squares multiple linear regression method, the analytical result with the minimum residual in the interpolated spectral library traversal results was output. The optimal number of interpolations for the standard spectral library within the range of M±2*S was studied, using the absolute value of the relative error (ARE) between the analytical result and the standard chlorophyll a concentration of the sample, and the average computation time for spectral analysis, as evaluation criteria. The analytical results for the 17 samples in the test set under different interpolation numbers are shown in Tables 2 and 3.
[0063] Table 2: Comparison of test set sample concentration analysis results under different interpolation numbers
[0064]
[0065]
[0066] Table 3. Comparison of ARE results for test set sample concentration analysis under different interpolation numbers.
[0067]
[0068] Tables 2 and 3 show that when the averaging method was used to normalize the spectral library, there were 3 completely misidentified samples (accounting for 17.6%). When the interpolation method was used to normalize the spectral library test set samples, the phenomenon of completely misidentification was completely avoided; and the mean measurement error gradually decreased. Given that the computational load increases exponentially with the number of interpolations as the number of phytoplankton species to be analyzed increases, computational efficiency must be considered while ensuring the accuracy of the analytical results. The average relative error and computation time of the test set experimental samples analysis results under different interpolation numbers are shown below. Figure 4 As shown.
[0069] Figure 4 The results show that when the average method is used to normalize the spectral library for the test set samples, the average ARE of the resolved results is 43.3%. When the number of interpolations is increased to 5 and 10 spectra, the ARE decreases to 31.0% and 28.9%, respectively, improving the accuracy of the measurement results by approximately 13%. When the number of interpolations increases to 20, 30, 40, and 50, the decreasing trend of ARE is not significant. In terms of computational efficiency, the average method has the shortest computation time, requiring only 10.3 ms. When the number of interpolations is 5 and 10, the computation time increases to 15.3 ms and 20.6 ms, respectively. Thereafter, as the number of interpolations increases to 60, the computation time increases approximately uniformly to 81.6 ms. This invention suggests that 10 spectra within the interpolation range is the optimal interpolation scheme for constructing the normalized spectral library.
[0070] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention, and no reference numerals in the claims should be construed as limiting the scope of the claims.
[0071] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.
Claims
1. A method for constructing a standardized spectral library of phytoplankton based on non-uniform interpolation, characterized in that, Includes the following steps: Step 1: Obtain the in vivo fluorescence spectra of different species of phytoplankton under different environmental conditions and of a certain category, and measure the standard concentration of chlorophyll a of the phytoplankton according to the national standard method. Step 2: Divide the obtained in vivo fluorescence spectra of specific phyla of phytoplankton by their respective chlorophyll a standard concentration to obtain the concentration-normalized spectra of specific phyla of phytoplankton, hereinafter referred to as the standard spectra. Step 3: Based on the standard spectra of a specific phylum of phytoplankton, calculate the mean fluorescence spectrum M and the corresponding relative standard deviation S of that specific phylum of phytoplankton. Step 4: The upper limit of the interpolation range for the standard spectral library of specific phyla of phytoplankton is M+2*S, and the lower limit is M-2*S. Step 5: Perform non-uniform interpolation within the upper and lower limits of the standard spectral library interpolation range to obtain a standard spectral library for a specific phylum of phytoplankton. Step six: Repeat steps one through five to obtain standard spectral libraries for other specific phyla of phytoplankton.
2. The method for constructing a standardized spectral library of phytoplankton based on non-uniform interpolation according to claim 1, characterized in that, Step five specifically includes: The i-th interpolated spectrum in the standard spectral library i for: the i =M-2*S*(1-2k′); Where i = 1, 2, ..., k; k represents the number of interpolated spectra, k > 1; k′ represents the interpolation coefficients: Z α This represents the α quantile of the standard normal distribution; when α = 0, the quantile is represented by Z. 0.01 Approximately, when α = 1, the quantile is represented by Z. 0.99 approximate.
3. The method for constructing a standardized spectral library of phytoplankton based on non-uniform interpolation according to claim 1, characterized in that, The number of interpolated spectra in the standard spectral library, k, is set to 10.
4. The method for constructing a standardized spectral library of phytoplankton based on non-uniform interpolation according to claim 1, characterized in that, It also includes methods for resolving standardized spectral libraries, specifically including: By combining the non-negative weighted least squares multiple linear regression method, the normalized spectral library is traversed, and the least squares results are output to realize the parsing of the normalized spectral library.