Rapid Detection Method and Device for Water Pollutants Based on Multi-Band Spectroscopy

Abstract

This invention discloses a rapid detection method and apparatus for water pollutants based on multi-band spectroscopy, belonging to the field of water pollution detection technology. The rapid detection method and apparatus for water pollutants based on multi-band spectroscopy includes: S1, preprocessing real-time spectral detection data, equipment configuration data, and historical environmental auxiliary data; S2, outputting a determination of the separability between scattering effects and pollutant absorption effects based on scattering and absorption aliasing analysis results; S3, selecting and forming a set of effective inversion spectral bands based on spectral band reliability allocation analysis results; and S4, outputting pollutant concentration estimates and result labels based on the issuance reliability assessment analysis results. This invention solves the problem that existing water pollutant detection methods, under conditions of high turbidity or water containing bubbles, struggle to accurately distinguish the differences in spectral response caused by scattering and pollutant absorption effects in different spectral bands, easily leading to distorted detection results.

Description

Technical Field

[0001] This invention relates to the field of water pollution detection technology, specifically to a rapid detection method and device for water pollutants based on multi-spectral spectroscopy. Background Technology

[0002] Water pollutant detection is a crucial technology in environmental monitoring, water resource management, and public safety, widely applied in scenarios such as drinking water safety assurance, industrial wastewater discharge monitoring, and surface and groundwater quality assessment. With the increasing demand for online and automated monitoring, optical-based water quality detection methods, due to their fast response, continuous operation, and suitability for on-site deployment, are gradually becoming a key technological approach for water pollutant detection. Existing optical detection technologies typically utilize the absorption, scattering, or transmission characteristics of different wavelengths of light in water to obtain spectral information reflecting changes in water composition, and then combine this with mathematical models to estimate pollutant concentrations. Under complex water conditions, changes in suspended particles, bubbles, and background turbidity can significantly affect optical signals. Therefore, improving the stability and reliability of detection results while ensuring real-time performance remains a key research focus in the field of optical detection of water pollutants.

[0003] For example, the invention patent with publication number CN118464871B discloses a method and system for detecting water pollutants based on surface-enhanced Raman spectroscopy. This method involves collecting water samples from a target water body and filtering them using a filter membrane with a preset pore size range. The filtered samples undergo preliminary testing. Based on the results from a turbidimeter, pH meter, and dissolved oxygen meter, the main pollutant type in the water body is determined to be either soluble or insoluble. The samples are then treated using either chemical or surface treatment methods. Specifically, soluble pollutant samples are obtained through chemical treatments such as enrichment, precipitation, or extraction to obtain the test samples, while insoluble pollutant samples are obtained through surface treatments such as solvent cleaning, ultrasonic cleaning, and drying. Subsequently, a surface-enhanced Raman spectroscopy substrate with a metal nanostructure or a surface-modified metal structure is used to perform spectral detection on the treated samples, thereby obtaining the detection results of water pollutants. This realizes a water pollutant detection process based on the differences in pollutant solubility characteristics.

[0004] For example, the invention patent with announcement number CN118465139B discloses a rapid screening and identification method for phosphorus-based disinfection byproducts based on mass spectrometry fragments. This method utilizes gas chromatography-triple quadrupole mass spectrometry to analyze the retention time and mass spectrometry fragment characteristics of various phosphorus-based compound standards, establishing a full-scan mass spectrometry screening method for phosphorus-based disinfection byproducts. By performing flocculation sedimentation, filtration, disinfection treatment, and organic solvent extraction on water samples from different sources, combined with the determination of total organic carbon and ammonia nitrogen concentrations, potential phosphorus-based disinfection byproducts in the disinfected water samples are pretreated. Based on this, full-scan mass spectrometry analysis is used, and unknown disinfection byproducts are located and screened according to the mass-to-charge ratio of characteristic mass spectrometry fragments. Derivatization treatment is used to improve the detection capability of non-volatile components, thereby achieving rapid screening and identification of phosphorus-based disinfection byproducts in water bodies.

[0005] Existing methods often rely on offline sampling, sample filtering, chemical treatment, or complex pretreatment processes. The detection process often requires manual intervention and multiple steps, making it difficult to adapt to continuous online monitoring scenarios. At the same time, some spectroscopic or mass spectrometry detection technologies are quite sensitive to the physical state of the water body. Under conditions of high turbidity, the presence of bubbles, or hydrodynamic fluctuations, scattering effects, baseline drift, and pollutant absorption characteristics are easily superimposed, leading to unstable spectral responses and increasing the risk of result distortion. Existing methods usually focus on a single detection step or a single discrimination dimension, lacking a systematic joint evaluation mechanism. In complex aquatic environments, it is difficult to simultaneously consider detection speed, stability, and result reliability, thus limiting their applicability in practical engineering and online applications.

[0006] Therefore, in order to address the above problems, there is an urgent need for rapid detection methods and devices for water pollutants based on multi-spectral spectroscopy. Summary of the Invention

[0007] Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a rapid detection method and device for water pollutants based on multi-band spectroscopy. This solves the problem that existing water pollutant detection methods, under conditions of high turbidity or water containing bubbles, are difficult to accurately distinguish the differences in spectral response caused by scattering effects and pollutant absorption effects in different spectral bands, which easily leads to distorted detection results.

[0008] Technical solution To achieve the above objectives, the present invention provides the following technical solution: a rapid detection method for water pollutants based on multi-band spectroscopy, comprising: S1, acquiring real-time spectral detection data, obtaining equipment configuration data and environmental historical auxiliary data; preprocessing the real-time spectral detection data, equipment configuration data, and environmental historical auxiliary data; S2, performing scattering and absorption aliasing analysis on the real-time spectral detection data, equipment configuration data, and environmental historical auxiliary data, and outputting the separability determination result of scattering effect and pollutant absorption effect based on the scattering and absorption aliasing analysis result; S3, performing spectral reliability allocation analysis on the residual stability characteristics and spectral morphology distribution consistency characteristics of each spectral band that has passed the aliasing determination, and selecting and forming an effective inversion spectral band set based on the spectral reliability allocation analysis result; S4, performing absorption model inversion and residual statistical analysis on the effective inversion spectral band set, and performing issuance reliability assessment analysis, and outputting pollutant concentration estimates and result labels based on the issuance reliability assessment analysis result.

[0009] Further, the specific process of acquiring real-time spectral band detection data, equipment configuration data, and historical environmental auxiliary data is as follows: Real-time spectral band detection data includes: multi-band raw spectral intensity sequences, spectral band center wavelength nominal value sequences, integration time setting value sequences, light source temperature data, dark field reference spectral intensity sequences, white reference spectral intensity sequences, flow cell inlet pressure data, flow cell outlet pressure data, flow cell volumetric flow rate data, water temperature data, turbidity sensor reading data, bubble scattering pulse count sequences, and frame timestamp sequences; Equipment configuration data includes: spectral band group boundary wavelength data, scattering-sensitive spectral band set index data, and absorption discrimination spectral band set index data; Historical environmental auxiliary data is acquired and an environmental historical auxiliary database is created. This data includes: historical dark field reference spectral mean sequences, historical white reference spectral mean sequences, historical turbidity statistical mean, historical turbidity statistical standard deviation, and historical flow cell pressure difference statistical quantile sequences.

[0010] Furthermore, the specific preprocessing process for real-time spectral band detection data, equipment configuration data, and environmental historical auxiliary data is as follows: outlier removal is performed on the real-time spectral band detection data; intra-spectral smoothing is performed on the original multi-band spectral intensity sequence; short-term missing data completion is performed on the real-time spectral band detection data using a piecewise linear interpolation completion algorithm and a data validity judgment method based on the maximum allowable continuous missing measurement length constraint; dark-field reference subtraction and white-reference normalization correction algorithms are used to perform dark-field reference subtraction and white-reference normalization on the original multi-band spectral intensity sequence to obtain a corrected spectral intensity sequence; the corrected spectral intensity sequence is mapped to a logarithmic spectral intensity sequence using logarithmic transformation and a logarithmic mapping processing method based on numerical lower limit constraints; the corrected spectral intensity sequence, logarithmic spectral intensity sequence, turbidity sensor reading data, bubble scattering pulse count sequence, flow tank volumetric flow rate data, and water temperature data are standardized and normalized using a zero-mean unit variance standardization algorithm based on sliding window statistics and a minimum-maximum normalization algorithm, respectively.

[0011] Furthermore, the specific process of performing scattering and absorption aliasing analysis on real-time spectral band detection data, equipment configuration data, and environmental historical auxiliary data is as follows: For the corrected spectral intensity sequence, within the spectral band sets corresponding to the scattering-sensitive spectral band set index data and the absorption-discriminative spectral band set index data, respectively, the spectral band intensity normalization ratio calculation algorithm is executed to obtain the scattering structure spectral band distribution and the absorption structure spectral band distribution; the Jensen-Shannon divergence calculation algorithm is executed on the scattering structure spectral band distribution and the absorption structure spectral band distribution to obtain the Jensen-Shannon divergence; for the bubble scattering pulse counting sequence, the Poisson excessive dispersion estimation and abrupt change point counting algorithm is executed within a sliding window to obtain the transient intensity of bubble perturbation; for the logarithmic spectral intensity sequence, a continuous spectrum removal algorithm is executed within the spectral band set indicated by the absorption-discriminative spectral band set index data to obtain the residual curve, and the robust local regression residual ratio calculation algorithm is executed on the residual curve to obtain the absorption structure. Interpret the metric; calculate the difference between the flow cell inlet pressure data and the flow cell outlet pressure data to obtain the flow cell pressure difference sequence, and then perform a quantile mapping algorithm on the historical flow cell pressure difference statistical quantile sequence to obtain the flow cell pressure difference perturbation value; calculate the Jensen-Shannon divergence between the scattering structure spectral distribution and the absorption structure spectral distribution; add one to the transient intensity of the bubble perturbation and take the natural logarithm to obtain the logarithmic modulation term; multiply the Jensen-Shannon divergence with the logarithmic modulation term to form the coupling difference term between the scattering structure difference and the bubble perturbation; calculate the exponential decay form of the absorption structure interpretation metric and add one to it to form the absorption interpretability suppression term; divide the coupling difference term by the absorption interpretability suppression term to obtain the aliasing basis term; multiply the aliasing basis term with the exponentially amplified form of the flow cell pressure difference perturbation value, perform exponential mapping on the result, and subtract one from the mapping result to obtain the scattering and absorption aliasing discrimination value.

[0012] Furthermore, the specific process for determining the separability of scattering effects and pollutant absorption effects based on the scattering absorption aliasing analysis results is as follows: For the corrected spectral intensity sequence within the spectral set corresponding to the scattering-sensitive spectral set index data, perform the spectral slope sign consistency test and local monotonicity test algorithms to obtain the scattering-dominant spectral set; For the logarithmic spectral intensity sequence within the spectral set corresponding to the absorption discrimination spectral set index data, perform the continuous spectrum removal absorption valley depth significance test algorithm to obtain the absorption-dominant spectral set; Real-time comparison of the scattering absorption aliasing discrimination value and the scattering absorption aliasing discrimination threshold: When the scattering absorption aliasing discrimination value is less than the scattering absorption aliasing threshold... At that time, the output sets of scattering-dominant and absorption-dominant spectral bands are kept constant while maintaining the pump control quantity corresponding to the flow cell volumetric flow rate data. When the scattering-absorption aliasing discrimination value is greater than or equal to the scattering-absorption aliasing discrimination threshold, defoaming and flow stabilization are performed: the bypass defoaming valve is opened, and the pump control quantity is adjusted in a closed loop based on the coefficient of variation calculation result of the flow cell volumetric flow rate data within the sliding time window, while the integral time setting value is increased. After the treatment is completed, the scattering-absorption aliasing discrimination value is re-acquired and recalculated. If the recalculated result is still greater than or equal to the scattering-absorption aliasing discrimination threshold, the bypass defoaming valve is kept open, the aliasing continuous marker is output, and the entry into the spectral band confidence allocation analysis is blocked.

[0013] Furthermore, the specific process of performing spectral reliability allocation analysis on the stability characteristics of the residuals and the consistency characteristics of the spectral morphology distribution of each spectral segment through aliasing discrimination is as follows: A robust local regression scattering baseline fitting algorithm is applied to the logarithmic spectral intensity sequence to obtain the scattering baseline; baseline residual calculation is performed on the spectral segments to obtain the spectral residuals, and a median absolute deviation calculation algorithm is applied to obtain the median absolute deviation of the residuals; a kernel density estimation distribution fitting algorithm is applied to the corrected spectral intensity sequence to obtain the spectral morphology distribution; a historical distribution fitting update algorithm is applied to the historical white reference spectral mean sequence and the historical dark field reference spectral mean sequence to obtain the historical spectral segment morphology distribution; and a Kohlbek-Leibler dispersion algorithm is applied to the spectral segment morphology distribution and the historical spectral segment morphology distribution. The algorithm calculates the Kulbeck-Leibler divergence; the absolute value of the spectral residual is calculated and divided by the sum of the median absolute deviation of the residual and the division-by-zero protection constant to obtain the degree of deviation of the spectral residual; an exponential decay mapping is performed on the degree of deviation of the spectral residual to form the spectral residual confidence decay factor; the Kulbeck-Leibler divergence between the spectral morphology distribution and the historical spectral morphology distribution is calculated; the Kulbeck-Leibler divergence is divided by the sum of the Kulbeck-Leibler divergence and a certain factor to obtain the spectral morphology distribution offset ratio; the spectral morphology distribution offset ratio is subtracted from the certain factor to obtain the spectral morphology consistency modulation factor; the spectral residual confidence decay factor is multiplied by the spectral morphology consistency modulation factor to obtain the spectral reliability allocation value.

[0014] Furthermore, the specific process of selecting and forming a valid inverted spectral band set based on the spectral band confidence allocation analysis results is as follows: Within the absorbing dominant spectral band set, the spectral band confidence allocation values ​​are jointly filtered by quantile truncation and spectral band confidence threshold to obtain a valid inverted spectral band set; the spectral band confidence allocation values ​​and the spectral band confidence allocation threshold are compared in real time: when the spectral band confidence allocation value is less than the spectral band confidence allocation threshold, it is determined that the inversion input is insufficient, and an inversion input insufficient flag is output; median statistical fusion processing is performed on the spectral band confidence allocation values ​​in adjacent acquisition windows; at the same time, the valid inverted spectral band set is tightened within the absorbing dominant spectral band set, and the spectral band confidence allocation values ​​are filtered by quantile; when the spectral band confidence allocation value is greater than or equal to the spectral band confidence allocation threshold, it is determined that the inversion input meets the requirements, and the spectral band confidence allocation value set and the valid inverted spectral band set are output.

[0015] Furthermore, the specific process of performing absorption model inversion and residual statistical analysis on the effective inversion spectral band set, and conducting issuance credibility assessment analysis, is as follows: The absorption model residual and scattering baseline residual are calculated on the effective inversion spectral band set, and the log-likelihood ratio is obtained using the Gaussian likelihood function log-likelihood ratio calculation algorithm; the flow cell pressure difference sequence is processed within a sliding window using the coefficient of variation calculation and quantile mapping algorithms, and then summed to obtain the flow cell pressure difference disturbance suppression amount; the median scaling estimation and quantile mapping algorithm are applied to the absorption model residual on the effective inversion spectral band set to obtain the inversion residual dispersion penalty; based on the effective inversion spectral band set, under the constraints of the corrected spectral intensity sequence and the log spectral intensity sequence, the pollutant concentration estimate is calculated using the absorption model parameter inversion algorithm. The absorption model parameter inversion algorithm is based on the physical mechanism of pollutant spectral absorption, and adopts an absorption model conforming to the Beer-Lambert law as the basic expression, allowing the introduction of nonlinear parameter estimation methods to solve the model parameters; the log-likelihood ratio is multiplied by the exponential decay mapping result of the scattering absorption aliasing discriminant value to obtain the inversion confidence gain term; the flow cell pressure difference disturbance suppression amount is calculated and added to form the flow disturbance suppression term; the inversion residual dispersion penalty is calculated and added to form the inversion stability penalty term; the inversion confidence gain term is divided by the product of the flow disturbance suppression term and the inversion stability penalty term to obtain the comprehensive issuance driving amount; the negative of the comprehensive issuance driving amount is subjected to exponential mapping and the mapping result is subtracted by one to obtain the concentration issuance confidence value.

[0016] Furthermore, the specific process of outputting the pollutant concentration estimate and result label based on the issuance credibility assessment analysis results is as follows: Real-time comparison of the concentration issuance credibility value and the concentration issuance credibility threshold: When the concentration issuance credibility value is less than the concentration issuance credibility threshold, an unissued concentration inversion label is output, and the unissued concentration inversion label is associated with the timestamp sequence of the corresponding collection window; When the concentration issuance credibility value is greater than or equal to the concentration issuance credibility threshold, the pollutant concentration estimate and the concentration inversion issuance label are output, and a pollutant concentration database is created and archived; At the same time, the concentration inversion issuance label is associated with the timestamp sequence of the corresponding collection window.

[0017] The second aspect of this invention provides a rapid detection device for water pollutants based on multi-band spectroscopy, comprising: an acquisition and preprocessing module for acquiring real-time spectral detection data, obtaining equipment configuration data and environmental historical auxiliary data; preprocessing the real-time spectral detection data, equipment configuration data, and environmental historical auxiliary data; a scattering and absorption aliasing discrimination module for performing scattering and absorption aliasing analysis on the real-time spectral detection data, equipment configuration data, and environmental historical auxiliary data, and outputting a determination result on the separability of scattering effect and pollutant absorption effect based on the scattering and absorption aliasing analysis results; a spectral band reliability allocation module for performing spectral band reliability allocation analysis on the residual stability characteristics and spectral band morphological distribution consistency characteristics of each spectral band that has passed the aliasing discrimination, and screening to form a set of effective inversion spectral bands based on the spectral band reliability allocation analysis results; and a pollutant concentration inversion and issuance module for performing absorption model inversion and residual statistical analysis on the effective inversion spectral band set, and performing issuance reliability assessment analysis, and outputting pollutant concentration estimates and result labels based on the issuance reliability assessment analysis results.

[0018] Beneficial effects The present invention has the following beneficial effects: (1) This invention separates and analyzes the scattering effect and the pollutant absorption effect of multi-band spectral information, thereby achieving the effect of stably distinguishing the pollutant absorption response and scattering disturbance under complex water optical conditions, effectively solving the problem of distorted detection results caused by the difficulty in distinguishing scattering changes and concentration changes in the prior art.

[0019] (2) This invention constructs a spectral reliability allocation mechanism based on spectral residual stability and spectral morphology consistency, thereby achieving an adaptive screening effect for effective inversion spectral segments, effectively solving the problem of introducing unreliable information in the prior art where all spectral segments participate in inversion with equal weight.

[0020] (3) This invention incorporates the statistical characteristics of water flow state, bubble disturbance and spectral residual into a unified analysis framework, thereby achieving explicit characterization and suppression of the influence of physical disturbance of water on the detection results, effectively solving the problem that changes in physical environment are implicitly superimposed into the detection results in the prior art.

[0021] (4) This invention forms a complete analysis link from data acquisition, aliasing discrimination, reliable spectral screening to result issuance, thereby improving the overall stability and engineering usability of the water pollutant detection process and effectively solving the problems of fragmented detection process and difficulty in supporting online applications in the prior art.

[0022] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0023] Figure 1 This is a flowchart of the rapid detection method for water pollutants based on multi-band spectroscopy of the present invention; Figure 2 This is a structural diagram of the rapid detection device for water pollutants based on multi-band spectroscopy according to the present invention; Figure 3 This is a graph showing the comparison between the Jensen-Shannon divergence and the transient intensity of bubble perturbation in this invention; Figure 4 This is a flowchart of the dynamic allocation and closed-loop screening control of spectral band reliability in this invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Please see Figures 1-4This invention provides a technical solution: a rapid detection method for water pollutants based on multi-band spectroscopy, comprising the following steps: S1, acquiring real-time spectral detection data, obtaining equipment configuration data and environmental historical auxiliary data; preprocessing the real-time spectral detection data, equipment configuration data, and environmental historical auxiliary data; S2, performing scattering and absorption aliasing analysis on the real-time spectral detection data, equipment configuration data, and environmental historical auxiliary data, and outputting the separability determination result of scattering effect and pollutant absorption effect based on the scattering and absorption aliasing analysis result; S3, performing spectral reliability allocation analysis on the residual stability characteristics and spectral morphology distribution consistency characteristics of each spectral band that has passed the aliasing judgment, and selecting and forming an effective inversion spectral band set based on the spectral reliability allocation analysis result; S4, performing absorption model inversion and residual statistical analysis on the effective inversion spectral band set, and performing issuance reliability assessment analysis, and outputting pollutant concentration estimates and result labels based on the issuance reliability assessment analysis result.

[0026] Specifically, the process of acquiring real-time spectral band detection data and obtaining equipment configuration data and environmental historical auxiliary data is as follows: real-time spectral band detection data includes: multi-band original spectral intensity sequence, spectral band center wavelength nominal value sequence, integration time setting value sequence, light source temperature data, dark field reference spectral intensity sequence, white reference spectral intensity sequence, flow cell inlet pressure data, flow cell outlet pressure data, flow cell volumetric flow rate data, water temperature data, turbidity sensor reading data, bubble scattering pulse count sequence, and frame timestamp sequence. The real-time spectral band detection data is directly acquired by the spectral acquisition unit, pressure sensor, flow meter, temperature sensor, turbidity sensor, bubble pulse counter, and system clock, and is a quantifiable numerical sequence.

[0027] Acquire device configuration data, which includes: spectral band grouping boundary wavelength data, scattering sensitive spectral band set index data, and absorption discrimination spectral band set index data. The scattering sensitive spectral band set index data and the absorption discrimination spectral band set index data are generated by comparing the spectral band center wavelength nominal value sequence and the spectral band grouping boundary wavelength data according to comparison rules, and are used to form a stable spectral band functional grouping aperture on the device side.

[0028] Historical environmental auxiliary data is acquired and an environmental historical auxiliary database is created. This data includes: historical dark-field reference spectral mean sequence, historical white reference spectral mean sequence, historical turbidity statistical mean, historical turbidity statistical standard deviation, and historical flow cell pressure differential statistical quantile sequence. The historical environmental auxiliary data is generated from historically acquired dark-field reference spectral intensity sequence, white reference spectral intensity sequence, turbidity sensor readings, flow cell inlet pressure data, flow cell outlet pressure data, and light source temperature data through sliding window statistics and quantile statistics. The sliding window length is determined based on the timescale under long-term stable operating conditions and is used to cover the acquisition cycles corresponding to multiple continuous integration time settings. The window is updated in a rolling manner based on the number of acquisition windows. Quantile statistical points are selected from multiple quantiles reflecting the steady-state distribution boundary and sensitivity to anomalous shifts, used to characterize the statistical range of flow and optical states under historical operating conditions, and are read from the environmental historical auxiliary database.

[0029] In this implementation plan, multi-band spectral observation data, flow pool operation status data, and environmental background information are uniformly collected and organized to form a complete observation description covering optical response, flow state, and environmental conditions. This ensures that the spectral data within each acquisition window has a clear physical context and a stable configuration caliber, providing the necessary information foundation for subsequent differentiation between scattering effects and pollutant absorption effects. At the equipment configuration level, the spectral band functional grouping rules are solidified to ensure that different spectral bands maintain consistent physical directivity and discriminative semantics during long-term operation. At the environmental history level, historical statistical references are introduced to construct a comparable background benchmark, enabling real-time observations to undergo scale correction and offset identification under historical steady-state constraints. This comprehensively improves the comparability, interpretability, and reliability of subsequent discriminant analysis of spectral data under complex water conditions.

[0030] Specifically, the preprocessing process for real-time spectral band detection data, equipment configuration data, and environmental historical auxiliary data is as follows: Anomaly removal is performed on the real-time spectral band detection data using a sliding window Hamper filter algorithm to suppress transient spikes that interfere with subsequent scattering baseline fitting and absorption residual calculation; intra-spectral smoothing is performed on the original multi-band spectral intensity sequence using a Savitsky-Gorye polynomial smoothing algorithm to suppress high-frequency noise while maintaining the absorption valley shape; short-term missing data completion is performed on the real-time spectral band detection data using a piecewise linear interpolation completion algorithm and a data validity judgment method based on the maximum allowable continuous missing length constraint. When the continuous missing length exceeds the allowable missing threshold, the data for the corresponding time period is judged as unusable and does not participate in subsequent statistical and discrimination calculations, thus preventing missing segments from entering subsequent discrimination and causing parameter estimation shifts; and dark field reference subtraction and white reference normalization correction algorithms are used to perform short-term missing data completion on the original multi-band spectral intensity sequence. Dark-field reference subtraction and white-field reference normalization are performed to obtain a corrected spectral intensity sequence, eliminating the differences in dark current and optical path gain, and forming comparable spectral observations across time periods. Through logarithmic transformation and logarithmic mapping based on numerical lower bound constraints, the corrected spectral intensity sequence is mapped to a logarithmic spectral intensity sequence. Before the logarithmic transformation, a minimum positive number constraint is introduced for spectral intensity values ​​close to zero to avoid numerical overflow and ensure the numerical stability of logarithmic operations, ensuring numerical stability for subsequent baseline fitting and divergence calculations. The corrected spectral intensity sequence, logarithmic spectral intensity sequence, turbidity sensor readings, bubble scattering pulse count sequence, flow cell volumetric flow rate data, and water temperature data are standardized and normalized using a zero-mean unit variance standardization algorithm based on sliding window statistics and a minimum-maximum normalization algorithm, respectively, to ensure that subsequent calculation inputs are dimensionless, forming a unified numerical scale input that can participate in subsequent divergence calculations, confidence allocation, and joint discrimination.

[0031] In this implementation scheme, by implementing quality constraints and scale unification processing on real-time spectral detection data and related operational status data, non-physical disturbances introduced by transient anomalies, random noise, short-term missing measurements, and differences in instrument response are significantly reduced, making the spectral observation results closer to the actual optical response state of the water body. The processed spectral and operational status data are consistent in terms of numerical distribution, dimensional scale, and temporal continuity, providing a stable input basis for scattering baseline characterization, absorption feature extraction, and structural difference quantification. Overall, the comparability, numerical stability, and physical interpretability of multi-spectral data under complex water body conditions are enhanced, laying a reliable data foundation for subsequent scattering and absorption aliasing discrimination, spectral confidence allocation, and pollutant concentration inversion.

[0032] Specifically, the process of performing scattering and absorption aliasing analysis on real-time spectral band detection data, equipment configuration data, and historical environmental auxiliary data is as follows: For the corrected spectral intensity sequence within the spectral band set corresponding to the scattering-sensitive spectral band set index data, a spectral band intensity normalization ratio calculation algorithm is executed to obtain the scattering structure spectral band distribution. By converting the energy distribution within the scattering-sensitive spectral bands into a proportional structure form, the influence of absolute intensity changes on discrimination is weakened, allowing the dominant scattering characteristics to be more reflected in the structural relationships between spectral bands rather than amplitude fluctuations. For the corrected spectral intensity sequence within the spectral band set indicated by the absorption discrimination spectral band set index data, spectral band intensity normalization is performed. The proportional calculation algorithm yields the absorption structure spectral distribution, allowing pollutant absorption characteristics to be presented in a relative distribution across multiple spectral bands, enhancing the ability to characterize the consistency of absorption structures at different concentration levels. The Jensen-Shannon divergence calculation algorithm is applied to both the scattering and absorption structure spectral distributions to obtain the Jensen-Shannon divergence. Utilizing its symmetry and boundedness, the overall difference between the two spectral distributions in the probability space is robustly quantified, preventing single-band anomalies from dominating the structure discrimination results. Finally, the Poisson excessive dispersion estimation and abrupt change point counting algorithm is applied to the bubble scattering pulse counting sequence within a sliding window to obtain the transient strength of bubble perturbations. The degree of deviation of bubble events from an ideal random process reflects the level of non-stationary scattering perturbation caused by bubble generation, aggregation, or collapse. A continuous spectrum removal algorithm is performed on the logarithmic spectral intensity sequence within the spectral set indicated by the absorption discrimination spectral set index data to obtain residual curves. A robust local regression residual ratio calculation algorithm is then applied to the residual curves to obtain an absorption structure interpretation metric, separating absorption features from the overall spectral shape and scattering background. The stability assessment of the residual structure quantifies the explanatory power of the absorption response for spectral variations under current observation conditions. A difference analysis is performed on the flow cell inlet pressure data and the flow cell outlet pressure data. The flow cell pressure difference sequence is calculated, and then combined with the historical flow cell pressure difference statistical quantile sequence to perform a quantile mapping algorithm to obtain the flow cell pressure difference perturbation value. The quantile mapping algorithm takes the relative quantile position of the real-time flow cell pressure difference sequence in the historical flow cell pressure difference statistical quantile sequence as input, and maps the relative quantile position to a normalized perturbation quantity. The perturbation quantity is located within a finite numerical range and is used to characterize the degree of deviation of the current flow state from the historical steady-state distribution range. By mapping the real-time flow state to the historical steady-state reference scale, a quantitative constraint on the optical path perturbation caused by flow instability is introduced, providing flow-level support information for aliasing discrimination.

[0033] The Jensen-Shannon divergence between the scattering and absorption spectral distributions is calculated to characterize the symmetric differences between the two distributions in the probability space, reflecting the overall deviation between the dominant scattering and absorption features at the structural level. The natural logarithm of the transient bubble perturbation intensity is obtained by adding one to it, and then using a logarithmic modulation term. This logarithmic mapping suppresses the numerical amplification effect of bubble pulse counting under high perturbation conditions, enhancing the numerical stability constraint on transient perturbations. Multiplying the Jensen-Shannon divergence by the logarithmic modulation term forms a coupled difference term between the scattering structure difference and the bubble perturbation, achieving a joint characterization of spectral structure differences and bubble perturbation intensity, enabling aliasing discrimination to simultaneously reflect structural mismatch and physical perturbation factors. The index of the absorption structure interpretation metric is calculated. The exponential decay form is added to form an absorption interpretability suppression term. This term, through an exponential decay mechanism, introduces suppression modulation for observations with clear absorption structures and strong physical interpretability, preventing misclassification as aliasing. The coupling difference term is divided by the absorption interpretability suppression term to obtain the aliasing basis term, ensuring automatic convergence of the aliasing discrimination result when the absorption characteristics stabilize, thus reducing the risk of false suppression of the true contamination absorption signal. The aliasing basis term is multiplied by the exponentially amplified form of the flow cell pressure difference disturbance value, and after performing an exponential mapping on the result, the mapping result is subtracted by one to obtain the scattering absorption aliasing discrimination value. Nonlinear compression mapping restricts the multi-factor coupling results to a bounded interval, forming a quantification index of aliasing risk suitable for threshold determination. The specific calculation formula is as follows: ; In the formula, It represents the scattering and absorption aliasing discrimination value, which comprehensively characterizes the coupling degree between scattering effect and pollutant absorption effect in multi-spectral observation within the current acquisition window; It represents the distribution of scattering structure spectral segments, reflecting the overall structural characteristics of spectral energy distribution with spectral segments within the set of scattering-sensitive spectral segments, and is used to characterize the spectral response morphology dominated by particle scattering and bubble scattering. It represents the distribution of absorption structure spectral segments, reflecting the relative distribution of spectral energy within the set of absorption discrimination spectral segments, and is used to characterize the spectral structure features formed by pollutant absorption. The Jensen-Shannon divergence is used to measure the symmetric distribution difference between the scattering structure spectral distribution and the absorption structure spectral distribution. It represents the transient intensity of bubble disturbance, characterizing the change in transient scattering disturbance intensity caused by bubble generation, desorption, or flow within the acquisition window; This represents the explanatory power of the absorption structure, quantifying its ability to explain spectral variations under current observation conditions. This represents the pressure difference disturbance value in the flow cell, characterizing the degree to which changes in the flow state within the flow cell affect the stability of spectral observations.

[0034] In this embodiment, Table 1 is a comparison table of scattering structure differences, bubble disturbance and scattering absorption aliasing discrimination values. It records in detail the Jensen-Shannon divergence, bubble disturbance transient intensity, absorption structure interpretation metric, flow cell pressure difference disturbance value and the finally calculated scattering absorption aliasing discrimination value under different acquisition windows. It is used to quantify the coupling degree of scattering effect and pollutant absorption effect in multi-band spectrum under different observation conditions. Specifically: for acquisition window t1, the Jensen-Shannon divergence is 0.08, the transient intensity of bubble disturbance is 0.10, the explanatory value of absorption structure is 1.60, the flow cell pressure difference disturbance is 0.05, and the corresponding scattering-absorption aliasing discrimination value is 0.0066; for acquisition window t2, the Jensen-Shannon divergence is 0.15, the transient intensity of bubble disturbance is 0.25, the explanatory value of absorption structure is 1.30, the flow cell pressure difference disturbance is 0.10, and the scattering-absorption aliasing discrimination value is 0.0287; for acquisition window t3, the Jensen-Shannon divergence is 0.30, the transient intensity of bubble disturbance is 0.60, the explanatory value of absorption structure is 0.90, the flow cell pressure difference disturbance is 0.20, and the scattering-absorption aliasing discrimination value is 0.1152. At acquisition window t4, the Jensen-Shannon divergence was 0.45, the transient intensity of bubble disturbance was 0.85, the explanatory value of absorption structure was 0.60, the flow cell pressure difference disturbance was 0.35, and the scattering-absorption aliasing discriminant value reached 0.2240. At acquisition window t5, the Jensen-Shannon divergence was 0.20, the transient intensity of bubble disturbance was 0.70, the explanatory value of absorption structure was 1.80, the flow cell pressure difference disturbance was 0.15, and the scattering-absorption aliasing discriminant value was 0.1004. At acquisition window t6, the Jensen-Shannon divergence was 0.12, the transient intensity of bubble disturbance was 0.15, the explanatory value of absorption structure was 2.10, the flow cell pressure difference disturbance was 0.05, and the corresponding scattering-absorption aliasing discriminant value was 0.0156.

[0035] Table 1 Comparison of Discrimination Values ​​for Scattering Structure Differences, Bubble Disturbance, and Scattering Absorption Aliasing like Figure 3The figure shows a comparison of the Jansen-Shannon divergence and the transient intensity of bubble perturbation. Combined with Table 1, it can be seen that the differences in scattering structure and the level of bubble perturbation exhibit obvious stage-wise changes under different acquisition windows. Specifically, the Jansen-Shannon divergence and the transient intensity of bubble perturbation are both at high levels corresponding to acquisition window t4, reflecting significant differences in scattering structure and active bubble perturbation within this window, representing the state with the most prominent risk of scattering absorption aliasing. The transient intensity of bubble perturbation is high in acquisition windows t3 and t5, but there are differences in Jansen-Shannon divergence, indicating that under similar physical perturbation conditions, differences in spectral structure have a significant modulating effect on the aliasing discrimination results. In acquisition windows t1 and t6, both the Jansen-Shannon divergence and the transient intensity of bubble perturbation are at low levels, the spectral structure is relatively stable, and the risk of scattering absorption aliasing is low. Overall, Figure 3 It intuitively reflects the changing trends of scattering structure differences and bubble disturbance intensity under different acquisition windows, providing a visual support for the formation of scattering absorption aliasing discrimination values.

[0036] In this implementation plan, by jointly quantifying the multi-band spectral structure characteristics, bubble disturbance behavior, and flow state of the flow cell, a scattering-absorption aliasing discriminant value is constructed that reflects the relative dominance of scattering effects and pollutant absorption effects, identifying the main physical sources of spectral response differences at the structural level. This discriminant result incorporates spectral distribution differences, transient scattering disturbance intensity, interpretability of absorption structure, and flow stability into a unified bounded quantification framework, enabling a preliminary assessment of spectral aliasing risk under conditions of highly turbid or bubble-containing water. Overall, it provides a clear aliasing risk criterion for subsequent spectral screening and concentration inversion, effectively avoiding misjudging scattering-dominant changes as pollutant concentration changes, and fundamentally reducing the possibility of distorted detection results under complex water conditions.

[0037] Specifically, the process of outputting the separability determination results of scattering effect and pollutant absorption effect based on the scattering absorption aliasing analysis results is as follows: For the corrected spectral intensity sequence within the spectral band set corresponding to the scattering sensitive spectral band set index data, the algorithm for spectral band slope sign consistency test and local monotonicity test is executed to obtain the scattering dominant spectral band set. By constraining the consistency and monotonicity characteristics of the spectral band intensity change trend in the wavelength dimension, the smooth and directional spectral response caused by particle scattering or bubble scattering is identified, so that the scattering dominant feature has a discriminable morphological basis at the spectral band level. For the logarithmic spectral intensity sequence within the spectral band set corresponding to the absorption discriminative spectral band set index data, the algorithm for the significance test of absorption valley depth after continuous spectrum removal is executed to obtain the absorption dominant spectral band set. By evaluating the significance of the absorption valley relative to the local baseline after removing the continuous spectrum background, the local energy attenuation characteristics caused by pollutant molecule absorption are highlighted, so that the absorption dominant spectral band can still maintain clear physical interpretability when the scattering background exists.

[0038] Real-time comparison of scattering and absorption aliasing discrimination value and scattering and absorption aliasing discrimination threshold: When the scattering and absorption aliasing discrimination value is less than the scattering and absorption aliasing discrimination threshold, it indicates that the changes in spectral structure caused by particle scattering, bubble scattering, etc. under the current water conditions have not formed a dominant interference on the pollutant absorption characteristics. The difference in spectral response mainly comes from the change in pollutant concentration rather than the fluctuation of scattering background, thus meeting the physical premise of stable inversion based on absorption characteristics. It is determined that the scattering effect and the pollutant absorption effect within the current acquisition window are separable. The set of scattering-dominant spectral bands and the set of absorption-dominant spectral bands are output. The pumping control quantity corresponding to the volumetric flow rate data of the flow tank remains unchanged. Under the condition that the current flow and optical conditions do not introduce a significant risk of aliasing, the existing flow conditions are maintained to ensure the temporal consistency of spectral acquisition conditions, and to prevent the introduction of new flow disturbances or scattering instabilities due to unnecessary flow rate adjustments. The corrected spectral intensity sequence, logarithmic spectral intensity sequence, set of scattering-dominant spectral bands and set of absorption-dominant spectral bands are input into the spectral band reliability allocation analysis.

[0039] When the scattering absorption aliasing discrimination value is greater than or equal to the scattering absorption aliasing discrimination threshold, it indicates that under the current conditions of highly turbid or bubble-containing water, the scattering effect has significantly masked or structurally distorted the absorption characteristics of pollutants across multiple spectral bands. This establishes a risk of scattering aliasing within the current acquisition window, prompting the implementation of defoaming and flow stabilization measures: The bypass defoaming valve is opened to actively release accumulated or transiently generated bubbles in the flow tank, reducing strong scattering and random optical path changes caused by the bubble interface. This weakens the interference of non-absorption scattering on the spectral structure at the source. Based on the coefficient of variation of the flow tank volumetric flow rate data within the sliding time window, the pumping control quantity is adjusted in a closed loop. The pumping control quantity aims to reduce the coefficient of variation of the flow tank volumetric flow rate data within the sliding time window. The adjustment process is limited by the flow range under the current operating conditions, and the flow stabilization criterion is that the flow coefficient of variation enters the historical steady-state statistical range, ensuring that the flow tank volumetric flow rate data remains within a continuous range. Within a time window of at least one integration period, the coefficient of variation is less than the target upper limit obtained from historical steady-state operation data, forming a stable flow state. Flow stability is used as a quantitative criterion for the controllability of the flow state, suppressing the scattering baseline drift caused by pumping fluctuations, local eddies, or uneven flow velocity, and gradually returning the spectral observation to repeatable steady-state conditions. At the same time, the integration time setting is increased, and the integration time setting is extended while ensuring the acquisition frequency and response capability. By extending the effective photon integration time, the random amplification effect of transient bubble scattering and high-frequency noise on the spectral intensity is weakened, so that the absorption characteristics obtain higher relative visibility in the sense of energy integration. After the processing is completed, the scattering absorption aliasing discrimination value is re-acquired and calculated. If the recalculated result is still greater than or equal to the scattering absorption aliasing discrimination threshold, the bypass debubbling valve is kept open, a scattering aliasing risk log is generated, an aliasing persistence marker is output, and entry into the spectral band confidence allocation analysis is blocked.

[0040] In this implementation plan, structural discrimination of multi-band spectral morphology characteristics is performed, and gating is implemented in conjunction with scattering and absorption aliasing discrimination values. This enables a clear distinction between scattering-dominant and absorption-dominant spectral bands. When aliasing risk is established, observation conditions are actively intervened through defoaming and flow stabilization to bring the spectral acquisition state back to a repeatable and interpretable physical range. This process constructs a closed-loop control mechanism at the workflow level, from discrimination to treatment to verification. It allows subsequent spectral band reliability allocation analysis only when the scattering effect and the pollutant absorption effect are separable. This effectively avoids misjudging changes in scattering background as changes in pollutant concentration under conditions of high turbidity or bubble-containing water, significantly improving the stability and accuracy of pollutant detection results.

[0041] Specifically, the process of performing spectral reliability allocation analysis on the stability characteristics of residuals and the consistency characteristics of spectral morphology distribution of each spectral band identified by aliasing is as follows: Using the set of scattering-dominant spectral bands as constraints, a robust local regression scattering baseline fitting algorithm is applied to the logarithmic spectral intensity sequence to obtain the scattering baseline. Then, baseline residual calculation is performed on the spectral bands to obtain the spectral band residuals. By constructing a baseline model using only the scattering-dominant spectral bands, the interference of absorption characteristics on baseline fitting is avoided, allowing the obtained scattering baseline to more realistically reflect changes in the scattering background. This ensures that the residuals mainly carry deviation information related to pollutant absorption. The median absolute deviation is then calculated on the spectral band residual sequence. The algorithm obtains the median absolute deviation of the residuals; it then performs a kernel density estimation distribution fitting algorithm on the window samples of the calibrated spectral intensity sequence within the spectral band to obtain the spectral band morphological distribution. This morphological distribution represents the normalized statistical distribution obtained after probability density estimation of the spectral band observations within the current acquisition window. The bandwidth parameter used in kernel density estimation is determined by the dispersion of the spectral band samples within the current acquisition window. The sample window range consists of all valid observation samples of the corresponding spectral band within the acquisition window. This non-parametric approach characterizes the statistical shape of the spectral band observations, enabling a continuous and smooth representation of the spectral band distribution characteristics within the current acquisition window, avoiding distorted representations of the distribution shape. The formula makes prior assumptions; it uses a historical distribution fitting and updating algorithm to obtain the morphological distribution of historical spectral segments by applying historical white field reference spectral mean sequences and historical dark field reference spectral mean sequences. Historical turbidity statistical mean is introduced as a background condition during the updating process. The historical distribution fitting and updating adopts a sliding window-based progressive updating strategy, updating the morphological distribution of historical spectral segments only when the historical steady-state criterion is met. The updated sample source is consistent with the current spectral segment statistical caliber, forming a spectral segment statistical reference that matches the turbidity state of the water body on a long-term operational scale, so that the historical morphological distribution can reflect the true optical characteristics under stable operating conditions; the spectral segment morphology... The Korbeck-Leibler divergence is obtained by performing the Korbeck-Leibler divergence calculation algorithm on the morphological distribution of the current spectral segment and the morphological distribution of the historical spectral segment. The information divergence measures the degree of deviation of the statistical morphology of the current spectral segment relative to the historical reference state, providing an objective criterion for identifying abnormal spectral segments caused by changes or drifts in scattering conditions. Before performing the Korbeck-Leibler divergence calculation, the morphological distribution of the spectral segment and the morphological distribution of the historical spectral segment are discretized within the same spectral segment range, and the discretized probability density is normalized to form a probability distribution defined on a unified support set, so as to ensure the uniqueness and feasibility of the information divergence calculation results.

[0042] The absolute value of the spectral band residual is calculated and divided by the sum of the median absolute deviation of the residual and the division-by-zero protection constant to obtain the degree of deviation of the spectral band residual. The amplification effect of abnormal residuals on the reliability assessment of a single spectral band is suppressed by using the median statistic as a scaling benchmark. An exponential decay mapping is performed on the degree of deviation of the spectral band residual to form a reliability decay factor for the spectral band residual. The exponential decay characteristic is used to maintain high reliability for small residuals while rapidly suppressing large deviations. The Koolbek-Leibler divergence between the spectral band morphological distribution and the historical spectral band morphological distribution is calculated. The Koolbek-Leibler divergence is then divided by the Koolbek-Leibler divergence. The Leibler divergence is summed with 1 to obtain the spectral band morphology distribution offset ratio. This bounded proportional mapping avoids the numerical dominance of distribution differences in the discrimination results under extreme conditions. Subtracting the spectral band morphology distribution offset ratio from 1 yields the spectral band morphology consistency modulation factor, which enhances the weight of spectral bands with higher consistency with historical distributions in the reliability assessment. Multiplying the spectral band residual reliability attenuation factor with the spectral band morphology consistency modulation factor yields the spectral band reliability assignment value, achieving joint constraint and comprehensive quantification of the spectral band's instantaneous stability and long-term statistical consistency. The specific calculation formula is as follows: ; In the formula, This represents the confidence assignment value of a spectral band, quantifying the reliability of the spectral band in pollutant retrieval under current observation conditions; This represents the spectral residual, reflecting the instantaneous deviation between model predictions and actual observations; This represents the absolute deviation of the median of the residuals, used to provide a robust estimation benchmark for the scale of the residual distribution; This represents the spectral band morphological distribution, used to describe the statistical structural characteristics of spectral band observations within the current acquisition window; It represents the distribution of the morphology of historical spectral segments, used to characterize the statistical reference morphology of the spectral segment under long-term stable operating conditions; represents the Kolbec-Leibler divergence, which measures the degree of information shift in the current spectral band morphological distribution relative to the historical morphological distribution; i represents the spectral band, distinguishing different spectral bands as independent evaluation objects in the confidence allocation process; This represents the division-to-zero protection constant, which is obtained by the denominator minimum robustness constraint algorithm for the absolute deviation of the residual median, and its value ranges from 0.0005 to 0.001.

[0043] In this implementation scheme, a reliable quantification mechanism oriented towards the spectral segment level is constructed by jointly modeling the residual behavior and statistical morphology of spectral segments under the constraint of the scattering-dominant spectral segment set. This provides a clear and calculable criterion for determining whether each spectral segment is suitable for pollutant inversion under the current observation conditions. On the one hand, this process uses the degree of deviation of the residual from the robust scale benchmark to characterize the stability of the spectral segment in instantaneous observation, suppressing occasional deviations caused by scattering perturbations or local anomalies. On the other hand, it measures the long-term consistency of the statistical characteristics of the spectral segment by comparing it with the historical spectral segment morphology distribution, avoiding the misleading effect of spectral segments with significant shifts in statistical structure on the inversion results. By jointly quantifying the instantaneous stability constraint and the historical consistency constraint, a spectral segment reliability allocation value is formed, so that subsequent inversions rely only on effective spectral segments that still maintain physical interpretability and statistical stability under complex water conditions, thereby significantly reducing the impact of scattering aliasing on the accuracy and stability of pollutant concentration inversion.

[0044] Specifically, the process of selecting and forming a set of valid inverted spectral segments based on the spectral segment reliability allocation analysis results is as follows: Figure 4 The diagram shows the dynamic allocation and closed-loop screening control process for spectral band confidence. Within the set of dominant absorption spectral bands, the quantile truncation and spectral band confidence threshold are used to jointly screen the spectral band confidence allocation values ​​to obtain an effective inversion spectral band set.

[0045] Real-time comparison of spectral segment confidence assignment values ​​and threshold values: When a spectral segment confidence assignment value is less than the threshold value, insufficient inversion input is identified, and an insufficient inversion input flag is output. Median statistical fusion processing is performed on the spectral segment confidence assignment values ​​within adjacent acquisition windows. This processing uses the spectral segment confidence assignment values ​​of corresponding spectral segments within the current acquisition window and its adjacent windows as the fusion objects. Robust fusion results are formed by taking the median value across windows. Robust statistical fusion across windows suppresses abnormal fluctuations in spectral segment confidence caused by transient bubbles and local turbidity abrupt changes within a single window. Simultaneously, the effective inversion spectral segment set is further tightened within the dominant spectral segment set. The spectral segment confidence assignment values ​​are quantile-filtered, retaining only the spectral segment confidence assignment values. The spectral bands whose values ​​are located within the upper quartile of the empirical distribution of the confidence allocation values ​​of the dominant absorption spectral band set are selected. The upper quartile is directly obtained based on the confidence allocation values ​​of the dominant absorption spectral band set within the current acquisition window. This makes the spectral bands participating in the inversion more concentrated in high-confidence spectral bands that still maintain stable absorption responses under the current complex water conditions, reducing the amplification effect of scattering aliasing on the inversion results. The acquisition window is a time interval divided by the frame timestamp sequence generated by the system clock. The time interval corresponds to the acquisition cycle of continuous multi-band spectral data under one integration time setting. The window range is jointly determined by the time span between adjacent frame timestamps and the corresponding integration time setting. After fusion and filtering, the processing results are used to determine the inversion availability of the next acquisition window.

[0046] When the spectral segment confidence assignment value is greater than or equal to the spectral segment confidence assignment threshold, the inversion input is deemed to meet the requirements. The set of spectral segment confidence assignment values ​​and the set of effective inverted spectral segments are output. The corrected spectral intensity sequence, logarithmic spectral intensity sequence, set of spectral segment confidence assignment values, set of effective inverted spectral segments, and scattering absorption aliasing discrimination value are input into the pollutant concentration rapid inversion and stable issuance module.

[0047] In this implementation scheme, a dynamic screening mechanism for inversion input quality is constructed by implementing quantile truncation and threshold constraints on the confidence allocation values ​​of spectral bands within the dominant absorption spectral band set. This ensures that only spectral bands that simultaneously meet the stability and consistency requirements under the current observation conditions are included in the pollutant concentration inversion process. When the confidence of a spectral band is insufficient, robust statistical fusion across acquisition windows and tightening of the spectral band set suppress spectral quality fluctuations caused by non-steady-state factors such as transient bubbles and local turbidity abrupt changes, thus avoiding amplified interference from low-confidence spectral bands on the inversion results. When the confidence of a spectral band meets the requirements, the selected high-confidence spectral band set is used as the inversion input, and scattering and absorption aliasing discrimination information is introduced simultaneously, allowing the concentration inversion to proceed under the premise of fully sensing scattering risks. Overall, adaptive control of the effectiveness of the inversion input is achieved, enhancing the stability, reliability, and anti-aliasing capability of the pollutant concentration inversion process under complex water conditions.

[0048] Specifically, the process of performing absorption model inversion and residual statistical analysis on the effective inverted spectral band set, and conducting issuance credibility assessment analysis, is as follows: On the effective inverted spectral band set, the absorption model residual and scattering baseline residual are calculated using both an absorption model fitting residual calculation algorithm with continuum removal constraints and a robust local regression scattering baseline fitting residual calculation algorithm. Specifically: the logarithmic spectral intensity sequence within the effective inverted spectral band set is fitted using a nonlinear least squares fitting algorithm with continuum removal constraints to estimate the absorption model parameters, and the absorption model residual is calculated using the difference between the fitted predicted value and the observed value; the logarithmic spectral intensity sequence within the scattering-dominant spectral band set is fitted using a robust local regression scattering baseline fitting algorithm to estimate the scattering baseline, and the scattering baseline is extrapolated to the effective inverted spectral band set and predicted using the baseline. The difference between the observed and the measured values ​​is used to calculate the scattering baseline residual, and the log-likelihood ratio is obtained using the Gaussian likelihood function and log-likelihood ratio calculation algorithm. The flow cell pressure difference sequence is processed within a sliding window using both the coefficient of variation calculation and quantile mapping algorithms, and the sum is taken to obtain the flow cell pressure difference disturbance suppression. The median scaling and quantile mapping algorithms are applied to the absorption model residuals on the effective inversion spectral set to obtain the inversion residual dispersion penalty. Based on the effective inversion spectral set, under the constraints of the corrected spectral intensity sequence and the logarithmic spectral intensity sequence, the pollutant concentration estimate is calculated using the absorption model parameter inversion algorithm. Water temperature data is used to characterize changes in the physical state of the water body and serves as an environmental condition reference during the pollutant absorption model parameter inversion process to avoid misinterpreting temperature-induced absorption changes as pollutant concentration changes.

[0049] Multiplying the log-likelihood ratio by the exponentially decaying mapping result of the scattering-absorption aliasing discriminant value yields the inversion credibility gain term. By introducing exponential suppression of aliasing risk, the explanatory advantage of the absorption model is automatically weakened when the scattering-absorption aliasing risk is high. The flow pool pressure difference disturbance suppression factor is calculated and added to form a flow disturbance suppression term, incorporating flow state changes into the inversion credibility assessment to prevent direct issuance of results under unstable flow conditions. The inversion residual dispersion penalty factor is calculated and added to form an inversion stability penalty term, constraining the discrete diffusion behavior of the inversion residual in the time and spectral dimensions. Dividing the inversion credibility gain term by the product of the flow disturbance suppression term and the inversion stability penalty term yields the comprehensive issuance driving factor. After performing an exponential mapping on the negative of the comprehensive issuance driving factor, subtracting the mapping result from it yields the concentration issuance credibility value. Bounded nonlinear mapping restricts the issuance result to a stable range, forming a credibility quantification index suitable for threshold determination. The specific calculation formula is as follows: ; In the formula, This indicates the reliability of the concentration issuance, representing the degree to which the pollutant concentration inversion results within the current collection window have stable output conditions; The log-likelihood ratio represents the relative interpretability of the absorption model compared to the scattering model for the observed spectrum. This represents the scattering absorption aliasing discrimination value, reflecting the level of coupling risk between scattering effects and pollutant absorption effects under current observation conditions; This represents the amount of pressure difference disturbance suppression in the flow cell, used to quantify the suppressive effect of flow state fluctuations on the stability of spectral inversion; This represents the penalty for inversion residual dispersion, used to characterize the degree of dispersion of the inversion residual in the spectral and time dimensions and to impose a penalty on unstable results.

[0050] In this implementation scheme, the residual behavior of the absorption model and the scattering baseline model is compared and modeled within the effective inversion spectral band set. A joint constraint of the flow state of the flow cell and the discrete characteristics of the inversion residuals is introduced to construct a concentration issuance confidence value that comprehensively reflects the model's interpretative advantage, scattering aliasing risk, and operational stability. This allows the pollutant concentration estimation results to undergo multi-dimensional consistency checks before output. This process modulates the interpretability of the absorption model under current observation conditions in conjunction with the scattering aliasing risk. When the flow is unstable or the residual dispersion is significant, the issuance confidence value is automatically reduced, avoiding misjudging inversion results affected by scattering interference or transient fluctuations as true concentration changes. Overall, this achieves integrated confidence control of pollutant concentration inversion and result issuance, ensuring that formal results are output only when the spectral structure, flow state, and statistical stability simultaneously meet the requirements. This significantly improves the stability, reliability, and physical interpretability of pollutant detection results under complex water conditions.

[0051] Specifically, the process of outputting pollutant concentration estimates and result labeling based on the credibility assessment analysis results is as follows: Real-time comparison of the credibility value and the credibility threshold of the concentration issuance: When the concentration issuance confidence value is less than the concentration issuance confidence threshold, it is determined that the pollutant concentration inversion result in the current collection window does not meet the stable issuance conditions, and a concentration inversion unissued mark is output to clearly distinguish the inversion result that does not yet have physical and statistical consistency from the formal detection result, so as to avoid unstable values ​​from entering the subsequent analysis or decision-making process. The concentration inversion unissued mark is associated with the timestamp sequence of the corresponding collection window to be used to determine the stability and continuity of the results in subsequent collection windows.

[0052] When the concentration issuance confidence value is greater than or equal to the concentration issuance confidence threshold, the pollutant concentration inversion result within the current acquisition window is determined to meet the stable issuance condition. The inversion result confirmed by multiple physical and statistical constraints is used as the confidence detection output to avoid mistaking scattering-dominated pseudo-changes for pollutant concentration fluctuations. The pollutant concentration estimate and concentration inversion issuance mark are output and created and archived in the pollutant concentration database. At the same time, the concentration inversion issuance mark is associated with the timestamp sequence of the corresponding acquisition window for subsequent result consistency verification.

[0053] This implementation scheme establishes a final issuance gating mechanism for pollutant concentration inversion results by implementing threshold-based adjudication of the concentration issuance credibility value. This mechanism can clearly distinguish between stable and interpretable detection results and unstable results affected by scattering disturbances or operational fluctuations during the output stage. When issuance conditions are insufficient, the stability of the inversion results is continuously tracked and delayed in judgment by associating the concentration inversion unissued marker with the timestamp, thus avoiding the misreporting of transient fluctuations as real pollution changes. When issuance conditions are met, only the concentration estimation results verified by multiple physical constraints and statistical consistency are formally issued and archived, and the continuity and reliability of the results are ensured by time series consistency verification. Overall, this scheme achieves robust output control of pollutant detection results under complex water conditions, significantly reducing the risk of detection distortion caused by scattering aliasing or operational fluctuations.

[0054] like Figure 2 As shown, the second aspect of the present invention provides a rapid detection device for water pollutants based on multi-band spectroscopy, comprising: an acquisition and preprocessing module for acquiring real-time spectral detection data, obtaining equipment configuration data and environmental historical auxiliary data; preprocessing the real-time spectral detection data, equipment configuration data, and environmental historical auxiliary data; a scattering and absorption aliasing discrimination module for performing scattering and absorption aliasing analysis on the real-time spectral detection data, equipment configuration data, and environmental historical auxiliary data, and outputting a determination result on the separability of scattering effect and pollutant absorption effect based on the scattering and absorption aliasing analysis result; a spectral band credibility allocation module for performing spectral band credibility allocation analysis on the residual stability characteristics and spectral band morphological distribution consistency characteristics of each spectral band that has passed the aliasing discrimination, and screening to form a set of effective inversion spectral bands based on the spectral band credibility allocation analysis result; and a pollutant concentration inversion and issuance module for performing absorption model inversion and residual statistical analysis on the effective inversion spectral band set, and performing issuance credibility assessment analysis, and outputting pollutant concentration estimates and result labels based on the issuance credibility assessment analysis result.

[0055] This implementation plan establishes a complete closed loop, encompassing multi-band spectral acquisition, operational status constraints, scattering and absorption aliasing discrimination, quantitative screening of spectral band reliability, and pollutant concentration inversion and result issuance. This enables the water pollutant detection process to maintain stable operation under complex water conditions such as high turbidity or the presence of air bubbles. It can identify the aliasing risk of scattering effects and pollutant absorption effects in the early stages of detection and selectively introduce effective spectral bands with physical interpretability and statistical stability into the inversion process at the spectral level, avoiding the systematic amplification of concentration estimation results by scattering disturbances. Simultaneously, the result issuance mechanism constrains the reliability of the inversion output, clearly distinguishing between detection results that can be stably output and inversion results that do not yet meet the conditions for issuance, thereby significantly improving the authenticity, stability, and engineering usability of pollutant detection results in complex water environments.

[0056] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0057] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A rapid detection method for water pollutants based on multi-band spectroscopy, characterized in that, Includes the following steps: S1: Collect real-time spectral band detection data, obtain equipment configuration data and historical environmental auxiliary data; preprocess the real-time spectral band detection data, equipment configuration data and historical environmental auxiliary data; S2 performs scattering and absorption aliasing analysis on real-time spectral detection data, equipment configuration data, and historical environmental auxiliary data, and outputs the results of the scattering and absorption aliasing analysis to determine the separability of scattering effect and pollutant absorption effect. S3, perform spectral segment reliability allocation analysis on the consistency characteristics of residual stability and spectral segment morphology distribution of each spectral segment identified by aliasing, and select and form an effective inversion spectral segment set based on the results of the spectral segment reliability allocation analysis; S4 performs absorption model inversion and residual statistical analysis on the effective inversion spectrum set, and conducts issuance credibility assessment analysis. Based on the issuance credibility assessment analysis results, it outputs pollutant concentration estimates and result labels.

2. The rapid detection method for water pollutants based on multi-band spectroscopy according to claim 1, characterized in that: The specific process of acquiring real-time spectral band detection data, obtaining device configuration data, and historical environmental auxiliary data is as follows: Real-time spectral band detection data is collected, including: multi-band raw spectral intensity sequence, spectral band center wavelength nominal value sequence, integration time setting value sequence, light source temperature data, dark field reference spectral intensity sequence, white reference spectral intensity sequence, flow cell inlet pressure data, flow cell outlet pressure data, flow cell volumetric flow rate data, water temperature data, turbidity sensor reading data, bubble scattering pulse count sequence, and frame timestamp sequence. Acquire device configuration data, which includes: spectral band grouping boundary wavelength data, scattering sensitive spectral band set index data, and absorption discrimination spectral band set index data; Acquire historical environmental auxiliary data and create an environmental historical auxiliary database. The historical environmental auxiliary data includes: historical dark field reference spectrum mean sequence, historical white reference spectrum mean sequence, historical turbidity statistical mean, historical turbidity statistical standard deviation, and historical flow cell pressure difference statistical quantile sequence.

3. The rapid detection method for water pollutants based on multi-band spectroscopy according to claim 1, characterized in that: The specific process for preprocessing real-time spectral band detection data, equipment configuration data, and historical environmental auxiliary data is as follows: Outlier removal is performed on real-time spectral band detection data; intra-spectral smoothing is performed on the original multi-band spectral intensity sequences; short-term missing data completion is performed on the real-time spectral band detection data using a piecewise linear interpolation completion algorithm and a data validity judgment method based on the maximum allowable continuous missing measurement length constraint; dark field reference subtraction and white reference normalization correction algorithms are used to perform dark field reference subtraction and white reference normalization on the original multi-band spectral intensity sequences to obtain the corrected spectral intensity sequences; the corrected spectral intensity sequences are mapped to logarithmic spectral intensity sequences using logarithmic transformation and a logarithmic mapping processing method based on numerical lower limit constraints; the corrected spectral intensity sequences, logarithmic spectral intensity sequences, turbidity sensor readings, bubble scattering pulse count sequences, flow tank volume flow rate data, and water temperature data are standardized and normalized using a zero-mean unit variance standardization algorithm based on sliding window statistics and a minimum-maximum normalization algorithm, respectively.

4. The rapid detection method for water pollutants based on multi-band spectroscopy according to claim 1, characterized in that: The specific process of performing scattering and absorption aliasing analysis on real-time spectral detection data, equipment configuration data, and historical environmental auxiliary data is as follows: For the corrected spectral intensity sequence, within the spectral sets corresponding to the scattering-sensitive spectral set index data and the absorption-discriminative spectral set index data, respectively, the spectral intensity normalization ratio calculation algorithm is executed to obtain the scattering structure spectral distribution and the absorption structure spectral distribution; the Jensen-Shannon divergence calculation algorithm is executed on the scattering structure spectral distribution and the absorption structure spectral distribution to obtain the Jensen-Shannon divergence; and the bubble scattering pulse counting sequence is subjected to the Poisson excessive dispersion estimation and abrupt change point counting algorithm within a sliding window to obtain the transient intensity of bubble perturbation. The logarithmic spectral intensity sequence is subjected to a continuous spectrum removal algorithm within the spectral band set indicated by the absorption discrimination spectral band set index data to obtain the residual curve. The residual curve is then subjected to a robust local regression residual ratio calculation algorithm to obtain the absorption structure interpretation metric. The difference between the flow cell inlet pressure data and the flow cell outlet pressure data is calculated to obtain the flow cell pressure difference sequence. The flow cell pressure difference perturbation value is then obtained by combining the historical flow cell pressure difference statistical quantile sequence with the quantile mapping algorithm. Calculate the Jensen-Shannon divergence between the scattering structure spectral distribution and the absorption structure spectral distribution; add one to the transient intensity of the bubble perturbation and take the natural logarithm to obtain the logarithmic modulation term; multiply the Jensen-Shannon divergence with the logarithmic modulation term to form the coupling difference term between the scattering structure difference and the bubble perturbation; calculate the exponential decay form of the absorption structure interpretability measure and add one to it to form the absorption interpretability suppression term; divide the coupling difference term by the absorption interpretability suppression term to obtain the aliasing basis term; multiply the aliasing basis term with the exponentially amplified form of the flow cell pressure difference perturbation value, perform an exponential mapping on the result, and subtract one from the mapping result to obtain the scattering and absorption aliasing discrimination value.

5. The rapid detection method for water pollutants based on multi-band spectroscopy according to claim 1, characterized in that: The specific process for outputting the separability determination result of scattering effect and pollutant absorption effect based on the scattering absorption aliasing analysis results is as follows: For the corrected spectral intensity sequence, within the spectral band set corresponding to the scattering-sensitive spectral band set index data, perform the spectral band slope sign consistency test and local monotonicity test algorithms to obtain the scattering-dominant spectral band set; The algorithm for testing the significance of absorption valley depth after continuous spectrum removal is performed on the logarithmic spectral intensity sequence within the spectral set corresponding to the absorption discrimination spectral set index data to obtain the absorption dominant spectral set. Real-time comparison of scattering absorption aliasing discrimination value and scattering absorption aliasing discrimination threshold: When the scattering and absorption aliasing discrimination value is less than the scattering and absorption aliasing discrimination threshold, the set of scattering dominant spectrum bands and the set of absorption dominant spectrum bands are output, while keeping the pumping control quantity corresponding to the flow cell volume flow rate data unchanged. When the scattering absorption aliasing discrimination value is greater than or equal to the scattering absorption aliasing discrimination threshold, defoaming and flow stabilization are performed: the bypass defoaming valve is opened, and the pumping control quantity is adjusted in a closed loop based on the coefficient of variation calculation result of the flow pool volume flow rate data within the sliding time window, while the integral time setting value is increased. After the treatment is completed, the scattering and absorption aliasing discrimination value is collected again and calculated. If the recalculated result is still greater than or equal to the scattering and absorption aliasing discrimination threshold, the bypass debubbling valve remains open, the aliasing is continuously marked, and entry into the spectral confidence allocation analysis is blocked.

6. The rapid detection method for water pollutants based on multi-band spectroscopy according to claim 1, characterized in that: The specific process of performing spectral reliability allocation analysis on the residual stability characteristics and spectral morphological distribution consistency characteristics of each spectral band identified by aliasing is as follows: A robust local regression scattering baseline fitting algorithm is applied to the logarithmic spectral intensity sequence to obtain the scattering baseline. Baseline residuals are calculated for each spectral segment to obtain the spectral segment residuals, and a median absolute deviation calculation algorithm is applied to obtain the median absolute deviation of the residuals. A kernel density estimation distribution fitting algorithm is applied to the corrected spectral intensity sequence to obtain the spectral segment morphological distribution. A historical distribution fitting update algorithm is applied to the historical white reference spectral mean sequence and the historical dark field reference spectral mean sequence to obtain the historical spectral segment morphological distribution. A Kolb-Leibler divergence calculation algorithm is applied to the spectral segment morphological distribution and the historical spectral segment morphological distribution to obtain the Kolb-Leibler divergence. Calculate the absolute value of the spectral residual and divide it by the sum of the absolute deviation of the residual median and the zero protection constant to obtain the degree of deviation of the spectral residual; An exponential decay mapping is performed on the deviation of the spectral band residuals to form a reliable decay factor for the spectral band residuals. Calculate the Kolb-Leibler divergence between the spectral band morphology distribution and the historical spectral band morphology distribution; divide the Kolb-Leibler divergence by the sum of the Kolb-Leibler divergence and a factor to obtain the spectral band morphology distribution offset ratio; subtract the spectral band morphology distribution offset ratio from the factor to obtain the spectral band morphology consistency modulation factor; multiply the spectral band residual confidence attenuation factor by the spectral band morphology consistency modulation factor to obtain the spectral band confidence assignment value.

7. The rapid detection method for water pollutants based on multi-band spectroscopy according to claim 1, characterized in that: The specific process of selecting and forming a set of effective inverted spectral bands based on the spectral band reliability allocation analysis results is as follows: Within the set of dominant absorption spectral bands, quantile truncation and spectral band confidence threshold are jointly screened for the spectral band confidence assignment values ​​to obtain an effective inversion spectral band set. Real-time comparison of spectral band confidence assignment values ​​and spectral band confidence assignment thresholds: When the spectral segment confidence allocation value is less than the spectral segment confidence allocation threshold, it is determined that the inversion input is insufficient, and an inversion input insufficient flag is output. Median statistical fusion processing is performed on the spectral segment confidence allocation values ​​in adjacent acquisition windows. At the same time, the effective inversion spectral segment set is tightened within the absorption dominant spectral segment set, and the spectral segment confidence allocation value is quantized and filtered. When the spectral segment confidence assignment value is greater than or equal to the spectral segment confidence assignment threshold, the inversion input is deemed to meet the requirements, and the set of spectral segment confidence assignment values ​​and the set of valid inverted spectral segments are output.

8. The rapid detection method for water pollutants based on multi-band spectroscopy according to claim 1, characterized in that: The specific process of performing absorption model inversion and residual statistical analysis on the effective inversion spectrum set, and conducting issuance credibility assessment analysis is as follows: The absorption model residuals and scattering baseline residuals are calculated on the effective inversion spectral set, and the log-likelihood ratio is obtained by the Gaussian likelihood function log-likelihood ratio calculation algorithm. The flow cell pressure difference sequence is subjected to the coefficient of variation calculation and quantile mapping algorithm within the sliding window, and the sum is taken to obtain the flow cell pressure difference perturbation suppression amount. The median scaling estimation and quantile mapping algorithm are performed on the absorption model residuals on the effective inversion spectral set to obtain the inversion residual dispersion penalty. Based on the effective inversion spectral set, under the constraints of the corrected spectral intensity sequence and the log spectral intensity sequence, the pollutant concentration estimate is calculated by the absorption model parameter inversion algorithm. The absorption model parameter inversion algorithm is based on the physical mechanism of pollutant spectral absorption, and adopts the absorption model in the form of Beer-Lambert law as the basic expression, allowing the introduction of nonlinear parameter estimation methods to solve the model parameters. Multiply the log-likelihood ratio by the exponential decay mapping result of the scattering-absorption aliasing discriminant value to obtain the inversion confidence gain term; calculate the flow cell pressure difference disturbance suppression amount and add it to form the flow disturbance suppression term; calculate the inversion residual dispersion penalty degree and add it to form the inversion stability penalty term; divide the inversion confidence gain term by the product of the flow disturbance suppression term and the inversion stability penalty term to obtain the comprehensive issuance driving amount; After performing an exponential mapping on the inverse of the comprehensive issuance driving quantity, subtract the mapping result from one to obtain the concentration issuance confidence value.

9. The rapid detection method for water pollutants based on multi-band spectroscopy according to claim 1, characterized in that: The specific process of outputting pollutant concentration estimates and result labeling based on the credibility assessment analysis results is as follows: Real-time comparison of concentration issuance credibility value and concentration issuance credibility threshold: When the concentration issuance confidence value is less than the concentration issuance confidence threshold, output the concentration inversion unissued flag and associate the concentration inversion unissued flag with the timestamp sequence of the corresponding acquisition window; When the concentration issuance confidence value is greater than or equal to the concentration issuance confidence threshold, the pollutant concentration estimate and concentration inversion issuance flag are output, and the pollutant concentration database is created and archived. At the same time, the concentration inversion issuance flag is associated with the timestamp sequence of the corresponding acquisition window.

10. A rapid detection device for water pollutants based on multi-band spectroscopy, employing the rapid detection method for water pollutants based on multi-band spectroscopy according to any one of claims 1-9, comprising: The acquisition and preprocessing module is used to acquire real-time spectral band detection data and obtain equipment configuration data and historical environmental auxiliary data. Preprocessing of real-time spectral band detection data, equipment configuration data, and historical environmental auxiliary data; The scattering and absorption aliasing discrimination module is used to perform scattering and absorption aliasing analysis on real-time spectral detection data, equipment configuration data and environmental historical auxiliary data, and output the separability judgment result of scattering effect and pollutant absorption effect based on the scattering and absorption aliasing analysis results. The spectral segment confidence allocation module is used to perform spectral segment confidence allocation analysis on the residual stability characteristics and spectral segment morphological distribution consistency characteristics of each spectral segment identified by aliasing, and to filter and form a set of effective inverted spectral segments based on the results of the spectral segment confidence allocation analysis. The pollutant concentration inversion and issuance module is used to perform absorption model inversion and residual statistical analysis on the effective inversion spectrum set, and to conduct issuance credibility assessment analysis. Based on the issuance credibility assessment analysis results, it outputs the pollutant concentration estimate and result label.

Images