A method, system, equipment, and medium for detecting water turbidity using multi-source scattering.

Abstract

This application provides a method, system, device, and medium for detecting water turbidity using multi-source scattering, relating to the field of turbidity detection technology. The method includes: evaluating the impact of wavelength combinations on pollutant identification performance based on a preset fitness function to obtain an optimal wavelength combination; learning light source switching and illumination strategies using a reward function for the optimal wavelength combination to determine a light source control strategy; using the light source control strategy to control multi-wavelength light sources to irradiate water samples and acquire scattering spectral data; performing wavelet multi-scale decomposition and soft threshold filtering on the scattering spectral data to obtain preprocessed scattering spectral data; and establishing a quantitative relationship model using the preprocessed scattering spectral data to determine pollutant concentrations. This invention significantly improves the accuracy and efficiency of water pollutant detection through wavelength combination optimization and intelligent learning of light source control strategies, providing an intelligent solution for water quality monitoring.

Description

Technical Field

[0001] This invention relates to the field of turbidity detection technology, and in particular to a method, system, equipment and medium for detecting water turbidity using multi-source scattering. Background Technology

[0002] Rapid and accurate identification and quantitative analysis of aquatic pollutants is a significant challenge in environmental monitoring. Traditional methods often require complex sample pretreatment and expensive equipment, making real-time online monitoring difficult. How to utilize multi-wavelength spectroscopy to achieve efficient identification and quantitative analysis of aquatic pollutants is a pressing technical problem. The core of this problem lies in extracting characteristic information of pollutants from complex water scattering spectra. This involves several key technical aspects: first, determining the optimal combination of excitation wavelengths to obtain the characteristic scattering spectra of pollutants; second, designing a light source control strategy to achieve efficient excitation of target pollutants; third, effectively preprocessing the acquired scattering spectral data to improve the signal-to-noise ratio and feature extraction effect; and finally, establishing an accurate quantitative analysis model to precisely determine pollutant concentrations. These technical aspects are interconnected and mutually influential, constituting a complex system optimization problem. The key to this problem lies in achieving synergistic optimization of these technical aspects under the constraints of limited hardware resources and computing power. Simultaneously, due to the complex and variable nature of the aquatic environment, ensuring the robustness and universality of the identification and quantitative analysis methods is also a crucial consideration. Furthermore, how to apply this technology to actual online monitoring systems and achieve long-term stable and reliable operation is also a problem that must be solved for this technology to move from the laboratory to practical application. Summary of the Invention

[0003] This invention provides a method for detecting water turbidity using multi-source scattering, mainly comprising:

[0004] The impact of wavelength combinations on pollutant recognition performance is evaluated based on a preset fitness function to obtain the optimal wavelength combination;

[0005] For the optimal wavelength combination, a reward function is used to learn the light source switching and illumination strategy, and the light source control strategy is determined.

[0006] The light source control strategy described above is used to control multiple wavelength light sources to irradiate water samples and obtain scattering spectrum data;

[0007] The scattering spectrum data is subjected to wavelet multi-scale decomposition and soft threshold filtering to obtain preprocessed scattering spectrum data;

[0008] A quantitative relationship model was established using the preprocessed scattering spectral data to determine the pollutant concentration.

[0009] As a preferred embodiment of the multi-source scattering water turbidity detection method of the present invention, the step of evaluating the influence of wavelength combinations on pollutant identification performance according to a preset fitness function to obtain the optimal wavelength combination includes:

[0010] Obtain a set of preset wavelength combinations as the initial candidate wavelength combinations;

[0011] For each candidate wavelength combination, a support vector machine model is used to identify pollutants in its corresponding spectral data to obtain the identification results;

[0012] Based on the preset fitness function, calculate the fitness value of each candidate wavelength combination. If the optimal fitness value is greater than the preset threshold, then the corresponding candidate wavelength combination is determined as the optimal wavelength combination.

[0013] Otherwise, a genetic algorithm is used to optimize the candidate wavelength combination until the optimal fitness value condition is met or the maximum number of iterations is reached.

[0014] The optimal wavelength combination is output for subsequent pollutant identification tasks.

[0015] As a preferred embodiment of the multi-source scattering water turbidity detection method of the present invention, wherein: for the optimal wavelength combination, a reward function is used to learn the source switching and illumination strategy to determine the source control strategy, including:

[0016] The Q-learning reinforcement learning algorithm is used to learn the light source switching and illumination strategies under different wavelength combinations, and to obtain the light source control strategy with the maximum reward value.

[0017] Based on the optimal wavelength combination, combined with the preset plant growth model and historical data, the optimal irradiation intensity parameters of each wavelength light source are determined.

[0018] If the power range of the light source cannot meet the optimal illumination intensity, then the type of light source with the optimal illumination intensity is selected according to the preset light source replacement rules.

[0019] Plant growth status parameters are collected in real time by environmental sensors and compared with preset plant growth models. If the actual growth status deviates from the model prediction value by more than a preset threshold, the light strategy re-optimization process is triggered.

[0020] As a preferred embodiment of the multi-source scattering water turbidity detection method of the present invention, the method includes: determining whether the deviation between the plant growth status parameters collected in real time by environmental sensors and the prediction of the preset growth model exceeds a threshold; if it does, triggering a re-optimization of the illumination strategy; and determining the optimal illumination intensity parameters of each wavelength light source through the optimal wavelength combination and historical data, specifically including:

[0021] Obtain a preset plant growth model, determine the optimal growth parameter prediction value for the current growth stage, compare it with the parameters collected in real time, and obtain the deviation value.

[0022] If the deviation value exceeds a preset threshold, the lighting strategy optimization process is triggered; if the deviation value does not exceed the preset threshold, the current lighting strategy is maintained.

[0023] To optimize the light strategy, plant physiology literature and historical planting data were obtained to determine the proportion of light requirements of plants at different growth stages and to obtain the optimal wavelength combination.

[0024] Based on the optimal wavelength combination, the optimized lighting strategy parameters are transmitted to the lighting control module to adjust the illumination intensity of each wavelength light source in the LED array, thereby creating a new lighting environment.

[0025] As a preferred embodiment of the multi-source scattering water turbidity detection method of the present invention, the method includes: using the light source control strategy to control multiple wavelength light sources to irradiate the water sample and obtain scattering spectrum data, including:

[0026] Multiple wavelength scattering spectrum data were acquired, and wavelet transform and least squares method were used for noise reduction. The data were then normalized using the maximum-minimum normalization method to obtain preprocessed scattering spectrum data.

[0027] The preprocessed scattering spectral data is input into a pre-trained support vector machine regression model. The model uses a radial basis kernel function, optimizes hyperparameters through grid search, and is trained using historical water quality data.

[0028] If the water quality parameters of multiple water samples exceed the standard continuously, the ARIMA time series model is used to analyze the trend of water quality parameter changes. Through model fitting and parameter optimization, the water quality changes in the future period can be predicted.

[0029] As a preferred embodiment of the water turbidity detection method using multi-source scattering described in this invention, the method involves performing wavelet multi-scale decomposition and soft threshold filtering on the scattering spectral data to obtain preprocessed scattering spectral data, including:

[0030] Acquire scattering spectral data, and perform denoising and normalization preprocessing on the scattering spectral data to obtain preprocessed scattering spectral data;

[0031] For the preprocessed scattering spectrum data, wavelet transform is used to perform multi-scale decomposition to obtain wavelet coefficients at multiple scales;

[0032] Obtain preset scattering spectrum template data, and use the Euclidean distance calculation method to compare the reconstructed scattering spectrum data with the scattering spectrum template data to calculate the similarity between the two.

[0033] If the similarity is greater than a preset similarity threshold, the reconstructed scattering spectrum data is determined as the preprocessed scattering spectrum data.

[0034] As a preferred embodiment of the multi-source scattering water turbidity detection method of the present invention, the method includes: establishing a quantitative relationship model using the preprocessed scattering spectral data to determine pollutant concentrations, including:

[0035] Preprocessed scattering spectral data is acquired, and based on the characteristics of the scattering spectral data, a support vector regression algorithm is used to establish a quantitative relationship model between the scattering spectral data and pollutant concentration.

[0036] Newly acquired scattering spectral data is obtained, preprocessed, and then input into the quantitative relationship model to obtain the corresponding predicted pollutant concentration.

[0037] The predicted pollutant concentration is compared with a preset environmental quality standard threshold. If the predicted pollutant concentration exceeds the threshold, the corresponding area is identified as an area with excessive pollutant concentration. Pollution control measures are then taken for the area with excessive pollutant concentration.

[0038] A water turbidity detection system based on multi-source scattering water turbidity detection method is characterized by: a wavelength combination optimization module, a light source control strategy module, an illumination data acquisition module, a data preprocessing module, and an exceedance judgment and treatment module.

[0039] The wavelength combination optimization module is used to evaluate the impact of wavelength combinations on pollutant recognition performance based on a preset fitness function, and obtain the optimal wavelength combination.

[0040] The light source control strategy module is used to learn the light source switching and illumination strategy using a reward function for the optimal wavelength combination, and to determine the light source control strategy.

[0041] The illumination data acquisition module is used to control a multi-wavelength light source to irradiate a water sample using a light source control strategy, and to acquire scattering spectrum data.

[0042] The data preprocessing module is used to perform wavelet multi-scale decomposition and soft threshold filtering on the scattering spectrum data to obtain preprocessed scattering spectrum data.

[0043] The Exceedance Judgment and Control Module is used to establish a quantitative relationship model using preprocessed scattering spectral data to determine pollutant concentrations.

[0044] A computer device includes: a memory and a processor; the memory stores a computer program, and the processor executes the computer program to implement the steps of the multi-source scattering water turbidity detection method.

[0045] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the multi-source scattering water turbidity detection method.

[0046] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0047] This invention discloses a method for detecting water pollutants based on multi-wavelength light sources. Addressing the challenges of wavelength selection and light source control strategy optimization in water pollutant detection, this method evaluates the impact of wavelength combinations on pollutant identification using a preset fitness function to obtain the optimal wavelength combination. A reward function is then used to learn and determine the light source switching and illumination strategy. Based on this, water samples are irradiated using multi-wavelength light sources according to the optimized control strategy to acquire scattering spectral data. Data preprocessing is performed using wavelet multi-scale decomposition and soft threshold filtering, ultimately establishing a quantitative relationship model to accurately determine pollutant concentrations. This invention significantly improves the accuracy and efficiency of water pollutant detection through wavelength combination optimization and intelligent learning of light source control strategies, providing an intelligent solution for water quality monitoring. Attached Figure Description

[0048] Figure 1 This is a flowchart of a method, device, and medium for detecting water turbidity using multi-source scattering according to the present invention.

[0049] Figure 2 This is a schematic diagram of a water turbidity detection method, equipment, and medium based on multi-source scattering according to the present invention.

[0050] Figure 3 This is another schematic diagram of a water turbidity detection method, equipment, and medium based on multi-source scattering according to the present invention. Detailed Implementation

[0051] To further understand the content of this invention, a detailed description of the invention is provided in conjunction with the accompanying drawings and embodiments. The specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0052] like Figures 1-3 This embodiment of a method for detecting water turbidity through multi-source scattering specifically includes:

[0053] Step S101: Evaluate the impact of wavelength combinations on pollutant recognition performance based on a preset fitness function to obtain the optimal wavelength combination.

[0054] A set of preset wavelength combinations is obtained as initial candidate wavelength combinations. For each candidate wavelength combination, a support vector machine model is used to identify pollutants in its corresponding spectral data, obtaining the identification results. Based on a preset fitness function, the fitness value of each candidate wavelength combination is calculated. If the optimal fitness value is greater than a preset threshold, the corresponding candidate wavelength combination is determined as the optimal wavelength combination; otherwise, a genetic algorithm is used to optimize the candidate wavelength combinations until the optimal fitness value condition is met or the maximum number of iterations is reached. The optimal wavelength combination is output for subsequent pollutant identification tasks.

[0055] Specifically, a set of preset wavelength combinations is obtained as initial candidate wavelength combinations. For each candidate wavelength combination, a Support Vector Machine (SVM) model is used to identify pollutants in its corresponding spectral data, obtaining the identification results. The fitness value of each candidate wavelength combination is calculated according to a preset fitness function. The fitness function comprehensively considers the identification accuracy of the SVM model and the number of wavelengths in the wavelength combination; the higher the identification accuracy and the fewer the number of wavelengths, the larger the fitness value. It is determined whether the current optimal fitness value is greater than a preset threshold (e.g., 9). If it is, the corresponding candidate wavelength combination is determined as the optimal wavelength combination; otherwise, the following steps are performed for further optimization. The candidate wavelength combinations are optimized using a genetic algorithm. Based on the fitness value, a roulette wheel selection strategy is used to select parent wavelength combinations, and then new offspring wavelength combinations are generated through random crossover and mutation operations. The crossover operation randomly selects some wavelengths from two parent combinations for exchange; the mutation operation randomly selects one wavelength for replacement. The generated offspring wavelength combinations are added to the candidate wavelength combination set, and the process returns to step 2. The updated candidate set is evaluated using the SVM model, and the fitness value is calculated. Repeat steps 4 through 6 until the optimal fitness value exceeds a threshold, or the preset maximum number of iterations is reached. Output the final optimal wavelength combination for subsequent pollutant identification tasks. Using the optimal wavelength combination, select the corresponding spectral bands to construct an SVM model for pollutant identification, applicable to pollutant detection and identification in real-world environments.

[0056] Step S102: For the optimal wavelength combination, a reward function is used to learn the light source switching and illumination strategy to determine the light source control strategy.

[0057] A Q-learning reinforcement learning algorithm is employed to learn light source switching and illumination strategies under different wavelength combinations, obtaining the light source control strategy with the highest reward value. Based on the optimal wavelength combination, combined with a preset plant growth model and historical data, the optimal illumination intensity parameters for each wavelength light source are determined. If the power range of the light source cannot meet the optimal illumination intensity, the light source type with the optimal illumination intensity is selected according to a preset light source replacement rule. Plant growth status parameters are collected in real time by environmental sensors and compared with the preset plant growth model. If the actual growth status deviates from the model prediction value by more than a preset threshold, a re-optimization process for the illumination strategy is triggered.

[0058] Specifically, based on a preset reward function, a Q-learning reinforcement learning algorithm is used to learn light source switching and illumination strategies under different wavelength combinations, obtaining the light source control strategy with the highest reward value. For the learned optimal control strategy, the corresponding optimal wavelength combination, light source switching time, and illumination time are obtained by looking up the strategy table, determining the switching sequence of each wavelength light source. Based on the optimal wavelength combination, combined with the plant growth model and historical data, the optimal illumination intensity parameters of each wavelength light source are obtained, and the actual light source illumination intensity is determined according to the power range of the light source type. If the current power range of the light source cannot meet the optimal illumination intensity, the light source type closest to the optimal illumination intensity is selected according to the preset light source replacement rules, and the illumination time is adjusted accordingly to ensure that the total light intensity meets the plant growth requirements. The optimal wavelength combination, light source switching time, illumination time, and illumination intensity are used as control parameters to generate a light source control strategy, which is then sent to the light source control unit. The light source control unit controls each wavelength light source to switch and illuminate according to the specified sequence and intensity according to the received control strategy, and collects the light source voltage, current, and other status parameters in real time. Based on preset rules for judging the working status of light sources, the system continuously monitors whether each light source is functioning properly. If parameters such as voltage or current exceed normal ranges, an alarm is triggered, and a backup light source is switched to according to a preset emergency response procedure to ensure uninterrupted plant illumination. During illumination, environmental sensors collect plant growth parameters in real time, such as plant height, leaf area, and chlorophyll content, and compare them with a preset plant growth model. If there is a significant deviation between the actual growth status and the model's predictions, a re-optimization process for the illumination strategy is triggered. By adjusting parameters such as illumination time and intensity, the illumination conditions are dynamically optimized to promote healthy plant growth and improve yield and quality. Once the preset plant growth cycle is reached, illumination is stopped, and the optimized light source control strategy is saved to the strategy library to guide subsequent plant cultivation. Simultaneously, the plant growth model is optimized and updated based on the yield and quality of the harvested plants, continuously improving the intelligence level of light source control.

[0059] Based on the plant growth status parameters collected in real time by environmental sensors, it is determined whether the deviation from the prediction of the preset growth model exceeds the threshold. If it does, the illumination strategy is re-optimized. The optimal illumination intensity parameters of each wavelength light source are determined by the optimal wavelength combination and historical data.

[0060] A preset plant growth model is acquired to determine the optimal growth parameter prediction values ​​for the current growth stage. These predictions are then compared with real-time collected parameters to obtain a deviation value. If the deviation value exceeds a preset threshold, a lighting strategy optimization process is triggered; otherwise, the current lighting strategy is maintained. For the lighting strategy optimization process, plant physiology literature and historical planting data are obtained to determine the plant's requirement for different wavelengths of light at different growth stages, resulting in the optimal wavelength combination. Based on this optimal wavelength combination, the optimized lighting strategy parameters are transmitted to the lighting control module to adjust the illumination intensity of each wavelength light source in the LED array, creating a new lighting environment.

[0061] Specifically, based on real-time collection of plant growth status parameters, including temperature, humidity, light intensity, and carbon dioxide concentration, the collected data is transmitted to the data processing module. The data processing module, based on a preset plant growth model, obtains the predicted optimal growth parameters for the current growth stage and compares them with the real-time collected parameters to calculate the deviation. It determines whether the deviation exceeds a preset threshold. If it does, the light strategy optimization process is triggered; otherwise, the current light strategy remains unchanged, and monitoring of plant growth status parameters continues. When triggering light strategy optimization, the optimal wavelength combination required for plant growth must first be obtained. By consulting plant physiology literature and historical planting data, the required proportions of light at different growth stages are determined, forming the optimal wavelength combination. Based on the optimal wavelength combination, the optimal irradiation intensity parameters for each wavelength light source are selected from historical irradiation data. Using data mining techniques, such as association rule analysis, the correlation patterns between light intensity and plant growth status in historical data are identified, determining the optimal irradiation intensity range for each wavelength light source. Finally, machine learning algorithms, such as support vector machines or backpropagation neural networks, are used to establish a correlation model between light intensity and plant growth status. Historical data is divided into training and testing sets. The model is trained using the training set and evaluated using the testing set. By continuously adjusting model parameters, such as the kernel type of the support vector machine and the number of hidden layers and nodes in the neural network, the model's fitting effect is optimized, improving the accuracy of the lighting strategy. The optimized lighting strategy parameters are transmitted to the lighting control module, which adjusts the illumination intensity of each wavelength light source in the LED array to create a new lighting environment. The lighting control module controls the switching and brightness of the LEDs based on the optimized parameters, executing the lighting strategy. Changes in plant growth status parameters are continuously monitored, and data visualization techniques, such as line graphs and radar charts, are used to display the trends in real time. The plant growth status parameters before and after optimization are compared and analyzed to evaluate the effectiveness of the lighting strategy optimization. If the optimization effect is unsatisfactory, a new round of lighting strategy optimization is triggered; if the optimization effect meets expectations, the current lighting strategy is maintained, and plant growth is continuously monitored, forming a closed-loop feedback control.

[0062] Step S103: Use the light source control strategy to control the multi-wavelength light source to irradiate the water sample and obtain scattering spectrum data.

[0063] Multiple wavelength scattering spectral data are acquired, and noise is denoised using wavelet transform and least squares method. The data is then normalized using the maximum-minimum normalization method to obtain preprocessed scattering spectral data. The preprocessed scattering spectral data is input into a pre-trained support vector machine regression model. The model uses a radial basis function kernel function, optimizes hyperparameters through grid search, and is trained using historical water quality data. If the water quality parameters of multiple water samples continuously exceed the standard, an ARIMA time series model is used to analyze the trend of water quality parameter changes. Through model fitting and parameter optimization, the water quality changes in the future period are predicted.

[0064] Specifically, based on a multi-wavelength light source control strategy, multiple wavelength light sources are sequentially used to illuminate water samples, acquiring scattering spectral data for each wavelength. For the acquired scattering spectral data across multiple wavelengths, wavelet transform and least squares methods are used for denoising, followed by normalization using a minimum-maximum normalization method to obtain preprocessed scattering spectral data. This preprocessed scattering spectral data is then input into a pre-trained support vector machine regression model. This model employs a radial basis function kernel function, optimizes hyperparameters through grid search, and is trained using historical water quality data. The model calculates water quality parameters such as turbidity and suspended solids concentration. These parameters are compared to preset thresholds; if a parameter exceeds the threshold, the water quality is deemed abnormal, triggering an early warning. When multiple water samples continuously exceed the water quality standards, an ARIMA time-series model is used to analyze the changing trends of these parameters. Through model fitting and parameter optimization, the water quality changes over a future period are predicted. Based on water quality prediction results, the control strategy for multi-wavelength light sources is automatically adjusted. If the predicted water quality continues to deteriorate, the frequency of water sample testing is increased within a specific time period, such as testing once per hour, to improve the timeliness of water quality testing. If the predicted water quality gradually improves, the normal testing frequency is restored. Water quality test results and early warning information are sent to a monitoring platform. Visualization libraries such as Echarts or Highcharts are used to design bar charts, line charts, and other graphs to display the real-time trends of water quality parameters. When water quality anomalies occur, a red warning icon is displayed on the interface along with a warning description, allowing management personnel to promptly grasp the dynamics of water quality.

[0065] Step S104: Perform wavelet multi-scale decomposition and soft threshold filtering on the scattering spectral data to obtain preprocessed scattering spectral data.

[0066] Scattering spectral data is acquired, and the scattering spectral data is preprocessed by denoising and normalization to obtain preprocessed scattering spectral data. For the preprocessed scattering spectral data, wavelet transform is used to perform multi-scale decomposition to obtain wavelet coefficients at multiple scales. A preset scattering spectral template data is acquired, and the reconstructed scattering spectral data is compared with the scattering spectral template data using Euclidean distance calculation to calculate the similarity between the two. If the similarity is greater than a preset similarity threshold, the reconstructed scattering spectral data is determined as the preprocessed scattering spectral data.

[0067] Specifically, the process involves acquiring scattering spectral data, performing denoising and normalization preprocessing to obtain preprocessed scattering spectral data. For this preprocessed data, a wavelet transform method is used, selecting a suitable wavelet basis function, such as the Daubechies wavelet, to decompose the scattering spectral data into multiple scales (5 levels), yielding wavelet coefficients at multiple scales. For these wavelet coefficients, an adaptive soft thresholding method is employed, automatically setting a soft threshold based on the statistical characteristics of the wavelet coefficients. Wavelet coefficients smaller than the soft threshold are set to zero, resulting in thresholded wavelet coefficients. Based on these thresholded wavelet coefficients, an inverse wavelet transform is used to reconstruct the wavelet coefficients at multiple scales, yielding reconstructed scattering spectral data. A preset scattering spectral template is acquired, and the Euclidean distance is used to compare the reconstructed scattering spectral data with the template data, calculating their similarity. If the similarity is greater than a preset similarity threshold (e.g., 9), the reconstructed scattering spectral data is identified as the preprocessed scattering spectral data. If the similarity is less than or equal to the preset similarity threshold, the soft threshold is adaptively adjusted based on the difference between the current similarity and the target similarity, and the process returns to step three to re-process the threshold until the similarity is greater than the preset similarity threshold or the maximum number of iterations is reached. The preprocessed scattering spectrum data is then output for subsequent scattering spectrum feature extraction and material composition identification analysis.

[0068] Step S105: Establish a quantitative relationship model using the preprocessed scattering spectral data to determine the pollutant concentration.

[0069] Preprocessed scattering spectral data is acquired. Based on the characteristics of the scattering spectral data, a support vector regression algorithm is used to establish a quantitative relationship model between the scattering spectral data and pollutant concentration. Newly acquired scattering spectral data is acquired, preprocessed, and input into the quantitative relationship model to obtain the corresponding predicted pollutant concentration value. The predicted pollutant concentration value is compared with a preset environmental quality standard threshold. If the predicted pollutant concentration value exceeds the threshold, the corresponding area is identified as a pollutant concentration exceeding the standard area, and pollution control measures are taken for the pollutant concentration exceeding the standard area.

[0070] A water turbidity detection system with multi-source scattering includes: a wavelength combination optimization module, a light source control strategy module, an illumination data acquisition module, a data preprocessing module, and an exceedance judgment and treatment module;

[0071] The wavelength combination optimization module is used to evaluate the impact of wavelength combinations on pollutant recognition performance based on a preset fitness function, and obtain the optimal wavelength combination.

[0072] The light source control strategy module is used to learn the light source switching and illumination strategy using a reward function for the optimal wavelength combination, and to determine the light source control strategy.

[0073] The illumination data acquisition module is used to control a multi-wavelength light source to irradiate a water sample using a light source control strategy, and to acquire scattering spectrum data.

[0074] The data preprocessing module is used to perform wavelet multi-scale decomposition and soft threshold filtering on the scattering spectrum data to obtain preprocessed scattering spectrum data.

[0075] The Exceedance Judgment and Control Module is used to establish a quantitative relationship model using preprocessed scattering spectral data to determine pollutant concentrations.

[0076] A computer device includes: a memory and a processor; the memory stores a computer program, and the processor executes the computer program to implement the steps of a method for detecting water turbidity through multi-source scattering.

[0077] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of a method for detecting water turbidity by multi-source scattering.

[0078] Specifically, after acquiring the raw scattering spectrum data, preprocessing operations are performed, including removing noise and outliers. Common denoising methods include wavelet transform and Fourier transform. Outliers can be removed by setting thresholds or using statistical methods such as the 3σ principle, resulting in preprocessed scattering spectrum data. Based on the characteristics of the preprocessed scattering spectrum data, Support Vector Regression (SVR) is selected as the machine learning algorithm to establish a quantitative relationship model between scattering spectrum data and pollutant concentration. A 10-fold cross-validation method is used, randomly dividing the preprocessed scattering spectrum data into 10 subsets. Nine subsets are selected as the training set each time, and the remaining subset is used as the test set to train the SVR model. The model's performance is evaluated on the test set. A grid search method is used to optimize the SVR hyperparameters, such as the penalty coefficient C and kernel function type, selecting the parameter combination with the minimum average prediction error as the optimal model. The established SVR model is then used to predict newly acquired scattering spectrum data. The new spectral data undergoes the same preprocessing operations as the training data and is then input into the trained SVR model to obtain the corresponding predicted pollutant concentration values. To further improve the accuracy of pollutant concentration prediction, more scattering spectral sensors and pollutant concentration monitoring points can be deployed to acquire more scattering spectral data and corresponding actual pollutant concentration values. The newly acquired data can be added to the existing training set to retrain the SVR model, continuously updating and optimizing it. The predicted pollutant concentration values ​​are then compared with environmental quality standards to determine if they exceed the limits. For areas where pollutant concentrations exceed the standards, timely pollution control measures should be implemented, such as restricting pollution source emissions and increasing pollutant treatment facilities. Simultaneously, the pollutant concentration prediction results should be presented to environmental management departments and the public in the form of charts and reports to provide a basis for environmental decision-making and enhance public environmental awareness and participation.

[0079] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for detecting turbidity in water bodies using multi-source scattering, characterized in that, The method includes: The impact of wavelength combinations on pollutant recognition performance is evaluated based on a preset fitness function to obtain the optimal wavelength combinations, including: Obtain a set of preset wavelength combinations as the initial candidate wavelength combinations; For each candidate wavelength combination, a support vector machine model is used to identify pollutants in its corresponding spectral data to obtain the identification results; Based on the preset fitness function, calculate the fitness value of each candidate wavelength combination. If the optimal fitness value is greater than the preset threshold, then the corresponding candidate wavelength combination is determined as the optimal wavelength combination. Otherwise, a genetic algorithm is used to optimize the candidate wavelength combination until the optimal fitness value condition is met or the maximum number of iterations is reached. The optimal wavelength combination is output for subsequent pollutant identification tasks; For the optimal wavelength combination, a reward function is used to learn the light source switching and illumination strategy, and a light source control strategy is determined, including: The Q-learning reinforcement learning algorithm is used to learn the light source switching and illumination strategies under different wavelength combinations, and to obtain the light source control strategy with the maximum reward value. Based on the optimal wavelength combination, combined with the preset plant growth model and historical data, the optimal irradiation intensity parameters of each wavelength light source are determined. If the power range of the light source cannot meet the optimal illumination intensity, then the type of light source with the optimal illumination intensity is selected according to the preset light source replacement rules. Plant growth status parameters are collected in real time by environmental sensors and compared with preset plant growth models. If the actual growth status deviates from the model prediction value by more than a preset threshold, the light strategy re-optimization process is triggered. The light source control strategy described above is used to control multiple wavelength light sources to irradiate water samples and obtain scattering spectral data, including: Multiple wavelength scattering spectrum data were acquired, and wavelet transform and least squares method were used for noise reduction. The data were then normalized using the maximum-minimum normalization method to obtain preprocessed scattering spectrum data. The preprocessed scattering spectral data is input into a pre-trained support vector machine regression model. The model uses a radial basis kernel function, optimizes hyperparameters through grid search, and is trained using historical water quality data. If the water quality parameters of multiple water samples exceed the standard continuously, the ARIMA time series model is used to analyze the trend of water quality parameter changes. Through model fitting and parameter optimization, the water quality changes in the future period are predicted. The scattering spectrum data is subjected to wavelet multi-scale decomposition and soft threshold filtering to obtain preprocessed scattering spectrum data; A quantitative relationship model was established using the preprocessed scattering spectral data to determine the pollutant concentration.

2. The method according to claim 1, characterized in that, Based on real-time plant growth status parameters collected by environmental sensors, it is determined whether the deviation from the prediction of the preset growth model exceeds a threshold. If it does, the illumination strategy is re-optimized. By using the optimal wavelength combination and historical data, the optimal illumination intensity parameters for each wavelength light source are determined, specifically including: Obtain a preset plant growth model, determine the optimal growth parameter prediction value for the current growth stage, compare it with the parameters collected in real time, and obtain the deviation value. If the deviation value exceeds a preset threshold, the lighting strategy optimization process is triggered; if the deviation value does not exceed the preset threshold, the current lighting strategy is maintained. To optimize the light strategy, plant physiology literature and historical planting data were obtained to determine the proportion of light requirements of plants at different growth stages and to obtain the optimal wavelength combination. Based on the optimal wavelength combination, the optimized lighting strategy parameters are transmitted to the lighting control module to adjust the illumination intensity of each wavelength light source in the LED array, thereby creating a new lighting environment.

3. The method according to claim 2, characterized in that, The scattering spectral data is subjected to wavelet multi-scale decomposition and soft thresholding to obtain preprocessed scattering spectral data, including: Acquire scattering spectral data, and perform denoising and normalization preprocessing on the scattering spectral data to obtain preprocessed scattering spectral data; For the preprocessed scattering spectrum data, wavelet transform is used to perform multi-scale decomposition to obtain wavelet coefficients at multiple scales; Obtain preset scattering spectrum template data, and use the Euclidean distance calculation method to compare the reconstructed scattering spectrum data with the scattering spectrum template data to calculate the similarity between the two. If the similarity is greater than a preset similarity threshold, the reconstructed scattering spectrum data is determined as the preprocessed scattering spectrum data.

4. The method according to claim 3, characterized in that, A quantitative relationship model is established using the preprocessed scattering spectral data to determine pollutant concentrations, including: Preprocessed scattering spectral data is acquired, and based on the characteristics of the scattering spectral data, a support vector regression algorithm is used to establish a quantitative relationship model between the scattering spectral data and pollutant concentration. Newly acquired scattering spectral data is obtained, preprocessed, and then input into the quantitative relationship model to obtain the corresponding predicted pollutant concentration. The predicted pollutant concentration is compared with a preset environmental quality standard threshold. If the predicted pollutant concentration exceeds the threshold, the corresponding area is identified as an area with excessive pollutant concentration. Pollution control measures are then taken for the area with excessive pollutant concentration.

5. A water turbidity detection system based on the water turbidity detection method based on the multi-source scattering method according to any one of claims 1-4, characterized in that: The system includes a wavelength combination optimization module, a light source control strategy module, an illumination data acquisition module, a data preprocessing module, and a exceedance judgment and control module. The wavelength combination optimization module is used to evaluate the impact of wavelength combinations on pollutant recognition performance based on a preset fitness function, and obtain the optimal wavelength combination. The light source control strategy module is used to learn the light source switching and illumination strategy using a reward function for the optimal wavelength combination, and to determine the light source control strategy. The illumination data acquisition module is used to control a multi-wavelength light source to irradiate a water sample using a light source control strategy, and to acquire scattering spectrum data. The data preprocessing module is used to perform wavelet multi-scale decomposition and soft threshold filtering on the scattering spectrum data to obtain preprocessed scattering spectrum data. The Exceedance Judgment and Control Module is used to establish a quantitative relationship model using preprocessed scattering spectral data to determine pollutant concentrations.

6. A computer device, comprising: memory and processor; The memory stores a computer program, characterized in that: when the processor executes the computer program, it implements the steps of the water turbidity detection method with multi-source scattering as described in any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the water turbidity detection method with multi-source scattering as described in any one of claims 1 to 4.

Images