Household garbage heavy metal content online detection method based on hyperspectral imaging

Abstract

The invention relates to the technical field of environmental protection detection, provides a household garbage heavy metal content online detection method based on hyperspectral imaging, and solves the problems that in an existing heavy metal recognition technology, online remote recognition cannot be carried out on a severe field environment; for the problem of large heavy metal content prediction error caused by neglect of field environment conditions, a hyperspectral image of household garbage is obtained through a hyperspectral sensor, temperature, humidity and dust concentration information in a test space are combined, and a simulation model under laboratory conditions is utilized to adjust spectral reflectivity to a standard condition, so that the content of heavy metal in the household garbage is predicted. According to the method, the light GBM algorithm is adopted to extract spectral features, CatBoost is adopted to carry out machine learning, prediction model parameters are optimized through the genetic algorithm, prediction accuracy is improved, and on-line rapid detection of the heavy metal content of the household garbage is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental protection detection, and in particular to an online detection method for heavy metal content in domestic waste based on hyperspectral imaging. Background Art

[0002] Hyperspectral imaging is a cutting-edge technology that combines imaging technology with spectral technology. It uses hyperspectral sensors to simultaneously obtain the spectral information of the target in multiple continuous and narrow bands to form a hyperspectral data cube. By analyzing and processing the data, the spectral characteristics of the target in different bands can be obtained, thereby achieving accurate identification and classification of the target. With its unique spectral resolution and spatial resolution, hyperspectral imaging technology has shown broad application prospects in food safety, medical diagnosis, aerospace and other fields, bringing many conveniences to scientific research, industrial production and social life.

[0003] In the process of online detection of heavy metal content in domestic waste, hyperspectral imaging technology can capture the reflection and absorption characteristics of different substances in the waste to light of different wavelengths by obtaining hyperspectral images of domestic waste. Since heavy metal elements have unique spectral characteristics at specific wavelengths, hyperspectral imaging technology can be used to identify and analyze these characteristics, thereby realizing rapid detection of heavy metal content in domestic waste.

[0004] When testing the heavy metal content of domestic waste incineration ash, the humidity, temperature and dust content in the on-site environment space are different from those in the laboratory. The air humidity will affect the content of water molecules in the air. Water molecules absorb and scatter infrared light of a specific wavelength. Temperature changes will affect the thermal radiation characteristics of the material and the vibration and rotation energy levels of the molecules, thereby affecting the infrared reflection spectrum of a specific band. Dust particles will scatter and absorb infrared light, resulting in large errors in the reflection spectrum information of the infrared band of the same type and content of heavy metals under different air conditions. Therefore, when using infrared reflection spectroscopy to predict the heavy metal content of ash, it is necessary to consider the impact of the on-site temperature, humidity and dust content on the spectral reflection information. Therefore, we propose an online detection method for the heavy metal content of domestic waste based on hyperspectral imaging. Summary of the invention

[0005] In view of the shortcomings of the existing technology, the present invention provides an online detection method for the heavy metal content of domestic waste based on hyperspectral imaging, which is used to solve the problem that the existing heavy metal identification technology is unable to perform online long-distance identification of harsh on-site environments and ignores on-site environmental conditions, resulting in large errors in heavy metal content prediction.

[0006] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows:

[0007] The present invention provides an online detection method for heavy metal content in domestic waste based on hyperspectral imaging, comprising:

[0008] Step S101, receiving configuration information, a start command and test space information, the optical system in the hyperspectral sensor loads the configuration information, after the configuration information is loaded, the optical system in the hyperspectral sensor executes the start command, the detector receives reflected light, saves the hyperspectral data at a preset time frequency and records the time, and obtains infrared information in the saved hyperspectral data;

[0009] Step S102, determining the test space position according to the test space information, obtaining the data acquisition device information within the test space position range, and receiving the space environment data acquired by the data acquisition device within the test space position range, the space environment data including the temperature information within the test space, the humidity information within the test space, and the dust concentration information within the test space;

[0010] Step S103, extracting infrared information from the stored hyperspectral data and calculating the spectral reflectance R;

[0011] Step S104, substituting the temperature information in the test space, the humidity information in the test space, and the dust concentration information in the test space into a simulation model under preset laboratory conditions to obtain a spectral reflectance R2 under standard conditions;

[0012] Step S105, taking the spectral reflectance R2 under the standard state as input into the prediction model to obtain the content of different types of heavy metals.

[0013] Furthermore, in the method for online detection of heavy metal content in domestic waste based on hyperspectral imaging of the present invention, step S101 comprises:

[0014] The configuration information includes a wavelength range of 1000-3000 nm and a spectral resolution set to 3 nm;

[0015] The preset time frequency includes saving the data collected by the hyperspectral sensor every 1 second.

[0016] Furthermore, in the method for online detection of heavy metal content in domestic waste based on hyperspectral imaging of the present invention, step S102 comprises:

[0017] The data acquisition device includes a temperature sensor, a humidity sensor and a light scattering dust monitor;

[0018] Obtain temperature information in the test space through a temperature sensor;

[0019] Obtain humidity information in the test space through a humidity sensor;

[0020] The dust concentration information in the space is obtained through the light scattering dust monitor.

[0021] Furthermore, in the method for online detection of heavy metal content in domestic waste based on hyperspectral imaging of the present invention, step S103 comprises:

[0022] The infrared information in the stored hyperspectral data is stored in the form of a three-dimensional data cube to obtain hyperspectral data cube information, wherein the hyperspectral data cube information includes a row space dimension and a spectral dimension;

[0023] The hyperspectral data cube is parsed to form a two-dimensional array with wavelength and corresponding reflection intensity. The spectral reflectance or radiance at each pixel position is calculated. After correction with a black and white plate, the infrared spectral reflectance R of the corresponding target at 1000-3000nm is obtained.

[0024] Furthermore, in the method for online detection of heavy metal content in domestic waste based on hyperspectral imaging of the present invention, step S104 comprises:

[0025] The spectral reflectance R is subjected to Gaussian smoothing to obtain the spectral reflectance R1;

[0026] Collect the optical parameters of different refractive index, absorption coefficient, scattering coefficient and extinction coefficient corresponding to the temperature information, humidity information and dust concentration information under laboratory conditions;

[0027] The temperature information, humidity information, and dust concentration information are used as boundary conditions, and the spectral reflectance R1 is used as input. The simulation model under laboratory conditions is substituted to obtain the spectral reflectance R2 under standard conditions.

[0028] Furthermore, in the method for online detection of heavy metal content in domestic waste based on hyperspectral imaging of the present invention, step S105 comprises:

[0029] The ash samples with different heavy metal contents under standard conditions were collected and analyzed for spectral reflectance. The spectral reflectance of the ash samples with different heavy metal contents under standard conditions was analyzed, and the spectral features were extracted using the LightGBM algorithm. The spectral reflectance corresponding to the first 20 bands with the largest scores in the spectral features was selected as the characteristic reflectance.

[0030] Taking the characteristic reflectance as input and the content of different types of heavy metals as output, CatBoost is selected for machine learning to obtain a prediction model. The prediction model is used to process the characteristic reflectance data to obtain the prediction results.

[0031] The prediction results obtained by the prediction model are imported into the recognition system, and R2 obtained by spectral image extraction and calculation is used as input to predict the content Q of different types of heavy metals.

[0032] Furthermore, in the method for online detection of heavy metal content in domestic waste based on hyperspectral imaging of the present invention, step S105 comprises:

[0033] Obtain the actual test results of the experiment and the data processing history, compare the actual test results of the experiment with the laboratory standard test results, and obtain the test result error data by comparing the difference between the two. If the test result error data is greater than the preset minimum error range, optimize the test result error data, the actual test results of the experiment, and the data processing history through a genetic algorithm;

[0034] Initialize the initial population of individuals, where individuals refer to different parameter combinations of the prediction model;

[0035] According to the population size, the initial population is randomly generated, and each individual represents the genotype of the chromosome, and the genotype of the chromosome includes the hyperparameters of the model;

[0036] Calculate the fitness of each individual, that is, the accuracy of the model's prediction results;

[0037] If the error is less than the set value, the best individual and the optimal solution it represents are output, and the algorithm ends;

[0038] If the error is less than the set value, the regenerated individuals are selected according to the fitness, and new individuals are created from the most suitable individuals by combining the genetic material. The genetic material of the individuals is randomly changed to produce a new generation of populations, and the fitness of each individual is recalculated until the optimal solution is obtained and the algorithm stops when the maximum number of iterations is reached;

[0039] After the algorithm is completed, the optimal individual and the optimal solution it represents, that is, the optimized prediction model parameters, are output, and the characteristic reflectivity is predicted using the optimized prediction model.

[0040] Beneficial effects of the present invention:

[0041] The present invention realizes real-time monitoring and rapid feedback of heavy metal content in the incineration equipment by saving hyperspectral data every 1 second, avoiding the lag of existing detection methods, so that users can timely understand the heavy metal content in the process of domestic waste incineration. The invention takes into account environmental factors such as temperature, humidity and dust concentration in the test space, and uses a simulation model under laboratory conditions for correction, which effectively improves the accuracy of spectral reflectance data, thereby improving the accuracy of heavy metal content prediction.

[0042] The prediction model is optimized using genetic algorithms. According to the differences between the actual test results of the experiment and the laboratory standard test results, the model parameters are continuously adjusted, which helps to obtain more accurate prediction results and improve the reliability and effectiveness of the detection method.

[0043] The online detection method of the present invention helps to timely monitor and control heavy metal pollution during the incineration of domestic waste, which is of great significance for environmental protection, can effectively reduce the harm of heavy metals to the environment and human body, and promote sustainable development.

[0044] In summary, the present invention has beneficial effects in terms of real-time monitoring and feedback, improved accuracy, optimized prediction models, and environmental protection significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on the drawings without paying any creative labor.

[0046] Figure 1 A schematic flow chart of a method for online detection of heavy metal content in domestic waste based on hyperspectral imaging provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. The technical solutions provided by the embodiments of the present invention are described in detail below in conjunction with the drawings.

[0048] In order to better understand the purpose of the present invention, the present invention is described in further detail below.

[0049] The present invention provides an online detection method for heavy metal content in domestic waste based on hyperspectral imaging, comprising:

[0050] Step S101, receiving configuration information, a start command and test space information, the optical system in the hyperspectral sensor loads the configuration information, after the configuration information is loaded, the optical system in the hyperspectral sensor executes the start command, the detector receives reflected light, saves the hyperspectral data at a preset time frequency and records the time, and obtains infrared information in the saved hyperspectral data;

[0051] The system first receives configuration information, a startup command, and test space information. The configuration information includes key parameters such as wavelength range and spectral resolution. In an embodiment of the present invention, the configuration information is specifically set to a wavelength range of 1000-3000nm, and a spectral resolution of 3nm. The selection of parameters is based on the effectiveness and accuracy of hyperspectral imaging technology in heavy metal detection, so that the unique spectral characteristics of heavy metal elements at specific wavelengths can be captured.

[0052] The optical system in the hyperspectral sensor loads these configuration information. When the configuration information is loaded, the optical system executes the start command and starts working. At this time, the detector begins to receive reflected light, which is reflected back from the domestic waste samples in the test space. The detector is very sensitive and can capture the reflection intensity of light of different wavelengths to generate hyperspectral data.

[0053] In order to monitor the heavy metal content in domestic waste in real time, the system needs to save these hyperspectral data at a certain time frequency and record the corresponding timestamp. In an embodiment of the present invention, the preset time frequency is set to save the data collected by the hyperspectral sensor every 1 second. Such a setting enables the system to capture any changes in heavy metal content and provide real-time feedback.

[0054] Finally, the system extracts infrared information from these saved hyperspectral data. Infrared information is the part of the hyperspectral data that corresponds to the infrared band. It is particularly important for the detection of heavy metals because heavy metal elements often have unique spectral characteristics in the infrared band. The extracted infrared information will be used for subsequent spectral analysis and heavy metal content prediction.

[0055] In summary, step S101 is a basic step in the method of the present invention, which enables the hyperspectral sensor to collect accurate infrared spectrum data according to a predetermined configuration and time frequency, and provides a reliable basis for subsequent heavy metal content prediction.

[0056] Step S102, determining the test space position according to the test space information, obtaining the data acquisition device information within the test space position range, and receiving the space environment data acquired by the data acquisition device within the test space position range, the space environment data including the temperature information within the test space, the humidity information within the test space, and the dust concentration information within the test space;

[0057] The specific content of step S102 is that in the online detection method for heavy metal content in domestic waste based on hyperspectral imaging, the position of the test space is determined according to the test space information. When the test space position is determined, the system will obtain information on all relevant data acquisition devices within the test space position. The data acquisition device is mainly responsible for collecting environmental data in the test space.

[0058] Specifically, the data acquisition device includes a temperature sensor, a humidity sensor, and a light scattering dust monitor. The temperature sensor is used to obtain the temperature information in the test space, and can monitor and record the temperature data of the current environment in real time. The humidity sensor is used to obtain the humidity information in the test space, and can also monitor and record the humidity data of the current environment in real time. The light scattering dust monitor is used to measure the dust concentration information in the test space, and can infer the dust particle concentration in the air by measuring the scattered light in the air.

[0059] Through the data acquisition device, the system can obtain environmental data such as temperature, humidity and dust concentration in the test space.

[0060] Step S103, extracting infrared information from the stored hyperspectral data and calculating the spectral reflectance R;

[0061] The system extracts infrared information from previously saved hyperspectral data. Hyperspectral data are collected and saved by hyperspectral sensors at a preset time frequency (for example, every 1 second). They contain the spectral characteristics of domestic waste in different bands.

[0062] The system stores infrared information in the form of a three-dimensional data cube. The hyperspectral data cube contains row space dimensions, column space dimensions, and spectral dimensions, and comprehensively records the spectral information of the target in different bands. The system parses this hyperspectral data cube.

[0063] Specifically, the system uses wavelength and corresponding reflection intensity to form a two-dimensional array, and calculates the spectral reflectance or radiance at each pixel position. The calculation process is calibrated with a black and white board to improve the accuracy of the measurement results. Black and white board calibration is a spectral measurement calibration method that uses black and white boards with known reflectance as references to correct the measurement results to eliminate system errors.

[0064] Finally, after the above processing, the system obtains the infrared spectral reflectance R of the corresponding target in the wavelength range of 1000-3000nm. This reflectance R is an important input parameter for heavy metal content prediction in subsequent steps.

[0065] In summary, the core content of step S103 is to extract infrared information from the stored hyperspectral data and obtain the infrared spectral reflectance of the target within a specific wavelength range through analysis and calculation. This process provides key data support for subsequent heavy metal content prediction.

[0066] Step S104, substituting the temperature information in the test space, the humidity information in the test space, and the dust concentration information in the test space into a simulation model under preset laboratory conditions to obtain a spectral reflectance R2 under standard conditions;

[0067] In the online detection method for heavy metal content in domestic waste based on hyperspectral imaging, step S104 involves adjusting the spectral reflectance under the actual test environment to the spectral reflectance under standard conditions for subsequent heavy metal content prediction. The following is an explanation of the specific content of step S104:

[0068] Gaussian smoothing is performed on the infrared information (i.e., spectral reflectance R) extracted from the hyperspectral data. Gaussian smoothing is a technique used to reduce data noise by smoothing the data by weighted averaging the data points. After Gaussian smoothing, a new spectral reflectance R1 is obtained, which is smoother and has less noise.

[0069] The simulation model was established under laboratory conditions to simulate the effects of different environmental parameters (such as temperature, humidity, and dust concentration) on spectral reflectance.

[0070] The steps of building the simulation model include: selecting specific temperature, humidity and dust concentration as standard conditions in the laboratory, simulating all possible working conditions, and collecting the corresponding spectral reflectance. Set four different optical parameters, namely, different refractive index, absorption coefficient, scattering coefficient and extinction coefficient, according to the temperature information, humidity information and dust concentration information, and build a physical model in which different optical parameters change with the temperature information, humidity information and dust concentration. Use the ray tracing algorithm to build a model of infrared light propagation in air containing dust and water vapor, including the absorption model and scattering model of water molecules for infrared light, and the absorption model and scattering model of dust particles for infrared spectrum.

[0071] Different parameter ranges of temperature field, humidity field and particle concentration field are set on ANSYS software, different boundary conditions are set, and the diffuse reflection data of infrared light under different temperature, humidity and dust particle concentration conditions are obtained by simulation and calculation. The influence of temperature, humidity and dust concentration on the reflection spectrum is analyzed, and the model prediction value is compared with the actual value to continuously optimize the model.

[0072] The temperature information, humidity information, and dust concentration information are used as boundary conditions, and the spectral reflectance R1 is used as input and brought into the simulation model to solve the spectral reflectance R2 under standard conditions.

[0073] Step S104: Substitute the temperature information in the test space, the humidity information in the test space, and the dust concentration information in the test space into a simulation model under preset laboratory conditions to obtain the spectral reflectance R2 under standard conditions.

[0074] Step S104 implementation steps:

[0075] The infrared information (i.e., spectral reflectance R) extracted from the hyperspectral data is subjected to Gaussian smoothing to reduce data noise and obtain a new spectral reflectance R1.

[0076] Under laboratory conditions, a model was constructed to simulate the effects of different environmental parameters (such as temperature, humidity, and dust concentration) on spectral reflectance. These parameters include refractive index, absorption coefficient, scattering coefficient, and extinction coefficient. The ray tracing algorithm was used to simulate the behavior of infrared light propagating in air containing dust and water vapor.

[0077] Different temperature field, humidity field and particle concentration field parameter ranges were set in the ANSYS software, and the infrared diffuse reflection data under different conditions were obtained by simulation and calculation. The influence of various factors on the reflection spectrum was analyzed, and the prediction accuracy of the model was continuously optimized.

[0078] The temperature information, humidity information and dust concentration information in the actual test space are used as boundary conditions.

[0079] The spectral reflectance R1 after Gaussian smoothing is used as input and substituted into the simulation model constructed under laboratory conditions.

[0080] Through the calculation of the simulation model, the spectral reflectance R2 under standard conditions is obtained. This R2 has taken into account the influence of factors such as temperature, humidity and dust concentration in the actual test environment, and is closer to the ideal value under laboratory conditions.

[0081] Gaussian smoothing is a common method for data smoothing, which reduces data noise by weighted averaging. The processed spectral reflectance R1 is smoother, which is conducive to subsequent processing.

[0082] The simulation model was built under laboratory conditions to simulate the effects of different environmental parameters on spectral reflectance. By continuously collecting experimental data and optimizing the model, it can accurately predict the spectral reflectance under standard conditions.

[0083] Substituting environmental parameters such as temperature, humidity and dust concentration in the actual test space into the simulation model can make the simulation model output closer to the actual situation and improve the accuracy of the prediction.

[0084] Through the above steps, step S104 successfully adjusts the spectral reflectance under the actual test environment to the spectral reflectance under standard conditions, providing a reliable data basis for subsequent heavy metal content prediction.

[0085] The actual test environment parameters are substituted into the simulation model, and the temperature information, humidity information and dust concentration information in the actual test space are used as boundary conditions. The spectral reflectance R1 after Gaussian smoothing is used as input and substituted into the simulation model under laboratory conditions.

[0086] Obtain the spectral reflectance R2 under standard conditions, and obtain the spectral reflectance R2 under standard conditions through simulation model calculation. The spectral reflectance R2 under standard conditions has taken into account the influence of factors such as temperature, humidity and dust concentration in the actual test environment, so it is closer to the ideal value under laboratory conditions.

[0087] In summary, the purpose of step S104 is to adjust the spectral reflectance under the actual test environment to the spectral reflectance under standard conditions for subsequent heavy metal content prediction. This step is achieved by Gaussian smoothing, the establishment of a simulation model under laboratory conditions, and the substitution of actual test environment parameters.

[0088] Step S105, taking the spectral reflectance R2 under the standard state as input into the prediction model to obtain the content of different types of heavy metals.

[0089] To prepare data and extract features, it is necessary to collect ash samples with different heavy metal contents under standard conditions and analyze the spectral reflectance of these samples. Then, the LightGBM algorithm is used to extract features from the collected spectral reflectance data, and the spectral reflectance corresponding to the top 20 bands with the largest scores in the spectral features is selected as the characteristic reflectance.

[0090] Secondly, a prediction model was established. With characteristic reflectance as input and the content of different types of heavy metals as output, the CatBoost algorithm was selected for machine learning and trained to obtain a prediction model.

[0091] Next, the model is predicted and optimized. The prediction results obtained by the prediction model are imported into the recognition system, and the R2 (i.e., spectral reflectance under standard conditions) obtained by spectral image extraction is used as input to predict the content of different types of heavy metals. In order to improve the accuracy of the prediction, the prediction model needs to be optimized and the optimal solution needs to be found through genetic algorithms.

[0092] Finally, the results are output. The optimized prediction model is used to predict the characteristic reflectance, obtain the content of different types of heavy metals, and output the results.

[0093] The different types of heavy metals in the technical solution of the present invention include:

[0094] Lead (Pb): Lead is a common heavy metal pollutant that is harmful to human health and the environment. Monitoring the lead content is of great significance for assessing the degree of environmental pollution and protecting human health.

[0095] Mercury (Hg): Mercury is a toxic heavy metal that easily accumulates in the environment and enters the human body through the food chain. Monitoring the content of mercury helps to detect and control its pollution in a timely manner.

[0096] Cadmium (Cd): Cadmium is a carcinogen that is toxic to multiple organ systems in the human body. Monitoring cadmium levels is essential for preventing and controlling its contamination.

[0097] Chromium (Cr): Although chromium is beneficial to the human body to a certain extent, excessive intake can cause health problems. Monitoring chromium levels can help ensure that they are within safe ranges.

[0098] Arsenic (As): Arsenic is a toxic element that can cause serious health problems if exposed or ingested for a long time. Monitoring arsenic levels is an important measure to protect the environment and human health.

[0099] Copper (Cu), zinc (Zn), nickel (Ni): These heavy metals are widely used in industrial production, but excessive emissions can have a negative impact on the environment and ecosystem. Monitoring their content helps control industrial pollution and protect the ecological environment.

[0100] Furthermore, in the method for online detection of heavy metal content in domestic waste based on hyperspectral imaging of the present invention, step S101 comprises:

[0101] The configuration information includes a wavelength range of 1000-3000 nm and a spectral resolution set to 3 nm;

[0102] The preset time frequency includes saving the data collected by the hyperspectral sensor every 1 second.

[0103] Furthermore, in the method for online detection of heavy metal content in domestic waste based on hyperspectral imaging of the present invention, step S102 comprises:

[0104] The data acquisition device includes a temperature sensor, a humidity sensor and a light scattering dust monitor;

[0105] Obtain temperature information in the test space through a temperature sensor;

[0106] Obtain humidity information in the test space through a humidity sensor;

[0107] The dust concentration information in the space is obtained through the light scattering dust monitor.

[0108] Furthermore, in the method for online detection of heavy metal content in domestic waste based on hyperspectral imaging of the present invention, step S103 comprises:

[0109] The infrared information in the stored hyperspectral data is stored in the form of a three-dimensional data cube to obtain hyperspectral data cube information, wherein the hyperspectral data cube information includes a row space dimension and a spectral dimension;

[0110] The hyperspectral data cube is parsed to form a two-dimensional array with wavelength and corresponding reflection intensity. The spectral reflectance or radiance at each pixel position is calculated. After correction with a black and white plate, the infrared spectral reflectance R of the corresponding target at 1000-3000nm is obtained.

[0111] Furthermore, in the method for online detection of heavy metal content in domestic waste based on hyperspectral imaging of the present invention, step S104 comprises:

[0112] The spectral reflectance R is subjected to Gaussian smoothing to obtain the spectral reflectance R1;

[0113] Collect the optical parameters of different refractive index, absorption coefficient, scattering coefficient and extinction coefficient corresponding to the temperature information, humidity information and dust concentration information under laboratory conditions;

[0114] The temperature information, humidity information, and dust concentration information are used as boundary conditions, and the spectral reflectance R1 is used as input. The simulation model under laboratory conditions is substituted to obtain the spectral reflectance R2 under standard conditions.

[0115] Furthermore, in the method for online detection of heavy metal content in domestic waste based on hyperspectral imaging of the present invention, step S105 comprises:

[0116] The ash samples with different heavy metal contents under standard conditions were collected and analyzed for spectral reflectance. The spectral reflectance of the ash samples with different heavy metal contents under standard conditions was analyzed, and the spectral features were extracted using the LightGBM algorithm. The spectral reflectance corresponding to the first 20 bands with the largest scores in the spectral features was selected as the characteristic reflectance.

[0117] Taking the characteristic reflectance as input and the content of different types of heavy metals as output, CatBoost is selected for machine learning to obtain a prediction model. The prediction model is used to process the characteristic reflectance data to obtain the prediction results.

[0118] The prediction results obtained by the prediction model are imported into the recognition system, and R2 obtained by spectral image extraction and calculation is used as input to predict the content Q of different types of heavy metals.

[0119] Furthermore, in the method for online detection of heavy metal content in domestic waste based on hyperspectral imaging of the present invention, step S105 comprises:

[0120] Obtain the actual test results of the experiment and the data processing history, compare the actual test results of the experiment with the laboratory standard test results, and obtain the test result error data by comparing the difference between the two. If the test result error data is greater than the preset minimum error range, optimize the test result error data, the actual test results of the experiment, and the data processing history through a genetic algorithm;

[0121] Initialize the initial population of individuals, where individuals refer to different parameter combinations of the prediction model;

[0122] According to the population size, the initial population is randomly generated, and each individual represents the genotype of the chromosome, and the genotype of the chromosome includes the hyperparameters of the model;

[0123] Calculate the fitness of each individual, that is, the accuracy of the model's prediction results;

[0124] If the error is less than the set value, the best individual and the optimal solution it represents are output, and the algorithm ends;

[0125] If the error is less than the set value, the regenerated individuals are selected according to the fitness, and new individuals are created from the most suitable individuals by combining the genetic material. The genetic material of the individuals is randomly changed to produce a new generation of populations, and the fitness of each individual is recalculated until the optimal solution is obtained and the algorithm stops when the maximum number of iterations is reached;

[0126] After the algorithm is completed, the optimal individual and the optimal solution it represents, that is, the optimized prediction model parameters, are output, and the characteristic reflectivity is predicted using the optimized prediction model.

[0127] The specific content of genetic algorithm optimization is as follows: obtain the actual test results of the experiment and the data processing history. This is to be able to compare the actual test results of the experiment with the laboratory standard test results, so as to evaluate the accuracy of the current prediction model. By comparing the difference between the two, the error data of the test results can be obtained.

[0128] If the error data of the test result is greater than the preset minimum error range, it means that the accuracy of the current prediction model needs to be improved, so it needs to be optimized. At this time, based on the error data of the test result, the actual test results of the experiment and the data processing history, the optimization is performed through the genetic algorithm.

[0129] The optimization process of the genetic algorithm is as follows: Initialize the initial population of individuals, where "individuals" refer to different parameter combinations of the prediction model. Each individual represents a possible prediction model configuration.

[0130] The initial population is randomly generated according to the population size. Each individual represents the genotype of the chromosome, and the genotype of the chromosome includes the hyperparameters of the model. These hyperparameters will determine the specific performance of the prediction model.

[0131] Calculate the fitness of each individual. Here, fitness refers to the accuracy of the model's prediction results. By comparing the difference between the predicted results and the actual results, the fitness of each individual can be evaluated.

[0132] If the error of an individual is less than the set value, it means that the prediction model of this individual performs well, and it can be output as the best individual and the algorithm ends.

[0133] If the error of all individuals is greater than the set value, the individuals to be regenerated need to be selected according to fitness. These individuals will serve as the basis for the new generation of population.

[0134] Combining genetic material to create new individuals from the fittest individuals. This usually includes operations such as crossover (i.e. exchanging some genes between two individuals) and mutation (i.e. changing the genes of an individual). These operations will produce a new generation of the population.

[0135] Recalculate the fitness of each new individual and repeat the above steps until the maximum number of iterations is reached or the best individual that meets the requirements is found.

[0136] After the algorithm is finished, the optimal individual and the optimal solution it represents are output. This optimal solution is the optimized prediction model parameters. Using these optimized parameters to predict the characteristic reflectance can obtain more accurate prediction results of heavy metal content.

[0137] In general, the genetic algorithm optimization in step S105 is an iterative process, which continuously adjusts and improves the parameter configuration of the prediction model to improve the accuracy of the model prediction.

[0138] The technical solution of the present invention solves the problem that the existing hyperspectral technology cannot perform corresponding data processing for different field environments through the following steps:

[0139] After receiving configuration information, startup commands and test space information, the hyperspectral sensor loads the configuration and executes the startup command, saves the hyperspectral data at a preset time frequency (such as every 1 second) and records the time. According to the test space information, determine the location and obtain the information of the data acquisition device (temperature sensor, humidity sensor, light scattering dust meter), and receive the space environment data (temperature, humidity, dust concentration) collected by these devices.

[0140] Extract the infrared information from the saved hyperspectral data and calculate the spectral reflectance R. Substitute the spatial environment data into the simulation model under the preset laboratory conditions to obtain the spectral reflectance R2 under standard conditions.

[0141] The spectral reflectance R2 under standard conditions is taken as input into the prediction model to predict the content of different types of heavy metals.

[0142] LightGBM algorithm is used to extract spectral features and select characteristic reflectance; CatBoost is used for machine learning to obtain the prediction model. If the actual test results of the experiment are significantly different from the laboratory standard test results, the prediction model parameters are optimized by genetic algorithm to improve the prediction accuracy.

[0143] The present invention can realize online rapid detection of heavy metal content in domestic garbage, and solves the problems of long cycle and inability to provide real-time feedback in existing detection methods.

Claims

1. An online detection method for heavy metal content in domestic waste based on hyperspectral imaging, characterized in that: include: Step S101, receiving configuration information, a start command and test space information, the optical system in the hyperspectral sensor loads the configuration information, after the configuration information is loaded, the optical system in the hyperspectral sensor executes the start command, the detector receives reflected light, saves the hyperspectral data at a preset time frequency and records the time, and obtains infrared information in the saved hyperspectral data; Step S102, determining the test space position according to the test space information, obtaining the data acquisition device information within the test space position range, and receiving the space environment data acquired by the data acquisition device within the test space position range, the space environment data including the temperature information within the test space, the humidity information within the test space, and the dust concentration information within the test space; Step S103, extracting infrared information from the stored hyperspectral data and calculating the spectral reflectance R; Step S104, substituting the temperature information in the test space, the humidity information in the test space, and the dust concentration information in the test space into a simulation model under preset laboratory conditions to obtain a spectral reflectance R2 under standard conditions; Step S105, taking the spectral reflectance R2 under the standard state as input into the prediction model to obtain the content of different types of heavy metals.

2. The method for online detection of heavy metal content in domestic waste based on hyperspectral imaging according to claim 1, characterized in that: The step S101 includes: The configuration information includes a wavelength range of 1000-3000 nm and a spectral resolution set to 3 nm; The preset time frequency includes saving the data collected by the hyperspectral sensor every 1 second.

3. The method for online detection of heavy metal content in domestic waste based on hyperspectral imaging according to claim 1, characterized in that: The step S102 includes: The data acquisition device includes a temperature sensor, a humidity sensor and a light scattering dust monitor; Obtain temperature information in the test space through a temperature sensor; Obtain humidity information in the test space through a humidity sensor; The dust concentration information in the space is obtained through the light scattering dust monitor.

4. The method for online detection of heavy metal content in domestic waste based on hyperspectral imaging according to claim 1, characterized in that: The step S103 includes: The infrared information in the stored hyperspectral data is stored in the form of a three-dimensional data cube to obtain hyperspectral data cube information, wherein the hyperspectral data cube information includes a row space dimension and a spectral dimension; The hyperspectral data cube is parsed to form a two-dimensional array with wavelength and corresponding reflection intensity. The spectral reflectance or radiance at each pixel position is calculated. After correction with a black and white plate, the infrared spectral reflectance R of the corresponding target at 1000-3000nm is obtained.

5. The method for online detection of heavy metal content in domestic waste based on hyperspectral imaging according to claim 1, characterized in that: The step S104 includes: The spectral reflectance R is subjected to Gaussian smoothing to obtain the spectral reflectance R1; Collect the optical parameters of different refractive index, absorption coefficient, scattering coefficient and extinction coefficient corresponding to the temperature information, humidity information and dust concentration information under laboratory conditions; The temperature information, humidity information, and dust concentration information are used as boundary conditions, and the spectral reflectance R1 is used as input. The simulation model under laboratory conditions is substituted to obtain the spectral reflectance R2 under standard conditions.

6. The method for online detection of heavy metal content in domestic waste based on hyperspectral imaging according to claim 1, characterized in that: The step S105 includes: The ash samples with different heavy metal contents under standard conditions were collected and analyzed for spectral reflectance. The spectral reflectance of the ash samples with different heavy metal contents under standard conditions was analyzed, and the spectral features were extracted using the LightGBM algorithm. The spectral reflectance corresponding to the first 20 bands with the largest scores in the spectral features was selected as the characteristic reflectance. Taking the characteristic reflectance as input and the content of different types of heavy metals as output, CatBoost is selected for machine learning to obtain a prediction model. The prediction model is used to process the characteristic reflectance data to obtain the prediction results. The prediction results obtained by the prediction model are imported into the recognition system, and R2 obtained by spectral image extraction and calculation is used as input to predict the content Q of different types of heavy metals.

7. The method for online detection of heavy metal content in domestic waste based on hyperspectral imaging according to claim 6, characterized in that: The step S105 includes: Obtain the actual test results of the experiment and the data processing history, compare the actual test results of the experiment with the laboratory standard test results, and obtain the test result error data by comparing the difference between the two. If the test result error data is greater than the preset minimum error range, optimize the test result error data, the actual test results of the experiment, and the data processing history through a genetic algorithm; Initialize the initial population of individuals, where individuals refer to different parameter combinations of the prediction model; According to the population size, the initial population is randomly generated, and each individual represents the genotype of the chromosome, and the genotype of the chromosome includes the hyperparameters of the model; Calculate the fitness of each individual, that is, the accuracy of the model's prediction results; If the error is less than the set value, the best individual and the optimal solution it represents are output, and the algorithm ends; If the error is less than the set value, the regenerated individuals are selected according to the fitness, and new individuals are created from the most suitable individuals by combining the genetic material. The genetic material of the individuals is randomly changed to produce a new generation of populations, and the fitness of each individual is recalculated until the optimal solution is obtained and the algorithm stops when the maximum number of iterations is reached; After the algorithm is completed, the optimal individual and the optimal solution it represents, that is, the optimized prediction model parameters, are output, and the characteristic reflectivity is predicted using the optimized prediction model.

8. The method for online detection of heavy metal content in domestic waste based on hyperspectral imaging according to claim 1, characterized in that: The step S104 includes: Obtain the standard conditions of temperature, humidity and dust concentration in the laboratory, collect the spectral reflectance corresponding to the standard conditions of temperature, humidity and dust concentration in the laboratory, set different refractive index, absorption coefficient, scattering coefficient and extinction coefficient according to the temperature information, humidity information and dust concentration information, build a physical model of how different optical parameters change with temperature information, humidity information and dust concentration, and build a simulation model of infrared light propagation in air containing dust and water vapor through a ray tracing algorithm. The simulation model includes the absorption model and scattering model of water molecules for infrared light, and the absorption model and scattering model of dust particles for infrared spectrum; Set different parameter ranges of temperature field, humidity field and particle concentration field on ANSYS software, set different boundary conditions, simulate and calculate the diffuse reflection data of infrared light under different temperature, humidity and dust particle concentration conditions, analyze the influence of temperature, humidity and dust concentration on the reflection spectrum, compare the model prediction value with the actual value, and continuously optimize the model; The temperature information, humidity information, and dust concentration information are used as boundary conditions, and the spectral reflectance R1 is used as input and brought into the simulation model to solve the spectral reflectance R2 under standard conditions.

Images