Soil environment quality monitoring system and method based on multi-source data

Abstract

The invention discloses a soil environment quality monitoring system and method based on multi-source data, and belongs to the technical field of soil quality monitoring. The method comprises the following steps: acquiring data of a plurality of pollutants, including heavy metals, pesticide residues and organic pollution; the collected data are preprocessed and integrated to form a soil pollution data set; calculating a single pollution index of each pollutant, and grading the soil environment quality according to the calculated comprehensive pollution index through the Neimero comprehensive pollution index; distributing weights by using a random forest algorithm to obtain land intensity factors of the region; based on a multi-layer sensor model, inputting factory emission data, pollutant treatment data and a soil pollution coefficient to obtain an industrial activity factor of the region; calculating the urbanization rate of the region to obtain a human activity influence factor; and performing hierarchical management of different levels by using a naive Bayesian algorithm, and allocating management levels according to the Bayesian theorem and the result of the maximum probability.

Description

Technical Field

[0001] The present invention relates to the technical field of soil quality monitoring, and in particular to a soil environment quality monitoring system and method based on multi-source data. Background Art

[0002] With the acceleration of industrialization and urbanization, soil pollution is becoming increasingly serious, directly threatening the sustainable development of agricultural production and the ecological environment. Soil pollutants include heavy metals, pesticide residues, and organic pollution, which usually come from human activities such as industrial emissions, agricultural activities, and urban construction. These pollutants affect the physical and chemical properties of the soil, and thus affect crop growth, agricultural product safety, and the health of the ecosystem. Therefore, soil environmental quality monitoring has become an important task in the field of environmental protection.

[0003] Traditional soil pollution monitoring usually relies on manual sampling and laboratory testing, which is complicated to operate and has a long cycle. It is difficult to achieve large-scale, real-time data collection, resulting in insufficient data timeliness and difficulty in reflecting changes in soil pollution in a timely manner. Existing monitoring methods usually only focus on the detection of a single pollutant, lack a comprehensive monitoring mechanism for multiple pollutants, and cannot fully reflect the overall situation of soil pollution. Traditional methods mainly focus on the concentration analysis of pollutants, while ignoring the profound impact of human activities on soil environmental quality, and cannot accurately assess the dynamic changes in the soil environment. Summary of the invention

[0004] The purpose of the present invention is to provide a soil environmental quality monitoring system and method based on multi-source data to solve the problems raised in the prior art.

[0005] To achieve the above object, the present invention provides the following technical solutions: A soil environmental quality monitoring method based on multi-source data, the method comprising the following steps: S100. Deploy sensors and detection equipment in the monitoring area to collect data on several types of pollutants, including heavy metals, pesticide residues, and organic pollution; remove outliers, fill in missing values, and normalize the data collected from different sensors to integrate and form a soil pollution data set; S200, based on the soil pollution data set, calculating the single pollution index of each pollutant, using the Nemerow comprehensive pollution index algorithm, taking the weighted average of the optimal value and the worst value of the single pollution index to obtain the comprehensive pollution index, and grading the soil environmental quality according to the calculated comprehensive pollution index; S300. Based on the land use mode, average fertilizer application amount, average irrigation amount and dominant vegetation type, the random forest algorithm is used to assign weights to obtain the land intensity factor of the region; based on the multi-layer perceptron model, the factory emission data, pollutant treatment data and soil pollution coefficient are input to obtain the industrial activity factor of the region; the construction land area, total land area, urban population and total population of the region are collected to calculate the urbanization rate of the region; the land intensity factor, industrial activity factor and urbanization rate are standardized and scored and weighted averaged to obtain the human activity impact factor; S400, based on the level of soil environmental quality and the influencing factors of human activities, the naive Bayes algorithm is used to carry out different levels of hierarchical management. According to the Bayesian theorem, the posterior probability of each management level is calculated, and the management level is assigned according to the result with the maximum probability.

[0006] According to step S100, a grid layout is adopted, and sensors of different densities are set according to different soil types and potential pollution sources to collect data on several types of pollutants, including heavy metals, pesticide residues and organic pollution; the data collection frequency is set to ensure that the soil pollution status can be monitored in real time, and the data is transmitted to the central database in real time through the wireless network; The z-score method is used to detect and eliminate outliers. The steps include: calculating the mean and standard deviation of each pollutant data, setting a threshold to determine outliers, and eliminating them; using mean filling to fill missing values; normalizing the data of different pollutants, and the normalization method used is Min-Max normalization; integrating the processed heavy metal, pesticide residue and organic pollution data to form a soil pollution data set, which is stored and managed using a database management system.

[0007] According to step S200, a single pollution index is calculated for each pollutant i. , the formula is: ,in, is the actual measured concentration of pollutant i, is the environmental quality standard limit of pollutant i, and The unit is mg / kg; if , it means that the concentration of the pollutant is lower than the standard limit and the pollution level is relatively light. , it means that the concentration of the pollutant has reached the standard limit and needs attention. , it means that the concentration of the pollutant exceeds the standard limit and the pollution level is high.

[0008] The Nemerow comprehensive pollution index P is calculated using the formula: ;in, , represents the maximum value of all single pollution indexes, n is the number of pollutant types, , represents the average value of all individual pollution indices; the calculated comprehensive pollution index P is used to evaluate soil environmental quality, and the classification standards are set as follows: The soil environment quality is level one, indicating that the environment quality is good. The soil environment quality is level 2, indicating slight pollution. The soil environment quality is level 3, indicating moderate pollution. The soil environmental quality is level 4, indicating severe pollution. The soil environmental quality is level five, indicating severe pollution.

[0009] According to step S300, land use patterns, average fertilizer application amount, average irrigation amount and dominant vegetation types in the region are collected, wherein the average fertilizer application amount and average irrigation amount are in kg / ha, and the influence of land use patterns and dominant vegetation types on soil environmental pollution is quantified; Define soil environmental quality as a label, evaluate the pollution level and ecological health of the soil through historical soil monitoring data, and quantify it into a soil environmental quality score; use land use mode, average fertilization amount, average irrigation amount and dominant vegetation type as input features, and input the soil environmental quality score as the output label into the random forest model; use the random forest algorithm to train the data, and analyze the impact of each feature on soil environmental quality through the integration of several decision trees. Each decision tree generates different feature combinations based on the input data and predicts the results of soil environmental quality; through several decision tree splits, gradually optimize the prediction accuracy of the model; During the training process, the random forest calculates the contribution of each feature to the final prediction result; by taking the weighted average of the results of multiple decision trees, the feature importance weights of land use mode, average fertilizer application amount, average irrigation amount and dominant vegetation type are automatically generated; based on the feature importance weights, the land use mode, average fertilizer application amount, average irrigation amount and dominant vegetation type are weighted averaged to obtain the land use intensity factor AA.

[0010] According to step S300, the factory emission data includes the factory's waste gas, wastewater and solid waste emissions, and the pollutant treatment data includes the pollutant treatment amount, treatment efficiency and treatment method; the factory emission data, pollutant treatment data and soil pollution coefficient are normalized to a range of 0 to 1, and the normalized factory emission data is input into the input layer of the multilayer perceptron model. , Pollutant treatment data and soil pollution coefficient ; It contains several hidden layers, each of which contains several neurons, which are used to learn the nonlinear relationship between input features; the output layer has one neuron, which represents the industrial activity factor IA, with a value between 0 and 1, representing the intensity of the impact of industrial activities on the soil environment; Each neuron calculates its output using the following formula: ,in, is the output of the jth neuron, is the input feature to the weight of neuron j, is the bias term, and f(·) is the ReLU activation function.

[0011] The multilayer perceptron model uses mean square error as the loss function to measure the error between the model prediction value and the true value. The loss function formula is as follows: , where m is the number of samples, is the predicted industrial activity factor of the ith sample, is the true value; The weight w and bias term b in the model are adjusted through the back-propagation algorithm to minimize the loss function. The formula for weight update is: ,in, is the learning rate, is the loss function L on the weight The gradient of is the updated weight, is the weight before updating; The trained model inputs new factory emission data, pollutant treatment data and soil pollution coefficient, and can output the industrial activity factor IA, ranging from 0 to 1.

[0012] Collect the construction land area of ​​the area , total land area , Urban Population and total population ; Based on the construction land area and total land area, the land urbanization rate of the region is obtained , the formula is: , It is the sum of the land areas of residential, commercial, industrial and public infrastructure in the region; based on the urban population and the total population, the urbanization rate of the region is obtained. , the formula is: , is the sum of urban and rural populations; the urbanization rate of the region is obtained by taking the weighted average of the land urbanization rate and the population urbanization rate; The land use intensity factor, industrial activity factor and urbanization rate are standardized and weighted averaged to obtain the human activity impact factor.

[0013] According to step S400, the naive Bayes algorithm is based on the Bayesian theorem, assuming that the soil environmental quality level and the human activity influencing factors are independent of each other, and the probability of each management level is calculated according to the relationship between the features and categories in the historical data, and the management level is selected according to the maximum probability principle; in the training process, the historical soil pollution data and human activity data are used, and the corresponding management level is used as label data; each historical data sample corresponds to the soil pollution level and the human activity influencing factor, as well as the actual management level adopted; the model learns the relationship between different features and management levels based on the known data, and constructs classification rules; In the process of learning the model, the naive Bayes algorithm learns the relationship between different features and management levels by calculating the conditional probability between each feature and management level according to the Bayesian theorem. The specific steps are as follows: First, calculate the prior probability of each management level. The prior probability represents the probability of each management level appearing when there is no feature information. For example, if the "low risk" management level accounts for 40% of the total number of samples in the training set, then its prior probability is 0.4. For each feature, calculate its conditional probability at different management levels. For example, for a specific soil pollution index level, calculate its probability of appearing at the "low risk", "medium risk" and "high risk" management levels. This probability is calculated based on the frequency of occurrence of a certain feature value at a certain management level in the training data. For continuous features, it is assumed that they follow a Gaussian distribution, so it is necessary to estimate the conditional probability of the feature at different management levels by calculating the mean and variance of the feature. For discrete features, such as management levels are discrete label data, the conditional probability is calculated directly by counting.

[0014] The Naive Bayes algorithm constructs a classification rule based on the conditional probabilities of all features and the prior probability of each management level. The rule is expressed as follows: when a data sample is input, its posterior probability at each management level is calculated, and the management level with the largest posterior probability is selected as the final prediction result. The posterior probability is expressed as: the probability of the management level under given feature conditions.

[0015] The posterior probability of each management level is calculated by multiplying the prior probability and the conditional probability of all features at that level. The model compares the posterior probabilities of all management levels and selects the management level with the largest posterior probability as the final output management strategy.

[0016] When applying the naive Bayes algorithm, for new soil environmental quality levels and human activity influencing factors, the model calculates the probability of characteristics under different management levels; according to the Bayesian theorem, the posterior probability of each management level is calculated, and the management level is assigned according to the result with the maximum probability; the management level with the largest posterior probability is selected as the soil environmental management strategy for the area.

[0017] A soil environmental quality monitoring system based on multi-source data, comprising: Data collection and processing module: including: data collection unit, data preprocessing unit and data integration unit; the data collection unit collects several types of pollutant data, including heavy metals, pesticide residues and organic pollution; the data preprocessing unit removes outliers, fills in missing values ​​and normalizes the data collected from different sensors; the data integration unit integrates the data to form a soil pollution data set; Soil environment quality assessment module: including: single pollution index calculation unit, Nemerow comprehensive pollution index calculation unit and soil environment grading unit; among them, the single pollution index calculation unit calculates the single pollution index of each pollutant based on the soil pollution data set, the Nemerow comprehensive pollution index calculation unit uses the Nemerow comprehensive pollution index algorithm to obtain the comprehensive pollution index by weighted average of the optimal value and the worst value of the single pollution index, and the soil environment grading unit grades the soil environment quality according to the calculated comprehensive pollution index; Human activity impact factor calculation module: including: land intensity factor calculation unit, industrial activity factor calculation unit, urbanization rate calculation unit and weighted average calculation unit; among them, the land intensity factor calculation unit uses the random forest algorithm to assign weights based on the average fertilizer application amount, the average irrigation amount and the dominant vegetation type to obtain the land intensity factor of the area; the industrial activity factor calculation unit is based on the multi-layer perceptron model, inputs factory emission data, pollutant treatment data and soil pollution coefficient to obtain the industrial activity factor of the area; the urbanization rate calculation unit collects the construction land area, total land area, urban population and total population of the area, and calculates the urbanization rate of the area; the weighted average calculation unit performs standardized scoring and weighted average calculation on the land intensity factor, industrial activity factor and urbanization rate to obtain the human activity impact factor; Hierarchical management and decision-making module: includes: Bayesian classification unit and management level allocation unit; among them, the Bayesian classification unit uses the naive Bayesian algorithm to perform different levels of hierarchical management based on the level of soil environmental quality and the influencing factors of human activities. The management level allocation unit calculates the posterior probability of each management level according to the Bayesian theorem, and allocates the management level according to the result with the maximum probability.

[0018] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention introduces indicators such as land use intensity factor, industrial activity factor and urbanization rate, combines the long-term impact of human activities on the soil environment, and uses algorithms such as random forest and multi-layer perceptron to scientifically evaluate the comprehensive impact of human activities on the soil. Compared with traditional technologies that ignore the impact of human activities, the present invention can more accurately evaluate the dynamic changes of the soil environment.

[0019] 2. The present invention uses the Nemerow comprehensive pollution index method to perform weighted average of the individual indexes of multiple pollutants to comprehensively reflect the degree of soil pollution, and scientifically grade and manage the soil environmental quality based on the calculation results. This method overcomes the limitations of single pollutant assessment in the prior art and better reflects the complexity and superposition effect of soil pollution.

[0020] 3. The present invention adopts the Naive Bayes algorithm to dynamically adjust different hierarchical management schemes according to the soil environmental quality level and the influencing factors of human activities, thereby realizing adaptive management of soil quality. Compared with the fixed standard evaluation method in the prior art, the intelligent management mechanism of the present invention has higher flexibility and pertinence, and can automatically adjust management measures according to the actual soil conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a schematic diagram of the steps of a soil environment quality monitoring system and method based on multi-source data of the present invention; Figure 2 It is a system structure diagram of a soil environment quality monitoring system and method based on multi-source data of the present invention. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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.

[0023] Example: Figure 1-Figure 2 As shown, the present invention provides a technical solution. According to one embodiment of the present invention, Figure 1 A soil environment quality monitoring system and method based on multi-source data is shown in a schematic diagram of steps. A soil environment quality monitoring method based on multi-source data includes the following steps: S100. Deploy sensors and detection equipment in the monitoring area to collect data on several types of pollutants, including heavy metals, pesticide residues, and organic pollution; remove outliers, fill in missing values, and normalize the data collected from different sensors to integrate and form a soil pollution data set; S200, based on the soil pollution data set, calculating the single pollution index of each pollutant, using the Nemerow comprehensive pollution index algorithm, taking the weighted average of the optimal value and the worst value of the single pollution index to obtain the comprehensive pollution index, and grading the soil environmental quality according to the calculated comprehensive pollution index; S300. Based on the land use mode, average fertilizer application amount, average irrigation amount and dominant vegetation type, the random forest algorithm is used to assign weights to obtain the land intensity factor of the region; based on the multi-layer perceptron model, the factory emission data, pollutant treatment data and soil pollution coefficient are input to obtain the industrial activity factor of the region; the construction land area, total land area, urban population and total population of the region are collected to calculate the urbanization rate of the region; the land intensity factor, industrial activity factor and urbanization rate are standardized and scored and weighted averaged to obtain the human activity impact factor; S400, based on the level of soil environmental quality and the influencing factors of human activities, the naive Bayes algorithm is used to carry out different levels of hierarchical management. According to the Bayesian theorem, the posterior probability of each management level is calculated, and the management level is assigned according to the result with the maximum probability.

[0024] According to step S100, a grid layout is adopted, and sensors of different densities are set according to different soil types and potential pollution sources to collect data on several types of pollutants, including heavy metals, pesticide residues and organic pollution; the data collection frequency is set to ensure that the soil pollution status can be monitored in real time, and the data is transmitted to the central database in real time through the wireless network; The z-score method is used to detect and eliminate outliers. The steps include: calculating the mean and standard deviation of each pollutant data, setting a threshold to determine outliers, and eliminating them; using mean filling to fill missing values; normalizing the data of different pollutants, and the normalization method used is Min-Max normalization; integrating the processed heavy metal, pesticide residue and organic pollution data to form a soil pollution data set, which is stored and managed using a database management system.

[0025] The z-score method is used to detect and eliminate outliers. The steps include: calculating the mean and standard deviation of each pollutant data, setting a threshold to determine outliers, and eliminating them; using mean filling to fill missing values; normalizing the data of different pollutants, and the normalization method used is Min-Max normalization; integrating the processed heavy metal, pesticide residue and organic pollution data to form a soil pollution data set, which is stored and managed using a database management system.

[0026] According to step S200, a single pollution index is calculated for each pollutant i. , the formula is: ,in, is the actual measured concentration of pollutant i, is the environmental quality standard limit of pollutant i, and The unit is mg / kg; if , it means that the concentration of the pollutant is lower than the standard limit and the pollution level is relatively light. , it means that the concentration of the pollutant has reached the standard limit and needs attention. , it means that the concentration of the pollutant exceeds the standard limit and the pollution level is high.

[0027] The Nemerow comprehensive pollution index P is calculated using the formula: ;in, , represents the maximum value of all single pollution indexes, n is the number of pollutant types, , represents the average value of all individual pollution indices; the calculated comprehensive pollution index P is used to evaluate soil environmental quality, and the classification standards are set as follows: The soil environment quality is level one, indicating that the environment quality is good. The soil environment quality is level 2, indicating slight pollution. The soil environment quality is level 3, indicating moderate pollution. The soil environmental quality is level 4, indicating severe pollution. The soil environmental quality is level five, indicating severe pollution.

[0028] In this embodiment, four pollutants were detected, including: heavy metal lead, cadmium, mercury and pesticide residues, and the actual measured concentration and environmental quality standard limit of each pollutant were as follows: the actual measured concentration of lead was 40, and the environmental quality standard limit was 50; the actual measured concentration of cadmium was 0.8, and the environmental quality standard limit was 0.5; the actual measured concentration of mercury was 0.03, and the environmental quality standard limit was 0.1; the actual measured concentration of pesticide residues was 0.05, and the environmental quality standard limit was 0.1; The individual pollution indexes of the four pollutants are calculated, and the results are as follows: For lead, , for cadmium, , for mercury, , for pesticide residues, .

[0029] Calculated using the Nemerow comprehensive pollution index formula: ; Therefore, according to the classification standards, the soil environmental quality assessment result is level two soil environmental quality, indicating mild pollution.

[0030] According to step S300, the land use pattern, average fertilizer application amount, average irrigation amount and dominant vegetation type in the area are collected, wherein the average fertilizer application amount and average irrigation amount are in kg / ha, and the influence of the dominant vegetation type on soil environmental pollution is quantified; Define soil environmental quality as a label, evaluate the pollution level and ecological health of the soil through historical soil monitoring data, and quantify it into a soil environmental quality score; use land use mode, average fertilization amount, average irrigation amount and dominant vegetation type as input features, and input the soil environmental quality score as the output label into the random forest model; use the random forest algorithm to train the data, and analyze the impact of each feature on soil environmental quality through the integration of several decision trees. Each decision tree generates different feature combinations based on the input data and predicts the results of soil environmental quality; through several decision tree splits, gradually optimize the prediction accuracy of the model; During the training process, the random forest calculates the contribution of each feature to the final prediction result; by taking the weighted average of the results of multiple decision trees, the feature importance weights of land use mode, average fertilizer application amount, average irrigation amount and dominant vegetation type are automatically generated; based on the feature importance weights, the land use mode, average fertilizer application amount, average irrigation amount and dominant vegetation type are weighted averaged to obtain the land use intensity factor AA.

[0031] In this embodiment, the study area is a certain urban green space, and the goal is to evaluate the effects of different land use patterns, average fertilizer application, average irrigation amount and dominant vegetation types on soil environmental quality; the collected data are as follows: average fertilizer application (kg / ha): area A: 150 kg / ha, area B: 200 kg / ha, area C: 100 kg / ha; average irrigation amount (kg / ha): area A: 50 kg / ha, area B: 80 kg / ha, area C: 30 kg / ha; dominant vegetation types: area A: shrubs, area B: lawns, area C: perennial plants; historical soil monitoring data (soil environmental quality score): area A: 75, area B: 60, area C: 85.

[0032] The model was trained using the random forest algorithm to generate several decision trees. The characteristic importance weight of fertilizer application was 0.50, the characteristic importance weight of pesticide application was 0.30, and the characteristic importance weight of dominant vegetation type was 0.20.

[0033] The soil environmental quality prediction results are as follows: the actual soil environmental quality score of area A is 75, and the predicted soil environmental quality score is 77; the actual soil environmental quality score of area B is 60, and the predicted soil environmental quality score is 62; the actual soil environmental quality score of area C is 85, and the predicted soil environmental quality score is 83.

[0034] According to step S300, the factory emission data includes the factory's waste gas, wastewater and solid waste emissions, and the pollutant treatment data includes the pollutant treatment amount, treatment efficiency and treatment method; the factory emission data, pollutant treatment data and soil pollution coefficient are normalized to a range of 0 to 1, and the normalized factory emission data is input into the input layer of the multilayer perceptron model. , Pollutant treatment data and soil pollution coefficient ; It contains several hidden layers, each of which contains several neurons, which are used to learn the nonlinear relationship between input features; the output layer has one neuron, which represents the industrial activity factor IA, with a value between 0 and 1, representing the intensity of the impact of industrial activities on the soil environment; Each neuron calculates its output using the following formula: ,in, is the output of the jth neuron, is the input feature to the weight of neuron j, is the bias term, and f(·) is the ReLU activation function.

[0035] The multilayer perceptron model uses mean square error as the loss function to measure the error between the model prediction value and the true value. The loss function formula is as follows: , where m is the number of samples, is the predicted industrial activity factor of the ith sample, is the true value; The weight w and bias term b in the model are adjusted through the back-propagation algorithm to minimize the loss function. The formula for weight update is: ,in, is the learning rate, is the loss function L on the weight The gradient of is the updated weight, is the weight before updating; The trained model inputs new factory emission data, pollutant treatment data and soil pollution coefficient, and can output the industrial activity factor IA, ranging from 0 to 1.

[0036] In this embodiment, the factory emission data are as follows: waste gas emissions: 300 tons / year, wastewater emissions: 150 tons / year, solid waste emissions: 50 tons / year. Pollutant treatment data are as follows: pollutant treatment capacity: 400 tons / year, treatment efficiency: 80%, treatment efficiency is represented by 0 to 1, that is, 0.8, treatment method code: biological treatment, coded as 1. The soil pollution coefficient is 0.5, the value is between 0 and 1, the higher the value, the higher the degree of pollution. The construction land area is 500 hectares, the total land area is 2000 hectares, the urban population is 50,000, and the total population is 80,000.

[0037] The maximum emission is 1000 tons / year, the waste gas emission is normalized to 300 / 1000 = 0.3, and the wastewater emission is normalized to 150 / 1000 = 0.15. The maximum waste emission is 100 tons / year, and the solid waste emission is normalized to 50 / 100 = 0.5. The maximum treatment capacity is 1000 tons / year, and the pollutant treatment capacity is normalized to 400 / 1000 = 0.4. The treatment efficiency is 0.8; the soil pollution coefficient has been normalized to 0.5. After training, the model can input new factory emission data, pollutant treatment data, and soil pollution coefficient, and output the industrial activity factor IA. The industrial activity factor output by the model is 0.65.

[0038] Collect the construction land area of ​​the area , total land area , Urban Population and total population ; Based on the construction land area and total land area, the land urbanization rate of the region is obtained , the formula is: , It is the sum of the land areas of residential, commercial, industrial and public infrastructure in the region; based on the urban population and the total population, the urbanization rate of the region is obtained. , the formula is: , is the sum of urban and rural populations; the urbanization rate of the region is obtained by taking the weighted average of the land urbanization rate and the population urbanization rate; The land use intensity factor, industrial activity factor and urbanization rate are standardized and weighted averaged to obtain the human activity impact factor.

[0039] In this embodiment, the urbanization rate is calculated to be 0.38, and the human activity impact factor is obtained to be 44.08.

[0040] According to step S400, the naive Bayes algorithm is based on the Bayesian theorem, assuming that the soil environmental quality level and the human activity influencing factors are independent of each other, and the probability of each management level is calculated according to the relationship between the features and categories in the historical data, and the management level is selected according to the maximum probability principle; in the training process, the historical soil pollution data and human activity data are used, and the corresponding management level is used as label data; each historical data sample corresponds to the soil pollution level and the human activity influencing factor, as well as the actual management level adopted; the model learns the relationship between different features and management levels based on the known data, and constructs classification rules; When applying the naive Bayes algorithm, for new soil environmental quality levels and human activity influencing factors, the model calculates the probability of characteristics under different management levels; according to the Bayesian theorem, the posterior probability of each management level is calculated, and the management level is assigned according to the result with the maximum probability; the management level with the largest posterior probability is selected as the soil environmental management strategy for the area.

[0041] According to another embodiment of the present invention, Figure 2 As shown in the system structure diagram of a soil environment quality monitoring system and method based on multi-source data, a soil environment quality monitoring system based on multi-source data includes: Data collection and processing module: including: data collection unit, data preprocessing unit and data integration unit; the data collection unit collects several types of pollutant data, including heavy metals, pesticide residues and organic pollution; the data preprocessing unit removes outliers, fills in missing values ​​and normalizes the data collected from different sensors; the data integration unit integrates the data to form a soil pollution data set; Soil environment quality assessment module: including: single pollution index calculation unit, Nemerow comprehensive pollution index calculation unit and soil environment grading unit; among them, the single pollution index calculation unit calculates the single pollution index of each pollutant based on the soil pollution data set, the Nemerow comprehensive pollution index calculation unit uses the Nemerow comprehensive pollution index algorithm to obtain the comprehensive pollution index by weighted average of the optimal value and the worst value of the single pollution index, and the soil environment grading unit grades the soil environment quality according to the calculated comprehensive pollution index; Human activity impact factor calculation module: including: land intensity factor calculation unit, industrial activity factor calculation unit, urbanization rate calculation unit and weighted average calculation unit; among them, the land intensity factor calculation unit uses the random forest algorithm to assign weights based on land use mode, average fertilizer application amount, average irrigation amount and dominant vegetation type to obtain the land intensity factor of the area; the industrial activity factor calculation unit is based on the multi-layer perceptron model, inputs factory emission data, pollutant treatment data and soil pollution coefficient to obtain the industrial activity factor of the area; the urbanization rate calculation unit collects the construction land area, total land area, urban population and total population of the area, and calculates the urbanization rate of the area; the weighted average calculation unit performs standardized scoring and weighted average calculation on the land intensity factor, industrial activity factor and urbanization rate to obtain the human activity impact factor; Hierarchical management and decision-making module: includes: Bayesian classification unit and management level allocation unit; among them, the Bayesian classification unit uses the naive Bayesian algorithm to perform different levels of hierarchical management based on the level of soil environmental quality and the influencing factors of human activities. The management level allocation unit calculates the posterior probability of each management level according to the Bayesian theorem, and allocates the management level according to the result with the maximum probability.

[0042] In this embodiment, we have collected the following data for training the Bayesian classification unit: soil environmental quality level (Y): Level 1 (good): 30 samples, Level 2 (mild pollution): 25 samples, Level 3 (moderate pollution): 20 samples, Level 4 (heavy pollution): 15 samples, Level 5 (severe pollution): 10 samples. According to historical data, the distribution of human activity impact factors at each level is statistically calculated: Level 1: average 0.2, standard deviation 0.05; Level 2: average 0.5, standard deviation 0.1; Level 3: average 0.7, standard deviation 0.15; Level 4: average 0.9, standard deviation 0.2; Level 5: average 1.2, standard deviation 0.25.

[0043] Calculate the prior probability P(Y) of each soil environmental quality level: , , , , .

[0044] The management level assignment unit will calculate the posterior probability according to Bayes' theorem and assign the management level. We have a new observation value, indicating that the human activity impact factor X is 0.75. The calculated conditional probability and posterior probability are as follows: , , , , Finally, the management level was assigned as Level 3 (moderate pollution) because maximum.

[0045] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A soil environmental quality monitoring method based on multi-source data, characterized in that: The method comprises the following steps: S100. Deploy sensors and detection equipment in the monitoring area to collect data on several types of pollutants, including heavy metals, pesticide residues, and organic pollution; remove outliers, fill in missing values, and normalize the data collected from different sensors to integrate and form a soil pollution data set; S200, based on the soil pollution data set, calculating the single pollution index of each pollutant, using the Nemerow comprehensive pollution index algorithm, taking the weighted average of the optimal value and the worst value of the single pollution index to obtain the comprehensive pollution index, and grading the soil environmental quality according to the calculated comprehensive pollution index; S300. Based on the land use mode, average fertilizer application amount, average irrigation amount and dominant vegetation type, the random forest algorithm is used to assign weights to obtain the land intensity factor of the region; based on the multi-layer perceptron model, the factory emission data, pollutant treatment data and soil pollution coefficient are input to obtain the industrial activity factor of the region; the construction land area, total land area, urban population and total population of the region are collected to calculate the urbanization rate of the region; the land intensity factor, industrial activity factor and urbanization rate are standardized and scored and weighted averaged to obtain the human activity impact factor; S400, based on the level of soil environmental quality and the influencing factors of human activities, the naive Bayes algorithm is used to carry out different levels of hierarchical management. According to the Bayesian theorem, the posterior probability of each management level is calculated, and the management level is assigned according to the result with the maximum probability.

2. The soil environmental quality monitoring method based on multi-source data according to claim 1 is characterized in that: According to step S100, a grid layout is adopted, and sensors of different densities are set according to different soil types and potential pollution sources to collect data on several types of pollutants, including heavy metals, pesticide residues and organic pollution; the data collection frequency is set to ensure that the soil pollution status can be monitored in real time, and the data is transmitted to the central database in real time through the wireless network; The z-score method is used to detect and eliminate outliers. The steps include: calculating the mean and standard deviation of each pollutant data, setting a threshold to determine outliers, and eliminating them; using mean filling to fill missing values; normalizing the data of different pollutants, and the normalization method used is Min-Max normalization; integrating the processed heavy metal, pesticide residue and organic pollution data to form a soil pollution data set, which is stored and managed using a database management system.

3. The soil environmental quality monitoring method based on multi-source data according to claim 1 is characterized in that: According to step S200, a single pollution index is calculated for each pollutant i. , the formula is: ,in, is the actual measured concentration of pollutant i, is the environmental quality standard limit of pollutant i, and The unit is mg / kg; if , it means that the concentration of the pollutant is lower than the standard limit and the pollution level is relatively light. , it means that the concentration of the pollutant has reached the standard limit and needs attention. , it means that the concentration of the pollutant exceeds the standard limit and the pollution level is high.

4. The soil environmental quality monitoring method based on multi-source data according to claim 3 is characterized in that: The Nemerow comprehensive pollution index P is calculated using the formula: ;in, , represents the maximum value of all single pollution indexes, n is the number of pollutant types, , represents the average value of all single pollution indexes; the calculated comprehensive pollution index P is used to evaluate soil environmental quality, and the classification standards are set as follows: The soil environment quality is level one, indicating that the environment quality is good. The soil environment quality is level 2, indicating slight pollution. The soil environment quality is level 3, indicating moderate pollution. The soil environmental quality is level 4, indicating severe pollution. The soil environmental quality is level five, indicating severe pollution.

5. The soil environmental quality monitoring method based on multi-source data according to claim 1 is characterized in that: According to step S300, land use patterns, average fertilizer application amount, average irrigation amount and dominant vegetation types in the region are collected, wherein the units of average fertilizer application amount and average irrigation amount are kg / ha, and the influence of land use patterns and dominant vegetation types on soil environmental pollution is quantified; Define soil environmental quality as a label, evaluate the pollution level and ecological health of the soil through historical soil monitoring data, and quantify it into a soil environmental quality score; use land use mode, average fertilization amount, average irrigation amount and dominant vegetation type as input features, and input the soil environmental quality score as the output label into the random forest model; use the random forest algorithm to train the data, and analyze the impact of each feature on soil environmental quality through the integration of several decision trees. Each decision tree generates different feature combinations based on the input data and predicts the results of soil environmental quality; through several decision tree splits, gradually optimize the prediction accuracy of the model; During the training process, the random forest calculates the contribution of each feature to the final prediction result; by taking the weighted average of the results of multiple decision trees, the feature importance weights of land use mode, average fertilizer application amount, average irrigation amount and dominant vegetation type are automatically generated; based on the feature importance weights, the land use mode, average fertilizer application amount, average irrigation amount and dominant vegetation type are weighted averaged to obtain the land use intensity factor AA.

6. The soil environmental quality monitoring method based on multi-source data according to claim 5 is characterized in that: According to step S300, the factory emission data includes the factory's waste gas, wastewater and solid waste emissions, and the pollutant treatment data includes the pollutant treatment amount, treatment efficiency and treatment method; Normalize the factory emission data, pollutant treatment data, and soil pollution coefficient to a range of 0 to 1, and input the normalized factory emission data into the input layer of the multilayer perceptron model. , Pollutant treatment data and soil pollution coefficient ; It contains several hidden layers, each of which contains several neurons, which are used to learn the nonlinear relationship between input features; the output layer has one neuron, which represents the industrial activity factor IA, with a value between 0 and 1, representing the intensity of the impact of industrial activities on the soil environment; Each neuron calculates its output using the following formula: ,in, is the output of the jth neuron, is the input feature to the weight of neuron j, is the bias term, and f(·) is the ReLU activation function.

7. The soil environment quality monitoring method based on multi-source data according to claim 6 is characterized by: The multilayer perceptron model uses mean square error as the loss function to measure the error between the model prediction value and the true value. The loss function formula is as follows: , where m is the number of samples, is the predicted industrial activity factor of the ith sample, is the true value; The weight w and bias term b in the model are adjusted through the back-propagation algorithm to minimize the loss function. The formula for weight update is: ,in, is the learning rate, is the loss function L on the weight The gradient of is the updated weight, is the weight before updating; The trained model inputs new factory emission data, pollutant treatment data and soil pollution coefficient, and can output the industrial activity factor IA, ranging from 0 to 1.

8. The soil environment quality monitoring method based on multi-source data according to claim 7 is characterized in that: Collect the construction land area of ​​the area , total land area , Urban Population and total population ; Based on the construction land area and total land area, the land urbanization rate of the region is obtained , the formula is: , It is the sum of the land areas of residential, commercial, industrial and public infrastructure in the region; based on the urban population and the total population, the urbanization rate of the region is obtained. , the formula is: , is the sum of urban and rural populations; the urbanization rate of the region is obtained by taking the weighted average of the land urbanization rate and the population urbanization rate; The land use intensity factor, industrial activity factor and urbanization rate are standardized and weighted averaged to obtain the human activity impact factor.

9. The soil environment quality monitoring method based on multi-source data according to claim 1 is characterized in that: According to step S400, the naive Bayes algorithm is based on the Bayesian theorem, assuming that the soil environmental quality level and the human activity influencing factors are independent of each other, and the probability of each management level is calculated according to the relationship between the features and categories in the historical data, and the management level is selected according to the maximum probability principle; in the training process, the historical soil pollution data and human activity data are used, and the corresponding management level is used as label data; each historical data sample corresponds to the soil pollution level and the human activity influencing factor, as well as the actual management level adopted; the model learns the relationship between different features and management levels based on the known data, and constructs classification rules; When applying the naive Bayes algorithm, for new soil environmental quality levels and human activity influencing factors, the model calculates the probability of characteristics under different management levels; according to the Bayesian theorem, the posterior probability of each management level is calculated, and the management level is assigned according to the result with the maximum probability; the management level with the largest posterior probability is selected as the soil environmental management strategy for the area.

10. A soil environment quality monitoring system based on multi-source data, using a soil environment quality monitoring method based on multi-source data according to any one of claims 1 to 9, characterized in that: include: Data collection and processing module: including: data collection unit, data preprocessing unit and data integration unit; the data collection unit collects several types of pollutant data, including heavy metals, pesticide residues and organic pollution; the data preprocessing unit removes outliers, fills in missing values ​​and normalizes the data collected from different sensors; the data integration unit integrates the data to form a soil pollution data set; Soil environment quality assessment module: including: single pollution index calculation unit, Nemerow comprehensive pollution index calculation unit and soil environment grading unit; among them, the single pollution index calculation unit calculates the single pollution index of each pollutant based on the soil pollution data set, the Nemerow comprehensive pollution index calculation unit uses the Nemerow comprehensive pollution index algorithm to obtain the comprehensive pollution index by weighted average of the optimal value and the worst value of the single pollution index, and the soil environment grading unit grades the soil environment quality according to the calculated comprehensive pollution index; Human activity impact factor calculation module: including: land intensity factor calculation unit, industrial activity factor calculation unit, urbanization rate calculation unit and weighted average calculation unit; among them, the land intensity factor calculation unit uses the random forest algorithm to assign weights based on land use mode, average fertilizer application amount, average irrigation amount and dominant vegetation type to obtain the land intensity factor of the area; the industrial activity factor calculation unit is based on the multi-layer perceptron model, inputs factory emission data, pollutant treatment data and soil pollution coefficient to obtain the industrial activity factor of the area; the urbanization rate calculation unit collects the construction land area, total land area, urban population and total population of the area, and calculates the urbanization rate of the area; the weighted average calculation unit performs standardized scoring and weighted average calculation on the land intensity factor, industrial activity factor and urbanization rate to obtain the human activity impact factor; Hierarchical management and decision-making module: includes: Bayesian classification unit and management level allocation unit; among them, the Bayesian classification unit uses the naive Bayesian algorithm to perform different levels of hierarchical management based on the level of soil environmental quality and the influencing factors of human activities. The management level allocation unit calculates the posterior probability of each management level according to the Bayesian theorem, and allocates the management level according to the result with the maximum probability.

Images