Constructed wetland invasive snail monitoring method based on space-time autocorrelation analysis

Abstract

The invention discloses a method for monitoring intruded snails in a constructed wetland based on space-time autocorrelation analysis, which comprises the following steps of: acquiring image data of the constructed wetland by using image acquisition equipment, and simultaneously collecting environment variables, water microorganisms and soil physicochemical property data; effective data are obtained by screening image samples and removing colinear environment variables, and features are extracted by using image recognition and principal component analysis algorithms. And constructing a monitoring model based on the cellular automaton, optimizing parameters by means of an ANN algorithm, and further evaluating the intrusion risk. Spatio-temporal self-correlation analysis is combined with a geographic information system, spatio-temporal distribution characteristics of snails are researched, and a diffusion path and range are predicted. Compared with a traditional monitoring method, multi-source data are integrated, advanced algorithms and models are applied, the invaded snails can be monitored more accurately, a scientific basis is provided for prevention and control of the invaded snails in the constructed wetland, and an ecological system of the constructed wetland is effectively protected.

Description

Technical Field

[0001] The invention relates to ecological protection and ecological distribution models, in particular to a method for monitoring invasive snails in artificial wetlands based on spatiotemporal autocorrelation analysis. Background Art

[0002] In the field of ecological protection, the invasion of alien species is a serious challenge, posing a serious threat to the stability of the ecosystem, biodiversity and economic development. As a unique and important ecosystem, artificial wetlands are not immune. For example, the invasion of Spartina alterniflora in Tianjin Wetland has seriously damaged the local ecological balance and affected the habitat of native plants and wetland birds.

[0003] Currently, a variety of technical means have been developed for monitoring invasive snails in artificial wetlands. Traditional monitoring methods, such as field surveys, require monitoring personnel to go directly to artificial wetlands and record information such as the type, quantity, and distribution range of snails by visual observation and simple tools. Although this method is intuitive, it is inefficient and easily affected by subjective factors of monitoring personnel. It is difficult to comprehensively monitor some hidden or widely distributed snails. Sample plot surveys are to set up a certain number and size of sample plots in artificial wetlands, conduct detailed surveys and statistics on the snails in the sample plots, and infer the snail situation in the entire area. However, the rationality and representativeness of the sample plot setting have a great impact on the accuracy of the results, and in large-scale artificial wetlands, the workload of sample plot surveys is huge.

[0004] With the development of science and technology, modern monitoring technologies continue to emerge. Remote sensing monitoring technology analyzes satellite images of different phases or high-resolution images obtained by drones to grasp the distribution range and changing trends of invasive snails. It uses the difference in spectral characteristics between snails and the surrounding environment to identify snails. For example, the reflectivity of some invasive snails in specific bands is different from that of local organisms. However, this technology is greatly affected by factors such as weather and clouds, and the accuracy of identifying some tiny snails or snails in obscured areas needs to be improved. eDNA monitoring technology extracts DNA from environmental media such as water and soil in artificial wetlands, and performs PCR amplification and high-throughput sequencing on specific DNA fragments of the genome to monitor invasive snails. This technology can detect trace amounts of snail DNA in the environment, and can monitor snails even if the number of snails is small or difficult to observe directly, but there are problems such as high detection costs and strict requirements on experimental environment and operation.

[0005] In recent years, artificial intelligence monitoring technology has used machine learning algorithms to analyze and identify large amounts of ecological and environmental data, and combined with geographic information systems, it can also predict the spread trend of snails. However, this technology relies on high-quality training samples and appropriate algorithm models, and the accuracy and generalization ability of the model still need to be further improved.

[0006] These existing technologies have certain limitations in monitoring invasive snails in artificial wetlands, making it difficult to monitor invasive snails comprehensively, accurately and efficiently. Therefore, it is of great practical significance and urgent need to develop a method that comprehensively considers temporal and spatial factors and can more accurately and comprehensively monitor invasive snails in artificial wetlands. Summary of the invention

[0007] The main purpose of the present invention is to construct an efficient and accurate artificial wetland invasive snail monitoring method based on spatiotemporal autocorrelation analysis to solve many technical problems existing in the current artificial wetland invasive snail monitoring methods. For example, traditional monitoring methods such as field investigations are inefficient and easily interfered by subjective factors, remote sensing monitoring is greatly affected by weather and has insufficient accuracy in identifying tiny snails, eDNA monitoring is costly and has strict operating requirements, and the model accuracy and generalization ability of artificial intelligence monitoring need to be improved. The present invention strives to comprehensively, accurately and efficiently monitor the distribution, changes and diffusion trends of invasive snails through multi-source data collection, scientific data processing and screening, advanced model construction and optimization, and in-depth spatiotemporal analysis, so as to provide strong support for the protection and management of artificial wetland ecosystems, thereby effectively preventing and controlling the damage of invasive snails to the ecological balance of artificial wetlands.

[0008] The present invention provides a method for monitoring invasive snails in artificial wetlands based on spatiotemporal autocorrelation analysis, comprising: S1, importing the topographic map into the computer system and rasterizing the monitoring area; using image acquisition equipment to collect image data of artificial wetlands including snails, and at the same time collecting environmental variable data, water quality microbial data and soil physical and chemical property data; S2, selecting representative image samples from images under the same environmental conditions in the same grid as valid image data, removing environmental variables whose collinearity exceeds a preset value from the environmental variable data by using correlation analysis, and obtaining valid environmental variable data; S3, using image recognition to process effective image data and extract image features of snails; using principal component analysis algorithm to extract fusion features of effective image data, effective environmental variable data, water quality microbial data and soil physical and chemical property data to obtain a comprehensive feature vector; S4, construct an invasive snail monitoring model based on cellular automata, define the state, neighborhood relationship and conversion rules of cells; use the ANN algorithm in machine learning to optimize the model parameters; S5, inputting the real-time collected image data, environmental variable data, water quality microbial data and soil physical and chemical property data into the invasive snail monitoring model to evaluate the risk level of invasive snails in the artificial wetland; S6. Use spatiotemporal autocorrelation analysis to determine the spatial distribution characteristics and temporal variation trends of invasive snails in artificial wetlands, combine with geographic information systems to display the dynamic changes in the spatiotemporal distribution of snails, and predict the future invasion and spread path and range.

[0009] As a further preferred solution, in step S2, when screening representative image samples, the clarity score is calculated by image gradient information; including: The difference between the pixel values ​​at each position in the image and the adjacent positions is squared and squared, and then accumulated; Select images with clarity scores in the top 30% and complete snail information as valid image data, and delete other images; When removing the collinearity of environmental variables, the Pearson correlation coefficient is used to calculate the correlation between the observed values ​​of two environmental variables; this coefficient is obtained by dividing the sum of the products of the differences between the observed values ​​of the variables and their means by the square root product of the sum of the squares of their respective differences; Environmental variables with correlation greater than 0.8 and little significance for snail monitoring were removed.

[0010] As a further preferred solution, in step S3, a convolutional neural network is used to extract snail image features; and: the convolution layer calculates the feature map by multiplying the convolution kernel with the corresponding position of the input image and adding a bias term; the pooling layer uses maximum pooling, that is, selecting the maximum value of pixels in the neighborhood.

[0011] As a further preferred solution, in step S3, when the principal component analysis algorithm is used to fuse the features, the data is first standardized by subtracting the mean of the variable from the original data and dividing it by the standard deviation; Next, the covariance matrix is ​​calculated, which is the average of the sum of the products of the differences between the sample data and its mean; The covariance matrix is ​​decomposed by eigenvalues, and the first several eigenvectors with larger eigenvalues ​​are selected to form a transformation matrix. The standardized data matrix is ​​multiplied by the transformation matrix to obtain a comprehensive eigenvector.

[0012] As a further preferred solution, in the cellular automaton model of step S4, the cellular states are divided into three types: no snails, a small number of snails, and a large number of snails; and: the neighborhood adopts Moore neighborhood, which includes 8 surrounding cells; The probability of cell state transition is determined by random terms, global transition probability, constraint conditions and neighborhood functions; the random terms are based on natural logarithm correlation operations, and the global transition probability is the ratio of the number of cells that undergo state transition to the total number of cells; The constraint condition is determined by the environmental suitability index of the cell location, which is 1 when the index is greater than 0.5, otherwise it is 0; and the neighborhood function is the sum of the state values ​​of the cells in the neighborhood divided by the number of cells in the neighborhood; When using the ANN algorithm to optimize model parameters, the back-propagation algorithm is used to calculate the gradient, which is calculated by the sum of the partial derivatives between the output layer neuron output, the hidden layer neuron input and the weights, and the weights are continuously adjusted to minimize the loss function.

[0013] As a further preferred solution, in step S5, when assessing the risk level of invasive snails in artificial wetlands, a risk assessment index system is constructed, including: The snail density index is the ratio of the number of snails per unit area to the area of ​​the monitoring area; The invasion and spread rate index is the ratio of the change in the number of snails within a certain period of time to the change in time; The environmental suitability index is obtained by multiplying the difference between the current values ​​of environmental factors such as temperature, humidity, pH, dissolved oxygen and the maximum values ​​suitable for snail survival by their respective weight coefficients, and the sum of the weight coefficients is 1; and: The weight of each indicator is determined by the hierarchical analysis method, and the final risk level is the sum of the product of each indicator score and its weight.

[0014] As a further preferred solution, in step S6, when using spatiotemporal autocorrelation analysis to study the spatiotemporal distribution characteristics and change trends of invasive snails in artificial wetlands, the following global Moran's I index and local Moran's I index are constructed; The global Moran's I index calculation formula is: ; Where n is the number of samples, is the sum of the elements of the spatial weight matrix, is the spatial weight matrix, , is the snail distribution data, is the mean; The local Moran's I index calculation formula is: ; in , ; When combining with the geographic information system to display the dynamic changes in the spatiotemporal distribution of snails, the calculated Moran's I index results are visualized on the map, with different colors and symbols indicating different degrees of autocorrelation; When predicting the future invasion and diffusion path and range, a prediction method based on the cellular automaton model is used. According to the current state of the cells and the conversion rules, the state of the cells at future moments is iteratively calculated to predict the diffusion of snails.

[0015] As a further preferred solution, when using the principal component analysis algorithm for feature extraction, the method for determining the number of principal components is: first calculate the cumulative contribution rate, the formula is: ; in, is the eigenvalue, is the number of original variables; Select The cumulative contribution rate reaches over 85%.

[0016] As a further preferred solution, in the invasive snail monitoring model based on cellular automata, the global transition probability According to historical data statistics, the formula is: .

[0017] As a further preferred solution, when using the analytic hierarchy process to determine the weights of risk assessment indicators, a judgment matrix is ​​constructed: ; in, Indicates the importance of the i-th indicator relative to the j-th indicator, which is determined by expert scoring; then the maximum eigenvalue of the judgment matrix is ​​calculated And the corresponding eigenvector W, the formula is: ; Finally, the feature vector is normalized to obtain the weight of each indicator. .

[0018] Compared with the existing technology, the present invention focuses on monitoring invasive snails in artificial wetlands and integrates multiple technologies to build a comprehensive monitoring system, which can comprehensively improve the monitoring level of invasive snails and protect the ecology of artificial wetlands. The details are as follows: First, the present invention uses advanced image acquisition equipment and multi-source data acquisition technology, combined with image screening and correlation analysis methods, to obtain high-quality images and environmental data. Through image recognition and principal component analysis algorithms, the image features and comprehensive feature vectors of snails are accurately extracted, and the types, quantity, distribution range, activity and other information of snails can be more accurately identified. For example, in a complex artificial wetland environment, traditional methods are susceptible to interference, resulting in large errors in snail identification, while the present invention can effectively eliminate interference, improve monitoring accuracy, and provide a reliable basis for subsequent prevention and control.

[0019] Secondly, the present invention constructs an invasive snail monitoring model based on cellular automata, and optimizes the model parameters in combination with the ANN algorithm. The cellular automata model takes into account the spatial heterogeneity and dynamic changes of snails in artificial wetlands, and the ANN algorithm enables the model to automatically adjust parameters to adapt to different monitoring scenarios by learning from a large amount of data. Compared with the traditional fixed parameter model, the model of the present invention can more accurately simulate the spatiotemporal distribution and change patterns of snails, improve the accuracy of prediction, and effectively deal with the complexity of the artificial wetland environment and the uncertainty of snail invasion.

[0020] Third, the present invention uses spatiotemporal autocorrelation analysis combined with geographic information systems to conduct in-depth research on the spatiotemporal distribution characteristics and changing trends of invasive snails. By calculating indicators such as the Moran's I index, the aggregation and correlation of snail distribution can be accurately judged, and the dynamic changes in the spatiotemporal distribution of snails can be intuitively displayed in combination with GIS technology to predict the future invasion and diffusion path and range. Potential invasion risk areas can be discovered in advance to buy time for timely prevention and control measures, thereby reducing the degree of damage to artificial wetland ecosystems caused by invasive snails.

[0021] The present invention combines multi-source data and historical prevention and control cases, finds similar scenarios from historical cases and optimizes prevention and control plans according to different snail invasion situations and environmental conditions, and makes decisions based on the set prevention and control rules. When the number of snails exceeds the threshold and the environmental suitability index is greater than 0.7, biological control combined with physical control methods is preferred to improve the scientificity and effectiveness of prevention and control decisions and avoid waste of resources and ecological damage caused by blind prevention and control.

[0022] The monitoring method of the present invention can timely detect the threat of invasive snails, effectively control the spread of invasive snails through precise monitoring, risk assessment and scientific prevention and control decisions, and protect the biodiversity and ecological balance of artificial wetlands. It can reduce the damage of invasive snails to wetland plants and water ecology, maintain the normal functions of wetland ecosystems, ensure the ecological service value of artificial wetlands in water purification, flood regulation, biological habitats, etc., promote the sustainable development of artificial wetlands, and provide strong support for ecological environment construction and protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 Shown is a workflow diagram of the present invention; Figure 2 A plan view of an artificial wetland in one embodiment of the present invention is shown; Figure 3 A schematic diagram of rasterizing terrain in one embodiment of the present invention is shown; Figure 4 The spatial distribution characteristics of invasive snails in an artificial wetland in one embodiment of the present invention are shown; Figure 5A schematic diagram showing the dynamic changes in the spatiotemporal distribution of invasive snails in one embodiment of the present invention is shown. DETAILED DESCRIPTION

[0024] The preferred embodiments of the present invention will be described in detail below so that the purpose, features and advantages of the present invention can be more clearly understood. It should be understood that the following embodiments are not intended to limit the scope of the present invention, but are only intended to illustrate the essential spirit of the technical solution of the present invention.

[0025] In the following description, certain specific details are set forth for the purpose of illustrating various disclosed embodiments to provide a thorough understanding of the various disclosed embodiments. However, those skilled in the relevant art will recognize that the embodiments may be practiced without one or more of these specific details. In other cases, well-known techniques associated with the present application may not be shown or described in detail to avoid unnecessarily obscuring the description of the embodiments.

[0026] References throughout the specification to "one embodiment" or "an embodiment" indicate that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. Thus, the appearances of "in one embodiment" or "in an embodiment" in various places throughout the specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any manner in one or more embodiments.

[0027] like Figure 1 As shown, a method for monitoring invasive snails in artificial wetlands based on spatiotemporal autocorrelation analysis includes: S1, importing the topographic map into the computer system and rasterizing the monitoring area; using image acquisition equipment to collect image data of artificial wetlands including snails, and at the same time collecting environmental variable data, water quality microbial data and soil physical and chemical property data; S2, selecting representative image samples from images under the same environmental conditions in the same grid as valid image data, removing environmental variables whose collinearity exceeds a preset value from the environmental variable data by using correlation analysis, and obtaining valid environmental variable data; S3, using image recognition to process effective image data and extract image features of snails; using principal component analysis algorithm to extract fusion features of effective image data, effective environmental variable data, water quality microbial data and soil physical and chemical property data to obtain a comprehensive feature vector; S4, construct an invasive snail monitoring model based on cellular automata, define the state, neighborhood relationship and conversion rules of cells; use the ANN algorithm in machine learning to optimize the model parameters; S5, inputting the real-time collected image data, environmental variable data, water quality microbial data and soil physical and chemical property data into the invasive snail monitoring model to evaluate the risk level of invasive snails in the artificial wetland; S6. Use spatiotemporal autocorrelation analysis to determine the spatial distribution characteristics and temporal variation trends of invasive snails in artificial wetlands, combine with geographic information systems to display the dynamic changes in the spatiotemporal distribution of snails, and predict the future invasion and spread path and range.

[0028] As a further preferred solution, in step S2, when screening representative image samples, the clarity score is calculated by image gradient information; including: The difference between the pixel values ​​at each position in the image and the adjacent positions is squared and squared, and then accumulated; Select images with clarity scores in the top 30% and complete snail information as valid image data, and delete other images; When removing the collinearity of environmental variables, the Pearson correlation coefficient is used to calculate the correlation between the observed values ​​of two environmental variables; this coefficient is obtained by dividing the sum of the products of the differences between the observed values ​​of the variables and their means by the square root product of the sum of the squares of their respective differences; Environmental variables with correlation greater than 0.8 and little significance for snail monitoring were removed.

[0029] The specific implementation is as follows: The method of selecting representative image samples from images under the same grid and the same environmental conditions is as follows: The clarity score of each image is calculated and measured by the gradient information of the image. The formula is: ; in Indicates that the image is The pixel value of the position, M and N are the number of rows and columns of the image respectively; Select images with clarity scores in the top 30% and complete snail information as valid image data, and automatically delete other redundant images; When the correlation analysis is used to remove environmental variables whose collinearity exceeds a preset value in the environmental variable data, the Pearson correlation coefficient is used for calculation, and the formula is: ; in and are the observed values ​​of two environmental variables, and is the mean of the corresponding variable, n is the sample size; Among the environmental variables with correlation greater than 0.8, variables with little significance for snail monitoring were removed.

[0030] In step S3, the convolutional neural network is used to extract the features of the snail image; and: the convolution layer calculates the feature map by multiplying the convolution kernel with the corresponding position of the input image and adding the bias term; the pooling layer uses maximum pooling, that is, selecting the maximum value of the pixels in the neighborhood. Among them, the convolution calculation process of the convolution layer is as follows: ; in, is the convolution kernel, m, n are the convolution kernel sizes, Image is the input image, is the bias term; The pooling layer uses the maximum pooling operation, the formula is: .

[0031] As a further preferred solution, when using the principal component analysis algorithm to fuse features, the data is first standardized by subtracting the mean of the variable from the original data and then dividing it by the standard deviation; Next, the covariance matrix is ​​calculated, which is the average of the sum of the products of the differences between the sample data and its mean; The covariance matrix is ​​decomposed by eigenvalues, and the first several eigenvectors with larger eigenvalues ​​are selected to form a transformation matrix. The standardized data matrix is ​​multiplied by the transformation matrix to obtain a comprehensive eigenvector.

[0032] The specific implementation is as follows: First, the data is standardized, and the formula is: ; in, is the original data, is the mean of the variable; Then calculate the covariance matrix: ; in, is the i-th sample data, is the sample mean; Perform eigenvalue decomposition on the covariance matrix to obtain eigenvectors and eigenvalues, select the first k eigenvectors whose eigenvalues ​​are greater than the set value to form the transformation matrix P, and finally obtain the comprehensive eigenvector Y=XP, where X is the standardized data matrix.

[0033] As a further preferred solution, in the cellular automaton model of step S4, the cellular states are divided into three types: no snails, a small number of snails, and a large number of snails; and: the neighborhood adopts Moore neighborhood, which includes 8 surrounding cells; The probability of cell state transition is determined by random terms, global transition probability, constraint conditions and neighborhood functions; the random terms are based on natural logarithm correlation operations, and the global transition probability is the ratio of the number of cells that undergo state transition to the total number of cells; The constraint condition is determined by the environmental suitability index of the cell location, which is 1 when the index is greater than 0.5, otherwise it is 0; and the neighborhood function is the sum of the state values ​​of the cells in the neighborhood divided by the number of cells in the neighborhood; When using the ANN algorithm to optimize model parameters, the back-propagation algorithm is used to calculate the gradient, which is calculated by the sum of the partial derivatives between the output layer neuron output, the hidden layer neuron input and the weights, and the weights are continuously adjusted to minimize the loss function.

[0034] The specific implementation process includes the following steps: Define the state of the cell as: ; The neighborhood relationship adopts Moore's neighborhood method, that is, the neighborhood of a cell includes the 8 cells around it; The transformation rule of the cell is determined by the following formula: ; in, For cells The probability of state transition at time t, represents a random item, is the global transition probability, Represents the constraint condition of the cell unit. When the environmental suitability index of the cell location is greater than 0.5, , otherwise 0; represents the neighborhood function, which is determined by the following formula; ; in, is the neighborhood of the cell, is the number of cells in the neighborhood; When the ANN algorithm is used to optimize the model parameters, the back propagation algorithm is used to calculate the gradient. The formula is: ; Where E is the loss function, is the output of the kth neuron in the output layer, is the input of the jth neuron in the hidden layer, is the weight connecting the i-th neuron in the hidden layer and the j-th neuron in the output layer, and L is the number of neurons in the output layer; Finally, the loss function is minimized by continuously adjusting the weights.

[0035] As a further preferred solution, in step S5, when assessing the risk level of invasive snails in artificial wetlands, a risk assessment index system is constructed, including: The snail density index is the ratio of the number of snails per unit area to the area of ​​the monitoring area; The invasion and spread rate index is the ratio of the change in the number of snails within a certain period of time to the change in time; The environmental suitability index is obtained by multiplying the difference between the current values ​​of environmental factors such as temperature, humidity, pH, dissolved oxygen and the maximum values ​​suitable for snail survival by their respective weight coefficients, and the sum of the weight coefficients is 1; and: The weight of each indicator is determined by the hierarchical analysis method, and the final risk level is the sum of the product of each indicator score and its weight.

[0036] The specific risk assessment index calculation formula is as follows: Snail density index: ; Where N is the number of snails per unit area, and A is the area of ​​the monitoring area; Invasion spread speed index: ; in, is Change in the number of snails over time; Environmental suitability index indicators: ; in, , , , They are the currently monitored temperature, humidity, pH and dissolved oxygen values. , , , , , , , are the minimum and maximum values ​​of the corresponding environmental factors suitable for the survival of snails. , , , is a weight coefficient determined according to the importance of each environmental factor on snails, and + + + =1; The weight of each indicator is determined by the hierarchical analysis method, and the final risk level is: ; is the score of each indicator, is the weight of the corresponding indicator.

[0037] In some embodiments, in step S6, when using spatiotemporal autocorrelation analysis to study the spatiotemporal distribution characteristics and change trends of invasive snails in artificial wetlands, the following global Moran's I index and local Moran's I index are constructed; The global Moran's I index calculation formula is: ; Where n is the number of samples, is the sum of the elements of the spatial weight matrix, is the spatial weight matrix, , is the snail distribution data, is the mean; The local Moran's I index calculation formula is: ; in , ; When combining with the geographic information system to display the dynamic changes in the spatiotemporal distribution of snails, the calculated Moran's I index results are visualized on the map, with different colors and symbols indicating different degrees of autocorrelation; When predicting the future invasion and diffusion path and range, a prediction method based on the cellular automaton model is used. According to the current state of the cells and the conversion rules, the state of the cells at future moments is iteratively calculated to predict the diffusion of snails.

[0038] As a further preferred solution, when calculating the image clarity score, the gradient information is Gaussian weighted, and the formula is: ; Among them, G(x,y) is a two-dimensional Gaussian function, , is the standard deviation.

[0039] As a further preferred solution, when using the principal component analysis algorithm for feature extraction, the method for determining the number of principal components is: first calculate the cumulative contribution rate, the formula is: ; in, is the eigenvalue, is the number of original variables; Select The cumulative contribution rate reaches over 85%.

[0040] As a further preferred solution, in the invasive snail monitoring model based on cellular automata, the global transition probability According to historical data statistics, the formula is: .

[0041] As a further preferred solution, when using the analytic hierarchy process to determine the weights of risk assessment indicators, a judgment matrix is ​​constructed: ; in, Indicates the importance of the i-th indicator relative to the j-th indicator, which is determined by expert scoring; then the maximum eigenvalue of the judgment matrix is ​​calculated And the corresponding eigenvector W, the formula is: ; Finally, the feature vector is normalized to obtain the weight of each indicator. .

[0042] like Figure 2 As shown in the figure, a sewage treatment plant is located on the edge of a natural wetland (Guizhou Caohai Nature Reserve), and a large artificial wetland has been constructed. There is a vegetation coverage area between the artificial wetland and the natural wetland, and the main vegetation species are Siberian iris, wild rice stem, water celery, etc. In 2018, monitoring found that snails and mollusks were breeding in the artificial wetland. After a long period of investigation, it was determined that they were alien invasive snails and mollusks, including the elegant sharp-margined snail. Oxyloma elegans (Risso, 1826), Cystoma acuminatum Physella acuta raparnaud ,1805) and Ryukyu snail Acusta despecta (Gray, 1839) etc. Allowing these threaded molluscs to invade will cause major ecological damage, such as breeding bacteria and endangering the drinking water safety of sewage treatment plants.

[0043] In this regard, the artificial wetland carried out experimental prevention and control work according to the scheme of the present invention and achieved good prevention and control effects. Figure 3 As shown, first, the terrain of the monitoring area is rasterized according to the grid unit of 10 meters × 10 meters, and the image data of the artificial wetland containing snails is collected by using image acquisition equipment. In this embodiment, the image acquisition is carried out during the day when the weather is clear and the light is sufficient. During the snail breeding season, the image acquisition is carried out for 3 months in a cycle of 7 days, using a combination of equipment such as a high-definition camera or a drone equipped with a camera. Most of the areas are continuously photographed by drones according to the set flight trajectory, and some areas are photographed manually.

[0044] In addition, we further combine multi-source information such as environmental variable data, water quality microbial data, and soil physical and chemical properties data. The distribution and characteristics of snails may be different under different environmental conditions. By comprehensively analyzing these data, we can assist in determining the type and state of snails and improve the accuracy of identification. For example, some snails prefer specific water pH or soil types, and the identification range can be narrowed based on these environmental data.

[0045] From the aspect of image screening under the same environmental conditions of the same grid, after completing image collection for 3 consecutive months with a cycle of 7 days, for each 10m×10m grid, the images collected under the same environmental conditions (such as light intensity, water pH, water temperature, etc. in the same time period) are used to calculate the clarity score of each image. The images with a clarity score in the top 30% and complete snail information are selected as valid image data, and other redundant images are automatically deleted. In terms of environmental variable data processing, the collected environmental variable data cover light intensity, water temperature, water pH, soil moisture, soil nutrient content, etc. Correlation analysis is used to remove environmental variables whose collinearity exceeds a preset value (set to 0.8 in this embodiment). The Pearson correlation coefficient is used for calculation. For environmental variables with a correlation greater than 0.8, their significance for snail monitoring is compared and analyzed, and the variables with less significance for snail monitoring are removed, thereby obtaining valid environmental variable data.

[0046] The valid image data of the same grid and the same environmental conditions selected in the previous step are input into the pre-built image recognition model based on convolutional neural network (CNN). The model has been trained on a large number of snail image data sets and has the ability to identify snail features. The image passes through the convolution layer of the model in turn. The convolution kernel in the convolution layer slides on the image, performs convolution operations on the local area of ​​the image, and extracts the primary features of the snail, such as the edge, texture, shape, etc., such as identifying the unique spiral texture of the shell and the general outline of the snail body. Then the image enters the pooling layer, and the feature map output by the convolution layer is reduced in dimensionality through the maximum pooling operation, retaining the key feature information while reducing the amount of calculation, so that the model can process data more efficiently, and to a certain extent improve the robustness of the model to prevent overfitting. After multiple alternating processes of the convolution layer and the pooling layer, a feature vector containing detailed image features of the snail is obtained.

[0047] The snail image feature vector extracted from the effective image data is integrated with the effective environmental variable data, water quality microbial data, and soil physical and chemical property data obtained in the previous step. These data cover various aspects of information related to the survival of snails in artificial wetlands, such as light intensity, water temperature, number of microbial species, soil pH, etc. The integrated multi-source data is standardized to ensure that data of different types and magnitudes are on the same comparable scale, eliminating the impact of data dimension differences on subsequent analysis.

[0048] The covariance matrix was calculated based on the standardized data to analyze the correlation between the variables. The eigenvalue decomposition of the covariance matrix was performed to obtain the eigenvectors and eigenvalues. The first several eigenvectors with larger eigenvalues ​​were selected according to the cumulative contribution rate to form a transformation matrix. The standardized multi-source data matrix was multiplied by the transformation matrix to obtain a comprehensive eigenvector. This comprehensive eigenvector integrates key information from multiple aspects such as images, environment, water quality microorganisms, and soil physical and chemical properties, and more comprehensively reflects the characteristics related to invasive snails in artificial wetlands.

[0049] According to the grid division of 10m×10m of the artificial wetland, each grid is regarded as a cell. According to the actual presence of snails in each cell, the cell state is defined as: no snails, snails are present but the number is small, and snails are present but the number is large. Through field investigation and image recognition statistics, if the number of snails in a cell is 0, its state is no snails; if the number is between 1-100, it is judged that snails are present but the number is small; if it exceeds this range, snails are present but the number is large.

[0050] The Moore neighborhood method is adopted, that is, the neighborhood of each cell includes 8 cells around it in the four diagonal directions. This neighborhood setting method can better simulate the spatial interaction between cells, taking into account the possible diffusion direction and range of snails in the wetland environment. The conversion rules of cells are determined in combination with the actual environmental conditions of artificial wetlands and the biological characteristics of snails. The conversion rules comprehensively consider factors such as random factors, global conversion probability, environmental suitability index of the cell location, and cell status in the neighborhood.

[0051] For example, the environmental suitability index is obtained by comprehensively evaluating the environmental factors such as light, water temperature, water quality, and soil at the location of the cell. When the environmental suitability index is greater than 0.5, the environment of the cell is considered to be more favorable for the survival and reproduction of snails, and the corresponding weight is given in the conversion rule. The global conversion probability is obtained based on historical data statistics, reflecting the overall possibility of cell state conversion in the entire monitoring area. The neighborhood function measures the impact of the neighborhood on the current cell state conversion by calculating the ratio of the sum of the cell states in the neighborhood to the number of cells in the neighborhood.

[0052] The comprehensive feature vector obtained in the previous step is used as input data and input into the artificial neural network (ANN) model. The ANN model has a hidden layer and an output layer composed of multiple neurons, and the neurons are connected by weights. The model is trained using the back propagation algorithm. During the training process, the model calculates based on the input comprehensive feature vector and outputs the predicted results of the cell state. The predicted results are compared with the actual cell state, and the loss function value is calculated, which reflects the degree of deviation between the model prediction results and the actual situation.

[0053] Through the back propagation algorithm, starting from the output layer, the gradient of the loss function is back-propagated along the connection path between neurons to calculate the influence of each weight on the loss function, that is, the gradient value. According to the calculated gradient value, the weights between neurons are adjusted so that the loss function value gradually decreases. This process is repeated continuously. After multiple iterative training, the ANN model can accurately predict the cellular state based on the input comprehensive feature vector, thereby optimizing the parameters of the invasive snail monitoring model based on cellular automata and improving the model's simulation accuracy of the spatiotemporal distribution and changes of invasive snails in artificial wetlands.

[0054] Continue to use the same image acquisition equipment as in the previous period, and collect images of the artificial wetland at a fixed period during the period of suitable weather to ensure that the latest snail distribution and activity are captured. At the same time, use various sensors to monitor environmental variable data in real time, such as light intensity sensors, water temperature sensors, pH sensors, etc., and record data regularly. Collect water quality microbial data through professional water quality sampling tools and microbial detection kits, sample the soil regularly, and use laboratory instruments to analyze soil physical and chemical properties data. Integrate these real-time collected data to ensure that the data timestamps are consistent so that they can be accurately input into the monitoring model.

[0055] Based on the previous research and monitoring experience of invasive snails, a risk assessment index system was constructed. The system includes snail density index, which is measured by counting the number of snails per unit area; invasion and diffusion speed index, which is determined based on the changes in the number of snails and the changes in the distribution range in different time periods; environmental suitability index index, which comprehensively considers the suitability of environmental factors such as light, water temperature, water quality, and soil for the survival and reproduction of snails. For example, by analyzing historical data and experimental results, the suitable range of different environmental factors is determined, and when the actual environmental data is within the suitable range, a higher suitability score is assigned.

[0056] The integrated real-time data is input into the optimized cellular automation-based invasive snail monitoring model. The model simulates the state changes and distribution of snails in the current environment based on the input data and the established transformation rules. At the same time, the scores of each indicator are calculated according to the risk assessment index system. The weight of each indicator is determined by the hierarchical analysis method, and the scores of each indicator are multiplied and accumulated with their weights to obtain the risk assessment results of invasive snails in artificial wetlands. According to the pre-set risk level threshold, the current risk is judged to be low, medium or high.

[0057] On the basis of the above, data preparation for spatiotemporal autocorrelation analysis was carried out. First, snail distribution data within a certain period of time (e.g., from the first discovery of invasive snails to the present) were collected, including snail status information of each cell at different time points, as well as corresponding environmental variable data, etc. These data were organized into a format suitable for spatiotemporal autocorrelation analysis to ensure data accuracy and completeness.

[0058] The global Moran's I index and the local Moran's I index were calculated using the spatiotemporal autocorrelation analysis method. In the calculation process, the snail distribution data and the spatial weight matrix were used as input, and the index values ​​were obtained through the corresponding algorithm. These index values ​​can reflect whether the spatial distribution of invasive snails is clustered and which areas have significant spatial autocorrelation. A global Moran's I index greater than 0 indicates positive autocorrelation, that is, the snails are clustered in space; the local Moran's I index can determine which specific cellular areas have abnormal aggregation or dispersion.

[0059] Integrate the calculated spatiotemporal autocorrelation analysis results with the geographic information system. Import the map data of the artificial wetland on the GIS platform. Associate the spatiotemporal distribution data of snails and the index values ​​obtained from the spatiotemporal autocorrelation analysis with the map, such as Figure 4 As shown. By setting different colors, symbols and transparency, the dynamic changes in the temporal and spatial distribution of snails can be intuitively displayed on the map. For example, red is used to represent areas with high concentrations of snails, and the darker the color, the higher the concentration. As time goes by, the expansion or contraction of the snail distribution range can be shown through dynamic demonstration. Figure 5 shown.

[0060] Based on the optimized invasive snail monitoring model and the results of spatiotemporal autocorrelation analysis, the spatial analysis function and model prediction function of the GIS platform are used to predict the future diffusion path and range of invasive snails. The model takes into account factors such as the reproductive characteristics of snails, environmental suitability, and spatial autocorrelation, and simulates the possible diffusion direction and speed of snails in the future. By drawing the predicted diffusion range and path on the GIS map, forward-looking information is provided for the management and prevention of artificial wetlands, so that corresponding response measures can be formulated in advance.

[0061] exist Figure 4 and Figure 5 It can be seen that there are two main directions of snail invasion, namely the F8 grid direction and the H16 grid direction. The H16 grid direction is the most serious, and it has spread widely to the J16 direction and has invaded the hinterland of the artificial wetland. Combined with other environmental indicators, it was found that Siberian iris provides a good medium for snail invasion. Therefore, the following measures are taken: according to Figure 5 Guided by the guidelines, the darker areas in the map are marked as key prevention and control areas. After all the Siberian irises in this area are removed, lime is evenly spread over the area to kill the mollusks. After a day of exposure to the sun, water is released into the area, and lime is evenly spread to kill snails. The pH value of the area is adjusted, and after disinfection with potassium permanganate, four native plants, water onion, wild rice stem, water celery and willow herb, are planted in a reasonable combination. The specific steps are as follows: (1) Disinfection of native plants. After the four native plants, water plantain, wild rice stem, water celery and willow herb, are taken back from the plant, potassium permanganate is diluted with water at a ratio of 2‰ and soaked in a large basin for about an hour. After disinfection, they can be planted.

[0062] (2) Planting of native plants. Four native plants, namely water plantain, wild rice stem, water celery and willow herb, were planted in the experimental plots at a spacing of 50 cm*50 cm, with each plant planted in an alternating pattern.

[0063] right Figure 5 The lighter colored area was marked as the secondary control area, and the plants were not replaced, only killing and disinfection were carried out. Through the above treatment, a good control effect was achieved.

[0064] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for monitoring invasive snails in artificial wetlands based on spatiotemporal autocorrelation analysis, characterized in that: include: S1, import the topographic map into the computer system and rasterize the monitoring area; Use image acquisition equipment to collect image data of artificial wetlands including snails, and collect environmental variable data, water quality microbial data, and soil physical and chemical property data; S2, selecting representative image samples from images under the same environmental conditions in the same grid as valid image data, removing environmental variables whose collinearity exceeds a preset value from the environmental variable data by using correlation analysis, and obtaining valid environmental variable data; S3, using image recognition to process effective image data and extract image features of snails; using principal component analysis algorithm to extract fusion features of effective image data, effective environmental variable data, water quality microbial data and soil physical and chemical property data to obtain a comprehensive feature vector; S4, construct an invasive snail monitoring model based on cellular automata, define the state, neighborhood relationship and conversion rules of cells; use the ANN algorithm in machine learning to optimize the model parameters; S5, inputting the real-time collected image data, environmental variable data, water quality microbial data and soil physical and chemical property data into the invasive snail monitoring model to evaluate the risk level of invasive snails in the artificial wetland; S6. Use spatiotemporal autocorrelation analysis to determine the spatial distribution characteristics and temporal variation trends of invasive snails in artificial wetlands, combine with geographic information systems to display the dynamic changes in the spatiotemporal distribution of snails, and predict the future invasion and spread path and range.

2. The method for monitoring invasive snails in artificial wetlands based on spatiotemporal autocorrelation analysis according to claim 1, characterized in that: In step S2, when selecting representative image samples, the clarity score is calculated by using the image gradient information; including: The difference between the pixel values ​​at each position in the image and the adjacent positions is squared and squared, and then accumulated; Select images with clarity scores in the top 30% and complete snail information as valid image data, and delete other images; When removing the collinearity of environmental variables, the Pearson correlation coefficient is used to calculate the correlation between the observed values ​​of two environmental variables; this coefficient is obtained by dividing the sum of the products of the differences between the observed values ​​of the variables and their means by the square root product of the sum of the squares of their respective differences; Environmental variables with correlation greater than 0.8 and little significance for snail monitoring were removed.

3. The method for monitoring invasive snails in artificial wetlands based on spatiotemporal autocorrelation analysis according to claim 1, characterized in that: In step S3, the snail image features are extracted using a convolutional neural network; In addition: the convolution layer calculates the feature map by multiplying the convolution kernel with the corresponding position of the input image and adding a bias term; the pooling layer uses maximum pooling, that is, selecting the maximum value of the pixel in the neighborhood.

4. The method for monitoring invasive snails in artificial wetlands based on spatiotemporal autocorrelation analysis according to claim 3, characterized in that: In step S3, when the principal component analysis algorithm is used to fuse features, the data is first standardized by subtracting the mean of the variable from the original data and dividing it by the standard deviation; Next, the covariance matrix is ​​calculated, which is the average of the sum of the products of the differences between the sample data and its mean; The covariance matrix is ​​decomposed by eigenvalues, and the first several eigenvectors with larger eigenvalues ​​are selected to form a transformation matrix. The standardized data matrix is ​​multiplied by the transformation matrix to obtain a comprehensive eigenvector.

5. The method for monitoring invasive snails in artificial wetlands based on spatiotemporal autocorrelation analysis according to claim 1, characterized in that: In the cellular automaton model of step S4, the cell states are divided into three types: no snails, a small number of snails, and a large number of snails; and: the neighborhood adopts Moore's neighborhood, which includes 8 surrounding cells; The probability of cell state transition is determined by random terms, global transition probability, constraint conditions and neighborhood functions; the random terms are based on natural logarithm correlation operations, and the global transition probability is the ratio of the number of cells that undergo state transition to the total number of cells; The constraint condition is determined by the environmental suitability index of the cell location, which is 1 when the index is greater than 0.5, otherwise it is 0; and the neighborhood function is the sum of the state values ​​of the cells in the neighborhood divided by the number of cells in the neighborhood; When using the ANN algorithm to optimize model parameters, the back-propagation algorithm is used to calculate the gradient, which is calculated by the sum of the partial derivatives between the output layer neuron output, the hidden layer neuron input and the weights, and the weights are continuously adjusted to minimize the loss function.

6. The method for monitoring invasive snails in artificial wetlands based on spatiotemporal autocorrelation analysis according to claim 1, characterized in that: In step S5, when assessing the risk level of invasive snails in artificial wetlands, a risk assessment index system is constructed, including: The snail density index is the ratio of the number of snails per unit area to the area of ​​the monitoring area; The invasion and spread rate index is the ratio of the change in the number of snails within a certain period of time to the change in time; The environmental suitability index is obtained by multiplying the difference between the current values ​​of environmental factors such as temperature, humidity, pH, dissolved oxygen and the maximum values ​​suitable for snail survival by their respective weight coefficients, and the sum of the weight coefficients is 1; and: The weight of each indicator is determined by the hierarchical analysis method, and the final risk level is the sum of the product of each indicator score and its weight.

7. The method for monitoring invasive snails in artificial wetlands based on spatiotemporal autocorrelation analysis according to claim 1, characterized in that: In step S6, when using spatiotemporal autocorrelation analysis to study the spatiotemporal distribution characteristics and changing trends of invasive snails in artificial wetlands, the following global Moran's I index and local Moran's I index are constructed; The global Moran's I index calculation formula is: ； Where n is the number of samples, is the sum of the elements of the spatial weight matrix, is the spatial weight matrix, , is the snail distribution data, is the mean; The local Moran's I index calculation formula is: ； in , ; Where n is the number of samples, is the normalized value of the current cell i, is the normalized value of the current cell j; is the standard deviation, which is used to standardize the original data. and ;pass The original data were transformed into a standard normal distribution so that the spatial autocorrelations of different variables were comparable; When combining with the geographic information system to display the dynamic changes in the spatiotemporal distribution of snails, the calculated Moran's I index results are visualized on the map, with different colors and symbols indicating different degrees of autocorrelation; When predicting the future invasion and diffusion path and range, a prediction method based on the cellular automaton model is used. According to the current state of the cells and the conversion rules, the state of the cells at future moments is iteratively calculated to predict the diffusion of snails.

8. The method for monitoring invasive snails in artificial wetlands based on spatiotemporal autocorrelation analysis according to claim 4, characterized in that: When using the principal component analysis algorithm for feature extraction, the method to determine the number of principal components is: first calculate the cumulative contribution rate, the formula is: ； in, is the eigenvalue, is the number of original variables; k represents the first k principal components selected from the principal component analysis results; Select The cumulative contribution rate reaches over 85%.

9. The method for monitoring invasive snails in artificial wetlands based on spatiotemporal autocorrelation analysis according to claim 5, characterized in that: In the invasive snail monitoring model based on cellular automata, the global transition probability According to historical data statistics, the formula is: 。 10. The method for monitoring invasive snails in artificial wetlands based on spatiotemporal autocorrelation analysis according to claim 6, characterized in that: When using the analytic hierarchy process to determine the weights of risk assessment indicators, a judgment matrix is ​​constructed: ； in, Indicates the importance of the i-th indicator relative to the j-th indicator, which is determined by expert scoring; then the maximum eigenvalue of the judgment matrix is ​​calculated And the corresponding eigenvector W, the formula is: ； Finally, the feature vector is normalized to obtain the weight of each indicator. .

Images