Prediction method, system and storage medium for chlorophyll content of invasive plants

Abstract

The present invention discloses a prediction method and system for the chlorophyll content of invasive plants, and a storage medium. By combining multispectral remote sensing technology and machine learning algorithms, high-precision prediction of the chlorophyll content of invasive plants is achieved. First, multispectral image acquisition technology is used to collect image data of a to-be-tested area; then, plants in the image are classified and identified by computer vision technology; then, the identified plant images are preprocessed and key spectral information, such as normalized vegetation index, ratio vegetation index and difference vegetation index, are extracted; finally, these spectral information are used to predict the chlorophyll content of the plant by a random forest model. The prediction model is subjected to accuracy evaluation and optimization to ensure that accurate chlorophyll content prediction can be provided, thereby realizing accurate spatial distribution prediction of chlorophyll content in a geographic information system. The present invention significantly improves the efficiency and accuracy of invasive plant monitoring, and is of great significance for ecological monitoring and management.

Description

Technical Field

[0001] The invention relates to the field of remote sensing technology and plant ecological monitoring technology, and in particular to a prediction method and system for chlorophyll content of invasive plants and a storage medium. Background Art

[0002] Invasive plants, such as Spartina alterniflora and Mikania micrantha, pose a major threat to the balance of ecosystems. They not only change soil structure and nutrient cycles, but may also suppress the growth of native species, thereby destroying biodiversity. Accurately predicting the chlorophyll content of invasive plants is crucial for monitoring their growth and formulating effective management measures. Chlorophyll content is an important indicator of plant health and photosynthesis capacity, so effective prediction of it can help environmental scientists and managers assess the growth trend and possible ecological impact of invasive plants.

[0003] Although multispectral and hyperspectral remote sensing technologies have been widely used to monitor vegetation and assess its physiological status, the prediction of chlorophyll content in invasive plants still faces many challenges. On the one hand, the growth response of invasive plants in different regions may vary from place to place, making it difficult for a single model to be universally applicable; on the other hand, conventional remote sensing technology may lack accuracy and resolution in local areas, making it difficult to capture subtle ecological changes. In addition, existing prediction models often fail to fully utilize the rich information contained in spectral data, resulting in limited accuracy and reliability of prediction results.

[0004] Existing technologies show obvious deficiencies in dealing with these problems. For example, traditional prediction methods rely on a large amount of ground sampling data to correct remote sensing data, which is not only time-consuming and labor-intensive, but also difficult to implement in inaccessible areas. In addition, the lack of effective data processing algorithms makes it challenging to extract effective information from complex remote sensing data, further limiting the adaptability and generalizability of prediction models in new environments. Summary of the invention

[0005] In view of the many problems existing in the above-mentioned prior art, the present invention provides a prediction method and system for the chlorophyll content of invasive plants, and a storage medium. The present invention uses multispectral imaging technology combined with a machine learning method to predict the chlorophyll content of invasive plants. First, a multispectral image acquisition device is used to collect images of invasive plants in the test area; then, computer vision technology is used to classify and identify these images, and the identified plant images are further preprocessed to extract key spectral information (such as the normalized difference vegetation index NDVI, the ratio vegetation index RVI and the difference vegetation index DVI). Based on these spectral indices, a random forest algorithm is used to construct a model to predict the chlorophyll content. The model is accurately evaluated and adjusted to ensure high-precision prediction results. The present invention can accurately predict the chlorophyll content, thereby effectively monitoring and managing invasive plants, and providing scientific support for ecological protection.

[0006] like Figure 1-2 As shown, a method for predicting the chlorophyll content of invasive plants comprises the following steps:

[0007] Obtain multispectral images of invasive plants in the area to be tested;

[0008] Using computer vision technology to identify and classify invasive plants in the multispectral image;

[0009] Preprocessing the multispectral image of the identified invasive plant part to extract spectral information, the spectral information includes: normalized vegetation index, ratio vegetation index and difference vegetation index;

[0010] Calculating a spectral index according to the spectral information, and constructing a prediction model for the chlorophyll content of invasive plants based on the spectral index and the chlorophyll content data of the invasive plants, wherein the prediction model adopts a random forest algorithm;

[0011] The prediction model is evaluated for accuracy. Once the prediction accuracy reaches the preset standard, it is used to predict the chlorophyll content of invasive plants in the test area and the spatial distribution of chlorophyll content.

[0012] The predicted chlorophyll content was spatially analyzed in the geographic information system, and the prediction residuals were processed using the Kriging interpolation method to improve the spatial accuracy of the chlorophyll content prediction and obtain the spatial distribution prediction map of the chlorophyll content of invasive plants.

[0013] The present invention uses a multispectral camera carried by an unmanned aerial vehicle to obtain a multispectral image of the area to be tested. Multispectral images can capture different bands of light reflected by plants, including infrared, near-infrared, etc. These bands have significant sensitivity to the physiological and biochemical characteristics of vegetation. For example, the spectral characteristics of Spartina alterniflora and Mikania micrantha have significant differences in the red edge band and the near-infrared band, which helps to distinguish them from other non-invasive plants.

[0014] Next, these images are preprocessed, including image stitching, cropping, and band synthesis. The purpose of this step is to improve the efficiency and accuracy of subsequent processing and remove noise and irrelevant information in the image.

[0015] Computer vision technology, especially support vector machine (SVM) classification algorithm, is used to identify and classify invasive plants in multispectral images. SVM is a powerful classification tool that can find the optimal segmentation hyperplane in multidimensional space to distinguish data points of different categories. In the present invention, SVM can effectively identify invasive plants such as Spartina alterniflora and Mikania micrantha, and distinguish them from native plants or other background features.

[0016] The identified invasive plant images are then used to extract spectral information and calculate indicators such as the Normalized Difference Vegetation Index (NDVI), Ratio Vegetation Index (RVI) and Difference Vegetation Index (DVI), which are key indicators for assessing plant health and chlorophyll content.

[0017] Based on the above spectral index and the chlorophyll content data collected in the field, the random forest algorithm is used to build a prediction model. Random forest is an ensemble learning algorithm that creates multiple decision trees and integrates their prediction results to improve the accuracy and stability of the prediction. In the model training stage, samples in the data set are randomly selected to build multiple decision trees. Each tree is trained on different data to reduce the risk of overfitting of the model, and the final prediction result is determined by a voting mechanism.

[0018] The method of the present invention can accurately predict the chlorophyll content of Spartina alterniflora or Mikania micrantha in a specific area, which is extremely important for ecological management and the formulation of invasive plant control strategies. The accuracy of the model is evaluated by the coefficient of determination (R 2 ) and root mean square error (RMSE), which can reflect the accuracy and reliability of model prediction. Once the prediction accuracy reaches the preset standard, the model can be used to predict the chlorophyll content and its spatial distribution in the test area.

[0019] In the Geographic Information System (GIS), the prediction residuals are processed using the Kriging interpolation method to improve the spatial accuracy of the prediction. Kriging is a spatial interpolation method based on a statistical model that can take into account the geographical location of spatial data and the spatial dependence of prediction residuals, thereby more accurately estimating the chlorophyll content of unsampled points. The application of this method ensures high spatial resolution and accuracy of the chlorophyll content prediction for the entire region, providing a scientific basis for management decisions.

[0020] Preferably, the identification and classification uses a support vector machine classification algorithm to distinguish invasive plants from non-invasive plants.

[0021] Support vector machine (SVM) is a supervised learning method used to solve classification and regression problems. It mainly distinguishes data points of different categories by finding an optimal hyperplane. In multispectral image processing, SVM can effectively distinguish invasive plants from non-invasive plants or other backgrounds, and maintain high classification accuracy even in the case of complex vegetation coverage or similar spectral characteristics.

[0022] SVM achieves maximum margin classification of categories by constructing an optimal hyperplane in feature space. In the context of invasive plant identification, each multispectral image pixel can be regarded as a point in a multidimensional space, where each dimension represents the spectral reflectance value of a specific band.

[0023] In practical applications, if the spectral characteristics of invasive plants and non-invasive plants are linearly separable, SVM will use a linear kernel function; if not, it will use a nonlinear kernel function such as a radial basis function (RBF) or a polynomial kernel function to process the nonlinear characteristics of the data, thereby finding the optimal segmentation surface of the data in a high-dimensional space.

[0024] During the SVM model training process, it is necessary to select a suitable kernel function and adjust the kernel function parameters (such as C value and γ value). The C value controls the penalty for misclassification, and the γ value affects the shape of the radial basis function. The optimization of these parameters is achieved through methods such as cross-validation to ensure that the model has good generalization ability.

[0025] The present invention uses the application of the SVM classification algorithm on multispectral data to allow accurate identification of invasive plants such as Spartina alterniflora and Mikania micrantha. This identification step is the basis for predicting the chlorophyll content of plants, because only by accurately distinguishing plant species can the physiological characteristics of specific plants be accurately measured and analyzed.

[0026] For example, in the case where both Spartina alterniflora and Mikania micrantha exist in a certain area, they may show similar reflectance characteristics in the visible and near-infrared bands. SVM maps the data to a higher-dimensional space using a kernel function, in which the spectral reflectance characteristics of the two plants can be clearly distinguished. In this way, SVM can effectively distinguish between Spartina alterniflora and Mikania micrantha even when the spectral characteristics are very close, thereby ensuring that the subsequent chlorophyll content prediction is performed for the correct plant species.

[0027] In addition, the image data after SVM classification will be directly used to calculate the Normalized Difference Vegetation Index (NDVI), Ratio Vegetation Index (RVI) and Difference Vegetation Index (DVI), which are further used to establish a prediction model for chlorophyll content. Therefore, the accuracy of SVM directly affects the accuracy and reliability of the chlorophyll content prediction model.

[0028] Through the method of the present invention, the support vector machine provides an efficient and reliable method to process complex data in multispectral images, providing strong technical support for accurate monitoring and management of invasive plants.

[0029] Preferably, the multispectral image preprocessing includes: stitching, cropping and band synthesis processing to generate a false color multispectral image for classification.

[0030] Multispectral image stitching is the process of merging multiple image fragments taken from different locations into a continuous large scene image. During drone monitoring, due to flight range and single-shot limitations, multispectral cameras may produce multiple images with partially overlapping coverage areas. Stitching is done by finding the overlapping areas between these images and seamlessly connecting the images in these areas to ensure the spatial continuity and integrity of the images. This ensures that the images for subsequent analysis cover the complete target area and eliminate boundary effects.

[0031] Cropping is to remove non-target areas in multispectral images and focus on areas of interest (such as invasive plant growth areas). This step is usually based on prior knowledge or automatically detected image edges. Removing irrelevant information improves processing efficiency and classification accuracy.

[0032] Band synthesis is the process of combining several selected bands to generate a new image. In a multispectral image, each band captures the reflection or emission of light at a specific wavelength. By selecting bands with specific characteristics for identifying and distinguishing invasive plants (such as red edge, near-infrared bands) and synthesizing them into a false color image, the difference between these plants and the surrounding environment can be more clearly shown. Enhanced visualization of specific plant features can assist classification algorithms in more accurately identifying target plants.

[0033] Through this series of preprocessing steps, the image data is optimized for subsequent machine learning model processing, especially in chlorophyll content prediction. For example, by accurately stitching and cropping images, it can be ensured that every part of the entire area is taken into account when making chlorophyll content predictions, without missing any possible invasive plant communities. Band synthesis is particularly important because it directly affects the calculation of spectral indices, which are key variables for subsequent chlorophyll prediction models (such as NDVI).

[0034] When pre-processing invasive plants such as Spartina alterniflora and Mikania micrantha, their strong reflective properties in specific bands (such as near infrared) can be selected. For example, by using the contrast between the high reflectivity of Spartina alterniflora in the near infrared band and the red light band, the contrast between it and the surrounding vegetation can be highlighted through band synthesis. In this way, when constructing a false color image for classification, the area of ​​Spartina alterniflora will be more obvious in the image, thereby improving the recognition accuracy of classification algorithms such as support vector machines.

[0035] Ultimately, this precise image preprocessing and effective use of bands provide high-quality input data for establishing a high-precision chlorophyll content prediction model for invasive plants, ensuring the accuracy and reliability of the prediction results. With this approach, managers can more effectively monitor and manage invasive plants and take appropriate control measures.

[0036] Preferably, when constructing a model, the random forest algorithm randomly selects a subset from all features, and performs node branching in the corresponding features based on the principle of minimum Gini coefficient to form a decision tree.

[0037] Random forest is an ensemble learning technique that improves the accuracy and stability of the overall prediction by building multiple decision trees and merging their prediction results. In a random forest, each decision tree is built independently, and randomness is introduced during the building process, making each tree different, thereby reducing the risk of overfitting of the model and improving generalization ability.

[0038] In the construction process of random forest, instead of using all the features to train each tree, a subset is randomly selected from the total set of features. This method of randomly selecting features not only increases the diversity of the model, but also helps improve the overall model's predictive power for unknown data.

[0039] The Gini coefficient is an indicator of the impurity of a data set. The smaller its value, the lower the impurity of the data set. In the process of building each decision tree, the random forest algorithm will find the best split point in the selected feature subset. This split point is selected based on the principle of minimizing the Gini coefficient. This means that the algorithm will select a feature and its threshold that can minimize the impurity to split the data set.

[0040] This segmentation strategy ensures that each segmentation divides different categories as clearly as possible, so that the constructed decision tree shows higher accuracy in classification problems.

[0041] By adopting the random forest algorithm, the invasive plant chlorophyll content prediction model can effectively process high-dimensional feature data and accurately classify and predict complex ecological data. The integration strategy of this method significantly reduces the risk of overfitting and enhances the model's prediction ability for new data.

[0042] Taking Spartina alterniflora and Mikania micrantha as examples, it is assumed that the spectral indices of these two plants, such as NDVI, RVI, and DVI, are obtained through multispectral images. These indices are provided as input features to the random forest algorithm.

[0043] When building each tree, the algorithm may select NDVI and RVI in one tree, and a combination of DVI and RVI in another tree, etc. This random selection of different feature subsets ensures that the model learns data from different perspectives and avoids relying on a single feature or feature combination, thereby improving the robustness and generalization ability of the model.

[0044] In each tree building process, for example, if NDVI and RVI are selected, the algorithm will calculate the Gini coefficient when splitting the data with NDVI or RVI, and select the feature with the smaller Gini coefficient for splitting. This operation ensures that the decision tree makes the clearest classification decision possible at each node.

[0045] Through the method of the present invention, the random forest model can accurately predict the chlorophyll content of invasive plants, provide a scientific basis for management decisions, and guide the control and management strategies of invasive plants.

[0046] Preferably, the spatial distribution prediction of chlorophyll content uses the block basis ratio of the theoretical semivariogram to test the spatial dependence of regional residuals. If the block basis ratio is less than 0.5, the ordinary kriging method is used to perform spatial interpolation of residuals.

[0047] Semivariogram is a function used in geostatistics to describe the relationship between distance and similarity between sample points in spatial data. It can show how the similarity of spatial variables (such as chlorophyll content) changes with increasing spatial distance. The theoretical semivariogram can help determine the spatial structure and continuity of data by analyzing the degree of variation between points with similar geographical locations.

[0048] Nugget-to-Sill Ratio is an indicator that measures the proportion of data in small-scale variation (nugget effect) and total variation (sill). If the nugget-to-sill ratio is less than 0.5, it means that the spatial continuity and dependence of the data are strong, and it is suitable for spatial interpolation.

[0049] Ordinary Kriging is a best unbiased estimation technique based on least squares estimation, which is used to perform spatial interpolation between known data points. This method uses the semivariogram to estimate the spatial association between different locations and uses this association to predict the value of unknown points. Ordinary Kriging assumes that the mean of the entire area is constant but unknown. It uses the value and location information of the known points to predict the value of the unknown points based on spatial correlation.

[0050] By performing Kriging interpolation on the residuals between the predicted and actual chlorophyll content values ​​at known points, a residual spatial distribution map of the entire area can be obtained. This interpolation result can be added to the original prediction result to obtain a more accurate prediction of the spatial distribution of chlorophyll content.

[0051] The spatial accuracy of chlorophyll content prediction can be significantly improved by using the theoretical semivariogram and ordinary kriging method. This method can ensure the continuity and smoothness of the prediction, especially in areas with complex terrain or uneven distribution of data sampling points.

[0052] Take the prediction of chlorophyll content of Spartina alterniflora and Mikania micrantha as an example. Assuming that in a large wetland reserve, the distribution data of these two plants and the preliminary prediction data of chlorophyll content were obtained by drone multispectral remote sensing technology. The measured values ​​of chlorophyll content at some key locations were obtained through field sampling. By analyzing the residuals between these measured points and the predicted values, the theoretical semivariogram function was used to calculate the spatial autocorrelation, and the block-to-base ratio was tested.

[0053] If the block basis ratio is less than 0.5, it indicates that the residuals are strongly correlated in space and are suitable for interpolation using ordinary kriging.

[0054] After ordinary kriging interpolation, a detailed spatial distribution map of the residuals in the entire area can be obtained. By adding this residual map to the preliminary prediction map, the final spatial distribution map of chlorophyll content will be closer to the actual situation.

[0055] Through this method, the spatial distribution of chlorophyll content of Spartina alterniflora and Mikania micrantha in the protected area can be described more accurately, providing more precise data support for ecological monitoring and management.

[0056] Preferably, the accuracy assessment of the prediction model uses the coefficient of determination and the root mean square error as evaluation indicators, wherein the larger the value of the coefficient of determination and the smaller the value of the root mean square error, the better the prediction accuracy of the model.

[0057] Coefficient of determination (R 2 ), also known as R-squared, is a statistical indicator that measures the predictive ability of a regression model. It indicates the degree of correlation between the model's predicted value and the actual value, and the value ranges from 0 to 1. 2 The closer the value is to 1, the closer the variation explained by the model is to the total variation in the data, and the higher the accuracy and reliability of the prediction.

[0058] The root mean square error (RMSE) is the square root of the average of the sum of squares of the deviations between the predicted values ​​and the actual observed values. It is a common method for evaluating the accuracy of model predictions and provides an absolute measure of the prediction error.

[0059] Using R 2 R and RMSE as evaluation indicators can comprehensively evaluate the prediction performance of the model. 2It provides a measure of the goodness of fit of the model, while RMSE provides a specific measure of the prediction error. The combination of the two can provide a more comprehensive understanding of the predictive effect of the model.

[0060] Suppose that a random forest model based on multispectral data is constructed to predict the chlorophyll content of Spartina alterniflora and Mikania micrantha. Through this model, the predicted chlorophyll content value is obtained. Next, the prediction accuracy of the model needs to be evaluated:

[0061] Coefficient of determination (R 2 ): By comparing the predicted value of the model with the measured value of the field sampling, R 2 If R 2 A value close to 1, say 0.85, indicates that the model is able to explain 85% of the data variation, showing a high model prediction effect.

[0062] Root mean square error (RMSE): Similarly, RMSE is calculated to measure the average error between the predicted values ​​and the measured values. Assuming the RMSE value is 2.5, it means that the average error between the predicted values ​​and the measured values ​​of the model is 2.5 units. Whether this error is acceptable depends on the actual range of variation and ecological importance of chlorophyll content.

[0063] Through these two indicators, the performance of the model in practical applications can be objectively evaluated, so as to decide whether it is necessary to further optimize the model or adjust the prediction strategy to ensure more accurate and effective monitoring and management of the chlorophyll content of invasive plants.

[0064] Preferably, the geographic information system is used to integrate and display the prediction results and their spatial distribution.

[0065] A geographic information system (GIS) is a system used to capture, store, analyze, and manage geographic spatial data. GIS can integrate spatial data from different sources and conduct comprehensive analysis, allowing users to intuitively view, understand, interpret, and visualize data on a map.

[0066] GIS provides a platform to integrate multiple spatial and non-spatial data such as multispectral remote sensing data, outputs of chlorophyll content prediction models, topography, climate conditions, etc. This integration is achieved through the geographic coordinate system, ensuring that all data are aligned in the same spatial reference system.

[0067] GIS can display the spatial distribution of data, such as using color-coded maps to indicate high and low chlorophyll content, or to show the trend of chlorophyll content changing with geographical location.

[0068] In GIS, prediction results can be converted into layers on the map, each layer representing a different data set, such as a chlorophyll content prediction layer. Users can analyze environmental factors that affect chlorophyll content changes by overlaying these layers with layers of other environmental factors (such as soil type, moisture conditions, etc.). This visualization not only enhances the understanding of the data, but also helps researchers and decision makers identify patterns of chlorophyll content changes, assess the impact of invasive plants on ecosystems, and monitor and manage the distribution of invasive plants.

[0069] The present invention can realize accurate monitoring and visual display of the spatial distribution of chlorophyll content of invasive plants through the application of GIS. This integration and display function makes plant ecology research and resource management decision-making more data-driven and efficient.

[0070] Suppose a large wetland area is being studied with Spartina alterniflora and Mikania micrantha as the main invasive plants. In this area, multispectral images were collected using a drone and a prediction model for chlorophyll content was constructed based on these images.

[0071] In GIS, topographic maps and hydrological maps are loaded first, followed by multispectral image data.

[0072] The results of the chlorophyll content prediction model are output as a layer that displays different colors to indicate the high and low chlorophyll content.

[0073] By overlaying this chlorophyll content layer onto the base map, researchers can visually see the distribution of chlorophyll content in different areas of the wetland.

[0074] By comparing the overlap of high chlorophyll content areas with hydrological conditions and soil types, researchers can explore possible environmental factors that influence changes in chlorophyll content.

[0075] The data were further analyzed using GIS tools, such as spatial analysis and statistical tools, such as calculating correlations between chlorophyll content and specific environmental factors.

[0076] Through these functions of GIS, managers can more effectively locate key areas for invasive plant management and develop targeted management measures to reduce the impact of invasive plants on native ecosystems. The application of this approach provides a powerful tool for real-time monitoring and response to changes in ecosystems.

[0077] Preferably, the spatial analysis comprises: using one of a spherical model, an exponential model or a Gaussian model as a theoretical semivariogram of the ordinary kriging method to optimize the accuracy of the residual interpolation.

[0078] Ordinary Kriging is a geostatistical interpolation method based on minimum variance estimation for predicting the value of a spatial continuous variable. It relies on a theoretical semivariogram to quantify the autocorrelation between spatial data points, that is, how the similarity between data points changes as the distance increases.

[0079] The theoretical semivariogram describes the relationship between the variability of spatial data and distance. In ordinary kriging, choosing an appropriate theoretical semivariogram model is key because it directly affects the accuracy of interpolation. Commonly used models include spherical models, exponential models, and Gaussian models, each of which has its specific applicable scenarios and assumptions.

[0080] At small distances, the semivariance value increases rapidly, then gradually stabilizes until it reaches a stable threshold (sill). Applicable to the spatial variation of most natural phenomena.

[0081] The semivariance value increases at a constant rate and theoretically reaches a stable threshold at infinite distance. It is suitable for characterizing some environmental and ecological data, in which spatial autocorrelation is still significant over a long distance.

[0082] It starts at the origin and grows slowly, then accelerates and finally approaches a stable threshold. This model is suitable for spatial processes with very smooth or continuous surfaces.

[0083] The present invention selects a suitable semivariogram model to significantly improve the accuracy and efficiency of kriging interpolation. By accurately describing the spatial dependence of chlorophyll content data, the chlorophyll content of unsampled locations can be more accurately predicted, thereby providing a scientific basis for the management and control of invasive plants.

[0084] Consider a wetland area where Spartina alterniflora and Mikania micrantha are widely distributed. Use drone multispectral remote sensing technology to obtain preliminary predictions of chlorophyll content, and verify the accuracy of the predictions through actual ground sampling. Before performing ordinary kriging interpolation, it is necessary to determine the most appropriate theoretical semivariogram:

[0085] Firstly, the spatial distribution characteristics of chlorophyll content data were analyzed and the semi-variance values ​​between actual sampling points were calculated.

[0086] If the data show that the spatial correlation of chlorophyll content decreases rapidly over short distances and reaches a stable value, a spherical model might be chosen.

[0087] If the correlation is significant even at longer distances, an exponential model may be more appropriate.

[0088] If the data varies very smoothly in space, a Gaussian model would be a better choice.

[0089] Through this method, the GIS tool can perform ordinary kriging interpolation using the selected theoretical semivariogram model, and the resulting spatial distribution map of chlorophyll content will more accurately reflect the actual situation and provide more effective information to support wetland management decisions.

[0090] like Figure 3 As shown, a system for implementing the method for predicting the chlorophyll content of invasive plants comprises:

[0091] An image acquisition unit configured to acquire multispectral images of invasive plants in the area to be tested; this unit is equipped with a multispectral camera system and is responsible for acquiring images of invasive plants in a specific area. Multispectral images can capture light of different wavelengths, which is very useful for analyzing the physiological state of plants. For example, Spartina alterniflora and Mikania micrantha can show their unique reflectance characteristics in different spectral bands, which is helpful for subsequent identification and classification. It can accurately capture images of invasive plants in the area to be tested and provide high-quality raw data for subsequent analysis.

[0092] The image processing unit includes a computer vision module configured to identify and classify invasive plants in the multispectral image; and includes a computer vision module for processing the acquired multispectral image to achieve automatic identification and classification of invasive plants. By applying deep learning or other machine learning algorithms, the system is able to distinguish different types of invasive plants. Automated identification and classification improves the efficiency and accuracy of data processing and lays an accurate foundation for the extraction of spectral information.

[0093] The spectral information processing unit is configured to pre-process the multispectral image of the identified invasive plant part and extract spectral information, the extracted spectral information includes the normalized vegetation index, the ratio vegetation index and the difference vegetation index; pre-process the identified image and extract key spectral information, including the normalized vegetation index (NDVI), the ratio vegetation index (RVI) and the difference vegetation index (DVI). These indices can effectively reflect the health status and physiological activities of plants. Through the calculation of spectral indices, the physiological state of plants, especially their chlorophyll content, can be evaluated more deeply, which is crucial for evaluating the growth intensity and potential ecological impact of invasive plants.

[0094] The model building unit is configured to calculate the spectral index based on the extracted spectral information, and use the spectral index and the chlorophyll content data of the invasive plants to build a prediction model for the chlorophyll content of the invasive plants; based on the extracted spectral information, a random forest algorithm is used to build a prediction model for the chlorophyll content. The random forest improves the stability and accuracy of the prediction by integrating the results of multiple decision trees. Each tree is trained on samples randomly drawn from the overall training set, and randomly selected features are considered during the training process. The random forest model can handle high-dimensional data and has strong resistance to outliers and overfitting, making the model more robust and reliable in practical applications.

[0095] The model evaluation unit is configured to perform accuracy evaluation on the constructed prediction model to ensure that the prediction accuracy meets the preset standard; the constructed prediction model is evaluated for accuracy to ensure that its prediction results meet the preset standard. It usually includes methods such as cross-validation to evaluate the performance of the model on unseen data. Through strict model evaluation, it is ensured that only models with high enough accuracy are used for actual chlorophyll content prediction, which improves the credibility of the prediction.

[0096] The prediction execution unit is configured to use the evaluated prediction model to predict the chlorophyll content of invasive plants in the test area and predict the spatial distribution of chlorophyll content; use the verified model to predict the chlorophyll content of invasive plants in the target area and expand it to spatial distribution prediction. This step is the key link in applying the model to actual situations. It can generate specific chlorophyll content distribution maps to provide a scientific basis for management decisions, especially in the control of invasive plants and ecological restoration plans.

[0097] A spatial analysis unit is configured to perform spatial analysis on the predicted chlorophyll content in a geographic information system and process the prediction residuals using the kriging interpolation method to improve the spatial accuracy of the chlorophyll content prediction, thereby generating a spatial distribution prediction map of the chlorophyll content of invasive plants. The prediction results are spatially analyzed in a geographic information system (GIS) and the prediction residuals are processed using the kriging interpolation method. This method can optimize the spatial distribution prediction of chlorophyll content by utilizing the statistical characteristics of geospatial data. Kriging interpolation can improve the spatial accuracy of chlorophyll content prediction, and the generated distribution map more accurately reflects the actual changes in chlorophyll content in geographic space, which is conducive to more refined ecological monitoring and management.

[0098] The system of the present invention can efficiently and accurately monitor and predict the chlorophyll content of invasive plants such as Spartina alterniflora and Mikania micrantha, providing strong technical support for ecological management. This not only helps the scientific community better understand the impact of invasive plants on local ecosystems, but also provides data support for the formulation of effective control strategies.

[0099] A storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for predicting the chlorophyll content of invasive plants.

[0100] Compared with the prior art, the advantages and beneficial effects of the present invention are:

[0101] By combining advanced multispectral image processing technology and machine learning algorithms, the present invention achieves high-precision prediction of chlorophyll content of invasive plants;

[0102] By using computer vision technology to accurately identify and classify invasive plants in multispectral images, the present invention can effectively improve the accuracy and efficiency of data processing;

[0103] The prediction model constructed using the random forest algorithm not only utilizes the detailed spectral information in the multispectral data, but also automatically adapts to different data characteristics, thereby improving the generalization ability and prediction accuracy of the model;

[0104] In summary, the present invention can achieve rapid and accurate prediction of chlorophyll content at a relatively low cost, providing strong scientific support for the management of invasive plants. BRIEF DESCRIPTION OF THE DRAWINGS

[0105] Figure 1 It is a schematic diagram of the process of the present invention;

[0106] Figure 2 It is a data processing schematic diagram of the present invention;

[0107] Figure 3 It is a structural block diagram of the system of the present invention;

[0108] Figure 4 This is a fitting diagram for predicting chlorophyll content of invasive plants in an embodiment of the present invention;

[0109] Figure 5 This is a prediction diagram of the spatial distribution of chlorophyll content of invasive plants in an embodiment of the present invention. DETAILED DESCRIPTION

[0110] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is obvious that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0111] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0112] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification, and should not be interpreted in an idealized or overly rigid manner.

[0113] In the case of using expressions such as "at least one of A, B, and C, etc.", it should generally be interpreted in accordance with the meaning of the expression generally understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.). In the case of using expressions such as "at least one of A, B, or C, etc.", it should generally be interpreted in accordance with the meaning of the expression generally understood by those skilled in the art (for example, "a system having at least one of A, B, or C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0114] Some block diagrams and / or flow charts are shown in the accompanying drawings. It should be understood that some boxes or combinations thereof in the block diagrams and / or flow charts can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that these instructions can create a device for implementing the functions / operations described in these block diagrams and / or flow charts when executed by the processor. The technology of the present disclosure can be implemented in the form of hardware and / or software (including firmware, microcode, etc.). In addition, the technology of the present disclosure can take the form of a computer program product on a computer-readable storage medium storing instructions, which can be used by an instruction execution system or used in conjunction with an instruction execution system.

[0115] Use drones to take multispectral images of invasive plants in the area to be tested, and use the support vector machine classification algorithm to classify the multispectral image data set of the best spectral band combination of invasive plants and non-invasive plants. Specifically, select the bands with better discrimination between invasive plants and other objects (taking Spartina alterniflora as an example, the red edge band, near infrared 1 band and near infrared 2 band) as the input bands of remote sensing data for the support vector machine classification method;

[0116] Spartina alterniflora samples and non-invasive plant features were selected for support vector machine classification. The support vector machine classification method was used to supervise the classification of invasive plants and non-invasive plants to obtain the distribution map of invasive plants without other land types.

[0117] The multispectral data of invasive plant samples were obtained and preprocessed. The spectral images were stitched and cropped and synthesized using ArcGIS bands to extract the spectral information in the multispectral images. The spectral index normalized vegetation index, ratio vegetation index, and difference vegetation index were calculated based on the spectral information. The specific calculation formula is as follows:

[0118]

[0119] DVI=ρ NIR -ρ Red

[0120] The chlorophyll content prediction model of invasive plants was constructed by collecting the chlorophyll content data of invasive plants and preprocessing the multispectral data. The model accuracy of the chlorophyll content prediction model of invasive plants was evaluated. If the prediction accuracy reached the expected standard, the model was feasible, and the constructed chlorophyll content prediction model of invasive plants was used to predict the chlorophyll content of invasive plants in the test area.

[0121] The preprocessed UAV multispectral images were substituted into the constructed prediction model, and ArcGIS software was used to predict the spatial distribution of chlorophyll content of invasive plants, which was then compared and analyzed with the actual conditions in the sampling area.

[0122] The specific method for collecting the chlorophyll content of the invasive plants required for constructing the model of the present invention is as follows: selecting the canopy leaves of Spartina alterniflora at the sampling point, testing them 5 times with a chlorophyll meter, taking the average value as the chlorophyll content of the sampling point and recording it in a register;

[0123] The present invention uses a random forest algorithm to construct a prediction model, selects N training sets from all data sets, and the data sets are divided into a training set and a validation set in a ratio of 7:3. Then, based on these selected data, a decision tree is created for each training set, and finally a forest consisting of n decision trees is formed. During the growth of the decision tree, m features (m≤M) are randomly selected from all M features, and the optimal attribute is selected from the m features according to the minimum Gini coefficient principle for node branching; finally, the n decision trees use voting to determine the category of the predicted sample based on the prediction results. The out-of-bag data estimation method in the random forest is used to analyze the importance of existing vegetation indices and chlorophyll content of invasive plants and sort them. The top vegetation indices are sequentially combined into a random forest model as decision trees for modeling and estimation.

[0124] The steps of constructing the random forest model of the present invention are as follows: first, a cost-sensitive random forest model is constructed, and the cost Gini index CGini is introduced to assign more reasonable weights to the initial feature variables to obtain cost feature variables, highlighting the strong and weak relationships between features, such as Figure 2 As shown. In this step, in order to ensure that the strong and weak relationships between features are highlighted when the random forest algorithm selects features, the cost Gini index is introduced and the cost feature variable is obtained. It effectively reduces the impact of the lack of strong and weak relationships between unbalanced data features and improves the overall recognition accuracy of minority class samples. Specifically, the cost Gini index CGini calculation formula is:

[0125]

[0126] Among them, r i is the initial feature variable in the data set, L i is the characteristic coefficient, n is the number of initial characteristic variables, and r i Take the inverse to get, q i For each initial variable weight.

[0127] Cost feature variable p i It is obtained by multiplying the cost Gini index by the initial characteristic variable. The calculation formula is:

[0128] p i =CGini(i) * r i

[0129] Prediction result y i The calculation formula is as follows:

[0130]

[0131] Among them, p i is the cost feature variable, w is p i The weight coefficient of the initial value defaults to 0.5.

[0132] The model accuracy evaluation of the present invention adopts the determination coefficient R2 and the root mean square error RMSE. The calculation method of the determination coefficient R2 and the root mean square error RMSE is as follows:

[0133]

[0134] Among them, x i and are the measured and predicted values ​​of chlorophyll, is the average value of chlorophyll samples, n is the number of samples; and the larger the determination coefficient R2, the smaller the root mean square error RMSE, indicating that the chlorophyll content prediction model is better.

[0135] like Figure 5As shown in the figure, a spatial distribution prediction map of the chlorophyll content of invasive plants was constructed. First, the predicted value of the prediction model was subtracted from the measured value of regional chlorophyll to obtain the residual of each area; the spatial dependence of the residual was tested; the block basis ratio of the theoretical semivariogram was used to test the spatial dependence of the regional residual. If the block basis ratio is not greater than 0.5, it indicates that the residual has good spatial dependence. Ordinary Kriging interpolation was performed on the residual of the predicted value of invasive plants. During the model fitting process, the model with the smallest error coefficient value was selected from the spherical model, exponential model and Gaussian model as the theoretical semivariogram of the ordinary Kriging method in this area; the predicted value and the Kriging prediction result were superimposed to obtain the spatial distribution map of lnSOC, and then the value was inversely transformed to obtain the spatial distribution prediction map of the chlorophyll content of invasive plants.

[0136] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0137] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0138] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0139] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0140] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0141] Memory includes non-permanent memory in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0142] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0143] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0144] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for predicting chlorophyll content of invasive plants, characterized in that: The following steps are involved: Obtain multispectral images of invasive plants in the area to be tested; Using computer vision technology to identify and classify invasive plants in the multispectral image; Preprocessing the multispectral image of the identified invasive plant part to extract spectral information, the spectral information includes: normalized vegetation index, ratio vegetation index and difference vegetation index; Calculating a spectral index according to the spectral information, and constructing a prediction model for the chlorophyll content of invasive plants based on the spectral index and the chlorophyll content data of the invasive plants, wherein the prediction model adopts a random forest algorithm; The prediction model is evaluated for accuracy. Once the prediction accuracy reaches the preset standard, it is used to predict the chlorophyll content of invasive plants in the test area and the spatial distribution of chlorophyll content. The predicted chlorophyll content was spatially analyzed in the geographic information system, and the prediction residuals were processed using the Kriging interpolation method to improve the spatial accuracy of the chlorophyll content prediction and obtain the spatial distribution prediction map of the chlorophyll content of invasive plants.

2. The method for predicting chlorophyll content of invasive plants according to claim 1, characterized in that: The identification and classification used the support vector machine classification algorithm to distinguish invasive plants from non-invasive plants.

3. The method for predicting chlorophyll content of invasive plants according to claim 1, characterized in that: The multispectral image preprocessing includes: splicing, cropping and band synthesis processing to generate a false color multispectral image for classification.

4. The method for predicting chlorophyll content of invasive plants according to claim 1, characterized in that: When building a model, the random forest algorithm randomly selects a subset from all features and branches nodes in the corresponding features based on the principle of minimum Gini coefficient to form a decision tree.

5. The method for predicting chlorophyll content of invasive plants according to claim 1, characterized in that: The spatial distribution prediction of chlorophyll content uses the block-basis ratio of the theoretical semivariogram to test the spatial dependence of regional residuals. If the block-basis ratio is less than 0.5, the ordinary kriging method is used to perform spatial interpolation of residuals.

6. The method for predicting chlorophyll content of invasive plants according to claim 1, characterized in that: The accuracy evaluation of the prediction model uses the determination coefficient and the root mean square error as evaluation indicators, wherein the larger the value of the determination coefficient and the smaller the value of the root mean square error, the better the prediction accuracy of the model.

7. The method for predicting chlorophyll content of invasive plants according to claim 1, characterized in that: The GIS is used to integrate and display the prediction results and their spatial distribution.

8. The method for predicting chlorophyll content of invasive plants according to claim 1, characterized in that: The spatial analysis includes: using one of a spherical model, an exponential model or a Gaussian model as a theoretical semivariogram of an ordinary kriging method to optimize the accuracy of residual interpolation.

9. A system for implementing the method for predicting the chlorophyll content of invasive plants according to any one of claims 1 to 8, characterized in that: include: an image acquisition unit configured to acquire a multispectral image of invasive plants in a detected area; an image processing unit, including a computer vision module, configured to identify and classify invasive plants in the multispectral image; A spectral information processing unit is configured to pre-process the multispectral image of the identified invasive plant part and extract spectral information, wherein the extracted spectral information includes a normalized vegetation index, a ratio vegetation index, and a difference vegetation index; A model building unit is configured to calculate a spectral index based on the extracted spectral information, and to build a chlorophyll content prediction model for invasive plants using the spectral index and chlorophyll content data of the invasive plants, wherein the prediction model uses a random forest algorithm; A model evaluation unit, configured to evaluate the accuracy of the constructed prediction model to ensure that the prediction accuracy reaches a preset standard; A prediction execution unit, configured to use the evaluated prediction model to predict the chlorophyll content of the invasive plants in the test area and to predict the spatial distribution of the chlorophyll content; The spatial analysis unit is configured to perform spatial analysis on the predicted chlorophyll content in a geographic information system and process the prediction residual using the Kriging interpolation method to improve the spatial accuracy of the chlorophyll content prediction, thereby generating a spatial distribution prediction map of the chlorophyll content of invasive plants.

10. A storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method for predicting the chlorophyll content of invasive plants according to any one of claims 1 to 8 are implemented.

Images