Carbon dioxide concentration monitoring method based on plant ecological response and related product

Abstract

The invention discloses a carbon dioxide concentration monitoring method based on plant ecological response and a related product, which can be applied to the technical field of carbon dioxide concentration monitoring, and the method comprises the following steps: collecting a target image and a target meteorological parameter corresponding to a target plant; the target plant is sensitive to carbon dioxide concentration; the target meteorological parameters comprise temperature, humidity, wind speed and illumination intensity; determining a target growth index corresponding to the target plant based on the target image; and determining the carbon dioxide concentration of the environment where the target plant is located by using a trained multiple linear regression model in combination with the target growth index and the target meteorological parameter. Thus, the carbon dioxide concentration in the area where the target plant is located can be accurately predicted for a long time only by observing the growth condition of the plant by utilizing the trained multiple linear regression model, and the monitoring cost is reduced.

Description

Technical Field

[0001] The present application relates to the technical field of carbon dioxide concentration monitoring, and in particular, to a carbon dioxide concentration monitoring method and related products based on plant ecological responses. Background Art

[0002] Long-term and continuous monitoring of the carbon dioxide concentration in the environment helps to better understand and manage the impacts brought about by climate change, providing strong support for ecological environment protection.

[0003] The existing methods for obtaining the carbon dioxide concentration mainly involve direct measurement using dedicated measuring instruments. Although this method can provide immediate and accurate data, the cost significantly increases during large-scale continuous monitoring because an instrument is required for measurement each time the carbon dioxide concentration is obtained.

[0004] Therefore, how to reduce the monitoring cost is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0005] Based on the above problems, the present application provides a carbon dioxide concentration monitoring method and related products based on plant ecological responses. By using a trained multiple linear regression model, the carbon dioxide concentration in the area where the target plant is located can be predicted long-term and accurately only by observing the growth of the plant, reducing the monitoring cost.

[0006] In a first aspect, a carbon dioxide concentration monitoring method based on plant ecological responses provided by an embodiment of the present application includes:

[0007] Collecting a target image corresponding to the target plant and target meteorological parameters; the target plant is sensitive to the carbon dioxide concentration; the target meteorological parameters include: temperature, humidity, wind speed, and light intensity;

[0008] Determining a target growth index corresponding to the target plant based on the target image;

[0009] Combining the target growth index and the target meteorological parameters, and using the trained multiple linear regression model to determine the carbon dioxide concentration in the environment where the target plant is located.

[0010] Optionally, the target growth index includes: leaf area index, leaf color index, plant height, new leaf growth rate, and branch expansion rate;

[0011] The determining the target growth index corresponding to the target plant based on the target image includes:

[0012] Based on the target image, use open-source image processing software to determine the leaf area of the target plant, and combine the number of leaves to determine the area index corresponding to the target plant;

[0013] Based on the target image, determine the leaf color index corresponding to the target plant through the red light and near-infrared light reflectance of the leaves;

[0014] Based on the reference object in the target image, determine the plant height corresponding to the target plant;

[0015] Based on the new leaves in the target image, determine the new leaf growth rate corresponding to the target plant;

[0016] Based on the branch extension length in the target image, determine the branch expansion speed corresponding to the target plant.

[0017] Optionally, the multiple linear regression model is trained by the following method:

[0018] Select a training plant sample set; the plants in the training plant sample set are located in the target area and are sensitive to carbon dioxide concentration; the soil type and historical pollution situation of the target area are known;

[0019] Collect the training images, training meteorological parameters, and training carbon dioxide concentrations corresponding to each plant in the training plant sample set at preset time intervals;

[0020] Based on the training images, determine the training growth indicators corresponding to each plant in the training plant sample set;

[0021] Train a multiple linear regression model based on the training growth indicators, the training meteorological parameters, and the training carbon dioxide concentrations.

[0022] Optionally, before determining the training growth indicators corresponding to each plant in the training plant sample set based on the training images, it further includes:

[0023] Remove the noise in the training images through a filter and perform color correction on the training images after removing the noise;

[0024] Through image segmentation technology, separate each plant in the training plant sample set from the corresponding training image, and label each part of each plant in the training plant sample set.

[0025] Optionally, the method further includes:

[0026] Perform standardization processing on each index in the training growth indicators through a data standardization formula;

[0027] Perform principal component analysis on each standardized index in combination with the first preset threshold to determine the key indexes in the growth indexes for training.

[0028] Optionally, training the multiple linear regression model based on the growth indexes for training, the meteorological parameters for training, and the carbon dioxide concentration for training includes:

[0029] Divide all the plants in the plant sample set for training into a training set and a test set;

[0030] According to the cross-validation method, train the multiple linear regression model through the key indexes corresponding to the plants in the training set, the meteorological parameters for training, and the carbon dioxide concentration for training.

[0031] Optionally, the method further includes:

[0032] Verify the trained multiple linear regression model based on the key indexes corresponding to the plants in the test set to determine the prediction error;

[0033] If the prediction error is greater than the second preset threshold, adjust the parameters of the multiple linear regression model according to the prediction error, and re-train and verify the multiple linear regression model with the adjusted parameters.

[0034] In a second aspect, an embodiment of the present application provides a carbon dioxide concentration monitoring device based on plant ecological response, including:

[0035] An acquisition module for acquiring a target image and target meteorological parameters corresponding to a target plant; the target plant is sensitive to the carbon dioxide concentration; the target meteorological parameters include: temperature, humidity, wind speed, and light intensity;

[0036] A first determination module for determining the target growth indexes corresponding to the target plant based on the target image;

[0037] A second determination module for determining the carbon dioxide concentration in the environment where the target plant is located by combining the target growth indexes and the target meteorological parameters and using the trained multiple linear regression model.

[0038] In a third aspect, an embodiment of the present application provides a carbon dioxide concentration monitoring device based on plant ecological response, including:

[0039] A memory for storing a computer program;

[0040] A processor for implementing the steps of the carbon dioxide concentration monitoring method based on plant ecological response as described above when executing the computer program.

[0041] Fourthly, an embodiment of the present application provides a readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the carbon dioxide concentration monitoring method based on plant ecological response as described above are implemented.

[0042] As can be seen from the above technical solutions, compared with the prior art, the present application has the following advantages:

[0043] The present application first collects a target image and target meteorological parameters corresponding to a target plant. Among them, the target plant is sensitive to the carbon dioxide concentration; the target meteorological parameters include: temperature, humidity, wind speed, and light intensity. Then, based on the target image, the target growth index corresponding to the target plant is determined. Finally, by combining the target growth index and the target meteorological parameters, the carbon dioxide concentration in the environment where the target plant is located is determined using the trained multiple linear regression model. In this way, using the trained multiple linear regression model, the carbon dioxide concentration in the area where the target plant is located can be predicted long-term and accurately only by observing the growth of the plant, reducing the monitoring cost. Description of the Drawings

[0044] Figure 1 It is a flowchart of a carbon dioxide concentration monitoring method based on plant ecological response provided by an embodiment of the present application;

[0045] Figure 2 It is a flowchart of a training method for a multiple linear regression model provided by an embodiment of the present application;

[0046] Figure 3 It is a schematic structural diagram of a target selection provided by an embodiment of the present application;

[0047] Figure 4 It is a schematic structural diagram of a carbon dioxide concentration monitoring device based on plant ecological response provided by an embodiment of the present application. Detailed Embodiments

[0048] As described above, the existing carbon dioxide concentration monitoring methods have the problem of high monitoring costs. Specifically, the existing methods for obtaining the carbon dioxide concentration mainly rely on direct measurement using special measuring instruments. However, since the instrument needs to be used for measurement each time the carbon dioxide concentration is obtained, the cost increases significantly during large-scale continuous monitoring. This is not only reflected in the high costs of equipment purchase and maintenance, but also involves the operation and maintenance work of professional personnel, increasing the overall labor cost and operation complexity. In addition, due to the limitations of the number of devices and their installation locations, it is difficult to achieve comprehensive and continuous monitoring over a large range, thus affecting the integrity and continuity of the data. Therefore, the existing direct measurement methods face multiple challenges such as high cost, complex operation, and limited monitoring range in practical applications, restricting their effective application in a wide area.

[0049] To solve the above problems, an embodiment of the present application provides a method for monitoring carbon dioxide concentration based on plant ecological response. This method first collects a target image and target meteorological parameters corresponding to the target plant. Among them, the target plant is sensitive to carbon dioxide concentration; the target meteorological parameters include: temperature, humidity, wind speed, and light intensity. Then, based on the target image, the target growth index corresponding to the target plant is determined. Finally, by combining the target growth index and the target meteorological parameters, the carbon dioxide concentration in the environment where the target plant is located is determined using a trained multiple linear regression model.

[0050] In this way, using the trained multiple linear regression model, the carbon dioxide concentration in the area where the target plant is located can be predicted long-term and accurately only by observing the growth of the plant, reducing the monitoring cost.

[0051] It should be noted that a method for monitoring carbon dioxide concentration based on plant ecological response and related products provided by an embodiment of the present application can be applied to the technical field of carbon dioxide concentration monitoring. The above is only an example and does not limit the application field of a method for monitoring carbon dioxide concentration based on plant ecological response and related products provided by the present application.

[0052] In order to make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0053] Figure 1 This is a flowchart of a method for monitoring carbon dioxide concentration based on plant ecological response provided by an embodiment of the present application. In combination with Figure 1 As shown, a method for monitoring carbon dioxide concentration based on plant ecological response provided by an embodiment of the present application may include:

[0054] S101: Collect a target image and target meteorological parameters corresponding to the target plant; the target plant is sensitive to carbon dioxide concentration; the target meteorological parameters include: temperature, humidity, wind speed, and light intensity.

[0055] In practical applications, plants are very sensitive to the concentration of carbon dioxide because the CO 2 concentration will interfere with the photosynthesis and other physiological processes of plants. Therefore, the CO 2 concentration monitoring method based on plant ecological response is to study the response of plants to CO 2An important means of responding to concentration, using chemical reactions or biological indicators in plants to monitor CO 2 concentration. For example, some plants produce a compound called "stress hormone" when the CO 2 concentration is high. This hormone can affect the growth and physiological processes of plants. By detecting the levels of these hormones, the response of plants to CO 2 concentration can be understood. Based on this, in the embodiments of the present application, the target plant can be photographed first through smart glasses or mobile phones with high-resolution cameras to ensure that the details of the plant can be clearly captured, the acquisition of the target image corresponding to the target plant is completed, and the plant name, shooting date, etc. are recorded. At the same time, the target meteorological parameters corresponding to the target plant need to be obtained synchronously. For example, a small meteorological sensor is used to measure humidity, temperature, wind speed, light intensity, etc. It can be understood that the target plant should be within the training plant sample set or have the same growth environment as a certain plant in the training plant sample set. For example, in the same climate zone, the same ecosystem, the same plant, the same soil type, and historical pollution conditions, etc.

[0056] S102: Determine the target growth index corresponding to the target plant based on the target image.

[0057] In practical applications, the carbon dioxide concentration in the environment will affect the growth of plants, and the carbon dioxide concentration in the environment can be determined by the growth of plants. The target growth index is a quantitative representation of the growth of the target plant. It can be understood that the growth of the target plant can be determined through the characteristics such as the shape and color of the target plant shown in the target image, and then the target growth index corresponding to the target plant can be quantified.

[0058] In addition, since the methods for determining the target growth index are not the same, the embodiments of the present application can illustrate a possible determination method.

[0059] In one case, the target growth index includes: area index, leaf color index, plant height, new leaf growth rate, and branch expansion speed;

[0060] The determining the target growth index corresponding to the target plant based on the target image includes:

[0061] Based on the target image, use open-source image processing software to determine the leaf area of the target plant, and combine the number of leaves to determine the area index corresponding to the target plant;

[0062] Based on the target image, determine the leaf color index corresponding to the target plant through the red light and near-infrared light reflectance of the leaves;

[0063] Determine the plant height corresponding to the target plant based on the reference object in the target image;

[0064] Determine the new leaf growth rate corresponding to the target plant based on the new leaves in the target image;

[0065] Determine the branch expansion speed corresponding to the target plant based on the branch extension length in the target image.

[0066] In practical applications, according to the target plant shown in the target image, the leaf area of the target plant can be calculated by OpenCV or ImageJ, and then the total area index can be calculated by combining the number of leaves. Further, the leaf color index can be calculated as the ratio of the difference between the red light and near-infrared light reflectance of the leaf to the sum of the two as the quantified leaf color index. Further, the plant height of the target plant can be calibrated by the reference object in the target image or the known height information. Further, the new leaf growth rate is obtained by recording the time and number of new leaves appearing, and then calculating the growth rate per unit time (by comparing the target image with the previous captured image and dividing the number of new leaves appearing by the time difference between the two image captures). Further, the branch expansion speed is determined by recording the branch extension length and calculating the extension speed per unit time (by comparing the target image with the previous captured image and dividing the branch extension length by the time difference between the two image captures).

[0067] S103: Combine the target growth index and the target meteorological parameters, and use the trained multiple linear regression model to determine the carbon dioxide concentration in the environment where the target plant is located.

[0068] In practical applications, the accuracy of the trained multiple linear regression model in predicting the carbon dioxide concentration can meet the preset requirements. Technical personnel can put forward reasonable accuracy requirements, such as 90%, 95%, etc., and then train the multiple linear regression model based on this requirement. In this way, by inputting the target growth index and target meteorological parameters corresponding to the target plant into the trained multiple linear regression model, the carbon dioxide concentration in the environment where the target plant is located that meets the requirements (the accuracy rate meets the preset requirements) can be predicted.

[0069] In summary, the present application first collects a target image and target meteorological parameters corresponding to a target plant. Among them, the target plant is sensitive to carbon dioxide concentration; the target meteorological parameters include: temperature, humidity, wind speed, and light intensity. Then, based on the target image, the target growth index corresponding to the target plant is determined. Finally, by combining the target growth index and the target meteorological parameters, the carbon dioxide concentration in the environment where the target plant is located is determined using the trained multiple linear regression model. In this way, by using the trained multiple linear regression model, the carbon dioxide concentration in the area where the target plant is located can be predicted long-term and accurately only by observing the growth of the plant, reducing the monitoring cost.

[0070] In addition, since the methods for training the multiple linear regression model are not all the same, the embodiments of the present application can illustrate one possible training method.

[0071] Figure 2 It is a flowchart of a training method for a multiple linear regression model provided by an embodiment of the present application. Combining Figure 2 As shown, the multiple linear regression model is trained through the following method:

[0072] S1: Select a training plant sample set; the plants in the training plant sample set are located in a target area and are sensitive to carbon dioxide concentration; the soil type and historical pollution situation of the target area are known.

[0073] Figure 3 It is a structural schematic diagram of a target selection provided by an embodiment of the present application. Combining Figure 3 As shown, first, select the plants for training the model (target selection). Specifically, the selected plants can be one or more specific types of trees, herbaceous plants, or algae selected in multiple different climate zones and ecosystem types (target areas), and they need to be sensitive to carbon dioxide concentration. The selected plants can be poplar, willow, and pine trees that are sensitive to carbon dioxide concentration; corn, wheat, and soybeans with a fast growth cycle and sensitive to changes in carbon dioxide concentration; spirulina and chlorella that are sensitive to changes in carbon dioxide concentration, etc. Plants with a shorter growth cycle are more convenient for quickly obtaining monitoring data. In addition, the target areas where the selected plants are located can be forests, grasslands, and wetlands in temperate and tropical zones, etc. It should be noted that the soil type and historical pollution situation of the target area need to be recorded in detail to exclude other interfering factors. In addition, for multiple selected plants in a target area, the planting density and cultivation methods of each selected plant need to be the same. In summary, assign a unique identifier and number to each of the selected plants, and record in detail the planting date, growth situation, and other relevant information of each plant to form a training plant sample set to ensure the continuity and integrity of the data.

[0074] S2: Collect the training images, training meteorological parameters, and training carbon dioxide concentrations corresponding to each plant in the training plant sample set at preset time intervals.

[0075] In practical applications, a regular shooting time interval (preset time interval) can be set, such as once a day, once a week, or once a month, depending on the training needs. For all plants in the training plant sample set, each image collection needs to be taken within the same time period to reduce the impact of light changes, ensure that the plants are photographed from multiple angles, including top views, side views, etc., in order to comprehensively record the growth of the plants, ensure that the images are clear, unobstructed, and have uniform light when shooting, avoid shadows and reflections, and record the shooting time and geographical location information during shooting for convenient later data management and analysis. The shooting device is selected as a smart glasses or smartphone with a high-resolution camera to ensure that the obtained training images are clear enough. The collection time of the training meteorological parameters and the training carbon dioxide concentration is synchronized with the training images. Specifically, record the temperature, air humidity, and wind speed at each shooting, and ensure that the data is accurate to one decimal place; further, the light intensity at each shooting can be measured and recorded using an illuminometer; further, multiple measurements can be made at the same location using a portable infrared gas analyzer to reduce errors caused by location differences and accurately record the carbon dioxide concentration in the air at each shooting in real time. Further, the training images, training meteorological parameters, and training carbon dioxide concentration data collected each time are uniformly stored in a database to ensure the integrity and traceability of the data, and add labels to the data collected each time, such as information like date, time, location, shooting device, etc., and regularly back up all the collected data to prevent data loss. In addition, to ensure data quality, the camera of the smart glasses or mobile phone needs to be cleaned regularly and the portable infrared gas analyzer needs to be calibrated to ensure that the collected data is accurate and abnormal-free.

[0076] In addition, since the image preprocessing methods are not all the same, the embodiments of the present application can illustrate one possible image preprocessing method.

[0077] In one case, the method further includes:

[0078] Remove the noise in the training images through a filter and perform color correction on the training images after removing the noise;

[0079] Separate each plant in the training plant sample set from the corresponding training image through image segmentation technology and label each part of each plant in the training plant sample set.

[0080] In practical applications, for the convenience of subsequent processing, it is necessary to convert the training images obtained by shooting into a unified format, ensure that the size of each image is the same, then use a filter to remove the noise in the image, improve the image quality and perform color correction on the image to eliminate the influence brought by light changes. In addition, the plants in each training image can be separated from the background through image segmentation technology, and then different parts of the plants can be marked, such as leaves, stems, etc., for subsequent quantitative processing. Commonly used image segmentation techniques include edge detection, threshold segmentation, etc.

[0081] S3: Determine the training growth indicators corresponding to each plant in the training plant sample set based on the training images.

[0082] In practical applications, the growth situation of plants can be described by several indicators, namely growth indicators, including area index, leaf color index, plant height, new leaf growth rate, and branch expansion rate. In order to facilitate subsequent analysis, it is necessary to quantitatively process the growth situation of each plant in the training plant sample set. Specifically, among the quantified growth indicators, the OpenCV or ImageJ is used to calculate the leaf area, and the total area index is calculated in combination with the number of leaves; the quantification of the leaf color index is obtained by calculating the ratio of the difference between the red light and near-infrared light reflectance of the leaves to the sum of the two; the quantification of the plant height is obtained by calibrating the height of the plant through the reference object in the image or known height information; the quantification of the new leaf growth rate is obtained by recording the time and number of new leaves appearing and calculating the growth rate per unit time; the quantification of the branch expansion rate is obtained by recording the extension length of the branches and calculating the extension speed per unit time. In this way, if the above growth indicators are calculated using the collected training images and the corresponding acquisition time of the images, the quantification results of the training growth indicators corresponding to each plant in the training plant sample set can be obtained.

[0083] In addition, since the processing methods for the training growth indicators are not the same, the embodiments of the present application can illustrate a possible processing method.

[0084] In one case, the method further includes:

[0085] Perform standardization processing on each indicator in the training growth indicators through a data standardization formula;

[0086] Combine the first preset threshold to perform principal component analysis on the standardized indicators to determine the key indicators in the training growth indicators.

[0087] In practical applications, the Z-score formula can be used to perform standardization processing on each indicator to eliminate the dimension difference. The Z-score formula is as follows:

[0088]

[0089] Among them, x is the original variable value, μ is the mean of the variable, σ is the standard deviation of the variable, and Z is the standardized result. Then, principal component analysis (PCA) is performed. A covariance matrix is constructed from each of the standardized training growth indicators, and the principal components are obtained by solving the eigenvalues and eigenvectors of the covariance matrix. Then, the principal components are sorted according to the magnitudes of the eigenvalues, and the principal components with a cumulative explained variance greater than a first preset threshold are selected as key variables (key indicators). It can be understood that the magnitude of the first preset threshold is determined by the technical personnel and can be freely designed, and is generally set to 80%.

[0090] S4: Train a multiple linear regression model based on the training growth indicators, the training meteorological parameters, and the training carbon dioxide concentration.

[0091] In practical applications, a prediction model is established using multiple linear regression on the existing training data (training growth indicators, training meteorological parameters, and training carbon dioxide concentration) to predict the carbon dioxide concentration in the target area, evaluate the performance of the model, calculate the prediction error, and use k-fold cross-validation to further verify the stability and generalization ability of the model to obtain a trained multiple linear regression model.

[0092] In addition, since the methods for training the multiple linear regression model are not all the same, an embodiment of the present application can illustrate one possible training method.

[0093] In one case, the training of the multiple linear regression model based on the training growth indicators, the training meteorological parameters, and the training carbon dioxide concentration includes:

[0094] Dividing all the plants in the training plant sample set into a training set and a test set;

[0095] According to the cross-validation method, train a multiple linear regression model using the key indicators corresponding to the plants in the training set, the training meteorological parameters, and the training carbon dioxide concentration.

[0096] In practical applications, all selected plants in the training plant sample set can be randomly divided into a training set and a test set, and the data volume division can be 7:3 or other values. In this way, the training growth indicators corresponding to each selected plant in the database, the training meteorological parameters and training carbon dioxide concentration corresponding to the training images, and the labels such as date, time, location, and shooting equipment when collecting each training image are also correspondingly divided into two parts, that is, one part corresponds to the training set and the other part corresponds to the test set. Combining the above principal component analysis results, non-critical indicators and their corresponding related data can be removed from the training set and the test set, and then further analysis can be carried out. Further, a prediction model is established and trained on the training set that only includes key indicators and their corresponding data using multiple linear regression. As a cross-validation method, the above training set is randomly divided into k subsets, one subset is selected as the validation set each time, and the remaining subsets are used as the training set. This is repeated k times, and different validation sets are used each time. The results of each validation are summarized, and the average prediction error is calculated, including the Root Mean Square Error (RMSE) and the Mean Absolute Error (MAE), to further verify the stability and generalization ability of the model. In this way, the regression coefficients of the multiple linear regression model are adjusted according to the results of cross-validation to optimize the model performance, and the first preset threshold of the principal component analysis is adjusted according to specific application requirements to ensure the applicability of the model.

[0097] In addition, since the verification methods for the multiple linear regression model are not the same, the embodiments of the present application can illustrate a possible verification method.

[0098] In one case, the method further includes:

[0099] Verifying the trained multiple linear regression model based on the key indicators corresponding to the plants in the test set to determine the prediction error;

[0100] If the prediction error is greater than the second preset threshold, the parameters of the multiple linear regression model are adjusted according to the prediction error, and the multiple linear regression model with adjusted parameters is retrained and verified.

[0101] In practical applications, the optimized model mentioned above is applied to the test set to continue predicting the carbon dioxide concentration, further verifying the effectiveness of the model. Specifically, based on the collected images of each selected plant in the test set at its respective time nodes, the key indicators obtained from the analysis of each collected image, the meteorological parameters at the same time nodes as the collected images, and the corresponding relationship with the carbon dioxide concentration, a one-to-one corresponding data pair is formed. Then, through the optimized multiple linear regression model mentioned above, predictions are continued for each key indicator in the test set to obtain the corresponding predicted carbon dioxide concentration values. Then, these predicted values are compared with the actual carbon dioxide concentration values (obtained by a portable infrared gas analyzer) at the corresponding time nodes to calculate the error values (prediction errors). Then, the prediction performance of the model is evaluated through prediction error evaluation.

[0102] Calculating the prediction error can use the root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R 2 ). The calculation formula for the root mean square error is as follows:

[0103]

[0104] In the formula, yi is the true value of the i-th observation, is the predicted value of the observation, n is the total number of observations. RMSE measures the square root of the average error between the predicted value and the true value, and can better reflect the overall magnitude of the prediction error, especially being more sensitive to larger errors. The unit of RMSE is the same as that of the predicted value, so its practical significance can be directly interpreted. At the same time, according to the error magnitude, it can be divided into: low error, medium error, and high error. For low error, its value is less than or equal to 10% of the fluctuation range of the data itself. If the fluctuation range of the carbon dioxide concentration is 0 to 1000 ppm, then RMSE < 100 ppm can be regarded as low error. For medium error, its value is between 10% and 30% of the data fluctuation range. That is, in the same situation, RMSE between 100 ppm and 300 ppm can be regarded as medium error. For high error, its value is greater than 30% of the data fluctuation range (greater than the second preset threshold). That is, when RMSE > 300 ppm, it is considered high error and the model needs to be further optimized. Of course, the second preset threshold can also be defined as other reasonable values other than 30% of the data fluctuation range or RMSE > 300 ppm.

[0105] The calculation formula for the mean absolute error is as follows:

[0106]

[0107] In the formula, yi is the true value of the i-th observation, is the predicted value of the observation, n is the total number of observations, MAE is the absolute value of the prediction error for all observations. It measures the average absolute error between the predicted value and the true value, without considering the direction of the error and only caring about the magnitude of the error. MAE is more robust than RMSE and is not affected by extreme values. Additionally, similar to the root mean square error mentioned above, the mean absolute error can also be divided into low error, medium error, and high error based on its magnitude. Based on the thresholds given above, correspondingly, it is also less than or equal to 10% of the data's own fluctuation range. When MAE < 100 ppm, it is considered low error; when it is between 10% and 30% of the data fluctuation range and MAE is between 100 ppm and 300 ppm, it is considered medium error; when it is greater than 30% of the data fluctuation range and MAE > 300 ppm, it is considered high error, and the model needs to be further optimized.

[0108] The calculation formula for the coefficient of determination is as follows:

[0109]

[0110] Where yi is the true value of the i-th observation, is the predicted value of the observation, n is the total number of observations, is the average value of the true values of all observations, and R 2 measures the improvement of the model's predicted value relative to the simple average. Its value range is usually between [0, 1]. The closer the value is to 1, the better the model fits, and it is used to evaluate the overall fitting effect of the model, but it does not directly reflect the specific error magnitude. If it is still divided into three levels: low goodness of fit, medium goodness of fit, and high goodness of fit, then high goodness of fit can be set to indicate that the model can well explain the variation in the data and the error is small; medium goodness of fit indicates that the model has a certain explanatory ability but there is still room for improvement; low goodness of fit indicates that the model has poor explanatory ability and needs to be further optimized.

[0111] In summary, if the prediction error is greater than the second preset threshold, analyze the bias and variance problems existing in the model based on the calculated prediction error index, adjust the regression coefficients in the multiple linear regression model, select the optimal principal component combination, or even adjust the cumulative explained variance threshold in the principal component analysis to improve the model's explanatory ability. Then use the adjusted parameters to retrain on the training set and re-evaluate on the test set, calculate the new prediction error, and if the preset effect is still not achieved, continue the above process until verification passes. In addition, the k-fold cross-validation can be used to verify again whether the adjusted model has stable prediction performance and evaluate the generalization ability of the model on an independent validation data set. If the verification passes, output the finally trained and verified adjusted model to obtain the trained multiple linear regression model.

[0112] In addition, for the convenience of users to consult and analyze, the predicted values of the model on the test set can be exported as files, such as in CSV or Excel format, and then the comparison chart of the predicted values and the actual values can be drawn using chart tools such as Matplotlib or Seaborn to visually display the prediction effect of the model.

[0113] Furthermore, the evaluation metrics (prediction errors) calculated when evaluating the model performance can be summarized into a table, and the specific values of each metric can be displayed using icons, such as bar charts or pie charts, for a visual understanding of the model performance.

[0114] Furthermore, the parameters used by the model can be recorded in real time during the training process, including the cumulative explained variance threshold of principal component analysis and the regression coefficients of multiple linear regression, etc., and at the same time, the meaning of each parameter and its impact on the model performance can be explained to help users understand the working principle of the model.

[0115] Furthermore, the relationship chart between the predicted values and the residuals can be drawn to check for systematic biases, mark potential outliers, and analyze their impact on the model performance, and the results of k-fold cross-validation can be displayed to ensure the stable performance of the model on different datasets.

[0116] Furthermore, detailed usage instructions can be provided, including how to import new data, how to run the prediction model, and how to interpret the prediction results, etc., list common problems and their solutions to help users solve the problems that may be encountered during use, and provide technical support contact information for users to seek help when encountering technical problems.

[0117] In summary, all the above output content can be integrated into a detailed report document, including: project overview, which is used to introduce the background, objectives, and research methods of the project; data collection and processing, which is used to describe the data collection process, data processing methods, and their results; model establishment and verification, which is used to introduce in detail the model establishment process, verification methods, and their results; result analysis, which is used to analyze the prediction results and their evaluation metrics, and summarize the advantages and disadvantages of the model; conclusions and suggestions, which are used to put forward conclusions and give suggestions for further improvement.

[0118] In addition, a Web application can be developed, through which users can upload data, view prediction results, and download reports, etc.; a mobile application can be developed, through which users can view prediction results in real time via smartphones for convenient outdoor detection; and a command-line tool can also be provided to facilitate technicians to batch process data and generate reports.

[0119] In summary, the present application first collects the target image and target meteorological parameters corresponding to the target plant. Among them, the target plant is sensitive to carbon dioxide concentration; the target meteorological parameters include: temperature, humidity, wind speed, and light intensity. Then, based on the target image, the target growth index corresponding to the target plant is determined. Finally, by combining the target growth index and the target meteorological parameters, the carbon dioxide concentration in the environment where the target plant is located is determined using the trained multiple linear regression model. In this way, using the trained multiple linear regression model, the carbon dioxide concentration in the area where the target plant is located can be predicted long-term and accurately only by observing the growth of the plant, reducing the monitoring cost.

[0120] Figure 4 FIG. is a schematic structural diagram of a carbon dioxide concentration monitoring device based on plant ecological response provided by an embodiment of the present application. In combination with Figure 4 As shown, the carbon dioxide concentration monitoring device 400 based on plant ecological response includes:

[0121] A collection module 401 for collecting the target image and target meteorological parameters corresponding to the target plant; the target plant is sensitive to carbon dioxide concentration; the target meteorological parameters include: temperature, humidity, wind speed, and light intensity;

[0122] A first determination module 402 for determining the target growth index corresponding to the target plant based on the target image;

[0123] A second determination module 403 for determining the carbon dioxide concentration in the environment where the target plant is located by combining the target growth index and the target meteorological parameters using the trained multiple linear regression model.

[0124] As an implementation manner, regarding how to determine the target growth index, the above target growth index includes: leaf area index, leaf color index, plant height, new leaf growth rate, and branch expansion speed;

[0125] The above first determination module 402 is specifically used for:

[0126] Based on the target image, using open-source image processing software to determine the leaf area of the target plant, and combining the number of leaves to determine the area index corresponding to the target plant;

[0127] Based on the target image, determining the leaf color index corresponding to the target plant through the red light and near-infrared light reflectance of the leaves;

[0128] Determining the plant height corresponding to the target plant based on the reference object in the target image;

[0129] Determining the new leaf growth rate corresponding to the target plant based on the new leaves in the target image;

[0130] Determine the branch expansion speed corresponding to the target plant based on the branch extension length in the target image.

[0131] As an implementation, regarding how to train a multiple linear regression model, the carbon dioxide concentration monitoring device 400 based on plant ecological response further includes: a training module; the training module includes: a selection module, a collection module, a third determination module, and a training sub-module;

[0132] The selection module is used to select a training plant sample set; the plants in the training plant sample set are located in the target area and are sensitive to the carbon dioxide concentration; the soil type and historical pollution situation of the target area are known;

[0133] The collection module is used to collect the training images, training meteorological parameters, and training carbon dioxide concentrations corresponding to each plant in the training plant sample set at preset time intervals.

[0134] The third determination module is used to determine the training growth indicators corresponding to each plant in the training plant sample set based on the training images.

[0135] The training sub-module is used to train a multiple linear regression model based on the training growth indicators, the training meteorological parameters, and the training carbon dioxide concentrations.

[0136] As an implementation, regarding how to preprocess images, the carbon dioxide concentration monitoring device 400 based on plant ecological response further includes: a preprocessing module;

[0137] The preprocessing module is used to remove the noise in the training images through a filter and perform color correction on the training images after removing the noise.

[0138] Separate each plant in the training plant sample set from the corresponding training image through image segmentation technology, and mark each part of each plant in the training plant sample set.

[0139] As an implementation, regarding how to determine key indicators, the carbon dioxide concentration monitoring device 400 based on plant ecological response further includes: a fourth determination module;

[0140] The fourth determination module is used to perform standardization processing on each indicator in the training growth indicators through a data standardization formula;

[0141] Combine the first preset threshold to perform principal component analysis on the standardized indicators to determine the key indicators in the training growth indicators.

[0142] As an implementation, for how to train a multiple linear regression model, the above-mentioned training sub-module is specifically used for:

[0143] Dividing all plants in the training plant sample set into a training set and a test set;

[0144] According to the cross-validation method, training a multiple linear regression model by using the key indicators corresponding to the plants in the training set, the training meteorological parameters, and the training carbon dioxide concentration.

[0145] As an implementation, for how to verify a multiple linear regression model, the above-mentioned carbon dioxide concentration monitoring device 400 based on plant ecological response further includes: a verification module;

[0146] The verification module is used to verify the trained multiple linear regression model based on the key indicators corresponding to the plants in the test set and determine the prediction error;

[0147] If the prediction error is greater than a second preset threshold, adjust the parameters of the multiple linear regression model according to the prediction error, and re-train and verify the multiple linear regression model with the adjusted parameters.

[0148] In summary, the present application first collects a target image and target meteorological parameters corresponding to a target plant. Among them, the target plant is sensitive to carbon dioxide concentration; the target meteorological parameters include: temperature, humidity, wind speed, and light intensity. Then, based on the target image, the target growth index corresponding to the target plant is determined. Finally, combining the target growth index and the target meteorological parameters, the carbon dioxide concentration in the environment where the target plant is located is determined by using the trained multiple linear regression model. In this way, by using the trained multiple linear regression model, the carbon dioxide concentration in the area where the target plant is located can be predicted long-term and accurately only by observing the growth of the plant, reducing the monitoring cost.

[0149] In addition, the present application also provides a carbon dioxide concentration monitoring device based on plant ecological response, including: a memory for storing a computer program; a processor for implementing the steps of the above-mentioned carbon dioxide concentration monitoring method based on plant ecological response when executing the computer program.

[0150] In addition, the present application also provides a readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above-mentioned carbon dioxide concentration monitoring method based on plant ecological response are implemented.

[0151] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for monitoring carbon dioxide concentration based on plant ecological response, characterized in that: The method comprises: Collecting target images and target meteorological parameters corresponding to target plants; the target plants are sensitive to carbon dioxide concentration; the target meteorological parameters include: temperature, humidity, wind speed and light intensity; Determining a target growth index corresponding to the target plant based on the target image; The target growth index and the target meteorological parameter are combined to determine the carbon dioxide concentration of the environment where the target plant is located using a trained multiple linear regression model.

2. The method according to claim 1, characterized in that The target growth indicators include: area index, leaf color index, plant height, new leaf growth rate and branch expansion speed; The determining the target growth index corresponding to the target plant based on the target image includes: Based on the target image, the leaf area of ​​the target plant is determined using open source image processing software, and the area index corresponding to the target plant is determined in combination with the number of leaves; Based on the target image, determining the leaf color index corresponding to the target plant through the red light and near-infrared light reflectance of the leaves; Determining a plant height corresponding to the target plant based on a reference object in the target image; Determining a growth rate of new leaves corresponding to the target plant based on the new leaves in the target image; The branch expansion speed corresponding to the target plant is determined based on the branch extension length in the target image.

3. The method according to claim 1, characterized in that The multivariate linear regression model is trained by the following method: A training plant sample set is selected; the plants in the training plant sample set are located in a target area and are sensitive to carbon dioxide concentration; the soil type and historical pollution conditions of the target area are known; Collecting training images, training meteorological parameters, and training carbon dioxide concentrations corresponding to each plant in the training plant sample set at preset time intervals; Determining, based on the training image, a training growth indicator corresponding to each plant in the training plant sample set; A multivariate linear regression model is trained based on the training growth index, the training meteorological parameter and the training carbon dioxide concentration.

4. The method according to claim 3, characterized in that Before determining the training growth indicator corresponding to each plant in the training plant sample set based on the training image, the method further includes: Removing noise from the training image by using a filter, and performing color correction on the training image after the noise is removed; Each plant in the training plant sample set is separated from the corresponding training image by using image segmentation technology, and each part of each plant in the training plant sample set is marked.

5. The method according to claim 3, characterized in that: The method further comprises: Standardizing each of the training growth indicators using a data standardization formula; The principal component analysis is performed on each index after the standardization process in combination with the first preset threshold value to determine the key index in the training growth index.

6. The method according to claim 5, characterized in that The multivariate linear regression model is trained based on the training growth index, the training meteorological parameter and the training carbon dioxide concentration, including: Dividing all plants in the training plant sample set into a training set and a test set; According to the cross-validation method, a multivariate linear regression model is trained using the key indicators corresponding to the plants in the training set, the training meteorological parameters and the training carbon dioxide concentration.

7. The method according to claim 6, characterized in that The method further comprises: Verifying the trained multivariate linear regression model based on the key indicators corresponding to the plants in the test set to determine the prediction error; If the prediction error is greater than a second preset threshold, the parameters of the multivariate linear regression model are adjusted according to the prediction error, and the multivariate linear regression model with adjusted parameters is retrained and verified.

8. A carbon dioxide concentration monitoring device based on plant ecological response, characterized in that: include: A collection module, used for collecting target images and target meteorological parameters corresponding to target plants; The target plant is sensitive to carbon dioxide concentration; The target meteorological parameters include: temperature, humidity, wind speed and light intensity; A first determination module, configured to determine a target growth index corresponding to the target plant based on the target image; The second determination module is used to determine the carbon dioxide concentration of the environment where the target plant is located by combining the target growth index and the target meteorological parameter using a trained multiple linear regression model.

9. A carbon dioxide concentration monitoring device based on plant ecological response, characterized in that: include: Memory for storing computer programs; A processor is used to implement the steps of the carbon dioxide concentration monitoring method based on plant ecological response as described in any one of claims 1 to 7 when executing the computer program.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the carbon dioxide concentration monitoring method based on plant ecological response as claimed in any one of claims 1 to 7 are implemented.

Images