Crop growth detection method and system based on 5G transmission

Abstract

The invention relates to a crop growth detection method based on 5G transmission. The method comprises the following steps: S1, acquiring a hyperspectral image: acquiring the hyperspectral image of a wheat field in real time; s2, establishing a regional customized model: establishing a regional customized characteristic wave band database, selecting a random forest, a gradient boosting tree and a support vector machine as base classifiers, training each base classifier, and inputting prediction results of the plurality of base classifiers as new characteristics into a logic regression model for final prediction, establishing a growth condition monitoring model; s3, data enhancement and model optimization: enhancing existing training data, generating wheat growth data under different environmental conditions by using a generative adversarial network, and mixing simulation data and actual data for use; and S4, growth stage identification and parameter adjustment: identifying the growth stage of the wheat by using the growth condition monitoring model in the S2, and adjusting monitoring parameters and management measures according to the requirements of the growth stage of the wheat.

Description

Technical Field

[0001] The present invention relates to the technical field of crop growth detection, and in particular to a crop growth detection method and system based on 5G transmission. Background Art

[0002] 5G technology, with its high speed, low latency, and large number of connections, can meet the needs of large numbers of devices online simultaneously and enable real-time data transmission during agricultural production, bringing new changes to agriculture. 5G-based crop growth monitoring technology can monitor wheat growth parameters (such as soil moisture, temperature, and light intensity) and growth status in real time, helping farmers to more scientifically implement planting, fertilization, and irrigation, thereby improving production efficiency.

[0003] In existing technologies, the generalization ability of crop growth monitoring models relies on the diversity of training data. If the training data lacks wheat growth data under certain specific environments or conditions, the model may inaccurately identify growth conditions under these conditions. Therefore, the development of a crop growth detection method and system based on 5G transmission is particularly important. Summary of the Invention

[0004] The purpose of the present invention is to solve the shortcomings of the existing technology that the training data lacks wheat growth data under certain specific environments or conditions, and the model may have errors when identifying the growth conditions under these environments. A crop growth detection method and system based on 5G transmission are provided.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A crop growth detection method based on 5G transmission includes the following steps:

[0007] S1: Hyperspectral image acquisition: Sensor nodes deployed at agricultural production sites and 5G wireless communication networks are used to collect hyperspectral images of wheat fields in real time to obtain spectral information. UAVs are used to capture visible light images to obtain crop growth information.

[0008] S2: Establishing a regionally customized model: Based on environmental factors such as climate conditions and soil types in different regions, a regionally customized feature band database is established. Random forests, gradient boosting trees, and support vector machines are selected as base classifiers. Each base classifier is trained and cross-validation is used to evaluate its performance. The prediction results of multiple base classifiers are used as new features and input into a logistic regression model for final prediction. This establishes a growth status monitoring model.

[0009] S3: Data Augmentation and Model Optimization: By augmenting existing training data and using generative adversarial networks to generate wheat growth data under different environmental conditions, simulated data is mixed with real data to increase the diversity of training data. During training, samples are randomly selected from the augmented dataset to improve the generalization ability of the model.

[0010] S4: Growth stage identification and parameter adjustment: Utilizing the growth status monitoring model in S2, the growth stage of wheat is automatically identified by analyzing the spectral features and growth parameters in hyperspectral and visible light images. Monitoring parameters and management measures, such as fertilization plans, irrigation strategies, and pest and disease control, are adjusted according to the growth stage requirements of the wheat.

[0011] S5: Data visualization and decision support: Wheat growth status information is visualized through an interactive front-end. The interactive front-end uses the WebSocket protocol to transmit data with the hyperspectral camera and cloud server, realizing real-time data update and display. Based on the growth status information, wheat planting strategies and management measures are formulated. For example, fertilization plans are adjusted according to the nutritional status of wheat, and pest and disease control plans are formulated according to the health status.

[0012] The above scheme further includes:

[0013] Furthermore, in S1, the sensor node is responsible for real-time monitoring and collection of various information about the farmland environment and crops, and uses a hyperspectral imaging sensor to capture the reflected light information of crops at different wavelengths, thereby generating a hyperspectral image. Through the 5G network, the hyperspectral image data collected by the sensor node is transmitted to the cloud server in real time.

[0014] Furthermore, in S2, the specific steps of establishing a regional customized characteristic band database based on environmental factors in different regions are as follows:

[0015] Feature extraction: Using hyperspectral imaging technology, the spectral characteristics of wheat fields at different growth stages are extracted. These characteristics are expressed as reflectance or absorption in specific bands.

[0016] Correlation analysis: The extracted spectral features were correlated with the growth conditions of wheat (such as plant height, leaf area index, biomass, etc.). The Pearson correlation coefficient was calculated using statistical software (such as SPSS, R language, etc.) to screen out the characteristic bands that were highly correlated with the growth conditions of wheat. The Pearson correlation coefficient calculation formula is: Among them, r is the correlation coefficient, x i and y i are the spectral characteristic value and growth status value of the i-th sample respectively, and are the means of spectral characteristic values and growth status values, respectively, and n is the number of samples;

[0017] Determine characteristic bands: Based on the results of the correlation analysis, determine the most relevant characteristic bands for each study area. The most relevant characteristic bands constitute a regionally customized characteristic band database.

[0018] Establish a database: Organize the determined characteristic bands and their corresponding spectral information into a database format. The database should include information such as band number, wavelength range, reflectance / absorbance value, and metadata such as the site and time of data collection.

[0019] Furthermore, in S2, the specific steps of establishing the growth status monitoring model are:

[0020] Feature extraction: Using hyperspectral imaging technology, we extract the spectral characteristics of the wheat field and select the most relevant characteristic bands as model input based on a regionally customized characteristic band database.

[0021] Base classifier training: Using historical wheat growth data as the training set, random forests, gradient boosting trees, and support vector machines are trained. Each base classifier learns how to predict wheat growth conditions from the input hyperspectral image features; Random Forest: Construct multiple decision trees as weak classifiers. Each decision tree is trained on a random subset of the training data. The prediction results of all decision trees are combined through a voting mechanism to obtain the final prediction; Gradient Boosting Tree: Initialize as a weak prediction model (such as a decision tree stump), and then gradually train new weak classifiers to correct the errors of the previous model. Finally, the prediction results of all weak classifiers are weighted and summed to obtain the final prediction; Support Vector Machine: Map the input data to a high-dimensional feature space and find a hyperplane to maximize the interval between two types of samples. The optimal hyperplane and classification decision function are obtained by solving a convex quadratic programming problem;

[0022] Cross-validation: Divide the dataset into k non-overlapping subsets, select k-1 subsets each time as training sets, and use the remaining subset as the test set for model evaluation. Repeat this process k times, selecting a different test set each time. The average of the k evaluation results is taken as the accuracy of each base classifier.

[0023] Weight allocation: weights are allocated based on the accuracy of each base classifier on the validation set. The base classifier with higher accuracy has a greater weight. The weight calculation formula is Among them, w i is the weight of the i-th base classifier, n is the total number of base classifiers, k i is the accuracy of the i-th base classifier;

[0024] Logistic regression model fusion: The prediction results of each base classifier are input into a logistic regression model as new features. The output of the logistic regression model is the final prediction result, which represents the growth status of wheat. The formula of the logistic regression model is where X i is the prediction result of the i-th base classifier, β i is the corresponding coefficient, and β0 is the intercept term.

[0025] Furthermore, in S3, the specific steps of enhancing the existing training data are as follows:

[0026] Basic image enhancement: Perform basic image enhancement operations such as rotation, scaling, and flipping on hyperspectral images to increase data diversity; Rotation: Rotate the image around the center point by a certain angle (such as 90 degrees, 180 degrees, 270 degrees); Scaling: Change the size of the image, usually enlarging or reducing it proportionally; Flip: Flip the image horizontally or vertically;

[0027] Color space conversion: convert the hyperspectral image from RGB color space to other color spaces (such as HSV, Lab, etc.), perform color adjustment, and then convert it back to RGB color space;

[0028] Noise addition: Add noise (such as Gaussian noise, salt and pepper noise, etc.) to the hyperspectral image to simulate the noise interference in the real environment.

[0029] Furthermore, in S3, the specific steps of using the generative adversarial network to generate wheat growth data under different environmental conditions are as follows:

[0030] Define the GAN network structure: Generator: Responsible for generating new data samples similar to real data. During the training process, the goal of the generator is to generate real data to deceive the discriminator; Input: Random noise vector z; Output: Synthesized image data G(z);

[0031] Discriminator: Responsible for distinguishing whether the input data is real data or fake data generated by the generator. During the training process, the discriminator's goal is to accurately distinguish real data from generated data; Input: real image data x or image data G(z) synthesized by the generator; Output: a probability value D(x), indicating the probability that the input image is a real image;

[0032] Define the loss function:

[0033] The training goal of the generative adversarial network is to minimize the loss of the generator while maximizing the loss of the discriminator, which is expressed by the following formula:

[0034]

[0035] Among them, Pdata(x) is the distribution of real data, p z (z) is the distribution of noise, D(x) is the judgment of the discriminator on the real image x, and D(G(z)) is the judgment of the discriminator on the synthetic image data G(z);

[0036] Training process:

[0037] Initialization: Randomly initialize the parameters of the generator and discriminator;

[0038] Iterative training: training the discriminator: fixing the parameters of the generator, and updating the parameters of the discriminator by optimizing the loss function of the discriminator; training the generator: fixing the parameters of the discriminator, and updating the parameters of the generator by optimizing the loss function of the generator;

[0039] Repeated iteration: Repeat the above iterative training until the preset number of training rounds is reached or other stopping conditions are met;

[0040] Synthesize image data: Use the trained generator and input the random noise vector z to generate new synthetic image data G(z);

[0041] Image data merging: Merge the synthesized image data G(z) with other image data to form the final image data set.

[0042] Furthermore, in S4, the monitoring parameters are adjusted according to the growth stage requirements of the wheat, and the monitoring parameters include wavelength range, sampling frequency and image resolution.

[0043] A crop growth detection system used in a crop growth detection method based on 5G transmission includes:

[0044] Acquisition equipment: responsible for collecting hyperspectral images and visible light images of crop fields, obtaining spectral information and crop growth information;

[0045] 5G transmission module: uses 5G technology to transmit the collected hyperspectral images to the cloud server in real time;

[0046] Cloud server: Receives and stores data from the intelligent hyperspectral camera, runs machine learning algorithms to process and analyze the data, and outputs growth status monitoring results;

[0047] Machine Learning Model Library: This includes random forest, gradient boosting tree, support vector machine, and logistic regression models for feature extraction, classification, and recognition of hyperspectral images. These models are customized and optimized based on environmental factors in different regions to improve monitoring accuracy and practicality.

[0048] Decision support system: Develop crop planting strategies and management measures based on the output of machine learning models;

[0049] User interaction interface: Provides a user interface that allows users to view crop growth information, receive decision-making suggestions, and perform corresponding operations and management.

[0050] The present invention has the following beneficial effects:

[0051] In the present invention, a generative adversarial network is used to simulate wheat growth data under different environmental conditions, thereby increasing the diversity of training data and improving the adaptability of the model under different environments. A random forest, a gradient boosting tree, and a support vector machine are selected as base classifiers, each base classifier is trained, and cross-validation is used to evaluate its performance. The prediction results of multiple base classifiers are input as new features into a logistic regression model for final prediction, and a growth status monitoring model is established, thereby improving the generalization ability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a method step diagram of a crop growth detection method and system based on 5G transmission proposed by the present invention;

[0053] Figure 2 、 Figure 3 as well as Figure 4 Visible light image of wheat taken by drone. DETAILED DESCRIPTION

[0054] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0055] See also Figures 1-4 As shown, the present invention is a crop growth detection method based on 5G transmission, comprising the following steps:

[0056] S1: Hyperspectral image acquisition: Sensor nodes deployed at agricultural production sites and 5G wireless communication networks are used to collect hyperspectral images of wheat fields in real time to obtain spectral information. UAVs are used to capture visible light images to obtain crop growth information.

[0057] S2: Establishing a regionally customized model: Based on environmental factors such as climate conditions and soil types in different regions, a regionally customized feature band database is established. Random forests, gradient boosting trees, and support vector machines are selected as base classifiers. Each base classifier is trained and cross-validation is used to evaluate its performance. The prediction results of multiple base classifiers are used as new features and input into a logistic regression model for final prediction. This establishes a growth status monitoring model.

[0058] S3: Data Augmentation and Model Optimization: By augmenting existing training data and using generative adversarial networks to generate wheat growth data under different environmental conditions, simulated data is mixed with real data to increase the diversity of training data. During training, samples are randomly selected from the augmented dataset to improve the generalization ability of the model.

[0059] S4: Growth stage identification and parameter adjustment: Utilizing the growth status monitoring model in S2, the growth stage of wheat is automatically identified by analyzing the spectral features and growth parameters in hyperspectral and visible light images. Monitoring parameters and management measures, such as fertilization plans, irrigation strategies, and pest and disease control, are adjusted according to the growth stage requirements of the wheat.

[0060] S5: Data visualization and decision support: Wheat growth status information is visualized through an interactive front-end. The interactive front-end uses the WebSocket protocol to transmit data with the hyperspectral camera and cloud server, realizing real-time data update and display. Based on the growth status information, wheat planting strategies and management measures are formulated. For example, fertilization plans are adjusted according to the nutritional status of wheat, and pest and disease control plans are formulated according to the health status.

[0061] In one embodiment, for the above S1, in S1, the sensor node is responsible for real-time monitoring and collecting various information about the farmland environment and crops, and uses a hyperspectral imaging sensor to capture the reflected light information of crops at different wavelengths, thereby generating a hyperspectral image. Through the 5G network, the hyperspectral image data collected by the sensor node is transmitted to the cloud server in real time.

[0062] In one embodiment, for the above S2, in S2, the specific steps of establishing a regional customized characteristic band database based on environmental factors in different regions are:

[0063] Feature extraction: Using hyperspectral imaging technology, the spectral characteristics of wheat fields at different growth stages are extracted. These characteristics are expressed as reflectance or absorption in specific bands.

[0064] Correlation analysis: The extracted spectral features were correlated with the growth conditions of wheat (such as plant height, leaf area index, biomass, etc.). The Pearson correlation coefficient was calculated using statistical software (such as SPSS, R language, etc.) to screen out the characteristic bands that were highly correlated with the growth conditions of wheat. The Pearson correlation coefficient calculation formula is: Among them, r is the correlation coefficient, x i and y i are the spectral characteristic value and growth status value of the i-th sample respectively, and are the means of spectral characteristic values and growth status values, respectively, and n is the number of samples;

[0065] Determine characteristic bands: Based on the results of the correlation analysis, determine the most relevant characteristic bands for each study area. The most relevant characteristic bands constitute a regionally customized characteristic band database.

[0066] Establish a database: Organize the determined characteristic bands and their corresponding spectral information into a database format. The database should include information such as band number, wavelength range, reflectance / absorbance value, and metadata such as the site and time of data collection.

[0067] Taking a wheat-growing area in the North China Plain as an example, assume that after correlation analysis, the following three characteristic bands are screened out:

[0068] Band 1: The wavelength range is 650-680nm (red light band), which is highly correlated with the chlorophyll content of wheat.

[0069] Band 2: The wavelength range is 800-850nm (near infrared band), which is related to the leaf area index of wheat.

[0070] Band 3: The wavelength range is 1600-1700nm (shortwave infrared band), which is related to the moisture content of wheat.

[0071] These three bands will form part of a database of characteristic bands for this wheat-growing region in the North China Plain. In practical applications, hyperspectral image data from these bands, combined with machine learning algorithms, will be used to establish a monitoring model for wheat growth.

[0072] In one embodiment, for the above S2, in S2, the specific steps of establishing the growth status monitoring model are:

[0073] Feature extraction: Using hyperspectral imaging technology, we extract the spectral characteristics of the wheat field and select the most relevant characteristic bands as model input based on a regionally customized characteristic band database.

[0074] Base classifier training: Using historical wheat growth data as the training set, random forests, gradient boosting trees, and support vector machines are trained. Each base classifier learns how to predict wheat growth conditions from the input hyperspectral image features; Random Forest: Construct multiple decision trees as weak classifiers. Each decision tree is trained on a random subset of the training data. The prediction results of all decision trees are combined through a voting mechanism to obtain the final prediction; Gradient Boosting Tree: Initialize as a weak prediction model (such as a decision tree stump), and then gradually train new weak classifiers to correct the errors of the previous model. Finally, the prediction results of all weak classifiers are weighted and summed to obtain the final prediction; Support Vector Machine: Map the input data to a high-dimensional feature space and find a hyperplane to maximize the interval between two types of samples. The optimal hyperplane and classification decision function are obtained by solving a convex quadratic programming problem;

[0075] Cross-validation: Divide the dataset into k non-overlapping subsets, select k-1 subsets each time as training sets, and use the remaining subset as the test set for model evaluation. Repeat this process k times, selecting a different test set each time. The average of the k evaluation results is taken as the accuracy of each base classifier.

[0076] Weight allocation: weights are allocated based on the accuracy of each base classifier on the validation set. The base classifier with higher accuracy has a greater weight. The weight calculation formula is Among them, w i is the weight of the i-th base classifier, n is the total number of base classifiers, k i is the accuracy of the i-th base classifier;

[0077] Logistic regression model fusion: The prediction results of each base classifier are input into a logistic regression model as new features. The output of the logistic regression model is the final prediction result, which represents the growth status of wheat. The formula of the logistic regression model is where X i is the prediction result of the i-th base classifier, β i is the corresponding coefficient, and β0 is the intercept term.

[0078] Suppose we have a hyperspectral image dataset of wheat fields containing 1,000 samples. Each sample has reflectance values in 20 characteristic bands and a corresponding plant height label. We plan to use random forests, gradient boosted trees, and support vector machines as base classifiers to build a growth monitoring model.

[0079] Data preprocessing and feature extraction: Perform preprocessing operations such as correction and denoising on the hyperspectral image data. Based on the regional customized feature band database, select the five most relevant feature bands as model input, and combine the most relevant feature bands and growth status labels into a training set.

[0080] Base classifier training: Random Forest: A random forest model consisting of 100 decision trees was constructed, with each tree trained on a random subset of the training data (70% of the samples); Gradient Boosted Tree: Decision tree stumps were used as weak classifiers, with 100 iterations and a learning rate of 0.1; Support Vector Machine: RBF kernel function was used, and the optimal hyperparameters C and γ were determined through grid search;

[0081] Cross-validation: The dataset was divided into five subsets (k=5) and five-fold cross-validation was performed. For each base classifier, the average accuracy of the five-fold cross-validation was calculated as an estimate of the model performance. The accuracy of the random forest was 85%, the accuracy of the gradient boosting tree was 88%, and the accuracy of the support vector machine was 82%.

[0082] Weight distribution: Assign weights based on accuracy: the weight of random forest is 0.32, the weight of gradient boosting tree is 0.33, and the weight of support vector machine is 0.35 (for simplicity, it is assumed that the sum of the weights of the three base classifiers is approximately 1).

[0083] Logistic regression model fusion: The prediction results of the three base classifiers are input into the logistic regression model as features. The logistic regression model is trained to obtain the final prediction results.

[0084] Application and Validation: A new wheat field sample was used for testing, extracting hyperspectral image features. These features were then fed into a trained model to generate predictions of wheat growth status. Based on these predictions, monitoring parameters and management measures, such as fertilization, irrigation, and pest control, were adjusted.

[0085] In one embodiment, for the above S3, in S3, the specific steps of enhancing the existing training data are:

[0086] Basic image enhancement: Perform basic image enhancement operations such as rotation, scaling, and flipping on hyperspectral images to increase data diversity; Rotation: Rotate the image around the center point by a certain angle (such as 90 degrees, 180 degrees, 270 degrees); Scaling: Change the size of the image, usually enlarging or reducing it proportionally; Flip: Flip the image horizontally or vertically;

[0087] Color space conversion: convert the hyperspectral image from RGB color space to other color spaces (such as HSV, Lab, etc.), perform color adjustment, and then convert it back to RGB color space;

[0088] Noise addition: Add noise (such as Gaussian noise, salt and pepper noise, etc.) to the hyperspectral image to simulate the noise interference in the real environment.

[0089] In one embodiment, for the above S3, in S3, the specific steps of using the generative adversarial network to generate wheat growth data under different environmental conditions are:

[0090] Define the GAN network structure: Generator: Responsible for generating new data samples similar to real data. During the training process, the goal of the generator is to generate real data to deceive the discriminator; Input: Random noise vector z; Output: Synthesized image data G(z);

[0091] Discriminator: Responsible for distinguishing whether the input data is real data or fake data generated by the generator. During the training process, the discriminator's goal is to accurately distinguish real data from generated data; Input: real image data x or image data G(z) synthesized by the generator; Output: a probability value D(x), indicating the probability that the input image is a real image;

[0092] Define the loss function:

[0093] The training goal of the generative adversarial network is to minimize the loss of the generator while maximizing the loss of the discriminator, which is expressed by the following formula:

[0094]

[0095] Among them, Pdata(x) is the distribution of real data, p z (z) is the distribution of noise, D(x) is the judgment of the discriminator on the real image x, and D(G(z)) is the judgment of the discriminator on the synthetic image data G(z);

[0096] Training process:

[0097] Initialization: Randomly initialize the parameters of the generator and discriminator;

[0098] Iterative training: training the discriminator: fixing the parameters of the generator, and updating the parameters of the discriminator by optimizing the loss function of the discriminator; training the generator: fixing the parameters of the discriminator, and updating the parameters of the generator by optimizing the loss function of the generator;

[0099] Repeated iteration: Repeat the above iterative training until the preset number of training rounds is reached or other stopping conditions are met;

[0100] Synthesize image data: Use the trained generator and input the random noise vector z to generate new synthetic image data G(z);

[0101] Image data merging: Merge the synthesized image data G(z) with other image data to form the final image data set.

[0102] In one embodiment, for the above S4, in S4, the monitoring parameters are adjusted according to the growth stage requirements of the wheat, and the monitoring parameters include wavelength range, sampling frequency and image resolution.

[0103] Wavelength range adjustment: During the heading stage, since chlorophyll and moisture content are crucial to wheat growth, sampling can be increased in the red (600-700nm) and near-infrared (700-1100nm) bands. These bands are more sensitive to chlorophyll and moisture content and can provide more accurate information, ensuring sufficient sampling points in the red and near-infrared bands.

[0104] Adjusting the sampling frequency: During critical wheat growth periods (such as heading and grain filling), more frequent data collection is required to monitor growth changes. Therefore, the sampling frequency can be increased, for example, from once a day to twice a day or more. Adjust the sampling frequency in the spectrometer settings to ensure sufficient data collection points during critical periods.

[0105] Image resolution adjustment: Image resolution requirements may vary at different wheat growth stages. For example, during the sowing and seedling stages, higher image resolution may be required to accurately identify wheat seedlings. During the maturity stage, however, due to the taller wheat plants and denser leaves, lower image resolution may be required to reduce data processing. Adjust the image resolution in the spectrometer settings based on the requirements of the wheat growth stage.

[0106] A crop growth detection system used in a crop growth detection method based on 5G transmission includes:

[0107] Acquisition equipment: responsible for collecting hyperspectral images and visible light images of crop fields, obtaining spectral information and crop growth information;

[0108] 5G transmission module: uses 5G technology to transmit the collected hyperspectral images to the cloud server in real time;

[0109] Cloud server: Receives and stores data from the intelligent hyperspectral camera, runs machine learning algorithms to process and analyze the data, and outputs growth status monitoring results;

[0110] Machine Learning Model Library: This includes random forest, gradient boosting tree, support vector machine, and logistic regression models for feature extraction, classification, and recognition of hyperspectral images. These models are customized and optimized based on environmental factors in different regions to improve monitoring accuracy and practicality.

[0111] Decision support system: Develop crop planting strategies and management measures based on the output of machine learning models;

[0112] User interaction interface: Provides a user interface that allows users to view crop growth information, receive decision-making suggestions, and perform corresponding operations and management.

[0113] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A crop growth detection method based on 5G transmission, characterized in that: The following steps are involved: S1: Hyperspectral image acquisition: Sensor nodes deployed at agricultural production sites and 5G wireless communication networks are used to collect hyperspectral images of wheat fields in real time to obtain spectral information. UAVs are used to capture visible light images to obtain crop growth information. S2: Establishing a regionally customized model: Based on the environmental factors of different regions, a regionally customized feature band database is established. Random forest, gradient boosting tree, and support vector machine are selected as base classifiers. Each base classifier is trained and cross-validation is used to evaluate its performance. The prediction results of multiple base classifiers are used as new features to input into a logistic regression model for final prediction, thus establishing a growth status monitoring model. S3: Data Augmentation and Model Optimization: This involves augmenting existing training data and using a generative adversarial network to generate wheat growth data under different environmental conditions. This process involves mixing simulated data with real data. During training, samples are randomly selected from the augmented dataset. S4: Growth stage identification and parameter adjustment: Using the growth status monitoring model in S2, the growth stage of wheat is automatically identified by analyzing the spectral characteristics and growth parameters in hyperspectral images and visible light images. Monitoring parameters and management measures are adjusted according to the growth stage requirements of the wheat. S5: Data visualization and decision support: The wheat growth status information is visualized through an interactive front-end. The interactive front-end uses the WebSocket protocol to transmit data with the hyperspectral camera and the cloud server. Based on the growth status information, wheat planting strategies and management measures are formulated.

2. The crop growth detection method based on 5G transmission according to claim 1, characterized in that: In S1, the sensor node is responsible for real-time monitoring and collection of various information about the farmland environment and crops. It uses a hyperspectral imaging sensor to capture the reflected light information of crops at different wavelengths, thereby generating a hyperspectral image. Through the 5G network, the hyperspectral image data collected by the sensor node and the visible light image data taken by the drone are transmitted to the cloud server in real time.

3. The crop growth detection method based on 5G transmission according to claim 1, characterized in that: In S2, the specific steps of establishing a regional customized characteristic band database based on environmental factors in different regions are: Feature extraction: Using hyperspectral imaging technology, the spectral characteristics of wheat fields at different growth stages are extracted; Correlation analysis: The extracted spectral features are correlated with the growth status of wheat, and the Pearson correlation coefficient is calculated to screen out the characteristic bands that are highly correlated with the growth status of wheat. The Pearson correlation coefficient calculation formula is: Among them, r is the correlation coefficient, x i and y i are the spectral characteristic value and growth status value of the i-th sample respectively, and are the means of spectral characteristic values and growth status values, respectively, and n is the number of samples; Determine characteristic bands: Based on the results of the correlation analysis, determine the most relevant characteristic bands for each study area. The most relevant characteristic bands constitute a regionally customized characteristic band database. Establish a database: organize the determined characteristic bands and their corresponding spectral information into a database format, and attach the metadata of the collected data.

4. The crop growth detection method based on 5G transmission according to claim 1, characterized in that: In S2, the specific steps of establishing the growth status monitoring model are: Feature extraction: Using hyperspectral imaging technology, we extract the spectral characteristics of the wheat field and select the most relevant characteristic bands as model input based on a regionally customized characteristic band database. Base classifier training: Using historical wheat growth data as a training set, random forests, gradient boosting trees, and support vector machines are trained. Each base classifier learns how to predict wheat growth from the input hyperspectral image features. Random forests: Multiple decision trees are constructed as weak classifiers. Each decision tree is trained on a random subset of the training data, and the predictions of all decision trees are combined through a voting mechanism to obtain the final prediction. Gradient boosting trees: Initialized as a weak prediction model, new weak classifiers are then gradually trained to correct the errors of the previous model. The final prediction is finally obtained by weighted summation of the predictions of all weak classifiers. Support Vector Machine: Maps input data into a high-dimensional feature space and finds a hyperplane to maximize the margin between two classes of samples. The optimal hyperplane and classification decision function are obtained by solving a convex quadratic programming problem. Cross-validation: Divide the dataset into k non-overlapping subsets, select k-1 subsets each time as training sets, and use the remaining subset as the test set for model evaluation. Repeat this process k times, selecting a different test set each time. The average of the k evaluation results is taken as the accuracy of each base classifier. Weight allocation: Assign weights based on the accuracy of each base classifier on the validation set. The weight calculation formula is Among them, w i is the weight of the i-th base classifier, n is the total number of base classifiers, k i is the accuracy of the i-th base classifier; Logistic regression model fusion: The prediction results of each base classifier are input into a logistic regression model as new features. The output of the logistic regression model is the final prediction result, which represents the growth status of wheat. The formula of the logistic regression model is where X i is the prediction result of the i-th base classifier, β i is the corresponding coefficient, and β0 is the intercept term.

5. The crop growth detection method based on 5G transmission according to claim 1, characterized in that: In S3, the specific steps of enhancing the existing training data are as follows: Basic image enhancement: perform basic image enhancement operations on hyperspectral images; Rotate: rotate the image around the center point by a certain angle; Zoom: change the size of the image; Flip: flip the image horizontally or vertically; Color space conversion: convert the hyperspectral image from RGB color space to other color spaces, perform color adjustment, and then convert it back to RGB color space; Noise addition: Add noise to the hyperspectral image to simulate the noise interference in the real environment.

6. The crop growth detection method based on 5G transmission according to claim 1, characterized in that: In S3, the specific steps of using the generative adversarial network to generate wheat growth data under different environmental conditions are: Define the GAN network structure: Generator: Responsible for generating new data samples similar to real data. During the training process, the goal of the generator is to generate real data to deceive the discriminator; Input: random noise vector z; Output: synthesized image data G(z); Discriminator: Responsible for distinguishing whether the input data is real data or fake data generated by the generator. During the training process, the goal of the discriminator is to accurately distinguish between real data and generated data; Input: real image data x or image data G(z) synthesized by the generator; Output: A probability value D(x), indicating the probability that the input image is a real image; Define the loss function: The training goal of the generative adversarial network is to minimize the loss of the generator while maximizing the loss of the discriminator, which is expressed by the following formula: Among them, Pdata(x) is the distribution of real data, p z (z) is the distribution of noise, D(x) is the judgment of the discriminator on the real image x, and D(G(z)) is the judgment of the discriminator on the synthetic image data G(z); Training process: Initialization: Randomly initialize the parameters of the generator and discriminator; Iterative training: training the discriminator: fixing the parameters of the generator, and updating the parameters of the discriminator by optimizing the loss function of the discriminator; training the generator: fixing the parameters of the discriminator, and updating the parameters of the generator by optimizing the loss function of the generator; Repeated iteration: Repeat the above iterative training until the preset number of training rounds is reached or other stopping conditions are met; Synthesize image data: Use the trained generator and input the random noise vector z to generate new synthetic image data G(z); Image data merging: Merge the synthesized image data G(z) with other image data to form the final image data set.

7. The crop growth detection method based on 5G transmission according to claim 1, characterized in that: In S4, the monitoring parameters are adjusted according to the growth stage requirements of the wheat, and the monitoring parameters include wavelength range, sampling frequency and image resolution.

8. The crop growth detection system used in the crop growth detection method based on 5G transmission according to claim 1, characterized in that: include: Acquisition equipment: responsible for collecting hyperspectral images and visible light images of crop fields, obtaining spectral information and crop growth information; 5G transmission module: uses 5G technology to transmit the collected hyperspectral images to the cloud server in real time; Cloud server: Receives and stores data from the intelligent hyperspectral camera, runs machine learning algorithms to process and analyze the data, and outputs growth status monitoring results; Machine Learning Model Library: This includes random forest, gradient boosting tree, support vector machine, and logistic regression models for feature extraction, classification, and recognition of hyperspectral images, customized and optimized based on environmental factors in different regions. Decision support system: Develop crop planting strategies and management measures based on the output of machine learning models; User interaction interface: Provides a user interface that allows users to view crop growth information, receive decision-making suggestions, and perform corresponding operations and management.

Images