Method for evaluating vegetative growth condition of lycium barbarum by using phenotypic characteristics
Through image data acquisition and deep learning technology, GAN and CNN models are used to construct a growth status recognition system for wolfberry, which solves the problems of low efficiency and lack of objectivity of traditional evaluation methods, and achieves efficient and accurate evaluation of wolfberry's vegetative growth status.
Patent Information
- Application Number
- CN202411263444.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-10
- Publication Date
- 2024-12-13
AI Technical Summary
The traditional method of evaluating growth status of wolfberry is inefficient and lacks objectivity. The existing automated monitoring technology fails to make full use of plant phenotypic characteristics, limiting an in-depth understanding of crop growth status.
Image data acquisition, preprocessing and expansion are used, data sets are constructed in combination with deep learning technology, neural network models are trained, and generative adversarial networks (GANs) and convolutional neural networks (CNNs) are used to identify and predict the nutritional growth status of wolfberry.
An automated, efficient and objective evaluation of wolfberry vegetative growth status has been achieved, which improves the identification accuracy and efficiency, and enhances the model's ability and generalization of different growth statuses.
Smart Images
Figure CN119131592A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of plant growth status evaluation, and in particular to a method for evaluating the nutritional growth status of wolfberry using phenotypic characteristics. Background Art
[0002] With the rise of precision agriculture, real-time monitoring and evaluation of crop growth has become particularly important. As a crop with important medicinal and economic value, the growth status of wolfberry (Lycium barbarum) directly affects the yield and quality of the fruit. Traditional growth status evaluation methods mainly rely on manual observation and empirical judgment. These methods are not only inefficient, but also greatly affected by human factors and lack objectivity and accuracy.
[0003] In recent years, some automated crop growth monitoring technologies have begun to emerge, but these technologies often focus on monitoring environmental factors or some physiological indicators, while ignoring the phenotypic characteristics of the plants themselves. Plant phenotypic characteristics, such as leaf shape, color, size, and the number and distribution of fruits, are an intuitive reflection of the plant's growth environment and nutritional status. At present, there is a relative lack of methods for comprehensive evaluation of the phenotypic characteristics of crops such as wolfberry, which limits the ability to deeply understand the growth status of crops.
[0004] With the development of computer vision and deep learning technology, it has become possible to use plant phenotypic characteristics for automated evaluation. Therefore, it is urgent to propose an automated, efficient, and objective method for evaluating the nutritional growth status of wolfberry using phenotypic characteristics. Summary of the invention
[0005] The purpose of the present invention is to provide an automated, efficient and objective method for evaluating the nutritional growth status of wolfberry using phenotypic characteristics, so as to solve the problems existing in the above-mentioned prior art.
[0006] To achieve the above object, the present invention provides the following solutions:
[0007] A method for evaluating the nutritional growth status of wolfberry using phenotypic characteristics, comprising:
[0008] Collect image data of wolfberry at different growth stages and under different nutritional conditions;
[0009] Preprocessing and expanding the image data, and constructing a data set based on the phenotypic characteristics of wolfberry;
[0010] The preset neural network model is trained using the data set to obtain a growth status recognition model;
[0011] The growth status recognition model is used to identify and evaluate the nutritional growth status of wolfberry.
[0012] Optionally, preprocessing the image data includes:
[0013] The image data is manually annotated to mark the phenotypic characteristics of wolfberry and the corresponding growth status evaluation; wherein the phenotypic characteristics are used as the basis for evaluating the nutritional growth status of wolfberry, including: leaf color, leaf shape, stem thickness, fruit size and quantity of wolfberry;
[0014] The annotated image data is denoised, contrast enhanced, and normalized.
[0015] Optionally, the preprocessed image data is expanded to construct a data set including:
[0016] Generative adversarial networks are used to expand the preprocessed image data, and synthetic images with a preset similarity to the real data are generated by learning the distribution of the preprocessed image data;
[0017] This facilitates quality assessment of the expanded synthetic image and generates the data set.
[0018] Optionally, using a generative adversarial network, the preprocessed image data is expanded including:
[0019] Constructing a GAN network structure based on a generator and a discriminator; wherein the generator is used to generate new images, and the discriminator is used to distinguish between real images and images generated by the generator;
[0020] It is beneficial to use binary cross entropy loss function to train the discriminator so that it can distinguish between real images and noisy data;
[0021] The reverse signal is used to train the generator to produce new synthetic images, using the same loss function as the discriminator.
[0022] During the training process of the generator and the discriminator, the discriminator and the generator are updated alternately;
[0023] It is beneficial for the trained generator to expand the preprocessed image data, and it is beneficial for the trained discriminator to evaluate the generated synthetic images and check whether the discriminator can distinguish between real images and generated images.
[0024] Optionally, using the data set to train a preset neural network model includes:
[0025] Build a CNN model;
[0026] Based on the data set, the CNN model is trained and the model parameters are continuously adjusted through a back-propagation algorithm;
[0027] During the training process, cross-validation is used to avoid overfitting, and the training effect is optimized by adjusting the learning rate and batch size hyperparameters.
[0028] Optionally, building a CNN model includes:
[0029] Set up an input layer for receiving input image data;
[0030] Set up several convolutional layers to extract spatial features;
[0031] Set the activation layer, use nonlinear activation functions, introduce nonlinearity, and help the model learn preset complex features;
[0032] Add a pooling layer between the convolutional layers to reduce the feature dimension and extract the preset main features;
[0033] Add a fully connected layer at the end of the CNN to map the features to the final classification label;
[0034] Set up the output layer and use the softmax activation function for multi-classification tasks.
[0035] Optionally, the identification and evaluation of the nutritional growth status of wolfberry includes:
[0036] Design a user interface to display images of wolfberry plants and recognition results;
[0037] Display the classification labels of wolfberry growth status on the interface;
[0038] Use a bar chart to show the confidence of each classification;
[0039] Record the identification results in a database;
[0040] Integrate real-time feedback mechanism to inform managers in time.
[0041] Optionally, the identification and evaluation of the nutrient growth status of wolfberry also includes:
[0042] Predict the nutritional growth status of wolfberry;
[0043] Based on the data set and the identification result of the nutritional growth status of wolfberry, a prediction data set is generated;
[0044] Using the prediction data to train a long short-term memory network to obtain a growth status prediction model;
[0045] The growth status prediction model is used to predict the nutritional growth status of wolfberry and generate corresponding evaluation results.
[0046] The beneficial effects of the present invention are:
[0047] The present invention provides an automated, efficient and objective method for evaluating the nutritional growth status of wolfberry by utilizing phenotypic characteristics. The method utilizes a growth status recognition model to perform identification and evaluation of the nutritional growth status of wolfberry, with a high degree of automation, reduced manual participation and improved evaluation efficiency. The method utilizes deep learning technology to improve the accuracy and objectivity of growth status identification. The image data is preprocessed and expanded to construct a data set, and through the expansion of the data set, the model's ability to recognize different growth conditions is enhanced, thereby improving the generalization of the model. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0049] Figure 1 The present invention is a flowchart of a method for evaluating the nutritional growth status of wolfberry using phenotypic characteristics according to an embodiment of the present invention. DETAILED DESCRIPTION
[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0051] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0052] like Figure 1 As shown, this embodiment proposes a method for evaluating the nutritional growth status of wolfberry using phenotypic characteristics, including:
[0053] Collect image data of wolfberry at different growth stages and under different nutritional conditions;
[0054] Preprocess and expand the image data, and construct a dataset based on the phenotypic characteristics of wolfberry;
[0055] The preset neural network model is trained using the data set to obtain a growth status recognition model;
[0056] The growth status recognition model is used to identify and evaluate the nutritional growth status of wolfberry.
[0057] Furthermore, preprocessing the image data includes:
[0058] The image data is manually annotated to mark the phenotypic characteristics of wolfberry and the corresponding growth status evaluation; wherein the phenotypic characteristics are used as the basis for evaluating the nutritional growth status of wolfberry, including: leaf color, leaf shape, stem thickness, fruit size and quantity of wolfberry;
[0059] The annotated image data is denoised, contrast enhanced, and normalized.
[0060] Specifically, in this embodiment, agricultural experts are invited to annotate the collected images, marking key phenotypic characteristics of wolfberry, such as leaf color, leaf shape, stem thickness, fruit size and quantity, as a basis for evaluating the nutritional growth status of wolfberry.
[0061] The image is then processed by denoising, contrast enhancement, and normalization to improve image quality and lay the foundation for subsequent processing.
[0062] Furthermore, the preprocessed image data is expanded to construct a data set including:
[0063] Generative adversarial networks are used to expand the preprocessed image data, and synthetic images with a preset similarity to the real data are generated by learning the distribution of the preprocessed image data;
[0064] It is helpful to evaluate the quality of the expanded synthetic images and generate data sets.
[0065] Using the generative adversarial network, the preprocessed image data is expanded including:
[0066] Construct a GAN network structure based on a generator and a discriminator; the generator is used to generate new images, and the discriminator is used to distinguish between real images and images generated by the generator;
[0067] It is beneficial to use binary cross entropy loss function to train the discriminator so that it can distinguish between real images and noisy data;
[0068] The reverse signal is used to train the generator to produce new synthetic images, using the same loss function as the discriminator.
[0069] During the training process of the generator and the discriminator, the discriminator and the generator are updated alternately;
[0070] It is beneficial for the trained generator to expand the preprocessed image data, and it is beneficial for the trained discriminator to evaluate the generated synthetic images and check whether the discriminator can distinguish between real images and generated images.
[0071] Specifically, in this embodiment, a generative adversarial network (GANs) is used to expand the initial data set. By learning the distribution of the original data, a large number of synthetic images that are similar to the real data but not exactly the same are generated, increasing the diversity and scale of the data set; specifically, it includes the following stages:
[0072] Prepare the initial dataset: Make sure the initial dataset has been preprocessed, has a unified format, and has been split into training and validation sets;
[0073] Configure the GAN architecture: define the network structure of the generator and the discriminator. The generator is responsible for generating new images, and the discriminator is responsible for distinguishing between real images and images generated by the generator;
[0074] Define the loss function: For the discriminator, use the binary cross entropy loss function so that it can distinguish between real and generated images; for the generator, use the reverse signal of the same loss function as the discriminator to encourage the generator to produce images that the discriminator has difficulty distinguishing;
[0075] Training the discriminator: With the generator fixed, first train the discriminator to distinguish between real images and noisy data;
[0076] Train the generator: Train the generator to produce new images that are realistic enough to be indistinguishable to the discriminator;
[0077] Alternating training: The discriminator and generator are updated alternately during the training process. This is the core process of GAN training and usually requires multiple iterations.
[0078] Evaluate the quality of generated images: Regularly evaluate the quality of generated images to ensure that they are visually similar to real images and have diversity;
[0079] Fine-tuning the model: Based on the evaluation results of the generated images, fine-tune the GAN model to optimize the network parameters and training process;
[0080] Ensemble generated data: Combine the generated images with the original dataset to form an augmented dataset.
[0081] Validate the augmented dataset: Use the validation set to evaluate the quality of the augmented dataset to ensure that the generated images increase the diversity of the dataset and do not introduce bias;
[0082] Iterative improvement: Based on the verification results, it may be necessary to go back to the previous steps and adjust the GAN model or training process.
[0083] Final evaluation: After the model training is completed, the model is finally evaluated using the entire expanded dataset to ensure the improvement of model performance.
[0084] Furthermore, using the data set to train the preset neural network model includes:
[0085] Build a CNN model;
[0086] Based on the data set, the CNN model is trained and the model parameters are continuously adjusted through the back-propagation algorithm;
[0087] During the training process, cross-validation is used to avoid overfitting, and the training effect is optimized by adjusting the learning rate and batch size hyperparameters.
[0088] Among them, building a CNN model includes:
[0089] Set up an input layer for receiving input image data;
[0090] Set up several convolutional layers to extract spatial features;
[0091] Set the activation layer, use nonlinear activation functions, introduce nonlinearity, and help the model learn preset complex features;
[0092] Add a pooling layer between the convolutional layers to reduce the feature dimension and extract the preset main features;
[0093] Add a fully connected layer at the end of the CNN to map the features to the final classification label;
[0094] Set up the output layer and use the softmax activation function for multi-classification tasks.
[0095] Specifically, in this embodiment, the designed CNN model includes the following contents:
[0096] Input layer: This is the entrance of the model, receiving input data, images of wolfberry plants;
[0097] Convolutional layer: applies learnable filters (or convolution kernels) to extract local features in the image;
[0098] Activation layer: Use nonlinear activation functions, such as ReLU, to introduce nonlinearity and help the model learn more complex features;
[0099] Pooling layer: used to reduce the spatial dimension of features and reduce the amount of calculation. Common pooling operations include maximum pooling and average pooling.
[0100] Normalization layers: Use normalization layers to speed up the training process and improve model stability.
[0101] Fully connected layer: Located at the end of the network, it maps the features extracted by the convolutional layer and the pooling layer to the final output, such as the classification label.
[0102] Dropout layer: Prevent overfitting by randomly discarding some network connections and improve the generalization ability of the model.
[0103] Output layer: The last layer uses softmax activation function for multi-classification tasks.
[0104] Further, the identification and evaluation of the nutritional growth status of wolfberry includes:
[0105] Design a user interface to display images of wolfberry plants and recognition results;
[0106] Display the classification labels of wolfberry growth status on the interface;
[0107] Use a bar chart to show the confidence of each classification;
[0108] Record the identification results in a database;
[0109] Integrate real-time feedback mechanism to inform managers in time.
[0110] In this embodiment, the recognition results can also be combined with the agricultural automation control system to achieve intelligent decision support.
[0111] Further, the identification and evaluation of the nutritional growth status of wolfberry also includes:
[0112] Predict the nutritional growth status of wolfberry;
[0113] Generate a prediction dataset based on the dataset and the identification results of the nutritional growth status of wolfberry;
[0114] The prediction data is used to train the long short-term memory network to obtain a growth status prediction model;
[0115] The growth status prediction model is used to predict the nutritional growth status of wolfberry and generate corresponding evaluation results.
[0116] Specifically, in this embodiment, a long short-term memory network is used to process time series data to capture the dynamic characteristics of growth trends; during the training process of the model, a cross-validation method is used to evaluate the performance of the model to avoid overfitting; and the impact of each feature in the model on the prediction results is analyzed to optimize the model and data collection strategy.
[0117] In this embodiment, real-time growth and environmental monitoring data can also be integrated into the prediction model to improve the real-time and adaptability of the prediction. The model can be updated regularly with new data to adapt to environmental changes and changes in growth conditions. The predicted growth trend can be displayed in the form of charts or graphs to facilitate understanding and decision-making. The prediction results can also be integrated into the decision support system to provide growers with recommendations on irrigation, fertilization and harvesting time; the uncertainty and potential risks of the prediction can be evaluated to provide growers with risk management recommendations.
[0118] Compared with the prior art, this embodiment has a high degree of automation, reduces manual participation, and improves evaluation efficiency. The use of deep learning technology improves the accuracy and objectivity of growth status identification. By expanding the data set, the model's ability to identify different growth conditions is enhanced, and the generalization of the model is improved.
[0119] The above embodiments are only descriptions of the preferred embodiments of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should fall within the protection scope determined by the claims of the present invention.
Claims
1. A method for evaluating the nutritional growth status of wolfberry using phenotypic characteristics, characterized in that: include: Collect image data of wolfberry at different growth stages and under different nutritional conditions; Preprocessing and expanding the image data, and constructing a data set based on the phenotypic characteristics of wolfberry; The preset neural network model is trained using the data set to obtain a growth status recognition model; The growth status recognition model is used to identify and evaluate the nutritional growth status of wolfberry.
2. The method for evaluating the nutritional growth status of wolfberry using phenotypic characteristics according to claim 1, characterized in that: Preprocessing the image data includes: The image data is manually annotated to mark the phenotypic characteristics of wolfberry and the corresponding growth status evaluation; wherein the phenotypic characteristics are used as the basis for evaluating the nutritional growth status of wolfberry, including: leaf color, leaf shape, stem thickness, fruit size and quantity of wolfberry; The annotated image data is denoised, contrast enhanced, and normalized.
3. The method for evaluating the nutritional growth status of wolfberry using phenotypic characteristics according to claim 1, characterized in that: The preprocessed image data is expanded to construct a data set including: Generative adversarial networks are used to expand the preprocessed image data, and synthetic images with a preset similarity to the real data are generated by learning the distribution of the preprocessed image data; This facilitates quality assessment of the expanded synthetic image and generates the data set.
4. The method for evaluating the nutritional growth status of wolfberry using phenotypic characteristics according to claim 3, characterized in that: Using the generative adversarial network, the preprocessed image data is expanded including: Constructing a GAN network structure based on a generator and a discriminator; wherein the generator is used to generate new images, and the discriminator is used to distinguish between real images and images generated by the generator; It is beneficial to use binary cross entropy loss function to train the discriminator so that it can distinguish between real images and noisy data; The reverse signal is used to train the generator to produce new synthetic images, using the same loss function as the discriminator. During the training process of the generator and the discriminator, the discriminator and the generator are updated alternately; It is beneficial for the trained generator to expand the preprocessed image data, and it is beneficial for the trained discriminator to evaluate the generated synthetic images and check whether the discriminator can distinguish between real images and generated images.
5. The method for evaluating the nutritional growth status of wolfberry using phenotypic characteristics according to claim 1, characterized in that: Using the data set to train the preset neural network model includes: Build a CNN model; Based on the data set, the CNN model is trained and the model parameters are continuously adjusted through a back-propagation algorithm; During the training process, cross-validation is used to avoid overfitting, and the training effect is optimized by adjusting the learning rate and batch size hyperparameters.
6. The method for evaluating the nutritional growth status of wolfberry using phenotypic characteristics according to claim 5, characterized in that: Building a CNN model includes: Set up an input layer for receiving input image data; Set up several convolutional layers to extract spatial features; Set the activation layer, use nonlinear activation functions, introduce nonlinearity, and help the model learn preset complex features; Add pooling layers between convolutional layers to reduce feature dimensions and extract preset main features; Add a fully connected layer at the end of the CNN to map the features to the final classification label; Set up the output layer and use the softmax activation function for multi-classification tasks.
7. The method for evaluating the nutritional growth status of wolfberry using phenotypic characteristics according to claim 1, characterized in that: The identification and evaluation of the nutritional growth status of wolfberry includes: Design a user interface to display images of wolfberry plants and recognition results; Display the classification labels of wolfberry growth status on the interface; Use a bar chart to show the confidence of each classification; Record the identification results in a database; Integrate real-time feedback mechanism to notify managers in time.
8. The method for evaluating the nutritional growth status of wolfberry using phenotypic characteristics according to claim 1, characterized in that: The identification and evaluation of the nutritional growth status of wolfberry also includes: Predict the nutritional growth status of wolfberry; Based on the data set and the identification result of the nutritional growth status of wolfberry, a prediction data set is generated; Using the prediction data to train a long short-term memory network to obtain a growth status prediction model; The growth status prediction model is used to predict the nutritional growth status of wolfberry and generate corresponding evaluation results.