Method and system for evaluating nitrogen uptake type of tea plant based on nitrogen transporter activity

Abstract

This invention discloses a method and system for evaluating nitrogen absorption types in tea trees based on nitrogen transporter activity, relating to the field of agricultural technology. The method includes collecting tea tree data to establish a total dataset; classifying the data by type using cluster analysis to establish corresponding training sets; training CNN sub-models using the corresponding training sets to obtain low-nitrogen high-efficiency prediction models, nitrogen-efficient prediction models, and high-nitrogen high-efficiency prediction models; calculating the fusion coefficients of the input parameters in each prediction model; determining the fusion weights of each input parameter based on the fusion coefficients to construct a comprehensive prediction model; calculating the nitrogen absorption rate of the tea trees to be identified using the comprehensive prediction model to obtain the measured nitrogen absorption rate; and classifying the tea trees to be identified by type using cluster analysis. This invention constructs a comprehensive prediction model by fusing the input parameter weights of the sub-models, considering the influence of nitrogen transporter activity on nitrogen absorption in tea trees, thus improving the comprehensiveness and accuracy of the comprehensive prediction model.

Description

Technical Field

[0001] This invention relates to the field of agricultural technology, and more specifically to a method and system for evaluating the nitrogen absorption type of tea plants based on nitrogen transporter activity. Background Technology

[0002] Currently, tea trees are harvested primarily for their fresh leaves. Nutrient management is fundamental to achieving high-quality, high-yield, and stable production, and nutrient diagnosis provides the theoretical and technical basis for this management. Plant nutrient diagnosis methods include soil diagnosis, leaf diagnosis, field trials, tracer atoms, and microbial assays. However, due to the large root system of tea trees, the high demand and storage of nutrients, and the complex site conditions, methods such as soil nutrient diagnosis and field trials often suffer from poor accuracy and low efficiency. The results are typically considered only as reference values ​​in production.

[0003] However, simply providing nutrients to tea trees through fertilization not only increases the input costs of the tea industry and the potential environmental risks associated with fertilizers, but also fails to meet the requirements for sustainable development of the tea industry. Therefore, breeding nitrogen-efficient tea tree varieties is of great significance. Tea trees are perennial economic crops, and breeding superior varieties is time-consuming. Developing a rapid and effective non-destructive identification technology for the nitrogen absorption capacity of tea trees can not only greatly shorten the breeding time for nitrogen-efficient tea tree varieties, but also ensure the complete preservation of the selected tea tree varieties for later field verification.

[0004] Therefore, how to evaluate the nitrogen absorption efficiency of tea trees is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a method and system for evaluating the nitrogen absorption type of tea trees based on nitrogen transporter activity. By integrating the input parameter weights of different sub-models, a comprehensive prediction model is constructed, which comprehensively considers the influence of various nitrogen transporter activities on nitrogen absorption in tea trees. This method can effectively evaluate the nitrogen absorption capacity of unknown tea tree species and improve the comprehensiveness and accuracy of the comprehensive prediction model.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] Methods for assessing nitrogen absorption patterns in tea plants based on nitrogen transporter activity include:

[0008] A total dataset was established by collecting the activities of various nitrogen transport proteins and nitrogen uptake rates in tea plants. The data in the total dataset were then classified by type using cluster analysis, and corresponding training sets were established to obtain low-nitrogen high-efficiency sets, nitrogen high-efficiency sets, and high-nitrogen high-efficiency sets.

[0009] We trained CNN sub-models with the same structure using low-nitrogen high-efficiency sets, nitrogen high-efficiency sets, and high-nitrogen high-efficiency sets respectively. We used the activity of various nitrogen transport proteins in tea trees as input data and the nitrogen absorption rate of tea trees as output data to obtain low-nitrogen high-efficiency prediction models, nitrogen high-efficiency prediction models, and high-nitrogen high-efficiency prediction models.

[0010] Calculate the fusion coefficient of the input parameters of each model in the low-nitrogen high-efficiency prediction model, nitrogen high-efficiency prediction model, and high-nitrogen high-efficiency prediction model; and determine the fusion weight of each input parameter based on the fusion coefficient, and construct a comprehensive prediction model based on the fusion weight of all input parameters;

[0011] The nitrogen uptake rate of the tea trees to be identified is calculated using the comprehensive prediction model to obtain the nitrogen uptake rate to be measured, and the tea trees to be identified are classified into types according to the cluster analysis method.

[0012] Preferably, the clustering analysis method is k-means, specifically including: initializing the classification, initializing three cluster centers, using Euclidean distance to characterize the similarity between samples, calculating the distance from all samples to each initial cluster center, and classifying the samples into categories according to the criterion of minimum distance; updating the cluster centers, calculating the mean of the samples belonging to each category as the new cluster center of that category, and then re-obtaining the sum of squared distances from all samples to the new cluster centers of their respective categories; determining whether the cluster centers have changed between the two clustering analyses, if they no longer change, then the clustering ends, otherwise the cluster centers are updated repeatedly until the clustering ends.

[0013] Preferably, the CNN sub-model specifically includes: selecting 6 input parameters related to the nitrogen absorption rate of tea trees, performing normalization processing, arranging the processed input parameter data in a time series as input data, and constructing 6 input data time series input layers; performing convolution operations on the input data, and pooling the input data after convolution operations, and performing fully connected operations on the output data after multiple convolution and pooling operations to complete the training of the low-nitrogen high-efficiency prediction model, the nitrogen high-efficiency prediction model, and the high-nitrogen high-efficiency prediction model.

[0014] Preferably, the input parameters specifically include ammonium nitrogen transporters: CsAMT1.1, CsAMT1.2; and nitrate nitrogen transporters: CsNRT2.1, CsNRT2.2, CsNRT2.3, CsNRT2.4.

[0015] Preferably, training the low-nitrogen high-efficiency prediction model, nitrogen high-efficiency prediction model, and high-nitrogen high-efficiency prediction model specifically includes: performing convolution operations on the input data in different directions at the input layer, extracting features between input data through vertical convolution, extracting data features of the same input data at different times through horizontal convolution, and performing pooling operations after each convolution calculation to reduce computational parameters.

[0016] Preferably, the convolution operation on the input data in different directions specifically includes:

[0017] S1: Vertical convolution is performed on the input data using s convolution kernels in the input layer. The vertical convolution is calculated using the following formula:

[0018] a i,j =f(w m x i,j +z m ), m=1,2,...,s;

[0019] Where, x i,j For the element in the i-th row and j-th column of the input data, w m Z represents the convolution kernel weights. m a is the bias term of the convolution kernel. i,j Let f be the element in the i-th row and j-th column of the convolutional data, and f be the activation function.

[0020] S2: The input data is convolved by s convolution kernels and then activated by the ReLU function. After activation, s neurons are output, and each neuron contains a data matrix.

[0021] S3: Perform average pooling on s neurons, using the following formula:

[0022]

[0023] Where q represents the size of the pooling region, D and F are the length and width of the data matrix for a single neuron. Since the pooling kernel is vertical pooling, only the length of the neuron matrix will decrease. i / q,j This represents the element in the i / q-th row and j-th column of the pooling layer's output neuron matrix;

[0024] S4: The neurons output from the pooling layer are then subjected to horizontal convolution by g convolution kernels and average pooling again, resulting in the output of g neurons;

[0025] S5: The neurons output after two convolutional pooling operations serve as the input to the connection layer. The connection layer integrates the feature information represented by the elements of all neuron matrices into the neurons of the connection layer. Let the number of neurons in the connection layer be T. Each neuron output after multiple convolutional pooling operations contains k rows and l columns of elements. Each neuron in the connection layer is convolved by g [k,l] convolution kernels on the elements in each neuron matrix. The specific formula is as follows:

[0026]

[0027] Where, x k,l For the element in the k-th row and l-th column of the input data, w n For the convolution kernel weights, zn y is the bias term of the convolution kernel. u This represents the value in each neuron of the fully connected layer.

[0028] Preferably, the output layer in S5 is calculated by a linear weighted sum of the output vectors of the connection layers, as shown in the following formula:

[0029]

[0030] Where the number of input neurons is T, and the output is... This is the output result of nitrogen uptake rate in tea plants.

[0031] Preferably, the calculation steps for the fusion coefficient specifically include:

[0032] The weight values ​​corresponding to each input parameter in each trained prediction model are divided into different value ranges according to their magnitude.

[0033] Calculate the weight value probability corresponding to each value range, and calculate the weight entropy corresponding to each input parameter in each prediction model based on the weight value probability;

[0034] The fusion coefficient corresponding to the input parameters is calculated based on the weight entropy.

[0035] Preferably, determining the fusion weight of each input parameter based on the fusion coefficient specifically includes: taking the same input parameter in all prediction models as the target, and merging the weight entropy of the input parameters through the fusion coefficient to obtain the fusion weight of the target parameter in the comprehensive prediction model.

[0036] A system for assessing nitrogen uptake levels in tea plants based on nitrogen transporter activity includes:

[0037] The training set establishment module collects the activity of various nitrogen transport proteins and nitrogen uptake rate of tea plants to establish a total dataset. The data in the total dataset are classified by type using cluster analysis and corresponding training sets are established to obtain low-nitrogen high-efficiency set, nitrogen high-efficiency set, and high-nitrogen high-efficiency set.

[0038] The sub-model training module uses low-nitrogen high-efficiency set, nitrogen high-efficiency set, and high-nitrogen high-efficiency set to train CNN sub-models with the same structure. The activity of each nitrogen transport protein in tea tree is used as input data, and the nitrogen absorption rate of tea tree is used as output data to obtain low-nitrogen high-efficiency prediction model, nitrogen high-efficiency prediction model, and high-nitrogen high-efficiency prediction model.

[0039] The prediction model construction module calculates the fusion coefficient of the input parameters of each model in the low-nitrogen high-efficiency prediction model, nitrogen high-efficiency prediction model, and high-nitrogen high-efficiency prediction model; and determines the fusion weight of each input parameter based on the fusion coefficient, and constructs a comprehensive prediction model based on the fusion weight of all input parameters.

[0040] The type classification module calculates the nitrogen absorption rate of the tea trees to be identified using the comprehensive prediction model, obtains the nitrogen absorption rate to be measured, and classifies the tea trees to be identified into types according to the cluster analysis method.

[0041] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a method and system for evaluating the nitrogen absorption type of tea trees based on nitrogen transporter activity. By utilizing deep learning models and comprehensive analysis methods, the accuracy and reliability of the evaluation results are improved. It comprehensively considers parameters such as nitrogen transporter activity of tea trees, and through the integrated prediction of multiple sub-models, it can comprehensively analyze the nitrogen absorption of tea trees, more fully reveal the characteristics of nutrient absorption of tea trees, and can effectively evaluate the nitrogen absorption capacity of unknown tea tree species. Through iterative optimization and model updates, it can obtain the latest nitrogen absorption status of tea trees in a timely manner, which helps to adjust nutrient management strategies in a timely manner and improve agricultural production efficiency. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0043] Figure 1 A flowchart of the method steps provided by the present invention;

[0044] Figure 2 This is a schematic diagram of the structure provided by the present invention. Detailed Implementation

[0045] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] This invention discloses a method for evaluating nitrogen absorption types in tea plants based on nitrogen transporter protein activity, such as... Figure 1 As shown, it includes:

[0047] A total dataset was established by collecting the activities of various nitrogen transport proteins and nitrogen uptake rates in tea plants. The data in the total dataset were then classified by type using cluster analysis, and corresponding training sets were established to obtain low-nitrogen high-efficiency sets, nitrogen high-efficiency sets, and high-nitrogen high-efficiency sets.

[0048] We trained CNN sub-models with the same structure using low-nitrogen high-efficiency sets, nitrogen high-efficiency sets, and high-nitrogen high-efficiency sets respectively. We used the activity of various nitrogen transport proteins in tea trees as input data and the nitrogen absorption rate of tea trees as output data to obtain low-nitrogen high-efficiency prediction models, nitrogen high-efficiency prediction models, and high-nitrogen high-efficiency prediction models.

[0049] Calculate the fusion coefficient of the input parameters of each model in the low-nitrogen high-efficiency prediction model, the nitrogen high-efficiency prediction model, and the high-nitrogen high-efficiency prediction model; determine the fusion weight of each input parameter based on the fusion coefficient, and construct a comprehensive prediction model based on the fusion weight of all input parameters;

[0050] The nitrogen uptake rate of the tea trees to be identified was calculated using a comprehensive prediction model to obtain the nitrogen uptake rate to be measured, and the tea trees to be identified were classified into different types based on cluster analysis.

[0051] In one specific embodiment, the clustering analysis method is k-means, which specifically includes: initializing the classification, initializing three cluster centers, using Euclidean distance to characterize the similarity between samples, calculating the distance from all samples to each initial cluster center, and classifying the samples into categories according to the criterion of minimum distance; updating the cluster centers, calculating the mean of the samples belonging to each category as the new cluster center for that category, and then re-obtaining the sum of squared distances from all samples to the new cluster centers of their respective categories; determining whether the cluster centers have changed between the two clustering analyses, if they no longer change, then the clustering ends, otherwise the cluster centers are updated repeatedly until the clustering ends.

[0052] In one specific embodiment, it includes:

[0053] Initialize the classification, initialize three cluster centers, and use Euclidean distance to characterize the similarity between samples. Calculate the similarity of all samples. To each initial cluster center C k The distance J(C) between (k=1,2,3) k The samples X are divided according to the minimum distance criterion. d =[Jv d UCS d Category C k c represents the total number of samples.

[0054] Update the cluster centers by calculating the mean {u1,u2,u3} of the samples belonging to each category {C1,C2,,C3} as the new cluster center C′={C1′,C2′,…,C3′} for that category, and then re-obtain the sum of squared distances J(C') from all samples to the new cluster center of their respective categories;

[0055]

[0056] In the formula, X dThis represents the d-th nitrogen transporter activity parameter sample, and X d =[Jv d UCS d ], u k C represents the cluster center of the k-th cluster. k The mean of the samples belonging to the middle, where k = 1, 2, 3;

[0057]

[0058] Determine whether the cluster centers C′ and J(C′) have changed between the two cluster analyses. If they no longer change, the clustering ends; otherwise, update the cluster centers repeatedly until the clustering ends.

[0059] In one specific embodiment, the CNN sub-model specifically includes: selecting six input parameters related to the nitrogen absorption rate of tea trees, normalizing them, arranging the processed input parameter data in a time series as input data, and constructing six input data time series input layers; performing convolution operations on the input data, pooling the input data after convolution operations, and performing fully connected operations on the output data after multiple convolution and pooling operations to complete the training of the low-nitrogen high-efficiency prediction model, the nitrogen high-efficiency prediction model, and the high-nitrogen high-efficiency prediction model.

[0060] In one specific embodiment, the CNN sub-model specifically includes an input layer, a convolutional layer, a pooling layer, a connection layer, and an output layer, which are sequentially interconnected.

[0061] The input layer is configured to take in six input parameters related to the nitrogen uptake rate of tea trees and to normalize these six input parameters.

[0062] The convolutional layer uses a sliding window technique to form a time series by correlating six parameter data from a past time interval with the nitrogen absorption rate of tea trees at future times, and then slides the window at unit time intervals.

[0063] The time series of the input data for the CNN sub-model is as follows:

[0064] Q I ={Q I (t),Q I (t+1),...,Q I (t+n)}, I=1,2,3,...,6;

[0065] The pooling layer selects a sliding window of width n from time t to time t+n, and inputs the 6 parameter data row by row to form a matrix.

[0066] The neurons output after two convolutional pooling operations serve as the input to the fully connected layer. The fully connected layer integrates the feature information represented by the elements of all neuron matrices into the neurons of the fully connected layer.

[0067] In one specific embodiment, the method further includes updating the weight parameters using backpropagation, updating the weights and biases of the convolutional layer by calculating the first-order moment estimate and the second-order moment estimate of the gradient, thereby achieving fine-tuning of the weight parameters.

[0068] In one specific embodiment, the input parameters specifically include ammonium nitrogen transporters: CsAMT1.1, CsAMT1.2; and nitrate nitrogen transporters: CsNRT2.1, CsNRT2.2, CsNRT2.3, CsNRT2.4.

[0069] In one specific embodiment, training the low-nitrogen high-efficiency prediction model, the nitrogen high-efficiency prediction model, and the high-nitrogen high-efficiency prediction model specifically includes: performing convolution operations on the input data in different directions at the input layer; vertical convolution extracts features between input data; horizontal convolution extracts data features of the same input data at different times; and pooling operations are performed after each convolution calculation to reduce computational parameters.

[0070] In one specific embodiment, performing convolution operations on the input data in different directions specifically includes:

[0071] S1: Vertical convolution is performed on the input data using s convolution kernels in the input layer. The vertical convolution is calculated using the following formula:

[0072] a i,j =f(w m x i,j +z m ), m=1,2,...,s;

[0073] Where, x i,j For the element in the i-th row and j-th column of the input data, w m Z represents the convolution kernel weights. m a is the bias term of the convolution kernel. i,j Let f be the element in the i-th row and j-th column of the convolutional data, and f be the activation function.

[0074] S2: The input data is convolved by s convolution kernels and then activated by the ReLU function. After activation, s neurons are output, and each neuron contains a data matrix.

[0075] S3: Perform average pooling on s neurons, using the following formula:

[0076]

[0077] Where q represents the size of the pooling region, D and F are the length and width of the data matrix for a single neuron. Since the pooling kernel is vertical pooling, only the length of the neuron matrix will decrease. i / q,j This represents the element in the i / q-th row and j-th column of the pooling layer's output neuron matrix;

[0078] S4: The neurons output from the pooling layer are then subjected to horizontal convolution by g convolution kernels and average pooling again, resulting in the output of g neurons;

[0079] S5: The neurons output after two convolutional pooling operations serve as the input to the connection layer. The connection layer integrates the feature information represented by the elements of all neuron matrices into the neurons of the connection layer. Let the number of neurons in the connection layer be T. Each neuron output after multiple convolutional pooling operations contains h rows and l columns of elements. Each neuron in the connection layer is convolved by g [h, l] convolution kernels on the elements in each neuron matrix. The specific formula is as follows:

[0080]

[0081] Where, x h,l The element in the h-th row and l-th column of the input layer data; w n For the convolution kernel weights, z n This is the bias term of the convolution kernel; the kernel size is the same as the neuron matrix size, and the output after convolution is a single numerical value, y. u This represents the numerical value in each neuron of the fully connected layer, which is the sum of the g values ​​obtained after convolution.

[0082] In one specific embodiment, the output layer in S5 is calculated by a linear weighted sum of the output vectors of the connection layers, as shown in the following formula:

[0083]

[0084] The number of input neurons is T. For the output results of nitrogen uptake rate in tea plants, the corresponding prediction models for low-nitrogen high-efficiency, nitrogen high-efficiency, and high-nitrogen high-efficiency are: They are respectively and

[0085] In one specific embodiment, the calculation steps for the fusion coefficient specifically include:

[0086] The weight values ​​corresponding to each input parameter in each trained prediction model are divided into different value ranges according to their magnitude.

[0087] Calculate the probability of the weight value o for each value range. r The weight entropy for each input parameter in each prediction model is calculated based on the probability of the weight value, as follows:

[0088]

[0089] Where R is the number of value ranges, o r This represents the probability of the weight values ​​distributed in the r-th range. It should be understood that H(w) corresponds to the weight. I The smaller the value, the smaller the change in that weight, and the less information it contains.

[0090] The fusion coefficient for the corresponding input parameters is calculated based on the weight entropy. The specific calculation of the fusion coefficient is as follows:

[0091]

[0092] Where A and B are hyperparameters, Is parameter I in the CNN prediction sub-model The fusion coefficient in.

[0093] In one specific embodiment, determining the fusion weight of each input parameter based on the fusion coefficient specifically includes: taking the same input parameter in all prediction models as the target, weighting and merging the weight entropies of the input parameters using the fusion coefficient to obtain the fusion weight of the target parameter in the comprehensive prediction model. The fusion weight calculation is as follows:

[0094]

[0095] in, The weights corresponding to the input parameter I in the fusion model. For the comprehensive prediction model, Y represents the number of CNN prediction sub-models, and Y = 3.

[0096] This invention employs a weighted fusion approach to integrate multiple CNN prediction sub-models. The weights are weighted and merged based on the weight entropy corresponding to each parameter in different models. This ensures that weights with high entropy and high information content have a greater impact on the fused weights, while weights with low entropy and low information content have a smaller impact on the fused weights. This avoids the practice in related technologies of removing weights with smaller influence factors, thereby improving the recognition accuracy and robustness of the integrated prediction model obtained by fusion.

[0097] The comprehensive prediction model is as follows:

[0098]

[0099] Where, N τInput parameters for the protein activity of tea plants to be determined include the activity data of six nitrogen transporters: ammonium nitrogen transporters CsAMT1.1 and CsAMT1.2; and nitrate nitrogen transporters CsNRT2.1, CsNRT2.2, CsNRT2.3, and CsNRT2.4.

[0100] A system for assessing nitrogen absorption levels in tea plants based on nitrogen transporter protein activity, such as... Figure 2 As shown, it includes:

[0101] The training set building module collects the activity of various nitrogen transport proteins and nitrogen uptake rate of tea plants to build a total dataset. The data in the total dataset are classified by type using cluster analysis and corresponding training sets are built to obtain low-nitrogen high-efficiency set, nitrogen high-efficiency set, and high-nitrogen high-efficiency set.

[0102] The sub-model training module uses low-nitrogen high-efficiency set, nitrogen high-efficiency set, and high-nitrogen high-efficiency set to train CNN sub-models with the same structure. The activity of each nitrogen transport protein in tea tree is used as input data, and the nitrogen absorption rate of tea tree is used as output data to obtain low-nitrogen high-efficiency prediction model, nitrogen high-efficiency prediction model, and high-nitrogen high-efficiency prediction model.

[0103] The prediction model building module calculates the fusion coefficients of the input parameters for each model in the low-nitrogen high-efficiency prediction model, the nitrogen high-efficiency prediction model, and the high-nitrogen high-efficiency prediction model; and determines the fusion weight of each input parameter based on the fusion coefficients, and builds a comprehensive prediction model based on the fusion weights of all input parameters.

[0104] The type classification module calculates the nitrogen uptake rate of the tea trees to be identified using a comprehensive prediction model, obtains the nitrogen uptake rate to be measured, and classifies the tea trees to be identified into types based on cluster analysis.

[0105] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0106] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for assessing nitrogen absorption types in tea plants based on nitrogen transporter protein activity, characterized in that, include: A total dataset was established by collecting the activities of various nitrogen transport proteins and nitrogen uptake rates in tea plants. The data in the total dataset were then classified by type using cluster analysis, and corresponding training sets were established to obtain low-nitrogen high-efficiency sets, nitrogen high-efficiency sets, and high-nitrogen high-efficiency sets. We trained CNN sub-models with the same structure using low-nitrogen high-efficiency sets, nitrogen high-efficiency sets, and high-nitrogen high-efficiency sets respectively. We used the activity of various nitrogen transport proteins in tea trees as input data and the nitrogen absorption rate of tea trees as output data to obtain low-nitrogen high-efficiency prediction models, nitrogen high-efficiency prediction models, and high-nitrogen high-efficiency prediction models. The CNN sub-model specifically includes: selecting 6 input parameters related to the nitrogen absorption rate of tea trees, normalizing them, arranging the processed input parameter data in a time series as input data, and constructing 6 input data time series input layers; performing convolution operations on the input data, pooling the input data after convolution operations, and performing fully connected operations on the output data after multiple convolution and pooling operations to complete the training of the low-nitrogen high-efficiency prediction model, the nitrogen high-efficiency prediction model, and the high-nitrogen high-efficiency prediction model. The input parameters specifically include ammonium nitrogen transporters: CsAMT1.1, CsAMT1.2; nitrate nitrogen transporters: CsNRT2.1, CsNRT2.2, CsNRT2.3, CsNRT2.4; Calculate the fusion coefficient of the input parameters of each model in the low-nitrogen high-efficiency prediction model, nitrogen high-efficiency prediction model, and high-nitrogen high-efficiency prediction model; and determine the fusion weight of each input parameter based on the fusion coefficient, and construct a comprehensive prediction model based on the fusion weight of all input parameters; The nitrogen uptake rate of the tea trees to be identified is calculated using the comprehensive prediction model to obtain the nitrogen uptake rate to be measured, and the tea trees to be identified are classified into types according to the cluster analysis method.

2. The method for evaluating nitrogen absorption type in tea plants based on nitrogen transporter activity according to claim 1, characterized in that, The clustering analysis method is k-means, which specifically includes: initializing the classification by initializing three cluster centers, using Euclidean distance to characterize the similarity between samples, calculating the distance from all samples to each initial cluster center, and classifying the samples into categories according to the criterion of minimum distance; updating the cluster centers by calculating the mean of the samples belonging to each category as the new cluster center for that category, and then re-obtaining the sum of squared distances from all samples to the new cluster centers of their respective categories; determining whether the cluster centers have changed between the two clustering analyses. If they no longer change, the clustering ends; otherwise, the cluster centers are updated repeatedly until the clustering ends.

3. The method for evaluating nitrogen absorption type in tea plants based on nitrogen transporter activity according to claim 1, characterized in that, The training of the low-nitrogen high-efficiency prediction model, nitrogen high-efficiency prediction model, and high-nitrogen high-efficiency prediction model specifically includes: performing convolution operations on the input data in different directions at the input layer; vertical convolution to extract features between input data; horizontal convolution to extract data features of the same input data at different times; and performing pooling operations after each convolution calculation to reduce the number of computational parameters.

4. The method for evaluating the nitrogen absorption type of tea plants based on nitrogen transporter activity according to claim 3, characterized in that, The specific steps of performing convolution operations on the input data in different directions include: S1: Adopt Each convolutional kernel performs a vertical convolution on the input data in the input layer. The vertical convolution is calculated using the following formula: ； in, For the input data Line number Column elements, Indicates the convolution kernel weights. The bias term for the convolution kernel. The first convolutional data Line number Column elements, For activation functions; S2: Input data passes through After convolution by multiple kernels, the system is activated by the ReLU function, and the output is then... There are 10 neurons, and each neuron contains a data matrix; S3: Yes The average pooling is performed on each neuron, and the specific formula is as follows: ； in, Indicates the size of the pooling region. and This represents the length and width of a single neuron's data matrix. Since the pooling kernel uses vertical pooling, only the length of the neuron matrix will decrease. The element representing the output neuron matrix element of the pooling layer Line number Column elements; S4: The neurons output from the pooling layer are then processed by... Each convolutional kernel performs horizontal convolution and then average pooling again, outputting... One neuron; S5: The neurons output after two convolutional pooling operations serve as the input to the connection layer. The connection layer integrates the feature information represented by the elements of all neuron matrices into the neurons of the connection layer. Let the number of neurons in the connection layer be... Each neuron output after multiple convolutional pooling processes contains OK Column elements, each neuron in the connection layer consists of indivual The convolution kernel performs convolution on the elements in each neuron matrix, as shown in the following formula: ； in, For the input data OK Column elements, For convolution kernel weights, The bias term for the convolution kernel. This represents the value in each neuron of the fully connected layer.

5. The method for evaluating the nitrogen absorption type of tea plants based on nitrogen transporter activity according to claim 4, characterized in that, The output layer in S5 is calculated by a linear weighted sum of the output vectors of the connection layers, as shown in the following formula: ； The number of input neurons is The output is , This is the output result of nitrogen uptake rate in tea plants.

6. The method for evaluating nitrogen absorption type in tea plants based on nitrogen transporter activity according to claim 1, characterized in that, The calculation steps for the fusion coefficient specifically include: The weight values ​​corresponding to each input parameter in each trained prediction model are divided into different value ranges according to their magnitude. Calculate the weight value probability corresponding to each value range, and calculate the weight entropy corresponding to each input parameter in each prediction model based on the weight value probability; The fusion coefficient corresponding to the input parameters is calculated based on the weight entropy.

7. The method for evaluating nitrogen absorption type in tea plants based on nitrogen transporter activity according to claim 6, characterized in that, Determining the fusion weight of each input parameter based on the fusion coefficient specifically includes: taking the same input parameter in all prediction models as the target, weighting and merging the weight entropy of the input parameters through the fusion coefficient to obtain the fusion weight of the target parameter in the comprehensive prediction model.

8. A system for assessing the nitrogen absorption level of tea plants based on nitrogen transporter activity, employing the method for assessing the nitrogen absorption type of tea plants based on nitrogen transporter activity as described in any one of claims 1-7, characterized in that, include: The training set establishment module collects the activity of various nitrogen transport proteins and nitrogen uptake rate of tea plants to establish a total dataset. The data in the total dataset are classified by type using cluster analysis and corresponding training sets are established to obtain low-nitrogen high-efficiency set, nitrogen high-efficiency set, and high-nitrogen high-efficiency set. The sub-model training module uses low-nitrogen high-efficiency set, nitrogen high-efficiency set, and high-nitrogen high-efficiency set to train CNN sub-models with the same structure. The activity of each nitrogen transport protein in tea tree is used as input data, and the nitrogen absorption rate of tea tree is used as output data to obtain low-nitrogen high-efficiency prediction model, nitrogen high-efficiency prediction model, and high-nitrogen high-efficiency prediction model. The prediction model construction module calculates the fusion coefficient of the input parameters of each model in the low-nitrogen high-efficiency prediction model, nitrogen high-efficiency prediction model, and high-nitrogen high-efficiency prediction model; and determines the fusion weight of each input parameter based on the fusion coefficient, and constructs a comprehensive prediction model based on the fusion weight of all input parameters. The type classification module calculates the nitrogen absorption rate of the tea trees to be identified using the comprehensive prediction model, obtains the nitrogen absorption rate to be measured, and classifies the tea trees to be identified into types according to the cluster analysis method.

Images