Comprehensive evaluation method for quality traits of plant breeding

Abstract

The present invention provides a comprehensive evaluation method for quality traits of breeding plants, which relates to the technical field of quality trait evaluation, including: obtaining plant growth information of target breeding plants, including a multi-dimensional growth parameter set of multiple growth cycles; performing feature classification and labeling to obtain plant growth feature parameters; constructing a quality trait comprehensive evaluation model library; matching and calling a target comprehensive evaluation model; performing real-time monitoring of the physiological state of target breeding plants to obtain real-time monitoring data of the physiological state; performing comprehensive evaluation to obtain a comprehensive evaluation result of the quality traits of target breeding plants. The present invention solves the technical problems that traditional evaluation methods usually focus on a single growth characteristic of plants, ignore the complex relationship between different plant traits, and that existing evaluation models are difficult to cover plants of different varieties, resulting in poor comprehensiveness and universality of the evaluation.

Description

Technical Field

[0001] The present invention relates to the technical field of quality trait evaluation, and in particular to a comprehensive evaluation method for quality traits of plant breeding. Background Art

[0002] The evaluation of plant quality traits is a key link in plant breeding and agricultural production. By accurately evaluating plant quality traits, the yield and quality of crops can be improved, thereby promoting the stability and sustainability of agricultural production. However, plant growth and quality traits involve growth parameters in multiple dimensions, which show highly dynamic changes in different growth cycles. How to effectively integrate these multi-dimensional data to provide comprehensive and accurate quality evaluation results is a technical difficulty. In addition, plants of different varieties and different growth environments have different growth characteristics. Existing evaluation models are difficult to cover all types of plants and cannot adapt to the diversity of plant species and changes in environmental conditions, resulting in unstable evaluation results. Summary of the invention

[0003] This application provides a comprehensive evaluation method for quality traits of breeding plants, aiming to solve the technical problems that traditional evaluation methods usually focus on a single growth characteristic of plants, ignore the complex relationship between different plant traits, and existing evaluation models are difficult to cover different varieties of plants, resulting in poor comprehensiveness and universality of the evaluation.

[0004] The present application discloses a method for comprehensive evaluation of quality traits of breeding plants, the method comprising: obtaining plant growth information of a target breeding plant, wherein the plant growth information comprises a multi-dimensional growth parameter set of multiple growth cycles; performing feature classification and labeling on the plant growth information to obtain plant growth feature parameters; constructing a quality trait comprehensive evaluation model library for comprehensive evaluation of quality traits of breeding plants; matching and calling a target comprehensive evaluation model from the quality trait comprehensive evaluation model library according to the plant growth feature parameters; performing real-time monitoring of the physiological state of the target breeding plant to obtain real-time monitoring data of the physiological state; and performing comprehensive evaluation on the real-time monitoring data of the physiological state based on the target comprehensive evaluation model to obtain a comprehensive evaluation result of the quality traits of the target breeding plant.

[0005] One or more technical solutions provided in this application have at least the following technical effects or advantages: By obtaining multi-dimensional growth information of target breeding plants in multiple growth cycles, comprehensive plant growth data is provided. This multi-dimensional data collection method can reflect the growth status of plants in detail and ensure the accuracy and comprehensiveness of the evaluation. Plant growth information is characterized and classified to further obtain plant growth characteristic parameters. By classifying and marking different growth characteristics, a detailed description of plants is achieved, which is helpful for subsequent quality evaluation and model construction. Through integrated learning technology, independent evaluation branches are established for different types of plants. Each branch is trained according to the historical data of different plants to obtain an accurate comprehensive evaluation model library of quality traits for different plant varieties, which can adapt to the quality evaluation of various plants. By monitoring the real-time physiological status of the target plants, the current health status of the plants can be obtained in real time. Based on the real-time monitoring data, the input of the evaluation model can be dynamically adjusted to ensure the timeliness and accuracy of the evaluation results. Based on the target comprehensive evaluation model, the real-time monitoring data of the physiological status is comprehensively evaluated, and finally a comprehensive comprehensive evaluation result of quality traits is output. The result can comprehensively reflect the growth status, health level and other characteristics of the plants, providing a scientific basis for breeders.

[0006] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 A schematic flow chart of a method for comprehensive evaluation of quality traits of plant breeding is provided for the embodiments of the present application.

[0008] Figure 2 A schematic flow chart of obtaining plant growth characteristic parameters in a comprehensive evaluation method for quality traits of breeding plants is provided for the embodiments of the present application. DETAILED DESCRIPTION

[0009] The embodiments of the present application provide a comprehensive evaluation method for quality traits of plant breeding, which solves the technical problems that traditional evaluation methods usually focus on a single growth characteristic of plants, ignore the complex relationship between different plant traits, and existing evaluation models are difficult to cover different varieties of plants, resulting in poor comprehensiveness and universality of the evaluation.

[0010] After introducing the basic principles of the present application, various non-limiting implementation methods of the present application will be specifically described below in conjunction with the drawings in the specification.

[0011] like Figure 1 As shown, the present application embodiment provides a method for comprehensive evaluation of quality traits of breeding plants, the method comprising: The plant growth information of the target breeding plant is obtained, wherein the plant growth information includes a multi-dimensional growth parameter set of multiple growth cycles.

[0012] The growth cycle of plants can vary according to different plant species and breeding goals. Usually, the growth cycle of plants is divided into several stages, such as germination, seedling, growth, and maturity. Each cycle reflects the different growth characteristics and physiological states of plants. Multidimensional growth parameters refer to the quantitative evaluation of the growth status of plants from multiple angles. Common growth parameters include plant height, number of leaves, chlorophyll content, root density, etc. Plant growth data can be automatically obtained through traditional physical measurements, such as using calipers to directly measure plant height, stem diameter, etc., or using sensors such as optical sensors, laser scanners, pressure sensors, etc. The multidimensional growth parameters in each acquisition cycle are integrated into a set to obtain a multidimensional growth parameter set.

[0013] The plant growth information is characterized by classification and labeling to obtain plant growth characteristic parameters.

[0014] The data in the plant growth information are classified and labeled according to different dimensions, so as to convert the original growth data into characteristic parameters that can be used to describe the growth status of the plant. These parameters can effectively express the key traits and physiological characteristics of plants at different growth stages. The characteristic parameters obtained after classification and labeling provide a basis for subsequent evaluation and analysis.

[0015] Construct a comprehensive evaluation model library for quality traits of breeding plants.

[0016] Collect historical growth information of different types of breeding plants, including plant growth parameters at different growth stages. Based on the ensemble learning method, randomly extract samples from the growth data of each type of breeding plants, and train the model in combination with different growth parameters. Different training sets are used to construct various evaluation branches to ensure that each branch can independently evaluate different types of plants. For example, multiple evaluation branch models can be constructed according to the different growth stages and varieties of plants. Each model judges the growth status and quality of the plant by integrating different features. After generating multiple separate evaluation models through the ensemble learning method, these models are integrated into a complete evaluation model library. Each model can perform quality evaluation based on the growth characteristics of different types of plants.

[0017] According to the plant growth characteristic parameters, a target comprehensive evaluation model is matched and called from the quality trait comprehensive evaluation model library.

[0018] According to the growth characteristic parameters of the target plant, such as plant height, number of leaves, chlorophyll content, etc., all sample plants in the model library are traversed, and the growth characteristic parameters of the target plant and the sample plant are compared. Similarity calculation methods, such as Euclidean distance and cosine similarity, are used to measure the similarity between the target plant and each sample plant, generate a similarity matrix, record the similarity information between the target plant and each sample plant, and select plants that meet the similarity threshold by comparing the similarities of different features, and extract their corresponding evaluation models as the target comprehensive evaluation model. In this way, a more accurate quality evaluation can be obtained based on the personalized characteristics of the plant.

[0019] The target breeding plant is subjected to real-time monitoring of its physiological state to obtain real-time monitoring data of its physiological state.

[0020] Spectral imaging technology or infrared thermal imaging technology is used to obtain plant physiological status data in real time. For example, by monitoring indicators such as plant height, number of leaves, root density, etc., the growth of plants can be evaluated. Monitoring data is collected at fixed time intervals, such as hours or days, so as to track plant growth changes in real time and capture any changes in physiological status. The real-time collected data is integrated to obtain real-time monitoring data of the physiological status.

[0021] Based on the target comprehensive evaluation model, the real-time monitoring data of the physiological state is comprehensively evaluated to obtain a comprehensive evaluation result of the quality traits of the target breeding plant.

[0022] The physiological status data monitored in real time is used as input and passed into the constructed target comprehensive evaluation model. The model generates evaluation results based on historical data and the current status of the plants. The evaluation results include a comprehensive score that reflects the overall health status and quality traits of the target plants at the current stage. These evaluation results provide a scientific basis for plant quality management and further breeding.

[0023] Furthermore, if Figure 2 As shown, the method of performing feature classification marking on the plant growth information to obtain plant growth feature parameters includes: Based on the multidimensional growth parameter set, multidimensional feature classification is performed to obtain plant growth feature classification information; based on the plant growth feature classification information, a plant growth feature coordinate system is constructed, wherein the plant growth feature coordinate system is a multidimensional coordinate system, and the coordinate axes correspond one-to-one to the plant growth feature classification information; the plant growth feature coordinate system is subjected to regional labeling classification to obtain a regional labeling classification result; the multidimensional growth parameter set is input into the plant growth feature coordinate system to obtain a plant growth feature vector; mapping and matching the plant growth feature vector and the regional labeling classification result are performed to obtain a plant growth classification label; and the plant growth feature parameters are determined according to the plant growth classification label.

[0024] The multidimensional growth parameter set includes multiple growth characteristics, such as plant height, number of leaves, chlorophyll content, root density, etc. Each parameter is a measure of the growth status of the plant. These parameters of each plant form a multidimensional data point. For each growth parameter, the continuous growth parameter is converted into discrete categories by setting a threshold to form plant growth characteristic classification information. For example, plant height can be divided into different categories such as low, medium, and high.

[0025] Based on the classification information of plant growth characteristics, each growth characteristic, such as height, number of leaves, chlorophyll content, root density, etc., is used as an axis in the coordinate system. In this way, each growth characteristic of the plant can correspond to a position in the coordinate system. For example, if three characteristics are selected: height, number of leaves, and root density, the coordinate system is a three-dimensional coordinate system, and each axis corresponds to height, number of leaves, and root density, respectively.

[0026] Based on the plant growth characteristic coordinate system, different regions are divided in multidimensional space. For example, the good growth area is the area with concentrated coordinates and healthy characteristics; the poor growth area is the area with discrete coordinates and poor characteristics; the abnormal area is the area where some characteristics obviously deviate from the normal range, such as abnormally high root density, etc. Clustering algorithms, such as K-means clustering, can be used to divide the growth characteristic coordinate system into regions, and the label of each region can be determined according to the distribution of plant growth characteristics.

[0027] The growth information of each plant is composed of parameters in multiple dimensions, such as plant height, number of leaves, chlorophyll content, root density, etc. Each dimension of these parameters has a value, representing the growth status of the plant in this aspect. These multi-dimensional growth parameters are input into the constructed growth characteristic coordinate system. Each dimension in the coordinate system represents a growth characteristic, and each value of the parameter set will correspond to the value of the corresponding dimension in the coordinate system. In this way, each growth characteristic of the plant is converted into a numerical value, thereby obtaining a plant growth characteristic vector, which represents the state of the plant in each growth characteristic dimension.

[0028] By calculating the position of the growth eigenvector in the coordinate system, we find out the area where it is located, and map it to a plant growth classification label based on the label of the area. The label of each area represents the growth status and health status of the plant in that area. The plant growth classification label is a qualitative assessment of the growth status of the plant, which can help us understand whether the plant is normal and healthy during its growth process.

[0029] Based on the plant growth classification labels, such as healthy growth and needs improvement, the performance of the plant at each growth stage can be determined, and its related growth characteristic parameters can be determined. These parameters will help evaluate the growth status of the plant and guide breeding and management decisions.

[0030] Furthermore, the method for constructing a comprehensive evaluation model library for quality traits of breeding plants includes: According to the historical records of breeding plants, multiple sample plant growth information sets of multiple types of breeding plants are collected; the first sample plant growth information set of the first type of breeding plants is randomly extracted, and based on ensemble learning, multiple first quality trait comprehensive evaluation branches are constructed to obtain a first quality trait comprehensive evaluation model; continue to use multiple sample plant growth information sets of other types of breeding plants, based on ensemble learning, respectively construct multiple quality trait comprehensive evaluation branches, obtain multiple quality trait comprehensive evaluation models, and construct the quality trait comprehensive evaluation model library.

[0031] Historical records of breeding plants usually come from long-term field trials, laboratory records, production practices, etc. The records include the growth status, variety type and other relevant physiological information of each plant. In order to ensure the diversity and representativeness of the data, it is necessary to collect information on multiple sample plants from multiple breeding plant categories. The records of each sample plant include data from multiple growth cycles, such as various stages from planting to maturity, forming a systematic data set.

[0032] From multiple sample plant growth information sets, sample information of the first type of breeding plants is randomly extracted. The growth information of each sample plant includes multiple growth characteristics, such as height, number of leaves, root density, chlorophyll content, etc. These characteristics provide a basis for subsequent model training.

[0033] Ensemble learning is a technique that improves the prediction performance of a model by building multiple models and combining their results. In this process, methods such as random forests and gradient boosting machines can be used. Multiple evaluation branches are constructed through ensemble learning methods. Each branch is trained based on different sample data and features. These branches will learn different growth patterns and laws. For example, multiple decision trees, support vector machines or other machine learning models are used to train the relationship between different growth characteristics and quality traits. Each evaluation branch is trained based on the growth data of sample plants. These evaluation branches will deal with the impact of different characteristics on quality traits. Multiple evaluation branches are combined to form a comprehensive first quality trait comprehensive evaluation model, which can comprehensively consider the impact of different plant growth characteristics on quality traits.

[0034] In addition to the first category of breeding plants, for each category of breeding plant data, we continue to build multiple evaluation branches based on the ensemble learning method. Each evaluation branch is trained based on specific sample data and features to predict the quality traits of each plant. Different categories of plants and their growth data will provide new perspectives for the evaluation branches and enhance the adaptability and accuracy of the evaluation model. By modeling the data of different categories of plants, we can obtain multiple comprehensive evaluation models for quality traits. Each model can be evaluated for a specific plant category. All evaluation models are integrated to form a complete evaluation model library. This library contains evaluation models for different plant categories, and the corresponding models can be called for evaluation based on the growth characteristics of the target plants.

[0035] Furthermore, the method of randomly extracting a first sample plant growth information set of the first type of breeding plants, constructing multiple first quality trait comprehensive evaluation branches based on ensemble learning, and obtaining a first quality trait comprehensive evaluation model includes: A first sample plant growth information set of a first type of breeding plants is randomly extracted, wherein the first sample plant growth information set includes a first multidimensional growth parameter set of multiple growth cycles, and the first multidimensional growth parameter set includes but is not limited to a first plant height set, a first leaf number set, a first chlorophyll content set, and a first root density set; a standard multidimensional growth parameter set of the multiple growth cycles is obtained; the first multidimensional growth parameter set is labeled and clustered according to the regional labeling classification results to obtain a growth characteristic clustering result; based on the growth characteristic clustering result, the first sample plant growth information set is divided to obtain a plant growth evaluation training data set; based on the standard multidimensional growth parameter set, each parameter type in the plant growth evaluation training data set is respectively trained with a model to obtain the multiple first quality trait comprehensive evaluation branches, and the first quality trait comprehensive evaluation model is obtained through integration.

[0036] Randomly select sample data of the first type of breeding plants. Random extraction helps reduce bias and ensure sample diversity. The extracted data contains information on multiple growth cycles. The growth parameters of each stage will provide a description of the state of the plant at that stage. The first sample plant growth information set includes data on multiple dimensions of plant growth, including plant height, number of leaves, chlorophyll content, root density, etc. The data of each cycle are integrated into a complete multi-dimensional growth parameter set to provide a basis for subsequent analysis and modeling.

[0037] The standard multi-dimensional growth parameter set is a data set based on multiple growth cycles of a standard or typical plant. These parameters represent the ideal or average performance of the plant at each growth stage. For example, the standard parameters may include: plant height set, which represents the height change of an ideal or typical plant in different cycles; leaf number set, which represents the number of leaves in each cycle, which represents the number of leaves of the plant at each stage. These standard parameter sets are summarized through a large amount of experimental data or theoretical research, and reflect the expected state of healthy plants in different growth cycles.

[0038] In the previous steps, by constructing a plant growth characteristic coordinate system and performing regional labeling classification, a label is assigned to each region. These labels represent the growth status of plants in different regions, and each plant is classified into a specific region according to its growth characteristic vector. A clustering algorithm, such as K-means clustering, is used to cluster the first multi-dimensional growth parameter set of the plant. The clustering algorithm classifies plants with similar characteristics into the same category based on the growth characteristics of each sample plant. The goal of clustering is to divide plants into different groups according to the similarity of growth characteristics, forming clustering results with common characteristics. The clustering results of each plant will correspond to its label.

[0039] By analyzing the clustering results, each group in the cluster is divided into different training data sets. Each group represents a type of plant with similar growth characteristics. The plant data in the clustering results are divided according to their characteristics (such as plant height, number of leaves, root density, etc.) and regional labels (such as healthy growth, abnormality, etc.) to form multiple different training data sets. These divided training data are used to train the evaluation model and provide data support for the establishment of the model. Through training on these data, the model can better learn the relationship between different growth characteristics and plant quality traits.

[0040] The standard multi-dimensional growth parameter set is used as a benchmark to evaluate the performance of the actual plant in each growth cycle. The evaluation results are used as training labels to train evaluation branches for each growth characteristic parameter. These evaluation branches learn the relationship between different growth parameters and quality traits. Through ensemble learning methods, such as weighted average and voting mechanism, these evaluation branches are integrated into a comprehensive first quality trait comprehensive evaluation model, which means that different evaluation models conduct comprehensive analysis on different growth parameters of the same plant, and finally obtain a comprehensive quality trait evaluation of the plant.

[0041] Furthermore, matching and calling a target comprehensive evaluation model from the quality trait comprehensive evaluation model library according to the plant growth characteristic parameters includes: Traverse all samples of the multiple types of breeding plants, and compare the similarities of plant height, leaf number, chlorophyll content, and root density with the plant growth characteristic parameters to obtain the similarities of multiple plant growth characteristic parameters; for the similarities of the multiple plant growth characteristic parameters, extract several comprehensive evaluation models of quality traits corresponding to several types of breeding plants whose similarities are greater than a preset similarity, as the target comprehensive evaluation model.

[0042] Extract all sample data of multiple types of breeding plants from the previous model library, traverse the data of each sample, and extract its growth characteristic data such as plant height, number of leaves, chlorophyll content, root density, etc. Use similarity calculation methods such as Euclidean distance and cosine similarity to compare the growth characteristics of the target plant with the growth characteristics of each sample plant. Specifically, compare the similarity of the target plant and the sample plant in each growth characteristic, and calculate the similarity to obtain the similarity value of each plant sample with the target plant. These similarity values ​​can help determine which sample plants are closest to the growth characteristics of the target plant.

[0043] Based on experience or data analysis, a preset similarity threshold is set, which determines which plant samples are considered to be similar enough to the target plant and are selected as inputs to the evaluation model. For example, the model corresponding to the plant sample with a similarity greater than a certain value, such as greater than 0.8, is selected as a candidate model. For each sample plant, check whether its similarity with the target plant in each growth characteristic is greater than a preset threshold. If so, the sample plant is considered to be a suitable match, and qualified plant samples are screened from all sample plants. The growth characteristics of these plants are similar to those of the target plant. The quality trait comprehensive evaluation model corresponding to each qualified plant sample is obtained as the target comprehensive evaluation model.

[0044] Furthermore, the method of performing a comprehensive evaluation on the real-time monitoring data of the physiological state based on the target comprehensive evaluation model to obtain a comprehensive evaluation result of the quality traits of the target breeding plant includes: Obtaining the similarities of several plant growth characteristic parameters of the several comprehensive evaluation models for quality traits; performing standardization processing on the similarities of the several plant growth characteristic parameters, and using the standardization processing results as weight coefficients of the several comprehensive evaluation models for quality traits; inputting the real-time monitoring data of the physiological state into the several comprehensive evaluation models for quality traits respectively, and outputting several comprehensive evaluation coefficients for quality traits; performing weighted summation on the several comprehensive evaluation coefficients for quality traits based on the weight coefficients of the several comprehensive evaluation models for quality traits, and obtaining the comprehensive evaluation result of quality traits.

[0045] For several comprehensive evaluation models of quality traits, each model corresponds to a similarity value, which indicates the similarity between the plant samples of the model and the growth characteristics of the target plant.

[0046] The similarities of several plant growth characteristic parameters are standardized. For example, the minimum-maximum standardization method is used to scale the similarity values ​​to the [0,1] interval. The standardized similarity values ​​are used as the weight coefficients of each model in subsequent calculations. The weight coefficients represent the importance of each model in the comprehensive evaluation. Models with high similarity will be given higher weights, and models with low similarity will be given lower weights.

[0047] The real-time monitoring data of physiological status is input into each evaluation model, and each model outputs its own comprehensive evaluation coefficient of quality traits, which represent the quality traits of the target plant in different dimensions. According to the weight coefficient of each evaluation model, the evaluation coefficient output by the corresponding model is weighted, and the weighted results are added and averaged to obtain the final comprehensive evaluation result of the quality traits of the target plant, which provides a comprehensive quality evaluation.

[0048] Furthermore, the method further comprises: Capture plant images of the target breeding plants to obtain plant image information; perform morphological feature analysis on the plant image information to obtain plant stem inclination information and plant stem diameter information; collect growth environment feature information of the target breeding plants, and evaluate the lodging resistance of the target breeding plants in combination with the growth environment feature information, the plant stem inclination information and the plant stem diameter information to obtain a lodging resistance coefficient; and add the lodging resistance coefficient to the comprehensive evaluation result of the quality traits.

[0049] Use high-resolution cameras or drones for image acquisition. Drones can be used to obtain large-scale plant images from the air, while high-resolution cameras can be used to photograph individual plants at close range to ensure that the growth characteristics and changes of the plants are captured, and plant image information is obtained through image acquisition.

[0050] Based on the pre-built plant morphological feature analyzer, the collected plant images are analyzed for morphological features to extract the plant's stem inclination and stem diameter information. The stem inclination reflects the stability of the plant, and plants with larger inclinations are more likely to fall over. The thickness of the stem is closely related to the plant's ability to resist lodging.

[0051] Collect relevant information about the plant growth environment, such as wind speed, rainfall, soil moisture, temperature, light, etc. These factors will affect the plant's resistance to lodging. Combined with the morphological characteristics of the stem and environmental characteristics data, a lodging resistance evaluation model is constructed through regression analysis and machine learning algorithms (such as decision trees and support vector machines). The lodging resistance coefficient of the target plant is calculated according to the model, which indicates the plant's resistance to lodging. Plants with a higher lodging resistance coefficient have a stronger ability to resist lodging caused by wind or other external forces.

[0052] Lodging resistance is an important aspect of plant stability and growth adaptability. Incorporating it into the comprehensive evaluation results of quality traits can more comprehensively reflect the overall quality performance of the plant. The updated evaluation results will help breeders make more accurate breeding decisions.

[0053] Furthermore, the performing morphological feature analysis on the plant image information to obtain plant stem inclination information and plant stem diameter information includes: According to the plant growth record, a sample plant image information set is obtained; the sample plant image information set is marked with stem inclination angles and stem diameters to obtain a sample plant stem inclination angle information set and a sample plant stem diameter information set; based on a convolutional neural network, the sample plant image information set is used as input, and the sample plant stem inclination angle information set and the sample plant stem diameter information set are used as outputs respectively to construct a plant stem inclination angle recognition branch and a plant stem diameter recognition branch to obtain a plant morphological feature analyzer; based on the plant morphological feature analyzer, a morphological feature analysis is performed on the plant image information to obtain the plant stem inclination angle information and the plant stem diameter information.

[0054] According to the existing plant growth records, sample plant data at different time points and growth stages are selected. Based on the collected sample plant growth records, high-definition cameras or drones are used to collect images of the sample plants to obtain a set of sample plant image information.

[0055] Image processing technology and computer vision algorithms, such as Hough transform or line detection, are used to identify the inclination angle of plant stems. The inclination angle reflects the stability of the plant. A larger inclination angle usually means a lower resistance to lodging. By calculating the inclination degree of the plant stems in the image, the stem inclination data of each sample plant is obtained and marked to form a set of sample plant stem inclination information.

[0056] Image processing techniques, such as image segmentation and edge detection, are used to determine the width of plant stems and then calculate the stem diameter. The stem diameter is usually closely related to the plant's resistance to lodging. By measuring the diameter of the sample plant stems, the stem diameter data of each sample plant is obtained and marked to form a sample plant stem diameter information set.

[0057] A convolutional neural network is used to automatically learn image features. The convolutional neural network model takes image data as input and extracts features through multiple convolutional layers, pooling layers, etc. It takes sample plant image information as input, and the sample stem inclination and stem diameter data as label input. It is trained using a labeled sample image dataset. The convolutional neural network model automatically predicts the stem inclination and diameter in new images by learning image features and corresponding labels, namely, stem inclination and stem diameter. The convolutional neural network is trained through back propagation and gradient descent algorithms so that the two branches can accurately predict the morphological characteristics of the stem. After training, the model can accurately predict the stem inclination and stem diameter of the plant based on the input image.

[0058] The new image of the target plant is input into the trained plant morphological feature analyzer, which extracts relevant morphological features from the image and outputs the plant's stem inclination based on the stem morphology in the image; based on the stem morphology in the image, it outputs the plant's stem diameter. The obtained stem inclination and stem diameter information provide support for subsequent lodging resistance evaluation and comprehensive quality evaluation.

[0059] In summary, the comprehensive evaluation method for quality traits of plant breeding provided in the embodiments of the present application has the following technical effects: By obtaining multi-dimensional growth information of target breeding plants in multiple growth cycles, comprehensive plant growth data is provided. This multi-dimensional data collection method can reflect the growth status of plants in detail and ensure the accuracy and comprehensiveness of the evaluation. Plant growth information is characterized and classified to further obtain plant growth characteristic parameters. By classifying and marking different growth characteristics, a detailed description of plants is achieved, which is helpful for subsequent quality evaluation and model construction. Through integrated learning technology, independent evaluation branches are established for different types of plants. Each branch is trained according to the historical data of different plants to obtain an accurate comprehensive evaluation model library of quality traits for different plant varieties, which can adapt to the quality evaluation of various plants. By monitoring the real-time physiological status of the target plants, the current health status of the plants can be obtained in real time. Based on the real-time monitoring data, the input of the evaluation model can be dynamically adjusted to ensure the timeliness and accuracy of the evaluation results. Based on the target comprehensive evaluation model, the real-time monitoring data of the physiological status is comprehensively evaluated, and finally a comprehensive comprehensive evaluation result of quality traits is output. The result can comprehensively reflect the growth status, health level and other characteristics of the plants, providing a scientific basis for breeders.

[0060] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for comprehensive evaluation of quality traits of plant breeding, characterized in that: The method comprises: Acquiring plant growth information of a target breeding plant, wherein the plant growth information includes a multi-dimensional growth parameter set of multiple growth cycles; Performing feature classification and marking on the plant growth information to obtain plant growth feature parameters; Construct a comprehensive evaluation model library for quality traits of breeding plants; According to the plant growth characteristic parameters, matching and calling a target comprehensive evaluation model from the quality trait comprehensive evaluation model library; Performing real-time monitoring of the physiological state of the target breeding plant to obtain real-time monitoring data of the physiological state; Based on the target comprehensive evaluation model, the real-time monitoring data of the physiological state is comprehensively evaluated to obtain a comprehensive evaluation result of the quality traits of the target breeding plant.

2. The method for comprehensive evaluation of quality traits of plant breeding according to claim 1, characterized in that: The method of performing feature classification and marking on the plant growth information to obtain plant growth feature parameters includes: Performing multi-dimensional feature classification based on the multi-dimensional growth parameter set to obtain plant growth feature classification information; Constructing a plant growth characteristic coordinate system based on the plant growth characteristic classification information, wherein the plant growth characteristic coordinate system is a multidimensional coordinate system, and the coordinate axes correspond one to one to the plant growth characteristic classification information; Performing regional labeling classification on the plant growth characteristic coordinate system to obtain a regional labeling classification result; Inputting the multi-dimensional growth parameter set into the plant growth characteristic coordinate system to obtain a plant growth characteristic vector; Mapping and matching the plant growth feature vector and the regional labeling classification result to obtain a plant growth classification label; The plant growth characteristic parameters are determined according to the plant growth classification label.

3. The method for comprehensive evaluation of quality traits of plant breeding according to claim 1, characterized in that: The method for constructing a comprehensive evaluation model library for quality traits of breeding plants includes: According to the historical records of breeding plants, a collection of plant growth information of multiple samples of various types of breeding plants is collected; Randomly extracting a first sample plant growth information set of the first type of breeding plants, constructing multiple first quality trait comprehensive evaluation branches based on ensemble learning, and obtaining a first quality trait comprehensive evaluation model; Continue to use multiple sample plant growth information sets of other types of breeding plants, and based on ensemble learning, respectively construct multiple quality trait comprehensive evaluation branches, obtain multiple quality trait comprehensive evaluation models, and construct the quality trait comprehensive evaluation model library.

4. The method for comprehensive evaluation of quality traits of plant breeding according to claim 3, characterized in that: The method of randomly extracting a first sample plant growth information set of the first type of breeding plants, constructing multiple first quality trait comprehensive evaluation branches based on ensemble learning, and obtaining a first quality trait comprehensive evaluation model includes: Randomly extracting a first sample plant growth information set of a first type of breeding plants, wherein the first sample plant growth information set includes a first multi-dimensional growth parameter set for multiple growth cycles, and the first multi-dimensional growth parameter set includes but is not limited to a first plant height set, a first leaf number set, a first chlorophyll content set, and a first root density set; Acquire a set of standard multi-dimensional growth parameters for the plurality of growth cycles; Performing labeling clustering on the first multi-dimensional growth parameter set according to the regional labeling classification result to obtain a growth feature clustering result; Based on the growth feature clustering result, dividing the first sample plant growth information set to obtain a plant growth assessment training data set; Based on the standard multi-dimensional growth parameter set, model training is performed on each parameter type in the plant growth evaluation training data set to obtain the multiple first quality trait comprehensive evaluation branches, and the first quality trait comprehensive evaluation model is obtained through integration.

5. The method for comprehensive evaluation of quality traits of plant breeding according to claim 3, characterized in that: The step of matching and calling a target comprehensive evaluation model from the quality trait comprehensive evaluation model library according to the plant growth characteristic parameters includes: Traversing all samples of the plurality of breeding plants, respectively comparing the similarity of plant height, leaf number, chlorophyll content, and root density with the plant growth characteristic parameters, to obtain the similarity of the plurality of plant growth characteristic parameters; For the similarities of the multiple plant growth characteristic parameters, several quality trait comprehensive evaluation models corresponding to several types of breeding plants with similarities greater than a preset similarity are extracted as the target comprehensive evaluation model.

6. The method for comprehensive evaluation of quality traits of plant breeding according to claim 5, characterized in that: The method of comprehensively evaluating the real-time monitoring data of the physiological state based on the target comprehensive evaluation model to obtain the comprehensive evaluation result of the quality traits of the target breeding plant includes: Obtaining similarities of several plant growth characteristic parameters of the several quality trait comprehensive evaluation models; Standardizing the similarities of the plurality of plant growth characteristic parameters, and using the standardized processing results as weight coefficients of the plurality of quality trait comprehensive evaluation models; Inputting the real-time monitoring data of the physiological state into the several comprehensive evaluation models of quality traits respectively, and outputting several comprehensive evaluation coefficients of quality traits; Based on the weight coefficients of the several comprehensive evaluation models of quality traits, the several comprehensive evaluation coefficients of quality traits are weighted and summed to obtain the comprehensive evaluation result of quality traits.

7. The method for comprehensive evaluation of quality traits of plants for breeding according to claim 1, characterized in that: The method further comprises: Collecting plant images of the target breeding plants to obtain plant image information; Performing morphological feature analysis on the plant image information to obtain plant stem inclination information and plant stem diameter information; Collecting growth environment characteristic information of the target breeding plant, and combining the growth environment characteristic information, the plant stem inclination information, and the plant stem diameter information to evaluate the lodging resistance of the target breeding plant to obtain a lodging resistance coefficient; The lodging resistance coefficient is added to the comprehensive evaluation results of the quality traits.

8. The method for comprehensive evaluation of quality traits of plant breeding according to claim 7, characterized in that: The step of performing morphological feature analysis on the plant image information to obtain plant stem inclination information and plant stem diameter information includes: According to the plant growth records, a sample plant image information set is obtained; Marking the sample plant image information set with a stem inclination angle and a stem diameter to obtain a sample plant stem inclination angle information set and a sample plant stem diameter information set; Based on a convolutional neural network, the sample plant image information set is used as input, and the sample plant stem inclination information set and the sample plant stem diameter information set are used as outputs respectively, to construct a plant stem inclination angle recognition branch and a plant stem diameter recognition branch, and obtain a plant morphological feature analyzer; Based on the plant morphological feature analyzer, the plant image information is subjected to morphological feature analysis to obtain the plant stem inclination information and the plant stem diameter information.

Images