A method for evaluating the traits of Lagerstroemia indica germplasm resources

Abstract

The present invention discloses a method for evaluating the traits of Lagerstroemia germplasm resources, which relates to the technical field of Lagerstroemia planting. A method for evaluating the traits of Lagerstroemia germplasm resources includes: extracting the morphological traits of the target Lagerstroemia variety, evaluating the morphological traits of the target Lagerstroemia variety, evaluating the genetic traits of the target Lagerstroemia variety, evaluating the environment of the target Lagerstroemia variety's germplasm resources, and comprehensively evaluating the traits of the target Lagerstroemia variety. By combining morphological trait extraction, genetic trait evaluation, and environmental evaluation, the present invention provides a comprehensive evaluation of Lagerstroemia germplasm resources, considering not only visible morphological characteristics but also genetic and environmental factors; using an image classification model and a multi-scale feature extraction model to automatically analyze a large number of images, significantly improving the efficiency and speed of trait extraction; through precise feature extraction and evaluation strategies, it can provide more refined and accurate evaluation results, helping to identify and distinguish the subtle differences between different Lagerstroemia varieties.

Description

Technical Field

[0001] The present invention relates to the technical field of Lagerstroemia indica cultivation, and particularly relates to a method for evaluating the traits of Lagerstroemia indica germplasm resources. Background Art

[0002] As an important ornamental plant, Lagerstroemia indica plays an important role in landscaping and ecological protection. The evaluation of the traits of its germplasm resources is of great significance for variety improvement, genetic diversity protection, and adaptability management. Traditional evaluation methods for Lagerstroemia indica germplasm resources often focus on the evaluation of single traits, such as only paying attention to the color or morphological characteristics of Lagerstroemia indica, while ignoring the comprehensive consideration of multiple traits; or relying on manual observation and evaluation, which is easily affected by subjective judgment, lacks quantification and standardization. There is a need for a method for evaluating the traits of Lagerstroemia indica germplasm resources by comprehensively considering multiple aspects such as the morphology, ecological adaptability, and genetic diversity of Lagerstroemia indica to provide a more comprehensive evaluation of Lagerstroemia indica germplasm resources. Summary of the Invention

[0003] The present invention aims to provide a method for evaluating the traits of Lagerstroemia indica germplasm resources, which diversely collects the trait characteristics of Lagerstroemia indica germplasm resources and accurately evaluates them.

[0004] A method for evaluating the traits of Lagerstroemia indica germplasm resources includes:

[0005] S1. Extracting the morphological traits of the target Lagerstroemia indica variety

[0006] Obtaining an image of the target Lagerstroemia indica variety to be analyzed; inputting the image of the target Lagerstroemia indica variety to be analyzed into a Lagerstroemia indica variety image classification model for analysis to obtain a classification result of the target Lagerstroemia indica variety; matching according to the classification result of the target Lagerstroemia indica variety in a Lagerstroemia indica germplasm resource trait evaluation library to obtain a trait evaluation strategy for the target Lagerstroemia indica variety; the trait evaluation strategy for the target Lagerstroemia indica variety includes morphological trait evaluation rules and genetic trait evaluation rules; the Lagerstroemia indica germplasm resource trait evaluation library contains several Lagerstroemia indica varieties and corresponding Lagerstroemia indica variety trait evaluation strategies;

[0007] Obtaining a set of morphological images of the target Lagerstroemia indica variety to be analyzed; the set of morphological images of the target Lagerstroemia indica variety to be analyzed contains M morphological images P of the target Lagerstroemia indica variety to be analyzed m , m = 1, 2,..., M; inputting the set of morphological images of the target Lagerstroemia indica variety to be analyzed into a Lagerstroemia indica variety multi-scale feature extraction model for feature extraction to obtain a set of morphological trait features X of the target Lagerstroemia indica variety, X = [X1, X2,..., X n ,..., X N ; X n represents the morphological trait feature X of the target Lagerstroemia indica variety in the set of morphological trait features X of the target Lagerstroemia indica variety n , n = 1, 2,..., N; N is the number of morphological classifications of the target Lagerstroemia indica variety;

[0008] S2. Evaluation of morphological traits of the target Lagerstroemia indica variety

[0009] Evaluate the morphological trait feature set X of the target Lagerstroemia indica variety based on the target Lagerstroemia indica variety trait evaluation strategy to obtain the morphological trait evaluation score set Y of the target Lagerstroemia indica variety, Y = [Y1, Y2, …, Y n , …, Y N ; Y n represents the morphological trait evaluation score Y of the target Lagerstroemia indica variety corresponding to the morphological trait feature X of the target Lagerstroemia indica variety n ; n

[0010] S3. Evaluation of genetic traits of the target Lagerstroemia indica variety

[0011] Obtain the genetic information of the target Lagerstroemia indica variety to be analyzed; evaluate the genetic information of the target Lagerstroemia indica variety to be analyzed based on the target Lagerstroemia indica variety trait evaluation strategy to obtain the genetic evaluation score of the target Lagerstroemia indica variety;

[0012] S4. Evaluation of the germplasm resources environment of the target Lagerstroemia indica variety

[0013] Obtain the germplasm environment information of the target Lagerstroemia indica variety to be analyzed; input the germplasm environment information of the target Lagerstroemia indica variety to be analyzed into the Lagerstroemia indica germplasm resources environment evaluation model for auxiliary analysis to obtain the environment evaluation score of the target Lagerstroemia indica variety;

[0014] S5. Comprehensive evaluation of the traits of the target Lagerstroemia indica variety

[0015] Comprehensively evaluate the genetic evaluation score of the target Lagerstroemia indica variety, the morphological trait evaluation score set Y of the target Lagerstroemia indica variety, and the environment evaluation score of the target Lagerstroemia indica variety by using the Lagerstroemia indica variety trait evaluation function to obtain the comprehensive trait evaluation score of the target Lagerstroemia indica variety; the Lagerstroemia indica variety trait evaluation function is determined by using the swarm optimization algorithm.

[0016] As a preferred technical solution of the present invention, the Lagerstroemia indica variety image classification model in step S1 includes an image preprocessing layer, an image feature extraction layer, an image classification layer, and a result output layer;

[0017] The image preprocessing layer is used to preprocess the image of the target Lagerstroemia indica variety to be analyzed to obtain the preprocessed image of the target Lagerstroemia indica variety;

[0018] The image feature extraction layer is used to extract features from the preprocessed image of the target Lagerstroemia indica variety to obtain the image features of the target Lagerstroemia indica variety;

[0019] The image classification layer is used to classify the Lagerstroemia indica variety according to the image features of the target Lagerstroemia indica variety to obtain the classification result of the target Lagerstroemia indica variety;

[0020] ​The result output layer is used to output the classification result of the target Lagerstroemia indica variety;

[0021] The specific steps for training the image classification layer in the Lagerstroemia indica variety image classification model include:

[0022] Collect several groups of Lagerstroemia indica variety image classification samples; each group of Lagerstroemia indica variety image classification samples contains Lagerstroemia indica variety images and the corresponding Lagerstroemia indica variety categories; use the corresponding Lagerstroemia indica variety category in each group of Lagerstroemia indica variety image classification samples as the target value; combine several groups of Lagerstroemia indica variety image classification samples to obtain the Lagerstroemia indica variety image classification training set;

[0023] Input the Lagerstroemia indica variety image classification training set into the Lagerstroemia indica variety image classification model to train the image classification layer, and obtain the initial image classification layer; conduct model evaluation on the initial image classification layer to obtain the model evaluation result of the initial image classification layer; if the model evaluation result of the initial image classification layer is passed, use the initial image classification layer as the image classification layer in the Lagerstroemia indica variety image classification model; otherwise, continue to perform model training using the Lagerstroemia indica variety image classification training set.

[0024] As a preferred technical solution of the present invention, the Lagerstroemia indica variety multi-scale feature extraction model in step S1 includes an image set classification layer, N morphological feature extraction layers T n and a result output layer;

[0025] The image set classification layer is used to classify the to-be-analyzed Lagerstroemia indica variety morphological image P m in the to-be-analyzed Lagerstroemia indica variety morphological image set, and obtain the Lagerstroemia indica variety morphological image classification F m ; the morphological feature extraction layer T n corresponds to the Lagerstroemia indica variety morphological image classification L n ;

[0026] For the to-be-analyzed Lagerstroemia indica variety morphological image P m , analyze the morphological feature extraction layer T m corresponding to the same Lagerstroemia indica variety morphological image classification L n as the Lagerstroemia indica variety morphological image classification F n , to obtain the target morphological feature extraction layer; input the to-be-analyzed Lagerstroemia indica variety morphological image P m into the target morphological feature extraction layer;

[0027] The N morphological feature extraction layers T n are used to extract features from the input to-be-analyzed Lagerstroemia indica variety morphological image, and obtain the target Lagerstroemia indica variety morphological trait feature X n ;

[0028] In the morphological feature extraction layer T nAmong them, the input target Lagerstroemia indica variety morphological image to be analyzed is stitched by the image stitching method to obtain the comprehensive target Lagerstroemia indica variety morphological image; the comprehensive target Lagerstroemia indica variety morphological image is subjected to feature analysis to obtain the target Lagerstroemia indica variety morphological trait feature X n ;

[0029] The result output layer is used to combine the target Lagerstroemia indica variety morphological trait features X n output in N morphological feature extraction layers T n to obtain the target Lagerstroemia indica variety morphological trait feature set X, X = [X1, X2,..., X n ,..., X N , and output the target Lagerstroemia indica variety morphological trait feature set X

[0030] As a preferred technical solution of the present invention, the specific steps of training the morphological feature extraction layer T n include:

[0031] For the morphological feature extraction layer T n :

[0032] Collect several groups of target Lagerstroemia indica variety morphological image training samples. Each group of target Lagerstroemia indica variety morphological image training samples includes the target Lagerstroemia indica variety morphological image classification L n corresponding target Lagerstroemia indica variety morphological image and the corresponding image features; combine several groups of target Lagerstroemia indica variety morphological image training samples to obtain the target Lagerstroemia indica variety morphological image training set

[0033] Input the target Lagerstroemia indica variety morphological image training set into the morphological feature extraction layer T n for training with the image features as the target to obtain the initial morphological feature extraction layer C n ; perform model evaluation on the initial morphological feature extraction layer C n to obtain the initial morphological feature extraction layer C n model evaluation result; if the initial morphological feature extraction layer C n model evaluation result is passed, then use the initial morphological feature extraction layer C n as the morphological feature extraction layer T n ; otherwise, continue model training using the target Lagerstroemia indica variety morphological image training set

[0034] Traverse all morphological feature extraction layers T n until N morphological feature extraction layers T n are constructed

[0035] As a preferred technical solution of the present invention, the specific steps of step S2 include:

[0036] Evaluate according to the morphological trait evaluation rules in the target Lagerstroemia indica variety trait evaluation strategy. There are I sets of morphological trait evaluation rules in the morphological trait evaluation rules.

[0037] For the morphological trait feature X of the target Lagerstroemia indica variety n :

[0038] Match the set of morphological trait evaluation rules with the maximum membership function for the morphological trait feature X of the target Lagerstroemia indica variety to obtain the set of morphological trait evaluation rules J n ; According to the set of morphological trait evaluation rules J n Assign the morphological trait evaluation score Y for the target Lagerstroemia indica variety n ; n ;

[0039] Evaluate for all morphological trait features X of the target Lagerstroemia indica variety n Combine all the morphological trait evaluation scores Y of the target Lagerstroemia indica variety n to obtain the set of morphological trait evaluation scores Y for the target Lagerstroemia indica variety, Y = [Y1, Y2,..., Y n ,..., Y N .

[0040] As a preferred technical solution of the present invention, the specific steps of step S3 include:

[0041] Evaluate according to the genetic trait evaluation rules in the target Lagerstroemia indica variety trait evaluation strategy. There are G ranges of genetic trait scores and corresponding gene evaluation scores in the genetic trait evaluation rules;

[0042] Judge the range of genetic trait scores of the gene information of the target Lagerstroemia indica variety to be analyzed to obtain the range of genetic trait scores of the target Lagerstroemia indica variety; Determine the gene evaluation score of the target Lagerstroemia indica variety according to the range of genetic trait scores of the target Lagerstroemia indica variety.

[0043] As a preferred technical solution of the present invention, the Lagerstroemia indica germplasm resource environment evaluation model in step S4 includes an environmental information preprocessing layer, an environmental prediction layer, and a result output layer;

[0044] The environmental information preprocessing layer is used to preprocess the germplasm environment information of the target Lagerstroemia indica variety to be analyzed to obtain the preprocessed germplasm environment information of the target Lagerstroemia indica variety;

[0045] The environmental prediction layer is used to make a prediction according to the preprocessed germplasm environment information of the target Lagerstroemia indica variety to obtain the germplasm environment prediction information of the target Lagerstroemia indica variety;

[0046] The result output layer is used to match the environmental evaluation score of the target Lagerstroemia indica variety according to the germplasm environment prediction information of the target Lagerstroemia indica variety;

[0047] The specific steps of the training environment prediction layer include:

[0048] Collect several sets of consecutive environmental prediction training samples, where each set of environmental prediction training samples contains Lagerstroemia indica variety germplasm environmental information; combine several sets of consecutive environmental prediction training samples to obtain an environmental prediction training set; use the latter set of environmental prediction training samples in two adjacent sets of environmental prediction training samples as the target value;

[0049] Use the environmental prediction training set to train the environmental prediction layer with the target value as the goal to obtain an initial environmental prediction layer; evaluate the initial environmental prediction layer to obtain the initial environmental prediction layer model evaluation result; if the initial environmental prediction layer model evaluation result passes, use the initial environmental prediction layer as the environmental prediction layer; otherwise, continue to perform model training using the environmental prediction training set.

[0050] As a preferred technical solution of the present invention, the specific steps of step S5 include:

[0051] Normalize the target Lagerstroemia indica variety gene evaluation score, the target Lagerstroemia indica variety morphological trait evaluation score set Y, and the target Lagerstroemia indica variety environmental evaluation score to obtain the standard target Lagerstroemia indica variety gene evaluation score, the standard target Lagerstroemia indica variety morphological trait evaluation score set Y', and the standard target Lagerstroemia indica variety environmental evaluation score;

[0052] Input the standard target Lagerstroemia indica variety gene evaluation score, the standard target Lagerstroemia indica variety morphological trait evaluation score set Y', and the standard target Lagerstroemia indica variety environmental evaluation score into the Lagerstroemia indica variety trait evaluation function for calculation to obtain the target Lagerstroemia indica variety trait comprehensive evaluation score;

[0053] The specific steps of constructing the Lagerstroemia indica variety trait evaluation function:

[0054] Construct K initial function individuals F k , and each initial function individual F k contains a simulated Lagerstroemia indica variety trait evaluation function for comprehensive evaluation; collect several sets of simulated Lagerstroemia indica variety comprehensive evaluation training samples; each set of simulated Lagerstroemia indica variety comprehensive evaluation training samples contains function input parameters and corresponding simulated target values;

[0055] Based on the initial function individual F k and several sets of simulated Lagerstroemia indica variety comprehensive evaluation training samples for simulation calculation to obtain the initial fitness A k ; the calculation method of the initial fitness A k is: according to the function input parameters in the simulated Lagerstroemia indica variety comprehensive evaluation training samples and the initial function individual F kPerform function calculations to obtain simulated score values; calculate the accuracy of the simulated score values and the simulated target values to obtain the simulated accuracy; take the mean of several groups of simulated accuracies as the initial fitness A k ;

[0056] Based on the initialized function individuals F k with the initial fitness A k , divide the K initialized function individuals F k into a rooster iteration population, a hen iteration population, and a chick iteration population; set the maximum number of iterations B max ; the current iteration number is b;

[0057] When performing iterations, mutate and update the initialized function individuals in the rooster iteration population, the hen iteration population, and the chick iteration population. When the current iteration number b is less than B max / 2, the chick iteration population mutates towards the initialized function individuals in the hen iteration population, and the hen iteration population mutates towards the initialized function individuals in the rooster iteration population; when the current iteration number b is greater than B max / 2, the chick iteration population and the hen iteration population mutate towards the initialized function individuals in the rooster iteration population;

[0058] When the maximum number of iterations is reached, output the initialized function individual F k with the maximum fitness, which is the optimal initialized function individual; use the simulated Lagerstroemia variety trait evaluation function in the optimal initialized function individual as the Lagerstroemia variety trait evaluation function.

[0059] The present invention has the following advantages:

[0060] 1. By combining morphological trait extraction, gene trait evaluation, and environmental evaluation, the present invention provides a comprehensive evaluation of Lagerstroemia germplasm resources, considering not only visible morphological characteristics but also genetic and environmental factors; automating the analysis of a large number of images using an image classification model and a multi-scale feature extraction model significantly improves the efficiency and speed of trait extraction; through precise feature extraction and evaluation strategies, it can provide more refined and accurate evaluation results, helping to identify and distinguish the subtle differences between different Lagerstroemia varieties; by establishing a trait evaluation library and an evaluation strategy, the standardization of the evaluation process is achieved, reducing the influence of human subjective judgment and improving the consistency and repeatability of the evaluation.

[0061] 2. By evaluating the genetic traits of Lagerstroemia indica varieties, the present invention helps to identify and protect genetic diversity, providing important information for breeding and germplasm resource protection; by evaluating the environmental resources of Lagerstroemia indica varieties, it provides information on the adaptability of Lagerstroemia indica varieties to specific environmental conditions, facilitating the selection of varieties suitable for specific environments; by using a population optimization algorithm to determine the trait evaluation function of Lagerstroemia indica varieties, the accuracy and reliability of the comprehensive evaluation score are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a schematic structural diagram of a method for evaluating the traits of Lagerstroemia indica germplasm resources adopted in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0063] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention.

[0064] A method for evaluating the traits of Lagerstroemia indica germplasm resources, as shown in Figure 1 includes:

[0065] S1. Extraction of morphological traits of the target Lagerstroemia indica variety

[0066] Obtain an image of the target Lagerstroemia indica variety to be analyzed; input the image of the target Lagerstroemia indica variety to be analyzed into a Lagerstroemia indica variety image classification model for analysis to obtain a classification result of the target Lagerstroemia indica variety; match according to the classification result of the target Lagerstroemia indica variety in the Lagerstroemia indica germplasm resource trait evaluation library to obtain a trait evaluation strategy for the target Lagerstroemia indica variety; the trait evaluation strategy for the target Lagerstroemia indica variety includes morphological trait evaluation rules and genetic trait evaluation rules; the Lagerstroemia indica germplasm resource trait evaluation library contains several Lagerstroemia indica varieties and corresponding trait evaluation strategies for Lagerstroemia indica varieties;

[0067] Obtain a set of morphological images of the target Lagerstroemia indica variety to be analyzed; the set of morphological images of the target Lagerstroemia indica variety to be analyzed contains M morphological images P m , m = 1, 2,..., M; input the set of morphological images of the target Lagerstroemia indica variety to be analyzed into a Lagerstroemia indica variety multi-scale feature extraction model for feature extraction to obtain a set of morphological trait features X of the target Lagerstroemia indica variety, X = [X1, X2,..., X n ,..., X N ; X n represents the morphological trait feature X of the target Lagerstroemia indica variety in the set of morphological trait features X of the target Lagerstroemia indica variety n , n = 1, 2,..., N; N is the number of morphological classifications of the target Lagerstroemia indica variety;

[0068] The image of the target Lagerstroemia indica variety to be analyzed and the image set of the morphological images of the target Lagerstroemia indica variety are obtained by photographing the target Lagerstroemia indica variety with a camera; the Lagerstroemia indica germplasm resource trait evaluation library is set by professional technicians according to the actual situation, and the Lagerstroemia indica germplasm resource trait evaluation library contains evaluation strategies for several Lagerstroemia indica varieties; the morphological classification of the target Lagerstroemia indica variety is set by professional technicians, such as characteristics like leaf size, petal color, stamen shape, etc.

[0069] The Lagerstroemia indica variety image classification model in step S1 includes an image preprocessing layer, an image feature extraction layer, an image classification layer, and a result output layer;

[0070] The image preprocessing layer is used to preprocess the image of the target Lagerstroemia indica variety to be analyzed to obtain a preprocessed target Lagerstroemia indica variety image;

[0071] The image feature extraction layer is used to extract features from the preprocessed target Lagerstroemia indica variety image to obtain target Lagerstroemia indica variety image features;

[0072] The image classification layer is used to classify Lagerstroemia indica varieties according to the target Lagerstroemia indica variety image features to obtain a target Lagerstroemia indica variety classification result;

[0073] The result output layer is used to output the target Lagerstroemia indica variety classification result;

[0074] The specific steps for training the image classification layer in the Lagerstroemia indica variety image classification model include:

[0075] Collect several groups of Lagerstroemia indica variety image classification samples; each group of Lagerstroemia indica variety image classification samples contains a Lagerstroemia indica variety image and the corresponding Lagerstroemia indica variety category; use the corresponding Lagerstroemia indica variety category in each group of Lagerstroemia indica variety image classification samples as the target value; combine several groups of Lagerstroemia indica variety image classification samples to obtain a Lagerstroemia indica variety image classification training set;

[0076] Input the Lagerstroemia indica variety image classification training set into the Lagerstroemia indica variety image classification model to train the image classification layer to obtain an initial image classification layer; conduct a model evaluation on the initial image classification layer to obtain an initial image classification layer model evaluation result; if the initial image classification layer model evaluation result is passed, use the initial image classification layer as the image classification layer in the Lagerstroemia indica variety image classification model; otherwise, continue to conduct model training using the Lagerstroemia indica variety image classification training set;

[0077] Through the image preprocessing layer, the automatic processing of images is achieved, including denoising, enhancing contrast, resizing, etc., laying a good foundation for subsequent feature extraction and classification; the image feature extraction layer is specifically responsible for extracting key features from the preprocessed images, and these features can effectively express the morphological differences of Lagerstroemia indica varieties and identify different Lagerstroemia indica varieties; by collecting a large number of labeled Lagerstroemia indica variety image samples and training the model using a data-driven method, the generalization ability of the model is enhanced, and the CNN model can be used as the basic model to identify the types of Lagerstroemia indica.

[0078] The multi-scale feature extraction model of Lagerstroemia indica varieties in step S1 includes an image set classification layer, N morphological feature extraction layers T n and a result output layer;

[0079] The image set classification layer is used to classify the morphological images P of the target Lagerstroemia indica variety to be analyzed in the morphological image set of the target Lagerstroemia indica variety to be analyzed m and obtain the classification F of the morphological images of the target Lagerstroemia indica variety m ; the morphological feature extraction layer T n corresponds to the classification L of the morphological images of the target Lagerstroemia indica variety n ;

[0080] For the morphological image P of the target Lagerstroemia indica variety to be analyzed m , analyze the morphological feature extraction layer T m corresponding to the same classification L of the morphological images of the target Lagerstroemia indica variety as the classification F of the morphological images of the target Lagerstroemia indica variety to be analyzed n to obtain the target morphological feature extraction layer; input the morphological image P of the target Lagerstroemia indica variety to be analyzed n into the target morphological feature extraction layer; m Input the morphological image P of the target Lagerstroemia indica variety to be analyzed into the target morphological feature extraction layer;

[0081] The N morphological feature extraction layers T n are used to extract features from the input morphological image of the target Lagerstroemia indica variety to be analyzed and obtain the morphological trait features X of the target Lagerstroemia indica variety n ;

[0082] In the morphological feature extraction layer T n , the input morphological image of the target Lagerstroemia indica variety to be analyzed is stitched using the image stitching method to obtain the comprehensive morphological image of the target Lagerstroemia indica variety; feature analysis is performed on the comprehensive morphological image of the target Lagerstroemia indica variety to obtain the morphological trait features X of the target Lagerstroemia indica variety n ;

[0083] The result output layer is used to combine the morphological trait features X n output from the N morphological feature extraction layers T n to obtain the morphological trait feature set X of the target Lagerstroemia indica variety, X = [X1, X2,..., X n, …, X N , output the morphological trait feature set X of the target Lagerstroemia indica variety;

[0084] Train the morphological feature extraction layer T n The specific steps include:

[0085] For the morphological feature extraction layer T n :

[0086] Collect several groups of morphological image training samples of the target Lagerstroemia indica variety. Each group of morphological image training samples of the target Lagerstroemia indica variety contains the morphological image classification L n of the corresponding morphological image of the target Lagerstroemia indica variety and the corresponding image features; Combine several groups of morphological image training samples of the target Lagerstroemia indica variety to obtain the morphological image training set of the target Lagerstroemia indica variety;

[0087] Input the morphological image training set of the target Lagerstroemia indica variety into the morphological feature extraction layer T n to train with the image features as the target, and obtain the initial morphological feature extraction layer C n ; Evaluate the initial morphological feature extraction layer C n to obtain the model evaluation result of the initial morphological feature extraction layer C n ; If the model evaluation result of the initial morphological feature extraction layer C n is passed, then use the initial morphological feature extraction layer C n as the morphological feature extraction layer T n ; Otherwise, continue to perform model training using the morphological image training set of the target Lagerstroemia indica variety;

[0088] Traverse all morphological feature extraction layers T n until N morphological feature extraction layers T n are constructed;

[0089] The combined use of the image set classification layer and the morphological feature extraction layer T n ensures the accurate classification and feature extraction of the morphological images of Lagerstroemia indica varieties, and helps to more accurately identify and distinguish the morphological features of different species of Lagerstroemia indica; Through the setting of N morphological feature extraction layers Tn, the model can extract features from different scales, increasing the ability to capture the morphological diversity of Lagerstroemia indica varieties; The use of the image stitching method improves the comprehensiveness of feature extraction, and helps to grasp the morphological features of Lagerstroemia indica varieties as a whole; The accurate morphological trait evaluation helps to identify Lagerstroemia indica varieties with excellent traits, providing a scientific basis for breeding and variety improvement;

[0090] S2. Evaluation of the morphological traits of the target Lagerstroemia indica variety

[0091] Evaluate the morphological trait feature set X of the target Lagerstroemia indica variety based on the trait evaluation strategy of the target Lagerstroemia indica variety to obtain the morphological trait evaluation score set Y of the target Lagerstroemia indica variety, Y = [Y1, Y2, …, Y n , …, Y N ; Y n represents the morphological trait evaluation score Y n corresponding to the morphological trait feature X n of the target Lagerstroemia indica variety;

[0092] The specific steps of step S2 include:

[0093] Evaluate according to the morphological trait evaluation rules in the trait evaluation strategy of the target Lagerstroemia indica variety. There are I morphological trait evaluation rule sets in the morphological trait evaluation rules; the I morphological trait evaluation rule sets are set by professional and technical personnel according to the actual situation;

[0094] For the morphological trait feature X n of the target Lagerstroemia indica variety:

[0095] Match the morphological trait evaluation rule set with the largest membership function for the morphological trait feature X n of the target Lagerstroemia indica variety to obtain the morphological trait evaluation rule set J n ; According to the morphological trait evaluation rule set J n allocate the morphological trait evaluation score Y n of the target Lagerstroemia indica variety;

[0096] Evaluate all the morphological trait features X n of the target Lagerstroemia indica variety, and combine all the morphological trait evaluation scores Y n to obtain the morphological trait evaluation score set Y of the target Lagerstroemia indica variety, Y = [Y1, Y2, …, Y n , …, Y N ;

[0097] For example, for the leaf morphological feature, if a certain morphological feature is "thick leaves", then calculate according to the membership function in the morphological trait evaluation rule set. The morphological trait evaluation rule set is: "excellent", "good", "medium". If the leaf trait is "thick leaves", then the score is "excellent", and the corresponding score is 95;

[0098] By according to the trait evaluation strategy of the target Lagerstroemia indica variety, the personalized evaluation of each feature is realized, ensuring that the evaluation result is more in line with the specific feature; by matching the evaluation rule set with the largest membership degree of the morphological trait feature of the target Lagerstroemia indica variety, the relevance and accuracy of the evaluation are improved; allocate specific evaluation scores Y n to each morphological trait feature X n, the refined scoring of different features is achieved;

[0099] S3. Evaluation of the Gene Traits of the Target Lagerstroemia indica Variety

[0100] Obtain the gene information of the target Lagerstroemia indica variety to be analyzed; evaluate the gene information of the target Lagerstroemia indica variety to be analyzed based on the trait evaluation strategy of the target Lagerstroemia indica variety to obtain the gene evaluation score of the target Lagerstroemia indica variety;

[0101] The specific steps of step S3 include:

[0102] Evaluate according to the genetic trait evaluation rules in the trait evaluation strategy of the target Lagerstroemia indica variety. The genetic trait evaluation rules contain G ranges of genetic trait scores and corresponding gene evaluation scores; the G ranges of genetic trait scores and corresponding gene evaluation scores are set by professional and technical personnel according to the actual situation;

[0103] Judge the range of genetic trait scores of the gene information of the target Lagerstroemia indica variety to be analyzed to obtain the range of genetic trait scores of the target Lagerstroemia indica variety; determine the gene evaluation score of the target Lagerstroemia indica variety according to the range of genetic trait scores of the target Lagerstroemia indica variety;

[0104] For example, the effective number of alleles of genetic diversity reaches 1.52, within a certain range of genetic trait scores, and the corresponding gene evaluation score is 90;

[0105] The ranges of genetic trait scores and gene evaluation scores set by professional and technical personnel according to the actual situation ensure the scientificity and professionalism of the evaluation; the quantitative scoring of gene information through specific evaluation rules improves the accuracy and measurability of genetic trait evaluation; setting score ranges for different genetic traits allows for personalized evaluation of different target Lagerstroemia indica varieties; through genetic trait evaluation, the adaptability of different Lagerstroemia indica varieties to specific environmental conditions can be analyzed, providing a basis for the selection of planting environments;

[0106] S4. Evaluation of the Germplasm Resources Environment of the Target Lagerstroemia indica Variety

[0107] Obtain the germplasm environment information of the target Lagerstroemia indica variety to be analyzed; input the germplasm environment information of the target Lagerstroemia indica variety to be analyzed into the Lagerstroemia indica germplasm resources environment evaluation model for auxiliary analysis to obtain the environment evaluation score of the target Lagerstroemia indica variety;

[0108] The Lagerstroemia indica germplasm resources environment evaluation model in step S4 includes an environment information preprocessing layer, an environment prediction layer, and a result output layer;

[0109] The environment information preprocessing layer is used to preprocess the germplasm environment information of the target Lagerstroemia indica variety to be analyzed to obtain the preprocessed germplasm environment information of the target Lagerstroemia indica variety;

[0110] The environmental prediction layer is used to make predictions based on the preprocessed germplasm environmental information of the target Lagerstroemia indica variety, and obtain the predicted environmental information of the target Lagerstroemia indica variety germplasm;

[0111] The result output layer is used to match the environmental assessment score of the target Lagerstroemia indica variety according to the predicted environmental information of the target Lagerstroemia indica variety germplasm;

[0112] The specific steps for training the environmental prediction layer include:

[0113] Collect several groups of consecutive environmental prediction training samples, each group of which contains the germplasm environmental information of the Lagerstroemia indica variety; Combine several groups of consecutive environmental prediction training samples to obtain an environmental prediction training set; Use the latter group of environmental prediction training samples in two adjacent groups of environmental prediction training samples as the target value;

[0114] Use the environmental prediction training set to train the environmental prediction layer with the target value as the goal, and obtain an initial environmental prediction layer; Evaluate the initial environmental prediction layer to obtain the evaluation result of the initial environmental prediction layer model; If the evaluation result of the initial environmental prediction layer model is passed, use the initial environmental prediction layer as the environmental prediction layer; Otherwise, continue model training using the environmental prediction training set;

[0115] Through the environmental information preprocessing layer, the quality and consistency of the input data are ensured, providing accurate environmental information for subsequent analysis; The environmental prediction layer makes predictions based on the preprocessed data, improving the accuracy of predicting the environmental adaptability of the Lagerstroemia indica variety germplasm; The environmental assessment score provides data-driven decision support for growers, helping them better understand the requirements of the Lagerstroemia indica variety for specific environmental conditions; The environmental assessment score is trained based on the environmental adaptation of the Lagerstroemia indica variety, and predicts the future growth conditions of the Lagerstroemia indica variety according to the predicted environmental conditions to judge whether environmental improvement is needed;

[0116] S5. Comprehensive evaluation of the traits of the target Lagerstroemia indica variety

[0117] Comprehensively evaluate the gene evaluation score of the target Lagerstroemia indica variety, the set Y of the morphological trait evaluation scores of the target Lagerstroemia indica variety, and the environmental evaluation score of the target Lagerstroemia indica variety using the Lagerstroemia indica variety trait evaluation function to obtain the comprehensive evaluation score of the traits of the target Lagerstroemia indica variety; The Lagerstroemia indica variety trait evaluation function is determined using a swarm optimization algorithm;

[0118] The specific steps of step S5 include:

[0119] Normalize the gene evaluation score of the target Lagerstroemia indica variety, the set Y of the morphological trait evaluation scores of the target Lagerstroemia indica variety, and the environmental evaluation score of the target Lagerstroemia indica variety to obtain the standard gene evaluation score of the target Lagerstroemia indica variety, the set Y' of the standard morphological trait evaluation scores of the target Lagerstroemia indica variety, and the standard environmental evaluation score of the target Lagerstroemia indica variety;

[0120] Input the gene evaluation score of the standard target Lagerstroemia indica variety, the set Y' of the morphological trait evaluation scores of the standard target Lagerstroemia indica variety, and the environmental evaluation score of the standard target Lagerstroemia indica variety into the Lagerstroemia indica variety trait evaluation function for calculation to obtain the comprehensive evaluation score of the target Lagerstroemia indica variety traits;

[0121] Specific steps for constructing the Lagerstroemia indica variety trait evaluation function:

[0122] Construct K initial function individuals F k , and each initial function individual F k contains a simulated Lagerstroemia indica variety trait evaluation function for comprehensive evaluation; collect several groups of simulated Lagerstroemia indica variety comprehensive evaluation training samples; each group of simulated Lagerstroemia indica variety comprehensive evaluation training samples contains function input parameters and corresponding simulated target values;

[0123] Based on the initial function individual F k and several groups of simulated Lagerstroemia indica variety comprehensive evaluation training samples for simulation calculation to obtain the initial fitness A k ; The calculation method of the initial fitness A k is: According to the function input parameters in the simulated Lagerstroemia indica variety comprehensive evaluation training samples and the initial function individual F k for function calculation to obtain the simulated score value; calculate the accuracy rate of the simulated score value and the simulated target value to obtain the simulated accuracy rate; take the mean of several groups of simulated accuracy rates as the initial fitness A k ;

[0124] Based on the initial fitness A k of the initial function individual F k , divide the K initial function individuals F k into a rooster iteration population, a hen iteration population, and a chick iteration population; set the maximum number of iterations B max ; The current number of iterations is b; The maximum number of iterations is set by professional and technical personnel according to the actual situation; Input some individuals with higher initial fitness into the rooster iteration population, some individuals with medium initial fitness into the hen iteration population, and some individuals with lower initial fitness into the chick iteration population; The population optimization algorithm used to determine the Lagerstroemia indica variety trait evaluation function is the chicken swarm optimization algorithm, which can simulate the chicken swarm optimization algorithm to update the population position. The rooster iteration population consists of rooster individuals, the hen iteration population consists of hen individuals, and the chick iteration population consists of chick individuals;

[0125] During iteration, perform mutation update on the initial function individuals in the rooster iteration population, the hen iteration population, and the chick iteration population. When the current number of iterations b is less than B maxWhen it is less than or equal to B / 2, the chicken iterative population mutates towards the initialization function individuals in the hen iterative population, and the hen iterative population mutates towards the initialization function individuals in the rooster iterative population; when the current iteration number b is greater than B max / 2, the chicken iterative population and the hen iterative population mutate towards the initialization function individuals in the rooster iterative population;

[0126] During the iteration process, in the early stage, the chicken individuals update their positions towards the hen individuals, and the hen individuals update their positions towards the rooster individuals, which can maintain the breadth and diversity of the population and avoid prematurely searching for the local optimal solution; in the later stage of iteration, it is necessary to accelerate the convergence speed, and both the chicken individuals and the hen individuals update in the direction of the rooster individuals, maintaining at a relatively high fitness level, which is beneficial to the convergence of the optimal solution;

[0127] When the maximum iteration number is reached, output the initialization function individual F corresponding to the maximum fitness k , which is the optimal initialization function individual; use the simulated crape myrtle variety trait evaluation function in the optimal initialization function individual as the crape myrtle variety trait evaluation function;

[0128] Through normalization processing and comprehensive evaluation functions, a comprehensive consideration of the genes, morphological traits, and environmental adaptability of crape myrtle varieties is achieved, providing a more comprehensive trait evaluation; by calculating the accuracy rate of the simulated score value and the simulated target value, the fitness is obtained, providing a basis for the selection and optimization of the evaluation function; the iterative method is used to continuously update and optimize the evaluation function, improving the accuracy and reliability of the evaluation function; the improved chicken swarm optimization algorithm is used for position update, improving the efficiency and effect of searching for the optimal solution. In the early stage of iteration, premature convergence to the local optimal solution is avoided by maintaining population diversity; in the later stage of iteration, the convergence speed is accelerated by concentrating on searching in the high-fitness region.

[0129] It should be understood that for those of ordinary skill in the art, improvements or transformations can be made according to the above description, and all such improvements and transformations should fall within the protection scope of the appended claims of the present invention. The parts not described in detail in this specification belong to the prior art well-known to those skilled in the art.

Claims

1. A method for evaluating the traits of Lagerstroemia indica germplasm resources, characterized in that: include: S1. Extraction of morphological traits of target crape myrtle varieties Acquire the image of the target crape myrtle variety to be analyzed; Input the target crape myrtle variety image to be analyzed into the crape myrtle variety image classification model for analysis to obtain the classification result of the target crape myrtle variety; According to the classification results of the target crape myrtle varieties, the crape myrtle germplasm resource trait evaluation database is matched to obtain the target crape myrtle variety trait evaluation strategy; the target crape myrtle variety trait evaluation strategy includes morphological trait evaluation rules and genetic trait evaluation rules; the crape myrtle germplasm resource trait evaluation database includes several crape myrtle varieties and corresponding crape myrtle variety trait evaluation strategies; Obtain a set of morphological images of target crape myrtle varieties to be analyzed; the set of morphological images of target crape myrtle varieties to be analyzed contains M morphological images P of target crape myrtle varieties to be analyzed. m , m=1,2,…,M; The morphological image set of the target crape myrtle variety to be analyzed is input into the multi-scale feature extraction model of crape myrtle varieties for feature extraction, and the morphological trait feature set X of the target crape myrtle variety is obtained, where X = [X1, X2, …, X n , …, X N ];X n represents the target crape myrtle variety morphological trait X in the target crape myrtle variety morphological trait set X n , n=1, 2, …, N; N is the number of morphological categories of the target crape myrtle varieties; S2. Evaluation of morphological traits of target crape myrtle varieties Based on the target crape myrtle variety trait evaluation strategy, the target crape myrtle variety morphological trait feature set X is evaluated to obtain the target crape myrtle variety morphological trait evaluation score set Y, Y = [Y1, Y2, …, Y n , …, Y N ]; Y n Indicates the morphological characteristics of the target crape myrtle variety X n Corresponding morphological trait evaluation score Y of target crape myrtle variety n ; S3. Evaluation of genetic traits of target crape myrtle varieties Obtaining the genetic information of the target crape myrtle variety to be analyzed; evaluating the genetic information of the target crape myrtle variety to be analyzed based on the target crape myrtle variety trait evaluation strategy, and obtaining the genetic evaluation score of the target crape myrtle variety; S4. Environmental assessment of target crape myrtle varieties germplasm resources Obtaining germplasm environmental information of the target crape myrtle variety to be analyzed; inputting the germplasm environmental information of the target crape myrtle variety to be analyzed into the crape myrtle germplasm resource environmental assessment model for auxiliary analysis to obtain an environmental assessment score of the target crape myrtle variety; S5. Comprehensive evaluation of the traits of target crape myrtle varieties The gene evaluation score of the target crape myrtle variety, the morphological trait evaluation score set Y of the target crape myrtle variety and the environmental evaluation score of the target crape myrtle variety are comprehensively evaluated using the crape myrtle variety trait evaluation function to obtain the target crape myrtle variety trait comprehensive evaluation score; the crape myrtle variety trait evaluation function is determined using a population optimization algorithm; The specific steps of step S5 include: The target crape myrtle variety gene evaluation score, the target crape myrtle variety morphological trait evaluation score set Y and the target crape myrtle variety environmental evaluation score are normalized to obtain the standard target crape myrtle variety gene evaluation score, the standard target crape myrtle variety morphological trait evaluation score set Y' and the standard target crape myrtle variety environmental evaluation score; The gene evaluation score of the standard target crape myrtle variety, the morphological trait evaluation score set Y' of the standard target crape myrtle variety and the environmental evaluation score of the standard target crape myrtle variety are input into the crape myrtle variety trait evaluation function for calculation to obtain the comprehensive evaluation score of the target crape myrtle variety trait; Specific steps to construct the evaluation function of crape myrtle variety traits: Construct K initialization function individuals F k , each individual initialization function F k The method includes a simulated crape myrtle variety trait evaluation function for comprehensive evaluation; a plurality of simulated crape myrtle variety comprehensive evaluation training samples are collected; each group of simulated crape myrtle variety comprehensive evaluation training samples includes function input parameters and corresponding simulation target values; Based on the initialization function individual F k The initial fitness A is obtained by performing simulation calculations with several groups of simulated comprehensive evaluation training samples of crape myrtle varieties. k ; Initial fitness A k The calculation method is as follows: according to the simulated crape myrtle varieties, the function input parameters in the comprehensive evaluation training sample and the initialization function individual F k Perform function calculation to obtain the simulated score value; calculate the accuracy of the simulated score value and the simulated target value to obtain the simulated accuracy; take the average of several groups of simulated accuracy as the initial fitness A k ; Based on the initialization function individual F k The initial fitness A k , initialize K function individuals F k Divide into rooster iteration population, hen iteration population and chick iteration population; set the maximum number of iterations B max ; The current iteration number is b; During iteration, the initialization function individuals in the rooster iteration population, hen iteration population, and chicken iteration population are mutated and updated, and the current iteration number b is less than B max / 2, the chicken iteration population mutates to the initialization function individuals in the hen iteration population, and the hen iteration population mutates to the initialization function individuals in the rooster iteration population; the current iteration number b is greater than B max / 2, the chicken iteration population and the hen iteration population mutate to the initialization function individuals in the rooster iteration population; When the maximum number of iterations is reached, the output is the initialization function individual F corresponding to the maximum fitness k , which is the optimal initialization function individual; the simulated crape myrtle variety trait evaluation function in the optimal initialization function individual is used as the crape myrtle variety trait evaluation function.

2. The method for evaluating the properties of crape myrtle germplasm resources according to claim 1, characterized in that: The crape myrtle variety image classification model in step S1 includes an image preprocessing layer, an image feature extraction layer, an image classification layer and a result output layer; The image preprocessing layer is used to preprocess the target crape myrtle variety image to be analyzed, and obtain the preprocessed target crape myrtle variety image; The image feature extraction layer is used to extract features from the preprocessed target crape myrtle variety image to obtain the target crape myrtle variety image features; The image classification layer is used to classify the target crape myrtle varieties according to the image features of the target crape myrtle varieties, and obtain the classification results of the target crape myrtle varieties; The result output layer is used to output the classification results of the target crape myrtle varieties; The specific steps for training the image classification layer in the image classification model for crape myrtle varieties include: Collect several groups of crape myrtle variety image classification samples; each group of crape myrtle variety image classification samples contains crape myrtle variety images and corresponding crape myrtle variety categories; use the corresponding crape myrtle variety categories in each group of crape myrtle variety image classification samples as target values; combine several groups of crape myrtle variety image classification samples to obtain a crape myrtle variety image classification training set; The crape myrtle variety image classification training set is input into the crape myrtle variety image classification model to train the image classification layer to obtain the initial image classification layer; the initial image classification layer is evaluated to obtain the initial image classification layer model evaluation result; if the initial image classification layer model evaluation result is passed, the initial image classification layer is used as the image classification layer in the crape myrtle variety image classification model; otherwise, the crape myrtle variety image classification training set is used to continue model training.

3. The method for evaluating the properties of crape myrtle germplasm resources according to claim 2, characterized in that: The multi-scale feature extraction model of crape myrtle varieties in step S1 includes an image set classification layer, N morphological feature extraction layers T n and the resulting output layer; The image set classification layer is used to classify the target crape myrtle variety morphological image P in the target crape myrtle variety morphological image set. m Classify and obtain the target crape myrtle variety morphological image classification F m ; Morphological feature extraction layer T n Corresponding to the target crape myrtle variety morphological image classification L n ; For the morphological image P of the target crape myrtle variety to be analyzed m , analysis and classification of target crape myrtle variety morphological images F m Classification of morphological images of the same target crape myrtle varieties n The corresponding morphological feature extraction layer T n , get the target morphological feature extraction layer; The morphological image P of the target crape myrtle variety to be analyzed m Input into the target morphological feature extraction layer; N morphological feature extraction layers T n It is used to extract features from the input morphological image of the target crape myrtle variety to be analyzed, and obtain the morphological trait feature X of the target crape myrtle variety. n ; In the morphological feature extraction layer T n In the process, the input morphological images of the target crape myrtle varieties to be analyzed are spliced ​​by using the image stitching method to obtain the comprehensive morphological images of the target crape myrtle varieties; the feature analysis of the comprehensive morphological images of the target crape myrtle varieties is performed to obtain the morphological trait feature X of the target crape myrtle varieties. n ; The result output layer is used to extract the N morphological features from the N morphological features. n Output of the target crape myrtle variety morphological characteristics X n Combination, to obtain the target crape myrtle variety morphological trait set X, X = [X1, X2, ..., X n , …, X N ], output the morphological trait feature set X of the target crape myrtle variety.

4. The method for evaluating the properties of crape myrtle germplasm resources according to claim 3, characterized in that: Train the morphological feature extraction layer T n The specific steps include: For the morphological feature extraction layer T n : Collect several sets of target crape myrtle variety morphological image training samples, each set of target crape myrtle variety morphological image training samples contains the target crape myrtle variety morphological image classification L n corresponding target crape myrtle variety morphological images and corresponding image features; combining a plurality of target crape myrtle variety morphological image training samples to obtain a target crape myrtle variety morphological image training set; The target crape myrtle variety morphological image training set is input into the morphological feature extraction layer T n In the training, the image features are used as the target to obtain the initial morphological feature extraction layer C n ; Extract the initial morphological features from layer C n Perform model evaluation to obtain the initial morphological feature extraction layer C n Model evaluation results; if the initial morphological feature extraction layer C n If the model evaluation result is passed, the initial morphological feature extraction layer C n As the morphological feature extraction layer T n ; Otherwise, continue model training using the target crape myrtle variety morphological image training set; Traverse all morphological feature extraction layers T n , until N morphological feature extraction layers T are constructed n .

5. The method for evaluating the properties of crape myrtle germplasm resources according to claim 4, characterized in that: The specific steps of step S2 include: Evaluate according to the morphological trait evaluation rules in the target crape myrtle variety trait evaluation strategy, where the morphological trait evaluation rules include a morphological trait evaluation rule set; Targeted crape myrtle variety morphological characteristics X n : Match the morphological characteristics of the target crape myrtle variety X n The morphological trait evaluation rule set with the largest membership function is obtained, and the morphological trait evaluation rule set J is obtained. n ; Based on the morphological trait evaluation rule set J n Assign the morphological trait evaluation score Y of the target crape myrtle variety n ; X is the morphological characteristics of all target crape myrtle varieties n Evaluate and score all target crape myrtle varieties with morphological traits Y n Combination, to obtain the target crape myrtle variety morphological trait evaluation score set Y, Y = [Y1, Y2, ..., Y n , …, Y N ].

6. The method for evaluating the properties of crape myrtle germplasm resources according to claim 5, characterized in that: The specific steps of step S3 include: The target crape myrtle variety is evaluated according to the genetic trait evaluation rules in the trait evaluation strategy, which includes G genetic trait score ranges and corresponding gene evaluation scores; Determine the genetic trait score range of the target crape myrtle variety's genetic information to be analyzed, and obtain the genetic trait score range of the target crape myrtle variety; determine the target crape myrtle variety's genetic evaluation score based on the genetic trait score range of the target crape myrtle variety.

7. The method for evaluating the properties of crape myrtle germplasm resources according to claim 6, characterized in that: The crape myrtle germplasm resource environmental assessment model in step S4 includes an environmental information preprocessing layer, an environmental prediction layer and a result output layer; The environmental information preprocessing layer is used to preprocess the germplasm environmental information of the target crape myrtle variety to be analyzed, and obtain the preprocessed germplasm environmental information of the target crape myrtle variety; The environmental prediction layer is used to make predictions based on the pre-processed target crape myrtle variety germplasm environmental information to obtain the target crape myrtle variety germplasm environmental prediction information; The result output layer is used to match the environmental assessment score of the target crape myrtle variety according to the target crape myrtle variety germplasm environmental prediction information; The specific steps of training the environment prediction layer include: Collecting several groups of continuous environmental prediction training samples, each group of environmental prediction training samples contains the germplasm environmental information of crape myrtle varieties; combining several groups of continuous environmental prediction training samples to obtain an environmental prediction training set; taking the latter group of environmental prediction training samples of two adjacent groups as the target value; The environment prediction layer is trained with the target value as the target using the environment prediction training set to obtain an initial environment prediction layer; the initial environment prediction layer is evaluated to obtain an initial environment prediction layer model evaluation result; if the initial environment prediction layer model evaluation result is passed, the initial environment prediction layer is used as the environment prediction layer; otherwise, the model training is continued using the environment prediction training set.

Images