A method and system for cultivating watermelon resistance breeding optimization with data identification traceability

Abstract

The present application relates to the technical field of data processing, in particular to a watermelon resistance breeding optimization method and system for cultivating data identification traceability. Through a user terminal, a first watermelon resistance breeding scheme is configured until an Nth watermelon resistance breeding scheme; relevant data of the first watermelon resistance breeding scheme is extracted, resistance evaluation is performed, and a first resistance score is obtained; until relevant data of the Nth watermelon resistance breeding scheme is extracted, an Nth resistance score is obtained; the first resistance score to the Nth resistance score is fitness, the first watermelon resistance breeding scheme to the Nth watermelon resistance breeding scheme is optimized, and a selected watermelon resistance breeding scheme is sent to the user terminal to perform breeding experiments. The resistance level of the breeding scheme can be quickly output, the breeding cycle is shortened, and the early experimental cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for selecting watermelon resistance breeds through data identification and traceability. Background Technology

[0002] Currently, in the field of crop resistance breeding, the mainstream technology relies on machine learning combined with genotype data to predict disease resistance. This requires first analyzing the genotype data of the target crop, collecting associated training data to build a model, and then conducting gene testing after breeding to predict seed disease resistance, followed by optimization based on the results. However, this technology has significant shortcomings. Specifically, it requires a large number of preliminary breeding experiments, resulting in low breeding efficiency and a long cycle; furthermore, the breeding scheme design is simplistic and lacks a dynamic optimization mechanism, easily overlooking superior schemes; and the data lacks reliable evidence and traceability, with some evaluations relying on manual methods, resulting in insufficient credibility and objectivity. Summary of the Invention

[0003] This invention addresses the technical problems in existing technologies, such as low breeding efficiency, long cycle, easy omission of high-quality schemes, and lack of reliable data storage and traceability, by providing a watermelon resistance breeding optimization method and system for breeding data identification and traceability.

[0004] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0005] In a first aspect, the present invention provides a method for selecting the optimal watermelon resistance breeding scheme through data identification and traceability, comprising: configuring a first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme via a user terminal; extracting the first watermelon germplasm type, the first rootstock type, and the first planting scheme of the first watermelon resistance breeding scheme, constructing a first index constraint, collecting a first watermelon breeding sample set stored on the blockchain, performing resistance assessment, and obtaining a first resistance score; up to extracting the Nth watermelon germplasm type, the Nth rootstock type, and the Nth planting scheme of the Nth watermelon resistance breeding scheme, constructing an Nth index constraint, collecting a Nth watermelon breeding sample set stored on the blockchain, performing resistance assessment, and obtaining an Nth resistance score; using the first resistance score up to the Nth resistance score as fitness, optimizing the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme, obtaining a selected watermelon resistance breeding scheme and sending it to the user terminal for breeding experiments.

[0006] Optionally, through the user terminal, configure the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme, including: through the user terminal, obtaining a watermelon constrained germplasm type set, a rootstock constrained type set, a planting parameter constrained type set, and a planting parameter constrained interval set; randomly extracting a value from each of the watermelon constrained germplasm type set, the rootstock constrained type set, the planting parameter constrained type set, and the planting parameter constrained interval set and combining them to obtain the first watermelon resistance breeding scheme; until the Nth watermelon resistance breeding scheme is obtained by randomly extracting a value from each of the watermelon constrained germplasm type set, the rootstock constrained type set, the planting parameter constrained type set, and the planting parameter constrained interval set and combining them.

[0007] Optionally, the first watermelon germplasm type, first rootstock type, and first planting scheme of the first watermelon resistance breeding program are extracted, a first index constraint is constructed, a first watermelon cultivation sample set stored on the blockchain is collected, resistance assessment is performed, and a first resistance score is obtained. This includes: extracting the first watermelon germplasm type, first rootstock type, and first planting scheme of the first watermelon resistance breeding program, constructing a first index constraint; collecting a first watermelon cultivation sample set stored on the blockchain and satisfying the first index constraint, performing genotype mode analysis to obtain several frequent genotype groups; traversing the several frequent genotype groups, performing resistance assessment through a resistance assessment network to obtain several initial resistance scores, wherein the resistance assessment network is generated by machine learning training using multiple sets of data, and any set of the multiple sets of data includes genotype group record data and a label identifying the resistance score; taking the set of values ​​of the several initial resistance scores as the first resistance score.

[0008] The process involves extracting the first watermelon germplasm type, first rootstock type, and first planting scheme from the first watermelon resistance breeding program, and constructing a first index constraint, including: extracting the first planting control attribute up to the Qth planting control attribute from the first planting scheme; performing genotype fluctuation association analysis on the first watermelon germplasm type, the first rootstock type, and the first planting control attribute up to the Qth planting control attribute to obtain weight distribution results; and, based on the weight distribution results, normalizing the deviation of the first watermelon germplasm type, the first rootstock type, and the first planting control attribute up to the Qth planting control attribute. The attribute normalization deviations are weighted and summed to obtain the index deviation coefficient. Specifically, when the watermelon resistance breeding type parameter of the sample to be analyzed differs from the first watermelon resistance breeding type parameter, the normalization deviation is equal to 1; otherwise, the normalization deviation is equal to 0. When the quantification deviation between the quantification parameter of the watermelon resistance breeding of the sample to be analyzed and the quantification parameter of the first watermelon resistance breeding is greater than or equal to the user-preset quantification deviation threshold of the corresponding attribute, the normalization deviation is equal to 1; otherwise, the normalization deviation is equal to 0. When the index deviation coefficient is less than or equal to the deviation coefficient threshold, the resistance breeding scheme of the sample to be analyzed is considered to satisfy the first index constraint; otherwise, it is considered not to satisfy it.

[0009] The process involves performing genotype fluctuation association analysis on the first watermelon germplasm type, the first rootstock type, the first planting control attribute up to the Qth planting control attribute, and obtaining weight distribution results. This includes: retrieving the first and second samples to be analyzed where the normalized deviation of the first watermelon germplasm type is 1 and the normalized deviation of other resistance breeding is 0; calculating the genotype crossover and union ratio (CRU) and setting it as a genotype fluctuation parameter, which is then added to the genotype fluctuation parameter set; when the number of the genotype fluctuation parameter set equals a preset statistical number, calculating the central tendency of the genotype fluctuation parameter set and setting it as the correlation degree of the first watermelon germplasm type; continuing until the correlation degree of the Qth planting control attribute is obtained; iterating through the correlation degrees of the first watermelon germplasm type up to the Qth planting control attribute and comparing them with the sum of the correlation degrees, setting them as the weights of the first watermelon germplasm type up to the Qth planting control attribute, and adding them to the weight distribution results.

[0010] The process involves traversing several frequent genotype groups, performing resistance assessment through a resistance assessment network to obtain several initial resistance scores. The resistance assessment network is generated using machine learning training on multiple sets of data. Each set of data includes genotype group records and labels identifying resistance scores. This includes: extracting a one-to-one corresponding dataset of genotype group records without germplasm differences and a set of labels identifying resistance scores from the multiple sets of data; using the set of labels identifying resistance scores as supervision, training a fully connected neural network using machine learning on the dataset of genotype group records without germplasm differences to obtain the backbone resistance assessment network, where germplasm characterizes watermelon germplasm during breeding; extracting a one-to-one corresponding dataset of target germplasm-difference genotype group records and a set of labels identifying resistance scores from the multiple sets of data; connecting a branch fully connected neural network to the backbone resistance assessment network, and outputting the backbone... The average output of the resistance assessment network and the branch fully connected neural network is used to obtain the first-level compensatory resistance assessment network architecture. The model parameters of the main resistance assessment network are frozen. Using the label set of the two types of resistance scores as supervision and the dataset of the genotype group without germplasm differences, the first-level compensatory resistance assessment network architecture is trained using machine learning to obtain the first-level compensatory resistance assessment network. When the validation accuracy of the first-level compensatory resistance assessment network is greater than or equal to the accuracy threshold, the first-level compensatory resistance assessment network is set as the resistance assessment network. When the validation accuracy of the first-level compensatory resistance assessment network is less than the accuracy threshold, a branch fully connected neural network is connected to the main resistance assessment network based on the first-level compensatory resistance assessment network architecture. The output is the average of the outputs of the main resistance assessment network and the two branch fully connected neural networks to obtain the second-level compensatory resistance assessment network architecture, and iterative training is performed.

[0011] The step of obtaining the label for the resistance score includes: obtaining a label for the watermelon's disease resistance score, a label for the watermelon's yield resistance score, and a label for the watermelon's quality score; extracting the minimum score of the labels for the watermelon's disease resistance score, yield resistance score, and quality score, and setting it as the label for the resistance score.

[0012] Optionally, using the first resistance score up to the Nth resistance score as fitness, the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme is optimized, and the selected watermelon resistance breeding scheme is sent to the user terminal for breeding experiments, including:

[0013] Step 1: Based on the first resistance score up to the Nth resistance score as fitness, extract the resistance breeding scheme with the maximum fitness, the resistance breeding scheme with the median fitness, and the resistance breeding scheme with a preset number of fitness tails from the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme;

[0014] Step 2: Using the resistance breeding scheme with the maximum fitness and the resistance breeding scheme with the median fitness as search targets, adjust the fitness preset number of resistance breeding schemes with the fitness tail to obtain an updated watermelon resistance breeding scheme;

[0015] Step 3: When the fitness of the updated watermelon resistance breeding program is less than or equal to that of the first watermelon resistance breeding program up to the minimum fitness value of the Nth watermelon resistance breeding program, return to Step 2 and execute the loop.

[0016] Step 4: When the fitness of the updated watermelon resistance breeding scheme is greater than the minimum fitness value of the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme, update the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme based on the updated watermelon resistance breeding scheme, and then return to Step 1 to execute the loop.

[0017] Secondly, the present invention provides a watermelon resistance breeding optimization system for cultivation data identification and traceability, comprising:

[0018] The watermelon resistance breeding program configuration module is used to configure the first watermelon resistance breeding program up to the Nth watermelon resistance breeding program through the user terminal;

[0019] The watermelon resistance breeding program evaluation module is used to extract the first watermelon germplasm type, first rootstock type, and first planting scheme of the first watermelon resistance breeding program, construct the first index constraint, collect the first watermelon cultivation sample set that has been stored on the blockchain, perform resistance evaluation, and obtain the first resistance score; until the Nth watermelon resistance breeding program extracts the Nth watermelon germplasm type, Nth rootstock type, and Nth planting scheme, constructs the Nth index constraint, collects the Nth watermelon cultivation sample set that has been stored on the blockchain, performs resistance evaluation, and obtains the Nth resistance score;

[0020] The watermelon resistance breeding program optimization module is used to optimize the first watermelon resistance breeding program up to the Nth watermelon resistance breeding program using the first resistance score up to the Nth resistance score as fitness, and send the selected watermelon resistance breeding program to the user terminal for breeding experiment execution.

[0021] By implementing this invention, it is possible to configure the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme through the user terminal, thereby breaking the limitations of a single scheme. Through multi-set random combination, covering the combination of different germplasm, rootstock and planting parameters, potential high-quality combinations that are easily overlooked in traditional single scheme design can be discovered, and sufficient sample size can be ensured in the subsequent evaluation and optimization process to avoid the screening results being one-sided due to insufficient samples, thereby improving the reliability of the final selected scheme.

[0022] The first watermelon germplasm type, first rootstock type, and first planting scheme of the first watermelon resistance breeding scheme are extracted, a first index constraint is constructed, a first watermelon breeding sample set stored on the blockchain is collected, resistance assessment is performed, and a first resistance score is obtained. This process continues until the Nth watermelon germplasm type, Nth rootstock type, and Nth planting scheme of the Nth watermelon resistance breeding scheme are extracted, an Nth index constraint is constructed, a Nth watermelon breeding sample set stored on the blockchain is collected, resistance assessment is performed, and an Nth resistance score is obtained. The immutability of the blockchain prevents the breeding sample data from being tampered with or forged, ensuring the authenticity and traceability of the sample data, laying the foundation for the accuracy of subsequent assessment results. Furthermore, with the help of a resistance assessment network trained by machine learning, the genotype data is transformed into specific resistance scores, realizing an objective and quantitative evaluation of the resistance level of the breeding scheme, avoiding the subjective bias of traditional manual assessment.

[0023] Using the first resistance score up to the Nth resistance score as fitness, the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme is optimized. The selected watermelon resistance breeding scheme is sent to the user terminal to execute the breeding experiment. Using the resistance score as fitness, the focus is on high fitness schemes. At the same time, inferior schemes are gradually eliminated by adjusting the tail schemes to ensure that the resistance level of the final output scheme is optimal. Furthermore, through a cyclical iteration mechanism, the scheme library can be continuously updated and re-evaluated for optimization, avoiding the omission of the optimal solution due to the limitations of the initial scheme.

[0024] In summary, by implementing this invention, it is possible to quickly output breeding programs with optimal resistance levels, shorten the breeding cycle, reduce early experimental costs, and improve the systematicness and efficiency of watermelon resistance breeding. Attached Figure Description

[0025] Figure 1 A flowchart illustrating a watermelon resistance breeding optimization method for data identification and traceability provided by this invention;

[0026] Figure 2 This is a schematic diagram of the structure of a watermelon resistance breeding optimization system for data identification and traceability provided by the present invention.

[0027] In the attached diagram, the components represented by each number are as follows:

[0028] Watermelon resistance breeding program configuration module 11, watermelon resistance breeding program evaluation module 12, watermelon resistance breeding program optimization module 13. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0031] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0032] Example 1, as Figure 1 As shown, this invention provides a method and system for selecting watermelon resistance breeds based on data identification and traceability, comprising:

[0033] S100: Configure the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme through the user terminal;

[0034] S200: Extract the first watermelon germplasm type, first rootstock type, and first planting scheme of the first watermelon resistance breeding scheme, construct the first index constraint, collect the first watermelon cultivation sample set that has been stored on the blockchain, perform resistance assessment, and obtain the first resistance score; until the Nth watermelon germplasm type, Nth rootstock type, and Nth planting scheme of the Nth watermelon resistance breeding scheme are extracted, the Nth index constraint is constructed, the Nth watermelon cultivation sample set that has been stored on the blockchain is collected, perform resistance assessment, and obtain the Nth resistance score;

[0035] S300: Using the first resistance score up to the Nth resistance score as fitness, optimize the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme, obtain the selected watermelon resistance breeding scheme and send it to the user terminal to execute the breeding experiment.

[0036] In step S100 of this application embodiment, configuring the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme via the user terminal includes:

[0037] Through the user terminal, the set of watermelon constrained germplasm types, rootstock constrained types, planting parameter constrained types, and planting parameter constrained intervals are obtained;

[0038] A value is randomly extracted from the watermelon constrained germplasm type set, the rootstock constrained type set, the planting parameter constrained type set, and the planting parameter constrained interval set, and combined to obtain the first watermelon resistance breeding scheme;

[0039] The Nth watermelon resistance breeding scheme is obtained by randomly extracting a value from the watermelon constrained germplasm type set, the rootstock constrained type set, the planting parameter constrained type set, and the planting parameter constrained interval set, and combining them.

[0040] In this embodiment of the application, the purpose of the above steps is to generate multiple watermelon resistance breeding schemes that cover diverse combinations, meet personalized needs, and support subsequent evaluation and optimization within the reasonable constraints set by the user, so as to provide a sufficient and targeted sample basis for the entire breeding optimization process.

[0041] Specifically, the user first needs to obtain the watermelon constraint germplasm type set, rootstock constraint type set, planting parameter constraint type set, and planting parameter constraint interval set through the user terminal.

[0042] Among them, the watermelon constrained germplasm type set includes watermelon germplasm categories that meet the requirements, such as germplasm resistant to Fusarium wilt and germplasm resistant to anthracnose.

[0043] The rootstock constraint type set contains suitable rootstock categories, such as pumpkin rootstock and gourd rootstock.

[0044] The planting parameter constraint type set contains key planting parameter categories that affect breeding, including two planting parameter subtypes: cultivation plan and disease control plan. The cultivation plan subtype includes planting parameters such as garden sanitation and disinfection, soil treatment, and seedling strengthening agents. The disease control plan subtype includes planting parameters such as pesticides for planting with pre-treated plants, the size of the attractant color board, and the type of natural enemy released.

[0045] The term "releasing natural enemies" refers to a method of killing pests, such as releasing ladybugs in greenhouses where aphids occur. "Planting with pesticides" refers to a technique that combines cultivation and pest control, in which specific pesticides are applied to crop seedlings before they are transplanted from the nursery bed to the field, greenhouse, or other planting environment, so that the seedlings "carry" the pesticides into the planting environment.

[0046] The set of planting parameter constraint intervals consists of the drug categories and parameter ranges typically selected for the aforementioned planting parameters. For example, for cultivation program parameter intervals, it may include the formalin concentration range used for orchard sanitation and disinfection, the duration of sanitation and disinfection, the deep tillage depth during soil treatment, the type of straw used for soil treatment, the concentration of plant growth promoters, etc.; for disease control program parameter intervals, it may include the number of bait sticks, the hanging height of the bait sticks, the proportion of natural enemies released, etc.

[0047] Furthermore, a value needs to be randomly extracted from the watermelon constrained germplasm type set, the rootstock constrained type set, the planting parameter constrained type set, and the planting parameter constrained interval set, and combined to obtain the first watermelon resistance breeding scheme.

[0048] First, one cultivation scheme subtype and one disease control scheme subtype, such as orchard disinfectant and medicated planting agent, need to be randomly extracted from the planting parameter constraint type set. For each extracted subtype, matching parameter values ​​are randomly extracted from the planting parameter constraint interval set. For example, if the cultivation scheme subtype is orchard disinfectant, then 1% formalin solution or 2% formalin solution is extracted from the planting parameter constraint interval set; if the disease control scheme subtype is medicated planting agent, then "Shenzinmycin root dipping" or "Oxyphenidyl root dipping" is extracted from the planting parameter constraint interval set as the medicated planting agent application scheme.

[0049] Then, the germplasm type, rootstock type, cultivation scheme subtype and parameters (1% formalin for orchard sanitation) and disease control scheme subtype and parameters (shenqinmycin for root dipping) are combined to form the first watermelon resistance breeding scheme containing complete planting scheme details.

[0050] Repeat the above random extraction-combination process, continuously randomly extracting and combining parameters from the four constraint sets until N sets of watermelon resistance breeding schemes are generated.

[0051] In step S200 of this application embodiment, the first watermelon germplasm type, the first rootstock type, and the first planting scheme of the first watermelon resistance breeding program are extracted, a first index constraint is constructed, a first watermelon cultivation sample set stored on the blockchain is collected, resistance assessment is performed, and a first resistance score is obtained, including:

[0052] Extract the first watermelon germplasm type, first rootstock type, and first planting scheme of the first watermelon resistance breeding scheme, and construct the first index constraint;

[0053] Collect the first watermelon breeding sample set that has been stored on the blockchain and meets the first index constraint, perform genotype mode analysis, and obtain several frequent genotype groups;

[0054] The resistance assessment network is used to perform resistance assessments by traversing the several frequent genotype groups and obtaining several initial resistance scores. The resistance assessment network is generated by machine learning training using multiple sets of data, and each set of data includes genotype group record data and labels that identify resistance scores.

[0055] Take the concentrated value of the aforementioned initial resistance scores and set it as the first resistance score.

[0056] In this embodiment of the application, the purpose of step S200 is to achieve a precise, objective, and quantitative evaluation of the resistance level of a single breeding scheme through the entire process of “constructing index constraints – screening matching watermelon breeding samples – analyzing genotypes – model evaluation – score aggregation”, so as to provide a reliable scoring basis for subsequent scheme optimization in S300.

[0057] Specifically, the first step is to extract the first watermelon germplasm type, the first rootstock type, and the first planting scheme of the first watermelon resistance breeding program, and then construct the first index constraint.

[0058] In step S200 of this application embodiment, the first watermelon germplasm type, the first rootstock type, and the first planting scheme of the first watermelon resistance breeding scheme are extracted, and a first index constraint is constructed, including:

[0059] Extract the first planting control attribute up to the Qth planting control attribute from the first planting scheme;

[0060] Genotypic fluctuation association analysis was performed on the first watermelon germplasm type, the first rootstock type, the first planting control attribute up to the Qth planting control attribute to obtain the weight distribution results;

[0061] Based on the weight distribution results, the normalization deviations of the first watermelon germplasm type, the first rootstock type, the first planting control attribute, and up to the Qth planting control attribute are weighted and summed to obtain the index deviation coefficient. Specifically, when the watermelon resistance breeding type parameter of the sample to be analyzed is different from the first watermelon resistance breeding type parameter, the normalization deviation is equal to 1; otherwise, the normalization deviation is equal to 0. When the quantification deviation between the watermelon resistance breeding quantification parameter of the sample to be analyzed and the first watermelon resistance breeding quantification parameter is greater than or equal to the user-preset quantification deviation threshold of the corresponding attribute, the normalization deviation is equal to 1; otherwise, the normalization deviation is equal to 0.

[0062] If the index deviation coefficient is less than or equal to the deviation coefficient threshold, the resistance breeding scheme of the sample to be analyzed is considered to meet the first index constraint; otherwise, it is considered not to meet the constraint.

[0063] In this embodiment, the purpose of the above steps is to achieve a refined and scientific construction of index constraints through the logic of "extracting planting control attributes - analyzing parameter association weights - calculating deviation coefficients - determining constraint satisfaction," ensuring that the breeding samples subsequently selected from the blockchain sample library are highly matched with the first watermelon resistance breeding scheme, and avoiding distortion of evaluation results due to sample bias.

[0064] The planting control attributes mentioned above refer to specific management parameters or operational indicators that are quantifiable, controllable, and directly or indirectly affect the resistance level of watermelons, extracted from planting schemes that include cultivation and disease control schemes.

[0065] First, it is necessary to extract the first planting control attribute up to the Qth planting control attribute from the first planting scheme. That is, to extract the key control attributes affecting watermelon resistance from the cultivation measures of the first planting scheme.

[0066] For example, the first planting control attribute can be orchard disinfection, including: disinfectant type and disinfectant concentration; the second planting control attribute can be soil treatment, including: straw type, deep plowing depth, and disinfection duration; the third planting control attribute can be planting with pesticides, including: root dipping pesticide type; and the fourth planting control attribute can be trapping with color boards, including: color board specifications, number of hanging boards, and hanging height, etc.

[0067] Based on the specific content of the first planting plan, the total number of planting control attributes is defined as Q. Following the above format, a list of planting control attributes from the first to the Qth is formed as the basis for subsequent analysis.

[0068] In step S200 of this application embodiment, genotype fluctuation association analysis is performed on the first watermelon germplasm type, the first rootstock type, the first planting control attribute up to the Qth planting control attribute to obtain the weight distribution results, including:

[0069] The first and second samples to be analyzed were retrieved, with a normalized deviation of 1 for the first watermelon germplasm type and a normalized deviation of 0 for other resistance breeding. The crossover ratio of genotypes was calculated and set as the genotype fluctuation parameter, and added to the genotype fluctuation parameter set.

[0070] When the number of the genotype fluctuation parameter set is equal to the preset statistical number, the set value of the genotype fluctuation parameter set is calculated and set as the first watermelon germplasm type correlation degree.

[0071] Until the correlation degree of the Qth planting control attribute is obtained;

[0072] Iterate through the correlation degree of the first watermelon germplasm type up to the correlation degree of the Qth planting control attribute, compare the sum of the correlation degrees, set it as the weight of the first watermelon germplasm type up to the weight of the Qth planting control attribute, and add it to the weight distribution result.

[0073] In step S200 of this application embodiment, the purpose of the above steps is to quantify the degree of influence of the core parameters of watermelon resistance breeding, and finally generate the weight distribution results to clarify the correlation strength of each parameter with the watermelon genotype, so as to provide a scientific weight basis for the calculation of deviation coefficients in subsequent index constraints.

[0074] The normalization bias is a standardized indicator used to quantify the difference between the "parameters of the sample to be analyzed" and the "parameters of the target breeding program". For type parameters, if the parameter value of the sample to be analyzed differs from the parameter value of the target breeding program (e.g., the germplasm of the sample to be analyzed is "anthrax-resistant" while the germplasm of the target breeding program is wilt-resistant), then the normalization bias = 1. For quantification parameters, the user can preset a "quantification bias threshold" for the parameter, such as a formalin concentration threshold of ±0.2%. If the difference between the quantification value of the sample to be analyzed and the quantification value of the target breeding program is greater than or equal to the quantification bias threshold (e.g., the formalin concentration of the sample to be analyzed is 1.5%, and the formalin concentration of the target breeding program is 1%, with a difference of 0.5% ≥ 0.2%), then the normalization bias = 1. If the difference between the quantification value of the sample to be analyzed and the quantification value of the target breeding program is less than the quantification bias threshold, then the normalization bias = 0.

[0075] First, we need to retrieve the first and second samples to be analyzed, where the normalization deviation of the first watermelon germplasm type is 1 and the normalization deviation of other resistance breeding is 0. We then calculate the genotype crossover ratio, set it as the genotype fluctuation parameter, and add it to the genotype fluctuation parameter set.

[0076] The first watermelon germplasm type is used as the first analysis parameter. The search rules are set as follows: the sample to be analyzed must meet the following conditions: the normalization deviation of the first watermelon germplasm type = 1, that is, the germplasm type of the sample to be analyzed is different from the germplasm type of the target breeding program, and the normalization deviation of all other parameters, such as rootstock type and the first to Q planting control attributes = 0, that is, the other parameters are completely consistent with the target breeding program.

[0077] The genotype crossover-union ratio (GUNR) is a core indicator used to quantify the genotype similarity and difference between two samples. It directly reflects the magnitude of genotype fluctuation caused by differences in target parameters by calculating the ratio of the intersection to the union of the genotype data of the two samples. The calculation method is: Genotype crossover-union ratio = (Number of intersections between the genotypes of the two samples) ÷ (Number of unions between the genotypes of the two samples). For example, if the number of intersections is 3 and the number of unions is 5, then the genotype crossover-union ratio is 0.6, indicating that there is a 40% fluctuation in the genotypes of the two samples due to differences in germplasm type. The genotype intersection refers to the completely identical genotype segments shared by the two samples, representing the similarity between their genotypes. The genotype union is the sum of all genotype segments between the two samples, representing the full range of genotypes between them.

[0078] Finally, the genotype crossover ratio is set as the genotype fluctuation parameter and added to the genotype fluctuation parameter set.

[0079] Furthermore, it is necessary to repeat the process of searching for the genotype to be analyzed and calculating the crossover ratio of genotypes, and continuously add fluctuation parameters to the genotype fluctuation parameter set. When the number of the genotype fluctuation parameter set is equal to the preset statistical number, calculate the set value of the genotype fluctuation parameter set and set it as the first watermelon germplasm type correlation degree. For example, if the preset statistical number is 50, that is, collect 50 sets of genotype crossover ratios caused by differences in germplasm types.

[0080] The calculation of the central tendency value of the genotype fluctuation parameter set involves first determining whether there are extreme values ​​in the genotype fluctuation parameter set, such as a genotype crossover ratio of 0.1 or 0.9, which far exceeds the range of 0.5-0.7 for other parameters. If such extreme values ​​exist, the "median" is used as the central tendency value of the genotype fluctuation parameter set; otherwise, the arithmetic mean is used. Then, the calculated central tendency value of the genotype fluctuation parameter set is defined as the correlation degree of the first watermelon germplasm type. For example, if the mean of the crossover ratios of the 50 genotypes in the genotype fluctuation parameter set is 0.55, then the corresponding correlation degree of the first watermelon germplasm type is 0.55.

[0081] Following the above logic, the correlation degree calculation of all planting control attributes is completed in sequence, and finally the correlation degree of the first watermelon germplasm type to the correlation degree of the Qth planting control attribute is obtained.

[0082] Furthermore, it is necessary to iterate through the correlation degree of the first watermelon germplasm type up to the correlation degree of the Qth planting control attribute, compare it with the sum of the correlation degrees, set it as the weight of the first watermelon germplasm type up to the weight of the Qth planting control attribute, and add it to the weight distribution result.

[0083] Specifically, firstly, the correlation degrees of all Q planting control attributes need to be summed. Then, the weight of a certain planting control attribute is calculated as: correlation degree of that planting control attribute / sum of the correlation degrees of the Q planting control attributes. Assuming the correlation degree of the first watermelon germplasm type is 0.55, and the sum of the correlation degrees of the Q planting control attributes is 1.85, then the weight of the first watermelon germplasm type is approximately 0.55 / 1.85 ≈ 0.30. Following this logic, the weights of all Q planting control attributes are calculated and added to the weight distribution result.

[0084] Furthermore, based on the weight distribution results, the normalized deviations of the first watermelon germplasm type, the first rootstock type, the first planting control attribute, and up to the Qth planting control attribute are weighted and summed to obtain the index deviation coefficient. The specific calculation method for the normalized deviation is as described in the preceding steps and will not be repeated here.

[0085] Here, we need to refer to the weights of each genotype fluctuation parameter obtained from the previous genotype fluctuation association analysis. Then, we multiply the normalized deviation of each parameter by its corresponding weight, and sum all the results to obtain the index deviation coefficient. That is, index deviation coefficient = (normalized deviation of the first watermelon germplasm type × weight of the first watermelon germplasm type) + (normalized deviation of the first rootstock type × weight of the first rootstock type) + ... + (deviation of the Qth planting control attribute × weight of the Qth planting control attribute).

[0086] The deviation coefficient threshold is set by the user according to the breeding accuracy requirements. For example, 0.3 represents an allowable weighted difference of up to 30%. If the index deviation coefficient of the sample to be analyzed is less than or equal to the deviation coefficient threshold, it is determined that the first index constraint is met and it is included in the analysis sample set; if the index deviation coefficient of the sample to be analyzed is greater than the deviation coefficient threshold, it is excluded.

[0087] Next, it is necessary to collect the first watermelon breeding sample set that has been stored on the blockchain and meets the first index constraint, perform genotype mode analysis, and obtain several frequent genotype groups.

[0088] In the embodiments of this application, frequent genotype groups are screened out through genotype mode analysis, that is, genotype groups that appear frequently and have high stability. The core purpose is to eliminate the interference of accidental and unrepresentative genotypes, focus on the core genotype characteristics that are compatible with the target breeding program, and provide a reliable genotype analysis basis for subsequent resistance assessment.

[0089] The aforementioned blockchain-based evidence storage refers to uploading key data from watermelon cultivation samples to a blockchain system and completing the entire process of "recording, verifying, adding to the chain, and storing" according to blockchain technology rules, ultimately forming an immutable, traceable, and transparent data certificate. This solves the problems of "easy to tamper with, difficult to trace, and low credibility" in traditional data storage.

[0090] Assuming the first watermelon breeding sample set, which has been certified by blockchain and satisfies the first index constraint, contains 5 matching watermelon breeding samples, and the genotype data is shown in Table 1:

[0091] Table 1. Example of genotype data

[0092]

[0093] Next, based on the (intersection number / union number) definition of the genotype crossover and union ratio mentioned above, the crossover and union ratio of all pairwise watermelon cultivation samples in the first watermelon cultivation sample set is calculated to quantify the genotype similarity.

[0094] For example, S1 and S2 both have the genotype A1B2C1D2. The number of intersections is 4 (A1, B2, C1, D2), the number of unions is 4, and the intersection / union ratio is 4 / 4 = 1.0. S1 and S3 (A1B2C1D2 vs A1B2C1D1) have the following intersections: the number of intersections is 3 (A1, B2, C1), the number of unions is 4 (A1, B2, C1, D1, D2), and the intersection / union ratio is 3 / 4 = 0.75.

[0095] The calculation results for other watermelon breeding sample combinations are summarized in Table 2:

[0096] Table 2 Examples of calculation results for other watermelon breeding sample combinations

[0097]

[0098] Next, the "outlier degree" of the first watermelon breeding sample needs to be quantified by calculating the density difference between a certain watermelon breeding sample and other watermelon breeding samples in the neighborhood based on the Local Outlier Factor (LOF) algorithm. The larger the LOF, the more significant the difference between the genotype and the genotype of other watermelon breeding samples, and the more likely it is to be an outlier genotype.

[0099] Set the number of neighborhood samples to k=3, and refer to the 3 nearest neighbors for each sample. Then, for each sample, calculate its average reachability distance with the k neighborhood samples, and take the reciprocal. Then, the LOF value of a sample = the average local reachability density of its neighborhood samples ÷ the local reachability density of the sample, where LOF=1 represents normal, and >1 represents possible outlier.

[0100] In the above examples, the LOF values ​​of each watermelon breeding sample are calculated as shown in Table 3:

[0101] Table 3. Examples of LOF value calculation results for watermelon cultivation samples.

[0102]

[0103] Next, an outlier threshold needs to be set, which can be determined by the user based on the overall genotype similarity in the first watermelon breeding sample set, for example, 1.5. Then, an outlier determination rule needs to be preset: genotype groups with an outlier factor ≤ the outlier threshold are considered frequent genotype groups; otherwise, they are outlier genotype groups and are excluded. In the example above, S5 is an outlier genotype group, and the remaining watermelon breeding samples are frequent genotype groups.

[0104] In step S200 of this embodiment, the plurality of frequent genotype groups are traversed, and resistance assessment is performed through a resistance assessment network to obtain a plurality of initial resistance scores. The resistance assessment network is generated using machine learning training on multiple sets of data. Each set of data includes genotype group record data and labels identifying the resistance score, including:

[0105] Extract one-to-one corresponding datasets of genotype records without germplasm differences and a set of labels that identify resistance scores from the multiple sets of data;

[0106] Using the label set that identifies resistance scores as supervision, and the dataset of the genotype group without germplasm differences, machine learning is used to train a fully connected neural network to obtain the backbone resistance evaluation network, wherein the germplasm characterizes the watermelon germplasm during breeding.

[0107] Extract one-to-one target germplasm differential genotype record datasets and label sets of two types of resistance scores from the multiple sets of data;

[0108] A branch fully connected neural network is connected to the main resistance assessment network. The output is the average of the outputs of the main resistance assessment network and the branch fully connected neural network to obtain the first-level compensatory resistance assessment network architecture. The model parameters of the main resistance assessment network are frozen. The label set of the two types of resistance scores is used as supervision. The dataset of the genotype group without germplasm differences is used to train the first-level compensatory resistance assessment network architecture using machine learning to obtain the first-level compensatory resistance assessment network.

[0109] When the verification accuracy of the first-level compensated resistance evaluation network is greater than or equal to the accuracy threshold, the first-level compensated resistance evaluation network is set as the resistance evaluation network.

[0110] When the validation accuracy of the first-level compensation resistance evaluation network is less than the accuracy threshold, a branch fully connected neural network is connected to the backbone resistance evaluation network based on the first-level compensation resistance evaluation network architecture. The output is the average of the outputs of the backbone resistance evaluation network and the two branch fully connected neural networks to obtain the second-level compensation resistance evaluation network architecture, and iterative training is performed.

[0111] In step S200 of this application embodiment, the step of obtaining the label for identifying the resistance score includes:

[0112] Obtain labels indicating watermelon disease resistance scores, watermelon yield resistance scores, and watermelon quality scores;

[0113] Extract the minimum score from the labels indicating watermelon disease resistance, watermelon yield resistance, and watermelon quality, and set it as the label indicating resistance.

[0114] In the resistance assessment network training and resistance score calculation in step S200 of this application embodiment, the "label for the resistance score" is the core supervisory data. This step integrates the scores of the three dimensions of disease resistance, yield and quality, and takes the minimum value as the final label. The core purpose is to ensure that the label can comprehensively reflect the overall resistance shortcomings of the watermelon breeding program, and avoid programs that are excellent in one dimension but seriously deficient in another dimension being misjudged as high-quality programs.

[0115] The watermelon disease resistance score focuses on resistance performance during key growth stages of watermelons, with core characteristics including parameters such as direct seeding germination time, emergence lodging rate, emergence seedling mortality rate, and vine growth rate. Specifically, direct seeding germination time is recorded using a timer; the emergence lodging rate is calculated as (number of lodged seedlings / total number of seedlings) × 100%; the emergence seedling mortality rate is calculated as (number of dead seedlings / total number of seedlings) × 100%; and the vine growth rate is obtained by measuring the daily difference in vine length.

[0116] The watermelon yield resistance score focuses on the final yield capacity of watermelons, with the core characteristic being the yield per plant. This score is obtained by weighing the total weight of a single watermelon plant after the sample has matured.

[0117] Watermelon quality scoring focuses on the commercial attributes of watermelons, with core characteristics including sugar content, appearance and texture, color, and sound when tapped. Sugar content can be obtained by measuring the Brix value of a watermelon sample using a refractometer.

[0118] For image features such as appearance texture and color, a convolutional neural network (CNN) can be used to train an image feature extractor to extract relevant features. Specifically, high-quality watermelon appearance images can be collected as sample data, and labeled with data such as "clear / blurry texture" and "standard / deviation in color" as supervision labels to form training data. The CNN network is trained using the training data until convergence to obtain the image feature extractor. The appearance image of the watermelon to be evaluated is input into the trained image feature extractor, which will output the texture feature vector and color feature vector of the watermelon to be evaluated. For the sound features of the tapping sound, audio signal processing algorithms such as Fourier transform can be combined with a lightweight neural network to extract the frequency and amplitude feature vectors of the sound.

[0119] Furthermore, the scores for all dimensions are calculated based on the similarity between sample features and standard features, using a pre-defined formula: Score SIM(feature A1, feature A2) = [2 × (feature A1) / (feature A2)] Feature A2)] / [(Feature A1) 2 + (Feature A2) 2 +ε]. Here, feature A1 is a feature vector of a certain dimension of the watermelon sample to be evaluated, such as germination time and lodging rate in disease resistance; feature A2 is the standard feature vector of that dimension, such as the standard germination time and standard lodging rate set by the breeding target; ε is a small constant, such as 0.0001, used to avoid a denominator of zero. The SIM value ∈ [0,1], and the closer the value is to 1, the more similar the sample features are to the standard features, and the higher the score. Furthermore, the SIM value can be multiplied by 100 to convert it to a 0-100 score scale for easier understanding.

[0120] Assuming the yield of a single watermelon plant in a certain sample is 8.8 kg, then the sample feature A1 = [8.8]. Assuming the standard feature of the yield of a single watermelon plant is A2 = [9], substituting into the above similarity formula, we can calculate the score SIM = [2 × 79.2] / [77.44 + 81 + 0.0001] ≈ 158.4 / 158.44 ≈ 0.9997. The percentage score = 0.9997 × 100 ≈ 100 points, which serves as a label to identify the yield resistance score of watermelons.

[0121] Using the same method, calculate the labels for watermelon disease resistance score, yield resistance score, and quality score. Take the minimum value of these three as the label for the resistance score. For example, if a sample has a disease resistance score of 100, a yield score of 100, and a quality score of 92, then the final label for the resistance score will be 92.

[0122] Furthermore, it is necessary to traverse the aforementioned frequent genotype groups and perform resistance assessments through a resistance assessment network to obtain several initial resistance scores. Specifically, a multi-level compensation assessment model capable of accurately assessing the resistance of different watermelon germplasm varieties is constructed through a progressive logic of training the backbone resistance assessment network, iteratively training the branch compensation network, and verifying accuracy.

[0123] First, it is necessary to extract a one-to-one corresponding dataset of genotype records without germplasm differences and a set of labels to identify resistance scores from the multiple datasets. For example, sample genotype records that are all resistant to Fusarium wilt are selected, such as the frequent genotype groups A1B2C1D2 (sample number S1) and A1B2C1D1 (sample number S3) in Table 1, ensuring that the data only contains genotype differences and not germplasm differences. It is important to note that although these samples belong to the same germplasm type, the differences in their genotypes will directly lead to fluctuations in their phenotypic characteristics such as disease resistance, yield, and quality, thus making the labels identifying resistance scores effectively different, providing valuable data for the backbone resistance assessment network.

[0124] Then, the corresponding resistance score labels for the above sample genotype groups, i.e., the minimum resistance score mentioned above, such as 92 points, are needed as supervision signals for training the backbone resistance assessment network.

[0125] For example, the dataset of genotype records without germplasm differences and the label set of a class of resistance scores contain 1,000 pairs of data, namely genotype (A1B2C1D2) - resistance score (92), genotype (A1B2C1D1) - resistance score (88), etc.

[0126] In addition, it is necessary to extract one-to-one target germplasm differential genotype record datasets and two-category label sets of resistance scores from the multiple sets of data. That is, to screen germplasm types as new target germplasms, such as anthrax-resistant germplasms, and sample genotype record data that are different from the baseline germplasm used for training the main resistance assessment network. These genotype sets have systematic differences from the baseline germplasm genotype sets due to germplasm differences. Then, the label of the resistance score corresponding to the target germplasm sample is used as a supervision signal for training the branch fully connected neural network. The dataset contains 500 records, each of which is the target germplasm differential genotype set (A3B1C2D4) - label resistance score (80), the target germplasm differential genotype set (A3B1C3D4) - label resistance score (76), etc.

[0127] Furthermore, using the aforementioned label set for identifying resistance scores as supervision, and the aforementioned dataset of genotype groups without germplasm differences, machine learning is employed to train a fully connected neural network to obtain the backbone resistance assessment network.

[0128] Specifically, the backbone resistance assessment network is constructed using a fully connected neural network. The input layer dimension is the genotype feature dimension. For example, if each genotype contains 4 gene loci, and each locus is represented by 2 features, then the input layer dimension = 8. There are 2-3 hidden layers, such as a 64- or 32-neuron structure. The output layer consists of 1 neuron, outputting the predicted resistance score.

[0129] The input to the backbone resistance assessment network is a dataset of genotype records without germplasm differences; the supervision signal is a set of labels that identify resistance scores, such as 92, 88, etc.; the loss function is the mean squared error (MSE), which measures the difference between the predicted score and the true label; the optimizer uses the Adam optimizer to iteratively adjust the network parameters and minimize the loss function; the training termination condition is that the training set loss converges, such as the loss change being <0.001 for 10 consecutive rounds; or the preset number of iterations is reached, such as 1000 rounds.

[0130] After training convergence, the backbone resistance assessment network is obtained. This network can predict the resistance score based on the genotype of the baseline germplasm. For example, if the input genotype is A1B2C1D2, the output resistance score prediction value is 91.5, which is close to the true label value of 92.

[0131] Furthermore, it is necessary to construct and train a first-level compensatory resistance assessment network. This involves connecting a branch fully connected neural network to the main resistance assessment network, and averaging the outputs of the main resistance assessment network and the branch fully connected neural network to obtain the first-level compensatory resistance assessment network architecture. The model parameters of the main resistance assessment network are then frozen. Using the label set of the two types of resistance scores as supervision, and the target germplasm differential genotype group recording dataset (i.e., sample data different from the benchmark germplasm used when training the main network), machine learning is used to train the first-level compensatory resistance assessment network architecture to obtain the first-level compensatory resistance assessment network.

[0132] Specifically, the trained backbone resistance assessment network needs to be retained; then, a branch fully connected neural network is connected next to the backbone resistance assessment network, with the same structure as the hidden layer of the backbone network, such as a 64- or 32-neuron structure, and the input is the feature vector of the target germplasm differential genotype group; the output layer takes the average of the output of the backbone resistance assessment network and the output of the branch fully connected neural network as the final prediction.

[0133] During the training of the primary compensatory resistance assessment network architecture, the parameters of the main resistance assessment network need to be frozen to ensure that the baseline germplasm patterns learned by the main resistance assessment network are not destroyed. The supervision signal is a set of labels for the second-class resistance scores. During training, only the parameters of the branch networks are updated. By adjusting the branch outputs, the mean predictions are made closer to the true labels.

[0134] For example, the main resistance assessment network predicts a resistance score of 75 for the target germplasm differential genotype A3B1C2D4. This predicted value deviates from the true label of 80 due to germplasm differences. The initial output of the branch fully connected neural network is 70, with a mean of 72.5. After training through the above steps, the output of the branch fully connected neural network is adjusted to 85, and the mean (75+85) / 2=80, which matches the true label.

[0135] Finally, the target germplasm samples with pre-reserved resistance score labels are used as the test set and input into the first-level compensation network to calculate the verification accuracy between the predicted resistance score and the real label. If the error is set to ≤5 points, the verification accuracy is calculated as: (Number of correct samples / Total number of samples) × 100%.

[0136] If the verification accuracy is greater than or equal to the accuracy threshold, such as 85%, it means that the first-level compensation can meet the evaluation requirements of the target germplasm, and the first-level compensation resistance evaluation network is set as the final resistance evaluation network.

[0137] If the verification accuracy is less than the accuracy threshold, it indicates that single-branch compensation is insufficient, and a new fully connected branch neural network needs to be added. That is, based on the first-level compensated resistance assessment network architecture, a new fully connected branch neural network is connected to the backbone resistance assessment network. This new fully connected branch neural network has the same structure and training method as the previous one. When training the new fully connected branch neural network, the backbone resistance assessment network and the already trained fully connected branch neural network need to be frozen; only the newly added fully connected branch neural network is trained. The input remains the feature vector of the target germplasm differential genotype group, and the supervision signal is the label set of two-class resistance scores.

[0138] The output is the average of the outputs of the main resistance assessment network and the two fully connected neural networks. If there are two fully connected neural networks, the output value is (output value of the main resistance assessment network + output value of the first fully connected neural network + output value of the second fully connected neural network) / 3.

[0139] Next, repeat the process of adding branches, training branches, and validating accuracy until the network's validation accuracy is greater than or equal to the accuracy threshold. Finally, set the network as the resistance evaluation network.

[0140] Furthermore, the concentrated value of the aforementioned initial resistance scores needs to be taken and set as the first resistance score. This process continues until the Nth watermelon germplasm type, Nth rootstock type, and Nth planting scheme of the Nth watermelon resistance breeding program are extracted, the Nth index constraint is constructed, the Nth watermelon breeding sample set stored on the blockchain is collected, resistance assessment is performed, and the Nth resistance score is obtained.

[0141] Specifically, several initial resistance scores for the first watermelon resistance breeding program are obtained. Based on the data distribution characteristics, an appropriate central tendency calculation method is selected: if there are no extreme values, the arithmetic mean is used; if extreme values ​​exist, the median is used. The calculated central tendency is set as the first resistance score for this program, serving as a representative indicator of the program's resistance level.

[0142] Then, following the same method as obtaining the first resistance score, continue calculating the resistance scores for the N watermelon resistance breeding programs until all N resistance scores for the N watermelon resistance breeding programs are obtained.

[0143] In step S300 of this application embodiment, using the first resistance score up to the Nth resistance score as fitness, the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme is optimized, and the selected watermelon resistance breeding scheme is sent to the user terminal to execute the breeding experiment, including:

[0144] Step 1: Based on the first resistance score up to the Nth resistance score as fitness, extract the resistance breeding scheme with the maximum fitness, the resistance breeding scheme with the median fitness, and the resistance breeding scheme with a preset number of fitness tails from the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme;

[0145] Step 2: Using the resistance breeding scheme with the maximum fitness and the resistance breeding scheme with the median fitness as search targets, adjust the fitness preset number of resistance breeding schemes with the fitness tail to obtain an updated watermelon resistance breeding scheme;

[0146] Step 3: When the fitness of the updated watermelon resistance breeding program is less than or equal to that of the first watermelon resistance breeding program up to the minimum fitness value of the Nth watermelon resistance breeding program, return to Step 2 and execute the loop.

[0147] Step 4: When the fitness of the updated watermelon resistance breeding scheme is greater than the minimum fitness value of the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme, update the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme based on the updated watermelon resistance breeding scheme, and then return to Step 1 to execute the loop.

[0148] In this embodiment of the application, step S300 involves obtaining the resistance scores of all N breeding schemes, and then using a backtracking search logic that extracts key schemes, adjusts inferior schemes in a targeted manner, and iteratively verifies and updates the schemes to achieve dynamic iterative optimization of the N groups of watermelon resistance breeding schemes. The core objective is to efficiently screen out the schemes with the best resistance level that are suitable for actual breeding needs.

[0149] In step one, it is first necessary to extract the resistance breeding scheme with the maximum fitness, the resistance breeding scheme with the median fitness, and the resistance breeding scheme with the tail of fitness from the first watermelon resistance breeding scheme to the Nth watermelon resistance breeding scheme, based on the first resistance score up to the Nth resistance score as fitness.

[0150] Specifically, the N groups of watermelon resistance breeding programs need to be sorted from high to low according to their resistance scores, i.e., fitness. For example, when N=100, a score of 95 is ranked 1st and a score of 60 is ranked 100th.

[0151] The scheme with the highest fitness is the watermelon resistance breeding scheme ranked first. The selection method for the median fitness scheme is divided into two cases: if N is an odd number, such as N=99, the scheme ranked in the middle is selected, that is, the scheme ranked 50th; if N is an even number, such as N=100, the two schemes ranked in the middle are selected, that is, the two schemes ranked 50th and 51st are selected; the preset number of tail schemes is the schemes at the end of the ranking according to the user-defined number, such as 10, such as the schemes ranked 91-100 with a score of 60-65.

[0152] In step two, the maximum fitness resistance breeding scheme and the median fitness resistance breeding scheme are used as search targets to adjust the preset fitness number of fitness tail resistance breeding schemes to obtain an updated watermelon resistance breeding scheme.

[0153] The search targets are the resistance breeding schemes with the highest fitness and the resistance breeding schemes with the median fitness. The tail schemes are adjusted in two ways: categorized by type parameters and numerical parameters, to generate and update watermelon resistance breeding schemes.

[0154] Specifically, for type parameters, such as germplasm and rootstock type, the type parameters of the target search solution are directly replaced with the type parameters of the tail solution that share the same attribute. For example, if the tail solution is "anthrax-resistant germplasm + gourd rootstock" (score 62), and the target search solution is "wilt-resistant germplasm + pumpkin rootstock" (score 95), then "anthrax-resistant germplasm" in the tail solution is replaced with "wilt-resistant germplasm," and "gourd rootstock" is replaced with "pumpkin rootstock."

[0155] For numerical parameters, such as drug concentration and number of color plates, the numerical parameters of the target solution are used as the endpoint, and the numerical parameters of the tail solution are used as the starting point. The updated value is generated by moving through an S-shaped fluctuation according to the user-defined fluctuation range. For example, if the fluctuation range is ±5%, the updated value is generated by first slowly adjusting, then accelerating in the middle, and finally fine-tuning in the later stage. For instance, if the tail solution is "formalin concentration 3% (starting point)" and the target solution is "formalin concentration 1% (end point)", with a fluctuation range of ±5%, the first adjustment is to 2.85% (3% × 95%), the second to 2.5% (2.85% × 87.7%, S-shaped acceleration), the third to 1.2% (2.5% × 48%), and finally approaching 1%, resulting in the updated value "1.1%".

[0156] Finally, the adjusted type and numerical parameters are integrated to form an updated watermelon resistance breeding program, such as "Fusarium wilt resistant germplasm + pumpkin rootstock + 1.1% formalin for orchard sanitation".

[0157] In step three, when the fitness of the updated watermelon resistance breeding scheme is less than or equal to the minimum fitness value of the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme, it is necessary to return to step two to execute the loop.

[0158] Specifically, the updated watermelon resistance breeding program needs to be substituted into the resistance assessment network to obtain its resistance score. If the fitness of the updated watermelon resistance breeding program is less than or equal to the minimum fitness of the original N watermelon resistance breeding programs, it indicates that the adjustment direction is incorrect. Return to step two and readjust based on the original tail program, such as changing the parameters of another search target program. If the fitness of the updated watermelon resistance breeding program is greater than the minimum fitness of the original N watermelon resistance breeding programs, it indicates that the adjustment is effective, and proceed to step four.

[0159] In step four, when the fitness of the updated watermelon resistance breeding program is greater than the minimum fitness value of the first watermelon resistance breeding program up to the Nth watermelon resistance breeding program, the first watermelon resistance breeding program up to the Nth watermelon resistance breeding program needs to be updated based on the updated watermelon resistance breeding program, and then the process returns to step one to execute the loop.

[0160] That is, when the fitness of the updated watermelon resistance breeding program is greater than the minimum fitness value of the first watermelon resistance breeding program up to the Nth watermelon resistance breeding program, one tail program in the original N watermelon resistance breeding programs is deleted. For example, the program with a resistance score of 60 is added to the updated program (score of 88), keeping the total number of watermelon resistance breeding programs still N.

[0161] Furthermore, repeat steps one through four. After three consecutive iterations, if the maximum fitness of the N watermelon resistance breeding schemes no longer increases, or if the user-defined number of iterations is reached (e.g., 20 rounds), then stop the loop and select the watermelon resistance breeding scheme with the highest fitness as the chosen watermelon resistance breeding scheme. Send the scheme to the user for breeding experiment execution.

[0162] Example 2, as Figure 2 As shown, based on the same inventive concept as the watermelon resistance breeding optimization method for cultivation data identification and traceability provided in Embodiment 1, this embodiment of the invention also provides a watermelon resistance breeding optimization system for cultivation data identification and traceability, comprising:

[0163] The watermelon resistance breeding program configuration module 11 is used to configure the first watermelon resistance breeding program up to the Nth watermelon resistance breeding program through the user terminal;

[0164] The watermelon resistance breeding program evaluation module 12 is used to extract the first watermelon germplasm type, the first rootstock type, and the first planting scheme of the first watermelon resistance breeding program, construct the first index constraint, collect the first watermelon cultivation sample set that has been stored on the blockchain, perform resistance evaluation, and obtain the first resistance score; until the Nth watermelon resistance breeding program extracts the Nth watermelon germplasm type, the Nth rootstock type, and the Nth planting scheme, constructs the Nth index constraint, collects the Nth watermelon cultivation sample set that has been stored on the blockchain, performs resistance evaluation, and obtains the Nth resistance score;

[0165] The watermelon resistance breeding program optimization module 13 is used to optimize the first watermelon resistance breeding program up to the Nth watermelon resistance breeding program with the first resistance score up to the Nth resistance score as fitness, and send the selected watermelon resistance breeding program to the user terminal for breeding experiment execution.

[0166] Furthermore, the watermelon resistance breeding program configuration module 11 includes the following execution steps:

[0167] Through the user terminal, the set of watermelon constrained germplasm types, rootstock constrained types, planting parameter constrained types, and planting parameter constrained intervals are obtained;

[0168] A value is randomly extracted from the watermelon constrained germplasm type set, the rootstock constrained type set, the planting parameter constrained type set, and the planting parameter constrained interval set, and combined to obtain the first watermelon resistance breeding scheme;

[0169] The Nth watermelon resistance breeding scheme is obtained by randomly extracting a value from the watermelon constrained germplasm type set, the rootstock constrained type set, the planting parameter constrained type set, and the planting parameter constrained interval set, and combining them.

[0170] Furthermore, the watermelon resistance breeding program evaluation module 12 includes the following execution steps:

[0171] Extract the first watermelon germplasm type, first rootstock type, and first planting scheme of the first watermelon resistance breeding scheme, and construct the first index constraint;

[0172] Collect the first watermelon breeding sample set that has been stored on the blockchain and meets the first index constraint, perform genotype mode analysis, and obtain several frequent genotype groups;

[0173] The resistance assessment network is used to perform resistance assessments by traversing the several frequent genotype groups and obtaining several initial resistance scores. The resistance assessment network is generated by machine learning training using multiple sets of data, and each set of data includes genotype group record data and labels that identify resistance scores.

[0174] Take the concentrated value of the aforementioned initial resistance scores and set it as the first resistance score.

[0175] Specifically, the first watermelon germplasm type, first rootstock type, and first planting scheme of the first watermelon resistance breeding program are extracted to construct the first index constraint, including:

[0176] Extract the first planting control attribute up to the Qth planting control attribute from the first planting scheme;

[0177] Genotypic fluctuation association analysis was performed on the first watermelon germplasm type, the first rootstock type, the first planting control attribute up to the Qth planting control attribute to obtain the weight distribution results;

[0178] Based on the weight distribution results, the normalization deviations of the first watermelon germplasm type, the first rootstock type, the first planting control attribute, and up to the Qth planting control attribute are weighted and summed to obtain the index deviation coefficient. Specifically, when the watermelon resistance breeding type parameter of the sample to be analyzed is different from the first watermelon resistance breeding type parameter, the normalization deviation is equal to 1; otherwise, the normalization deviation is equal to 0. When the quantification deviation between the watermelon resistance breeding quantification parameter of the sample to be analyzed and the first watermelon resistance breeding quantification parameter is greater than or equal to the user-preset quantification deviation threshold of the corresponding attribute, the normalization deviation is equal to 1; otherwise, the normalization deviation is equal to 0.

[0179] If the index deviation coefficient is less than or equal to the deviation coefficient threshold, the resistance breeding scheme of the sample to be analyzed is considered to meet the first index constraint; otherwise, it is considered not to meet the constraint.

[0180] Specifically, genotypic fluctuation association analysis was performed on the first watermelon germplasm type, the first rootstock type, the first planting control attribute up to the Qth planting control attribute to obtain the weight distribution results, including:

[0181] The first and second samples to be analyzed were retrieved, with a normalized deviation of 1 for the first watermelon germplasm type and a normalized deviation of 0 for other resistance breeding. The crossover ratio of genotypes was calculated and set as the genotype fluctuation parameter, and added to the genotype fluctuation parameter set.

[0182] When the number of the genotype fluctuation parameter set is equal to the preset statistical number, the set value of the genotype fluctuation parameter set is calculated and set as the first watermelon germplasm type correlation degree.

[0183] Until the correlation degree of the Qth planting control attribute is obtained;

[0184] Iterate through the correlation degree of the first watermelon germplasm type up to the correlation degree of the Qth planting control attribute, compare the sum of the correlation degrees, set it as the weight of the first watermelon germplasm type up to the weight of the Qth planting control attribute, and add it to the weight distribution result.

[0185] Specifically, the process involves traversing several frequent genotype groups, performing resistance assessment through a resistance assessment network, and obtaining several initial resistance scores. The resistance assessment network is generated using machine learning training on multiple sets of data. Each set of data includes genotype group records and labels identifying the resistance score, including:

[0186] Extract one-to-one corresponding datasets of genotype records without germplasm differences and a set of labels that identify resistance scores from the multiple sets of data;

[0187] Using the label set that identifies resistance scores as supervision, and the dataset of the genotype group without germplasm differences, machine learning is used to train a fully connected neural network to obtain the backbone resistance evaluation network, wherein the germplasm characterizes the watermelon germplasm during breeding.

[0188] Extract one-to-one target germplasm differential genotype record datasets and label sets of two types of resistance scores from the multiple sets of data;

[0189] A branch fully connected neural network is connected to the main resistance assessment network. The output is the average of the outputs of the main resistance assessment network and the branch fully connected neural network to obtain the first-level compensatory resistance assessment network architecture. The model parameters of the main resistance assessment network are frozen. The label set of the two types of resistance scores is used as supervision. The dataset of the genotype group without germplasm differences is used to train the first-level compensatory resistance assessment network architecture using machine learning to obtain the first-level compensatory resistance assessment network.

[0190] When the verification accuracy of the first-level compensated resistance evaluation network is greater than or equal to the accuracy threshold, the first-level compensated resistance evaluation network is set as the resistance evaluation network.

[0191] When the validation accuracy of the first-level compensation resistance evaluation network is less than the accuracy threshold, a branch fully connected neural network is connected to the backbone resistance evaluation network based on the first-level compensation resistance evaluation network architecture. The output is the average of the outputs of the backbone resistance evaluation network and the two branch fully connected neural networks to obtain the second-level compensation resistance evaluation network architecture, and iterative training is performed.

[0192] The step of obtaining the label for the resistance score includes:

[0193] Obtain labels indicating watermelon disease resistance scores, watermelon yield resistance scores, and watermelon quality scores;

[0194] Extract the minimum score from the labels indicating watermelon disease resistance, watermelon yield resistance, and watermelon quality, and set it as the label indicating resistance.

[0195] Furthermore, the watermelon resistance breeding program optimization module 13 includes the following execution steps:

[0196] Step 1: Based on the first resistance score up to the Nth resistance score as fitness, extract the resistance breeding scheme with the maximum fitness, the resistance breeding scheme with the median fitness, and the resistance breeding scheme with a preset number of fitness tails from the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme;

[0197] Step 2: Using the resistance breeding scheme with the maximum fitness and the resistance breeding scheme with the median fitness as search targets, adjust the fitness preset number of resistance breeding schemes with the fitness tail to obtain an updated watermelon resistance breeding scheme;

[0198] Step 3: When the fitness of the updated watermelon resistance breeding program is less than or equal to that of the first watermelon resistance breeding program up to the minimum fitness value of the Nth watermelon resistance breeding program, return to Step 2 and execute the loop.

[0199] Step 4: When the fitness of the updated watermelon resistance breeding scheme is greater than the minimum fitness value of the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme, update the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme based on the updated watermelon resistance breeding scheme, and then return to Step 1 to execute the loop.

[0200] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0201] Those skilled in the art will understand that embodiments of the present invention can provide methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0202] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0203] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0204] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0205] Although preferred embodiments of the invention have been described, those skilled in the art, once they have learned the basic inventive concept, can make other changes and modifications to these embodiments.

[0206] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of this invention and its equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for selecting watermelon resistance-resistant varieties based on data identification and traceability, characterized in that, include: Configure the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme through the user terminal; Extract the first watermelon germplasm type, first rootstock type, and first planting scheme of the first watermelon resistance breeding scheme, construct the first index constraint, collect the first watermelon cultivation sample set that has been stored on the blockchain, perform resistance assessment, and obtain the first resistance score; Until the Nth watermelon germplasm type, Nth rootstock type and Nth planting scheme of the Nth watermelon resistance breeding scheme are extracted, the Nth index constraint is constructed, the Nth watermelon cultivation sample set that has been stored on the blockchain is collected, resistance assessment is performed, and the Nth resistance score is obtained; Using the first resistance score up to the Nth resistance score as fitness, the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme is optimized, and the selected watermelon resistance breeding scheme is sent to the user terminal to execute the breeding experiment. Specifically, the first watermelon germplasm type, first rootstock type, and first planting scheme of the first watermelon resistance breeding program are extracted, a first index constraint is constructed, a first watermelon cultivation sample set stored on the blockchain is collected, resistance assessment is performed, and a first resistance score is obtained, including: Extract the first watermelon germplasm type, first rootstock type, and first planting scheme of the first watermelon resistance breeding scheme, and construct the first index constraint; Collect the first watermelon breeding sample set that has been stored on the blockchain and meets the first index constraint, perform genotype mode analysis, and obtain several frequent genotype groups; The resistance assessment network is used to perform resistance assessments by traversing the several frequent genotype groups and obtaining several initial resistance scores. The resistance assessment network is generated by machine learning training using multiple sets of data, and each set of data includes genotype group record data and labels that identify resistance scores. Take the concentrated value of the aforementioned initial resistance scores and set it as the first resistance score; The fitness of the first watermelon resistance breeding program up to the Nth resistance score is determined by using the first resistance score up to the Nth resistance score. The selected watermelon resistance breeding program is then sent to the user terminal for breeding experiments, including: Step 1: Based on the first resistance score up to the Nth resistance score as fitness, extract the resistance breeding scheme with the maximum fitness, the resistance breeding scheme with the median fitness, and the resistance breeding scheme with a preset number of fitness tails from the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme; Step 2: Using the resistance breeding scheme with the maximum fitness and the resistance breeding scheme with the median fitness as search targets, adjust the fitness preset number of resistance breeding schemes with the fitness tail to obtain an updated watermelon resistance breeding scheme; Step 3: When the fitness of the updated watermelon resistance breeding program is less than or equal to that of the first watermelon resistance breeding program up to the minimum fitness value of the Nth watermelon resistance breeding program, return to Step 2 and execute the loop. Step 4: When the fitness of the updated watermelon resistance breeding scheme is greater than the minimum fitness value of the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme, update the first watermelon resistance breeding scheme up to the Nth watermelon resistance breeding scheme based on the updated watermelon resistance breeding scheme, and then return to Step 1 to execute the loop.

2. The method as described in claim 1, characterized in that, Through the user terminal, configure the first watermelon resistance breeding program up to the Nth watermelon resistance breeding program, including: Through the user terminal, the set of watermelon constrained germplasm types, rootstock constrained types, planting parameter constrained types, and planting parameter constrained intervals are obtained; A value is randomly extracted from the watermelon constrained germplasm type set, the rootstock constrained type set, the planting parameter constrained type set, and the planting parameter constrained interval set, and combined to obtain the first watermelon resistance breeding scheme; The Nth watermelon resistance breeding scheme is obtained by randomly extracting a value from the watermelon constrained germplasm type set, the rootstock constrained type set, the planting parameter constrained type set, and the planting parameter constrained interval set, and combining them.

3. The method as described in claim 1, characterized in that, Extract the first watermelon germplasm type, first rootstock type, and first planting scheme from the first watermelon resistance breeding program, and construct the first index constraint, including: Extract the first planting control attribute up to the Qth planting control attribute from the first planting scheme; Genotypic fluctuation association analysis was performed on the first watermelon germplasm type, the first rootstock type, the first planting control attribute up to the Qth planting control attribute to obtain the weight distribution results; Based on the weight distribution results, the normalization deviations of the first watermelon germplasm type, the first rootstock type, the first planting control attribute, and up to the Qth planting control attribute are weighted and summed to obtain the index deviation coefficient. Specifically, when the watermelon resistance breeding type parameter of the sample to be analyzed is different from the first watermelon resistance breeding type parameter, the normalization deviation is equal to 1; otherwise, the normalization deviation is equal to 0. When the quantification deviation between the watermelon resistance breeding quantification parameter of the sample to be analyzed and the first watermelon resistance breeding quantification parameter is greater than or equal to the user-preset quantification deviation threshold of the corresponding attribute, the normalization deviation is equal to 1; otherwise, the normalization deviation is equal to 0. If the index deviation coefficient is less than or equal to the deviation coefficient threshold, the resistance breeding scheme of the sample to be analyzed is considered to meet the first index constraint; otherwise, it is considered not to meet the constraint.

4. The method as described in claim 3, characterized in that, Genotypic fluctuation association analysis was performed on the first watermelon germplasm type, the first rootstock type, the first planting control attribute up to the Qth planting control attribute to obtain the weight distribution results, including: The first and second samples to be analyzed were retrieved, with a normalized deviation of 1 for the first watermelon germplasm type and a normalized deviation of 0 for other resistance breeding. The crossover ratio of genotypes was calculated and set as the genotype fluctuation parameter, and added to the genotype fluctuation parameter set. When the number of the genotype fluctuation parameter set is equal to the preset statistical number, the set value of the genotype fluctuation parameter set is calculated and set as the first watermelon germplasm type correlation degree. Until the correlation degree of the Qth planting control attribute is obtained; Iterate through the correlation degree of the first watermelon germplasm type up to the correlation degree of the Qth planting control attribute, compare the sum of the correlation degrees, set it as the weight of the first watermelon germplasm type up to the weight of the Qth planting control attribute, and add it to the weight distribution result.

5. The method as described in claim 1, characterized in that, The resistance assessment network is used to traverse several frequent genotype groups and perform resistance assessment to obtain several initial resistance scores. The resistance assessment network is generated using machine learning training on multiple sets of data. Each set of data includes genotype group records and labels identifying the resistance score, including: Extract one-to-one corresponding datasets of genotype records without germplasm differences and a set of labels that identify resistance scores from the multiple sets of data; Using the label set that identifies resistance scores as supervision, and the dataset of the genotype group without germplasm differences, machine learning is used to train a fully connected neural network to obtain the backbone resistance evaluation network, wherein the germplasm characterizes the watermelon germplasm during breeding. Extract one-to-one target germplasm differential genotype record datasets and label sets of two types of resistance scores from the multiple sets of data; A branch fully connected neural network is connected to the main resistance assessment network. The output is the average of the outputs of the main resistance assessment network and the branch fully connected neural network to obtain the first-level compensatory resistance assessment network architecture. The model parameters of the main resistance assessment network are frozen. The label set of the two types of resistance scores is used as supervision. The dataset of the genotype group without germplasm differences is used to train the first-level compensatory resistance assessment network architecture using machine learning to obtain the first-level compensatory resistance assessment network. When the verification accuracy of the first-level compensated resistance evaluation network is greater than or equal to the accuracy threshold, the first-level compensated resistance evaluation network is set as the resistance evaluation network. When the validation accuracy of the first-level compensation resistance evaluation network is less than the accuracy threshold, a branch fully connected neural network is connected to the backbone resistance evaluation network based on the first-level compensation resistance evaluation network architecture. The output is the average of the outputs of the backbone resistance evaluation network and the two branch fully connected neural networks to obtain the second-level compensation resistance evaluation network architecture, and iterative training is performed.

6. The method as described in claim 5, characterized in that, The steps for obtaining the label indicating the resistance score include: Obtain labels indicating watermelon disease resistance scores, watermelon yield resistance scores, and watermelon quality scores; Extract the minimum score from the labels indicating watermelon disease resistance, watermelon yield resistance, and watermelon quality, and set it as the label indicating resistance.

7. A watermelon resistance breeding and selection system for data identification and traceability, characterized in that, The system is used to implement the watermelon resistance breeding optimization method for breeding data identification and traceability as described in any one of claims 1-6, including: The watermelon resistance breeding program configuration module is used to configure the first watermelon resistance breeding program up to the Nth watermelon resistance breeding program through the user terminal; The watermelon resistance breeding program evaluation module is used to extract the first watermelon germplasm type, first rootstock type, and first planting scheme of the first watermelon resistance breeding program, construct the first index constraint, collect the first watermelon cultivation sample set that has been stored on the blockchain, perform resistance evaluation, and obtain the first resistance score; until the Nth watermelon resistance breeding program extracts the Nth watermelon germplasm type, Nth rootstock type, and Nth planting scheme, constructs the Nth index constraint, collects the Nth watermelon cultivation sample set that has been stored on the blockchain, performs resistance evaluation, and obtains the Nth resistance score; The watermelon resistance breeding program optimization module is used to optimize the first watermelon resistance breeding program up to the Nth watermelon resistance breeding program using the first resistance score up to the Nth resistance score as fitness, and send the selected watermelon resistance breeding program to the user terminal for breeding experiment execution.

Images