Digital storage library management method and system
By introducing a progressive hierarchical extraction model with hierarchical perception bias terms, the problem of feature information differentiation and collaborative utilization in tissue culture seedling management was solved, realizing precise management and environmental configuration of tissue culture seedlings, and improving the quality of seedlings leaving the nursery and storage efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-30
- Publication Date
- 2026-04-07
AI Technical Summary
Existing technologies struggle to effectively distinguish and synergistically utilize feature information at different abstraction levels in the management of plant tissue culture seedlings. This results in insufficient precision and poor interpretability of environmental parameter control vectors, affecting the growth status and quality of tissue culture seedlings upon delivery.
A progressive hierarchical extraction model is adopted. By introducing a hierarchical perception bias term, the proprietary and shared features of tissue culture seedlings are extracted and fused hierarchically to generate an accurate environmental response vector. Based on this vector, the optimal scheduling of storage resources and personalized environmental configuration are performed.
It has enabled precise, automated, and continuous optimization of tissue culture seedling management, significantly improving the quality of seedlings leaving the nursery and the efficiency of storage operations.
Smart Images
Figure CN121809964A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of plant tissue culture seedling management, and in particular to a digital repository management method and system. Background Technology
[0002] In large-scale production of plant tissue culture, meticulous storage management of tissue culture seedlings is crucial to ensuring their quality and uniformity. Traditional management methods rely heavily on the experience of technicians, using fixed formulas to roughly set environmental parameters. This approach is ill-suited to the individualized needs of different varieties and growth stages of tissue culture seedlings, easily leading to differentiation in growth status and affecting the quality of seedlings ready for sale.
[0003] With the widespread adoption of IoT and sensor technologies, obtaining real-time phenotypic data of tissue culture seedlings and warehouse environmental data has become possible. Some intelligent attempts utilize machine learning models to correlate formulations with environmental data to predict or recommend environmental parameters. However, most existing methods treat formulations and phenotypic data as flattened feature inputs to general models. These models have inherent limitations when processing hierarchical biological data: their internal feature extraction process lacks guidance and cannot effectively distinguish and synergistically utilize feature information at different levels of abstraction—that is, they fail to fully exploit shallow, variety-specific features and efficiently integrate deep, general physiological laws across varieties and tasks.
[0004] Therefore, control vectors generated based on general models often lack accuracy and interpretability, leading to significant deviations in subsequent storage environment matching and initialization operations. How to provide a digital repository management method has become a pressing issue in this field. Summary of the Invention
[0005] This application provides a digital storage management method and system. The method introduces a hierarchical perception bias term, which enables a progressive hierarchical extraction model to extract and integrate the proprietary and shared features of tissue culture seedlings in a hierarchical manner, generating an accurate and interpretable environmental response vector. Based on this environmental response vector, the optimal scheduling of storage resources and personalized environmental configuration are realized, thereby achieving precise, automated and continuous optimization of tissue culture seedling management, significantly improving the quality of seedlings leaving the nursery and the efficiency of storage operations.
[0006] Firstly, a digital repository management method is provided, the method comprising:
[0007] S1: Obtain digital formula data of plant tissue culture seedlings and collect phenotypic data of the plant tissue culture seedlings;
[0008] S2: Input the digital formula data and the phenotypic data into the improved progressive hierarchical extraction model to generate the environmental response vector of the plant tissue culture seedling. The improvement of the progressive hierarchical extraction model includes: introducing a hierarchical awareness bias term for the task gating network of each layer in the hierarchical extraction mechanism.
[0009] S3: Match the environmental response vector with the preset environmental parameters of each storage layer, determine the target storage layer of the plant tissue culture seedling based on the matching result, and generate scheduling instructions;
[0010] S4: According to the scheduling instruction, the tissue culture seedlings are transferred to the target storage layer, and the environmental parameters of the target storage layer are adjusted according to the environmental response vector.
[0011] It should be understood that the Progressive Layered Extraction (PLE) model is a high-level neural network architecture specifically designed for multi-task learning, which can effectively overcome the "negative transfer" or "seesaw effect" in traditional multi-task learning models. This application chooses the Progressive Layered Extraction model as its foundation mainly because its multi-task architecture, which combines shared and exclusive experts in parallel, is well-suited to the practical needs of simultaneously predicting multiple environmental parameters in tissue culture seedling management. By sharing experts to learn common physiological characteristics, each task's exclusive expert focuses on extracting key information specific to parameters such as light intensity and stress resistance. The hierarchical perception bias term introduced on this basis further guides the model to focus on task-related details at a shallow level and to integrate cross-task abstract patterns at a deep level. This mechanism effectively avoids interference between multiple tasks, ensuring that the final generated environmental response vector is both accurate and interpretable, providing a reliable basis for subsequent precise regulation.
[0012] In conjunction with the first aspect, in some implementations of the first aspect, the hierarchical perception bias term is calculated as follows:
[0013] B{i,k}=α·(k / K)·P other +β·((Kk) / K)·P self ,
[0014] Where k is the current layer index, K is the total number of layers, i is the task index, and P other Let P be the other's preference vector. self Let be the self-preference vector, and α and β be the parameters.
[0015] It should be understood that the hierarchical perception bias term introduced in this application enables shallow networks to enhance task-specific features to capture details, while deep networks focus on shared features to integrate general patterns. This mechanism effectively mitigates multi-task interference, enabling the generated environmental response vector to possess both higher accuracy and stronger robustness, providing a reliable basis for precise control.
[0016] In conjunction with the first aspect, in some implementations of the first aspect, the environmental response vector includes at least assessment values of light parameters, temperature and humidity parameters, and growth potential.
[0017] In conjunction with the first aspect, in some implementations of the first aspect, step S3 includes:
[0018] S301: Obtain the current environmental status data and physical space occupancy rate of each candidate storage layer;
[0019] S302: Perform an initial match between the environmental response vector and the preset environmental parameters of each candidate storage layer, and calculate the first matching degree S1;
[0020] S303: Based on the current environmental status data and the physical space occupancy rate, the first matching degree is corrected to generate the final matching degree S. f ;
[0021] S304: Determine the target storage layer based on the final matching degree.
[0022] In conjunction with the first aspect, in certain implementations of the first aspect, the S f The calculation method is: Sf = S1 × (1 - W) env ×D env )×(1-W occ ×R),
[0023] Among them, D env The comprehensive environmental deviation coefficient, R is the physical space occupancy rate, and W is the physical space deviation rate. env W occ These are the environmental weighting coefficient and the spatial weighting coefficient.
[0024] It should be understood that the comprehensive environmental deviation coefficient D env It is calculated from the normalized deviation of the current state of each environmental parameter from its preset value. This application summarizes all environmental deviations into a single coefficient D. env And it is listed alongside the physical space occupancy rate R, through two dynamically adjustable coefficients W. env W occ To control their weights.
[0025] In conjunction with the first aspect, in some implementations of the first aspect, the method includes: assigning an initial horizontal position to the plant tissue culture seedling according to the environmental response vector within the target storage layer; and adjusting the environmental parameters within the target storage layer according to new phenotypic data in subsequent storage.
[0026] In conjunction with the first aspect, in some implementations of the first aspect, the method further includes:
[0027] Collect and construct a sample dataset with the aforementioned digital formula data and initial phenotypic data as input and the final quality data at the time of leaving the nursery as the supervisory label;
[0028] Using the sample dataset, the hierarchical perception bias term parameters in the improved progressive hierarchical extraction model are iteratively optimized to improve the accuracy of the model's prediction of growth results.
[0029] Secondly, a digital repository management system is provided, the system comprising:
[0030] The data acquisition module is used to acquire the digital formula data and the phenotypic data;
[0031] The control module is used to execute the method described in any implementation of the first aspect and generate the scheduling instruction; the storage execution module stores the plant tissue culture seedlings to the target storage layer based on the scheduling instruction. Attached Figure Description
[0032] Figure 1 A flowchart illustrating the implementation of a digital repository management method provided in this application embodiment. Detailed Implementation
[0033] The terminology used in the following embodiments is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” and “this” are intended to also include expressions such as “one or more,” unless the context clearly indicates otherwise. It should also be understood that in the following embodiments of this application, “at least one” and “one or more” refer to one, two, or more than two. The term “and / or” is used to describe the relationship between related objects, indicating that three relationships can exist; for example, A and / or B can indicate: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character “ / ” generally indicates that the preceding and following related objects are in an “or” relationship.
[0034] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.
[0035] In large-scale plant tissue culture production, precise storage management of tissue-cultured seedlings is crucial to ensuring quality uniformity. Traditional methods rely on fixed formulas and human experience, making it difficult to meet the personalized needs of different varieties and growth stages, easily leading to differentiation in growth states. Although IoT technology can collect phenotypic and environmental data in real time, and existing intelligent methods attempt to use machine learning models to predict environmental parameters, these models typically flatten the data, lacking effective modeling of the hierarchical structure of biological data. Their internal feature extraction process fails to distinguish between shallow, specific features and deep, general rules, resulting in insufficient precision and poor interpretability of the generated regulatory vectors, making it difficult to support accurate storage decisions.
[0036] This application provides a digital repository management method and system, which can effectively overcome the above-mentioned problems.
[0037] The technical solutions of the embodiments of this application will be described below with reference to the accompanying drawings.
[0038] Figure 1 A flowchart illustrating the implementation of a digital repository management method provided in this application embodiment.
[0039] refer to Figure 1 In some examples, the method includes:
[0040] S1: Obtain digital formula data of plant tissue culture seedlings and collect phenotypic data of the plant tissue culture seedlings;
[0041] S2: Input the digital formula data and the phenotypic data into the improved progressive hierarchical extraction model to generate the environmental response vector of the plant tissue culture seedling. The improvement of the progressive hierarchical extraction model includes: introducing a hierarchical awareness bias term for the task gating network of each layer in the hierarchical extraction mechanism.
[0042] S3: Match the environmental response vector with the preset environmental parameters of each storage layer, determine the target storage layer of the plant tissue culture seedling based on the matching result, and generate scheduling instructions;
[0043] S4: According to the scheduling instruction, the tissue culture seedlings are transferred to the target storage layer, and the environmental parameters of the target storage layer are adjusted according to the environmental response vector.
[0044] In some examples, the hierarchical awareness bias term is calculated as follows:
[0045] B{i,k}=α·(k / K)·P other +β·((Kk) / K)·P self ,
[0046] Where k is the current layer index, K is the total number of layers, i is the task index, and P other Let P be the other's preference vector. self Let be the self-preference vector, and α and β be the parameters.
[0047] In some examples, the environmental response vector includes at least assessments of light parameters, temperature and humidity parameters, and growth potential.
[0048] In some examples, step S3 includes:
[0049] S301: Obtain the current environmental status data and physical space occupancy rate of each candidate storage layer;
[0050] S302: Perform an initial match between the environmental response vector and the preset environmental parameters of each candidate storage layer, and calculate the first matching degree S1;
[0051] S303: Based on the current environmental status data and the physical space occupancy rate, the first matching degree is corrected to generate the final matching degree S. f ;
[0052] S304: Determine the target storage layer based on the final matching degree.
[0053] In some examples, the S f The calculation method is as follows:
[0054] Sf=S1×(1-W env ×D env )×(1-W occ ×R),
[0055] Among them, D env The comprehensive environmental deviation coefficient, R is the physical space occupancy rate, and W is the physical space deviation rate. env W occ These are the environmental weighting coefficient and the spatial weighting coefficient.
[0056] In one possible implementation, the comprehensive environmental deviation coefficient D envIt is calculated from the normalized deviation between the current state of each environmental parameter and the preset value.
[0057] In some examples, the method includes: assigning an initial horizontal position to the plant tissue culture seedling based on the environmental response vector within the target storage layer; and adjusting the environmental parameters within the target storage layer based on new phenotypic data in subsequent storage.
[0058] In one possible implementation, a gradient-based optimal location matching algorithm is executed based on the priority of each dimension (such as light, temperature, and humidity requirements) in the environmental response vector and the measured gradient maps of the microenvironment at different horizontal positions within the target storage layer. This algorithm maps the vector requirements of the tissue culture seedlings to the physical space of the storage layer, assigning them an initial location that maximally satisfies their key requirements. In subsequent storage, new phenotypic data are periodically collected, and environmental parameter adjustments are calculated using a lightweight incremental prediction model. These adjustments are then converted by a PID controller into fine-tuning instructions for the macro-environmental settings of the storage layer (such as light intensity and humidification), thereby achieving dynamic and adaptive optimization of the storage environment for this batch of tissue culture seedlings.
[0059] In some examples, the method further includes:
[0060] Collect and construct a sample dataset with the aforementioned digital formula data and initial phenotypic data as input and the final quality data at the time of leaving the nursery as the supervisory label;
[0061] Using the sample dataset, the hierarchical perception bias term parameters in the improved progressive hierarchical extraction model are iteratively optimized to improve the accuracy of the model's prediction of growth results.
[0062] This application embodiment also provides a digital repository management system, the system comprising:
[0063] The data acquisition module is used to acquire the digital formula data and the phenotypic data;
[0064] The control module is used to execute the method described in any of the foregoing examples and generate the scheduling instructions;
[0065] The storage execution module stores the plant tissue culture seedlings to the target storage layer based on the scheduling instructions. The above are merely preferred embodiments of the present invention, and the scope of protection of the present invention is not limited to the above embodiments. Any equivalent modifications or variations made by those skilled in the art based on the disclosure of the present invention should be included within the scope of protection set forth in the claims.
Claims
1. A method for managing a digital repository, characterized in that, The method includes: S1: Obtain digital formula data of plant tissue culture seedlings and collect phenotypic data of the plant tissue culture seedlings; S2: Input the digital formula data and the phenotypic data into the improved progressive hierarchical extraction model to generate the environmental response vector of the plant tissue culture seedling. The improvement of the progressive hierarchical extraction model includes: introducing a hierarchical awareness bias term for the task gating network of each layer in the hierarchical extraction mechanism. S3: Match the environmental response vector with the preset environmental parameters of each storage layer, determine the target storage layer of the plant tissue culture seedling based on the matching result, and generate scheduling instructions; S4: According to the scheduling instruction, the tissue culture seedlings are transferred to the target storage layer, and the environmental parameters of the target storage layer are adjusted according to the environmental response vector.
2. The method according to claim 1, characterized in that, The calculation method for the hierarchical perception bias term is as follows: B{i,k}=α·(k / K)·P other +β·((K-k) / K)·P self , Where k is the current layer index, K is the total number of layers, i is the task index, and P other Let P be the other's preference vector. self Let be the self-preference vector, and α and β be the parameters.
3. The method according to claim 1, characterized in that, The environmental response vector includes at least assessment values for light parameters, temperature and humidity parameters, and growth potential.
4. The method according to claim 1, characterized in that, Step S3 includes: S301: Obtain the current environmental status data and physical space occupancy rate of each candidate storage layer; S302: Perform an initial match between the environmental response vector and the preset environmental parameters of each candidate storage layer, and calculate the first matching degree S1; S303: Based on the current environmental status data and the physical space occupancy rate, the first matching degree is corrected to generate the final matching degree S. f ; S304: Determine the target storage layer based on the final matching degree.
5. The method according to claim 4, characterized in that, The S f The calculation method is as follows: Sf=S1×(1-W env ×D env )×(1-W occ ×R), Among them, D env The comprehensive environmental deviation coefficient, R is the physical space occupancy rate, and W is the physical space deviation rate. env W occ These are the environmental weighting coefficient and the spatial weighting coefficient.
6. The method according to claim 1, characterized in that, The method includes: assigning an initial horizontal position to the plant tissue culture seedling according to the environmental response vector within the target storage layer; and adjusting the environmental parameters within the target storage layer according to new phenotypic data in subsequent storage.
7. The method according to claim 1, characterized in that, The method further includes: Collect and construct a sample dataset with the aforementioned digital formula data and initial phenotypic data as input and the final quality data at the time of leaving the nursery as the supervisory label; Using the sample dataset, the hierarchical perception bias term parameters in the improved progressive hierarchical extraction model are iteratively optimized to improve the accuracy of the model's prediction of growth results.
8. A digital repository management system, characterized in that, The system includes: The data acquisition module is used to acquire the digital formula data and the phenotypic data; A control module is used to execute the method as described in any one of claims 1 to 7 to generate the scheduling instruction; a storage execution module is used to store the plant tissue culture seedlings to the target storage layer based on the scheduling instruction.