Plant seed felt management method and system based on blockchain

Abstract

The present invention provides a plant seed felt management method and system based on blockchain, which relates to the field of agricultural management technology, including obtaining growth parameter information and identification information of plant seed felt, uploading the growth parameter information and the identification information to a blockchain network, and generating a growth file of the plant seed felt; extracting a smart contract through a visual feature in the blockchain network, parsing the growth image and the growth video, acquiring growth features, identifying growth abnormalities of the plant seed felt based on an abnormality recognition smart contract, and triggering a quality traceability process when growth abnormalities occur, and issuing early warning prompts to other plant seed felts of the same batch; generating a three-dimensional digital twin model of the plant seed felt based on a digital twin smart contract in the blockchain network, simulating the growth state of the plant seed felt based on different growth parameter information, generating simulated growth results, and dynamically adjusting the growth parameters of the plant seed felt based on the simulated growth results.

Description

Technical Field

[0001] The present invention relates to the field of agricultural management technology, and in particular to a plant seed blanket management method and system based on blockchain. Background Art

[0002] Plant seed felt is a new type of seedling raising technology. By evenly fixing seeds on biodegradable materials to form seed felt, the seedling raising process can be standardized and automated.

[0003] The traditional plant seed felt management method has the following problems: data is scattered and there is a lack of unified data management and sharing mechanisms, which leads to data island problems; the storage and transmission of planting data may face security risks such as tampering and forgery, and there is a lack of reliable data security assurance mechanisms; it relies on manual records and inspections, and the management efficiency is low; there is a lack of a complete traceability mechanism throughout the life cycle, which affects product quality and safety; it is difficult to achieve accurate monitoring and dynamic optimization of the growth process, which affects planting results and resource utilization efficiency.

[0004] In summary, by applying the technologies of blockchain, Internet of Things, and digital twins to plant seed felt management, data interoperability, sharing, security, reliability, and intelligent optimization can be achieved. The present invention can solve the problems existing in traditional management methods and improve the efficiency, credibility, and intelligence level of plant seed felt management. Summary of the invention

[0005] The embodiments of the present invention provide a plant seed felt management method and system based on blockchain, which can solve the problems in the prior art.

[0006] According to a first aspect of the embodiments of the present invention,

[0007] A plant seed blanket management method based on blockchain is provided, comprising:

[0008] Acquire growth parameter information and identification information of the plant seed felt, upload the growth parameter information and the identification information to the blockchain network, generate a growth profile of the plant seed felt, and bind the growth profile to the electronic tag of the plant seed felt;

[0009] Through the Internet of Things devices in the blockchain network, the growth images and growth videos of the plant seed felt are collected and uploaded in real time. Through the visual feature extraction smart contract in the blockchain network, the growth images and the growth videos are analyzed to obtain the growth features. Based on the abnormality identification smart contract, the abnormal growth of the plant seed felt is identified. When the abnormal growth occurs, the information of the plant seed felt manufacturer is obtained from the growth file. Through the supply chain management smart contract in the blockchain network, an abnormality alarm is sent to the manufacturer of the plant seed felt to trigger the quality traceability process of the plant seed felt; the batch number of the plant seed felt is obtained from the growth file, and based on the supply chain management smart contract, an early warning prompt is issued to other plant seed felts in the same batch;

[0010] Based on the digital twin smart contract in the blockchain network, a three-dimensional digital twin model of the plant seed felt is generated. Through the three-dimensional digital twin model, the growth state of the plant seed felt based on different growth parameter information is simulated to generate simulated growth results. Based on the simulated growth results, the growth parameters of the plant seed felt are dynamically adjusted.

[0011] In an optional embodiment,

[0012] By extracting the smart contract of visual features in the blockchain network, parsing the growth image and the growth video, and obtaining the growth features include:

[0013] Based on computer vision algorithms, the growth images and growth videos of plant seed felts are analyzed to obtain growth features, wherein the growth features include scale-invariant shape feature vectors, color aggregation vectors, and motion boundary histogram feature vectors;

[0014] The growth image is input into the region-based convolutional neural network model, and the leaf region candidate frame is automatically extracted through the region proposal network. The convolutional neural network is used to classify and regress the leaves in the leaf region candidate frame to determine the leaf segmentation result, and the coordinate sequence of the leaf contour is obtained. The coordinate sequence is Fourier transformed, and the first K low-frequency coefficients are taken as Fourier descriptors to perform compact encoding of the leaf shape, and the invariant moment features of the leaf contour are calculated to obtain the scale-invariant shape feature vector.

[0015] Convert the RGB image of the leaf area in the growth image into the HSV color space, extract the color histogram, calculate the statistics of the color histogram, determine the color features, combine the first-order moment and the second-order moment of the leaf area, determine the color change features, divide the leaf area into a color aggregation area and a color non-aggregation area, count the corresponding pixel numbers respectively, obtain the color aggregation features, and determine the color aggregation vector in combination with the color change features;

[0016] The continuous video frames corresponding to the growth video are input into the optical flow algorithm, the pixel motion vector between adjacent video frames is calculated, the motion field of each part of the plant is determined, the average speed and acceleration of each part of the plant are calculated, the growth rate characteristics are determined, the horizontal gradient and vertical gradient of the motion field are calculated respectively, the motion boundary image is obtained, and the grid is divided. The gradient direction histogram in each grid is calculated to obtain the motion boundary histogram feature vector.

[0017] In an optional embodiment,

[0018] Based on the abnormal identification smart contract, the abnormal growth of plant seed felt is identified, including:

[0019] Receive growth features, input into the anomaly recognition smart contract, read anomaly feature data from a pre-built anomaly feature library, map the high-dimensional feature vector corresponding to the growth feature into a low-dimensional hash code based on the local sensitive hashing algorithm, search for approximate features through hash collision, and determine candidate anomaly features;

[0020] Among the candidate abnormal features, combining R-tree search, determining the abnormal feature that is closest to the growth feature in the feature space, and determining the abnormal feature comparison result;

[0021] Calculate the similarity score between the growth feature and the abnormal feature comparison results, set a multi-level similarity threshold based on the preset severity of the abnormal type, compare the similarity score with the corresponding threshold, convert the preset abnormal judgment rule into a conditional statement, match the conditional statement on the growth feature, and determine the abnormal type.

[0022] In an optional embodiment,

[0023] Based on the digital twin smart contract in the blockchain network, a three-dimensional digital twin model of the plant seed felt is generated, including:

[0024] Perform multi-site scanning on the plant seed felt to obtain point cloud data containing geometric coordinates and color information, register and splice the point cloud data, generate a rigid body transformation matrix between different sites through an ICP algorithm, unify the point cloud data corresponding to all sites into the same coordinate system, and generate a seed felt point cloud model;

[0025] Based on the seed felt point cloud model, the point cloud data is downsampled using a voxel grid downsampling algorithm, and the downsampled point cloud is denoised based on a statistical filtering algorithm to obtain preprocessed point cloud data. The preprocessed point cloud data is classified point by point using a PointNet++ semantic segmentation model, and each point is divided into different semantic categories, and a semantic label is determined to obtain a semantic segmentation result;

[0026] Based on the semantic segmentation result, the point cloud data is grouped according to the semantic labels to obtain point cloud subsets of different parts, the point cloud subsets of each part are reconstructed by Poisson surface, and a mesh model composed of triangular facets is generated. Based on the semantic information corresponding to the semantic segmentation result, semantic attributes are assigned to the triangular facets, the mesh model is smoothed and simplified, and the mesh models corresponding to different parts are assembled to obtain a three-dimensional mesh model of the plant seed felt.

[0027] The three-dimensional grid model is parameterized and mapped into a parameter domain. Combined with the growth parameter function defined in the parameter domain, the growth parameter function is sampled onto the vertices of the three-dimensional grid model. The growth parameter value at any position is calculated using an interpolation method to obtain a three-dimensional digital twin model.

[0028] In an optional embodiment,

[0029] Through the three-dimensional digital twin model, the growth state of the plant seed felt based on different growth parameter information is simulated, and the simulated growth results are generated including:

[0030] Based on predetermined plant growth prior knowledge, determining a growth model, wherein the growth model includes a growth rate function, a growth direction function, and a branching and bifurcation rule;

[0031] The growth rate function simulates the auxin concentration distribution based on the pre-constructed auxin polar transport model, maps the auxin concentration distribution to each vertex of the three-dimensional digital twin model, constructs a growth rate adjustment factor in combination with environmental factors, combines the environmental factors with the auxin concentration distribution through a function interpolation method, and dynamically adjusts the growth rate of each vertex;

[0032] The growth direction function is based on the vertices of various parts of the plant seed felt, takes the position vector of the vertex as the initial vector, combines the predetermined growth stress strain field and gravity influence factor, generates a growth direction vector through vector synthesis, combines the random disturbance factor, generates a random unit vector through random sampling, and performs weighted synthesis with the growth direction vector to determine the growth direction;

[0033] The branching and bifurcation rule simulates the distribution of plant hormone concentration and maps the plant hormone concentration distribution to each vertex of the three-dimensional digital twin model. When the hormone concentration of the vertex exceeds the preset branching and bifurcation threshold, a branching and bifurcation event is triggered through random sampling, and the growth rate function and the growth direction function are called to dynamically generate new branches.

[0034] Based on the pre-constructed nutrient transport and distribution model, the nutrient transport and distribution process in different growth stages is determined, the nutrient transport and distribution results are fed back to the growth rate function, the growth rate of different parts is dynamically adjusted, and the growth state under different nutritional conditions is determined;

[0035] Initialize the three-dimensional digital twin model according to the preset initial growth conditions, determine the growth parameters of each vertex, calculate the growth direction and growth rate of each vertex in each time step by solving the growth model and the nutrient transport and distribution model according to the preset time step, update the vertex coordinates, dynamically add new vertices in combination with the branching and bifurcation rules, and update the topological structure of the three-dimensional digital twin model;

[0036] Repeating the iteration until a preset growth time is reached, recording the state data of the three-dimensional digital twin model at each time step, and generating a simulated growth result of the plant seed felt;

[0037] Based on the random forest algorithm, the mapping relationship between the growth parameters of the seed felt and the growth state is established by training the three-dimensional digital twin model to simulate the growth results under different parameter settings.

[0038] In an optional embodiment, it also includes:

[0039] The growth rate function has the following formula:

[0040] ;

[0041] in, t Indicates time, i represents the vertex number, R i ( t ) represents the vertex i In time t The growth rate, k i Represents a vertex i The maximum growth rate, r i Represents a vertex i The growth rate coefficient, A i ( t ) represents the vertex i The auxin concentration distribution function is j Indicates the ordinal number of environmental factors, m represents the total number of environmental factors, α j Indicates the growth rate j The influence of environmental factors, E j ( t ) indicates thej The regulatory factor of an environmental factor at time t, β represents the nutrient influence coefficient, N ( t ) represents the nutrient partitioning factor;

[0042] The growth direction function has the following formula:

[0043] ;

[0044] in, D i ( t ) represents the vertex i In time t The growth direction vector, normalize(·) represents the vector unit length adjustment operation, v i Represents a vertex i The current position vector of or represents the random perturbation factor, RV () represents a random unit vector, c represents the stress vector influence factor, SV (·) represents the stress vector, l represents the gravity vector influence factor, GV (·) represents the gravity vector, d j Indicates the growth direction j The influence of environmental factors.

[0045] According to a second aspect of the embodiments of the present invention,

[0046] Provide a plant seed felt management system based on blockchain, including:

[0047] The first unit is used to obtain growth parameter information and identification information of the plant seed felt, upload the growth parameter information and the identification information to the blockchain network, generate a growth file of the plant seed felt, and bind the growth file to the electronic tag of the plant seed felt;

[0048] The second unit is used to collect and upload the growth images and growth videos of the plant seed felt in real time through the Internet of Things devices in the blockchain network, analyze the growth images and the growth videos through the visual feature extraction smart contract in the blockchain network, obtain the growth features, and identify the abnormal growth of the plant seed felt based on the abnormality identification smart contract. When the abnormal growth occurs, the information of the plant seed felt manufacturer is obtained from the growth file, and the abnormality alarm is sent to the manufacturer of the plant seed felt through the supply chain management smart contract in the blockchain network to trigger the quality traceability process of the plant seed felt; the batch number of the plant seed felt is obtained from the growth file, and based on the supply chain management smart contract, an early warning prompt is issued to other plant seed felts in the same batch;

[0049] The third unit is used to generate a three-dimensional digital twin model of the plant seed felt based on the digital twin smart contract in the blockchain network, simulate the growth state of the plant seed felt based on different growth parameter information through the three-dimensional digital twin model, generate simulated growth results, and dynamically adjust the growth parameters of the plant seed felt based on the simulated growth results.

[0050] According to a third aspect of the embodiments of the present invention,

[0051] An electronic device is provided, comprising:

[0052] processor;

[0053] a memory for storing processor-executable instructions;

[0054] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0055] According to a fourth aspect of the embodiments of the present invention,

[0056] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0057] In the embodiments of the present invention, real-time monitoring of the plant growth environment and status is achieved to improve the timeliness and accuracy of data; a unique identifier is assigned to each plant seed felt through an electronic tag to facilitate management and traceability; blockchain technology is used to ensure permanent storage and non-tamperability of data, thereby ensuring data integrity and security; a growth archive of the plant seed felt is created and maintained to record the complete growth history and achieve full traceability; the growth archive is associated and bound to the electronic tag to facilitate quick query and data access and improve management efficiency; the consensus mechanism and non-tamperability of the blockchain are used to ensure the integrity and security of uploaded image and video data; by deploying smart contracts and applying deep learning models, image and video data are preprocessed and feature extracted to achieve accurate plant growth feature analysis; based on smart contracts and predefined abnormal feature models, Automatically identify abnormal conditions in plant growth and improve the accuracy and efficiency of anomaly detection; when an abnormal condition is detected, automatically trigger the abnormal alarm mechanism, promptly notify the manufacturer and start the quality traceability process to ensure rapid response and processing; by generating a three-dimensional digital twin model of the plant seed felt, it can accurately simulate the growth state of the plant under different growth parameters and provide realistic growth simulation; based on the simulated growth results, the growth parameters of the plant seed felt can be dynamically adjusted to optimize environmental conditions such as temperature, humidity, and light, and promote the healthy growth of plants; the digital twin smart contract runs in the blockchain network, ensuring the transparency and immutability of the data, and improving the efficiency and reliability of plant seed felt growth management; by simulating and predicting the impact of different growth parameters on plant growth, potential problems can be discovered in advance to prevent the occurrence of growth abnormalities. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 This is a flow chart of a plant seed felt management method based on blockchain according to an embodiment of the present invention;

[0059] Figure 2 This is a schematic diagram of the structure of a plant seed felt management system based on blockchain according to an embodiment of the present invention. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0062] Figure 1 Schematic diagram of the process of the plant seed blanket management method based on blockchain in an embodiment of the present invention. Figure 1 As shown, the method includes:

[0063] S101. Acquire growth parameter information and identification information of the plant seed felt, upload the growth parameter information and the identification information to the blockchain network, generate a growth file of the plant seed felt, and bind the growth file to the electronic tag of the plant seed felt;

[0064] Through various sensors installed in the growth environment of the plant seed felt, the growth parameter information of the plant seed felt is collected in real time, including environmental parameters such as temperature, humidity, light intensity, carbon dioxide concentration, and plant growth status parameters such as plant height, number of leaves, leaf color, etc.; the collected growth parameter information is transmitted through wired or wireless communication modules and sent to the data processing center.

[0065] Assigning a unique identification number to each plant seed felt can be achieved using RFID tags, QR codes, or other forms of electronic tags; by scanning or reading the electronic tags, the identification information of the plant seed felt can be obtained, including the identification number, production batch, production date, etc.

[0066] The acquired plant seed felt growth parameter information and identification information are packaged according to a predefined data format to form a data packet; the data packet is uploaded to the blockchain network through the access node of the blockchain network, waiting for verification and confirmation.

[0067] The consensus nodes in the blockchain network verify the received data packets to ensure the integrity and accuracy of the data. After reaching an agreement through the consensus mechanism, the data packets are written into a new block of the blockchain to achieve permanent storage and non-tamperability of the data.

[0068] In the blockchain network, a unique growth file is created for each plant seed felt based on the identification information of the plant seed felt; the growth file contains the basic information of the plant seed felt, such as identification number, production batch, production date, etc., as well as the growth parameter information of the plant seed felt; each uploaded growth parameter information will be appended to the growth file of the corresponding plant seed felt to form a complete growth history record; the growth file is stored and managed through the blockchain network to ensure the traceability and non-tamperability of the data.

[0069] In the blockchain network, the growth file of the plant seed felt is associated and bound with its corresponding electronic tag; through the unique identification information of the electronic tag, the growth file of the corresponding plant seed felt can be quickly queried and accessed in the blockchain network; the association and binding ensures a one-to-one correspondence between the plant seed felt and its growth file, facilitating subsequent data traceability and query.

[0070] In this embodiment, real-time monitoring of the plant growth environment and status is achieved to improve the timeliness and accuracy of data; a unique identifier is assigned to each plant seed felt through an electronic tag to facilitate management and traceability; blockchain technology is used to ensure permanent storage and non-tamperability of data, thereby ensuring data integrity and security; a growth file of the plant seed felt is created and maintained to record the complete growth history and achieve full traceability; the growth file is associated and bound to the electronic tag to facilitate quick query and data access, thereby improving management efficiency; the development of precision agriculture and intelligent management is promoted to improve plant growth quality and agricultural production efficiency.

[0071] S102. Through the Internet of Things devices in the blockchain network, the growth images and growth videos of the plant seed felt are collected and uploaded in real time. Through the visual feature extraction smart contract in the blockchain network, the growth images and the growth videos are analyzed to obtain the growth features. Based on the abnormality identification smart contract, the abnormal growth of the plant seed felt is identified. When the abnormal growth occurs, the information of the plant seed felt manufacturer is obtained from the growth archive. Through the supply chain management smart contract in the blockchain network, an abnormality alarm is sent to the manufacturer of the plant seed felt to trigger the quality traceability process of the plant seed felt; the batch number of the plant seed felt is obtained from the growth archive, and based on the supply chain management smart contract, an early warning prompt is issued to other plant seed felts in the same batch;

[0072] Install high-definition cameras and other IoT devices in the growth environment of the plant seed felt to collect growth images and videos of the plant seed felt in real time; upload the collected image and video data to the blockchain network in real time through the communication module in the IoT device; the consensus nodes in the blockchain network verify the received image and video data to ensure the integrity and accuracy of the data; after reaching an agreement through the consensus mechanism, write the image and video data into a new block of the blockchain to achieve permanent storage and non-tamperability of the data.

[0073] Deploy visual feature extraction smart contracts in the blockchain network to parse growth images and growth videos; the smart contracts use computer vision algorithms to pre-process images and videos, such as image enhancement, denoising, and segmentation; apply deep learning models, such as convolutional neural networks or recurrent neural networks, to extract key features in images and videos, such as plant height, number of leaves, leaf color and other growth features; store the extracted growth features in a structured data format and associate them with the corresponding plant seed felt growth archives.

[0074] An anomaly identification smart contract is deployed in the blockchain network to identify abnormal growth of plant seed felt. The smart contract analyzes and compares the extracted growth features based on predefined abnormal feature models and thresholds. Through feature similarity calculation, pattern matching and other algorithms, the growth abnormalities of plant seed felt are identified, such as abnormal growth rate, abnormal leaf color, etc. The identified abnormalities are recorded in the form of events and associated with the corresponding plant seed felt growth files.

[0075] When the anomaly identification smart contract identifies abnormal growth of plant seed felt, the anomaly alarm mechanism is automatically triggered; the manufacturer information, such as manufacturer name, contact information, etc., is obtained from the growth file of the plant seed felt; the supply chain management smart contract is deployed in the blockchain network to manage the production and sales process of the plant seed felt; the supply chain management smart contract sends an anomaly alarm to the manufacturer of the plant seed felt, notifying it that the plant seed felt has an abnormal growth; after receiving the anomaly alarm, the manufacturer starts the quality traceability process to investigate and handle the abnormal plant seed felt.

[0076] The batch number is obtained from the growth file of the plant seed felt to identify other plant seed felts from the same batch; the supply chain management smart contract queries the growth files of other plant seed felts from the same batch in the blockchain network based on the batch number; the growth characteristics of other plant seed felts from the same batch are analyzed to assess their risk of abnormal growth; if the assessment results indicate that other plant seed felts from the same batch have a risk of abnormal growth, the supply chain management smart contract sends an early warning to the relevant parties; the early warning includes the batch number and the abnormal risk level.

[0077] In this embodiment, growth images and videos of plant seed felt are collected in real time through high-definition cameras and Internet of Things devices, and uploaded to the blockchain network in real time to ensure the timeliness of the data; the consensus mechanism and immutability of the blockchain are used to ensure the integrity and security of the uploaded image and video data; by deploying smart contracts and applying deep learning models, image and video data are preprocessed and feature extracted to achieve accurate plant growth feature analysis; based on smart contracts and predefined abnormal feature models, abnormal conditions in plant growth are automatically identified to improve the accuracy and efficiency of abnormality detection; when an abnormal condition is detected, the abnormal alarm mechanism is automatically triggered to promptly notify the manufacturer and start the quality traceability process to ensure rapid response and processing; risk assessment and analysis are performed on the same batch of plant seed felts to warn of potential risks in advance, help relevant parties take preventive measures, and reduce production and sales risks; through supply chain management smart contracts, the production and sales process of plant seed felt is optimized to improve supply chain transparency and management efficiency.

[0078] In an optional embodiment, based on a computer vision algorithm, a growth image and a growth video of a plant seed felt are analyzed to obtain growth features, wherein the growth features include a scale-invariant shape feature vector, a color aggregation vector, and a motion boundary histogram feature vector;

[0079] The growth image is input into the region-based convolutional neural network model, and the leaf region candidate frame is automatically extracted through the region proposal network. The convolutional neural network is used to classify and regress the leaves in the leaf region candidate frame to determine the leaf segmentation result, and the coordinate sequence of the leaf contour is obtained. The coordinate sequence is Fourier transformed, and the first K low-frequency coefficients are taken as Fourier descriptors to perform compact encoding of the leaf shape, and the invariant moment features of the leaf contour are calculated to obtain the scale-invariant shape feature vector.

[0080] Convert the RGB image of the leaf area in the growth image into the HSV color space, extract the color histogram, calculate the statistics of the color histogram, determine the color features, combine the first-order moment and the second-order moment of the leaf area, determine the color change features, divide the leaf area into a color aggregation area and a color non-aggregation area, count the corresponding pixel numbers respectively, obtain the color aggregation features, and determine the color aggregation vector in combination with the color change features;

[0081] The continuous video frames corresponding to the growth video are input into the optical flow algorithm, the pixel motion vector between adjacent video frames is calculated, the motion field of each part of the plant is determined, the average speed and acceleration of each part of the plant are calculated, the growth rate characteristics are determined, the horizontal gradient and vertical gradient of the motion field are calculated respectively, the motion boundary image is obtained, and the grid is divided. The gradient direction histogram in each grid is calculated to obtain the motion boundary histogram feature vector.

[0082] The growth image is input into the region-based convolutional neural network model, which consists of a region proposal network and a convolutional neural network. The region proposal network scans the growth image in a sliding window manner to generate a large number of region candidate boxes, covering the potential leaf area; the region candidate boxes are input into the convolutional neural network, and the feature representation of the leaf area is extracted through the convolution layer and the pooling layer; the output of the convolutional neural network is passed through the classifier and the position regressor to classify and refine the position of the leaves in the leaf region candidate box to determine the final leaf segmentation result; the contour of the segmented leaf area is extracted to obtain the coordinate sequence of the leaf contour to represent the shape of the leaf; the coordinate sequence of the leaf contour is Fourier transformed to convert the coordinate sequence in the spatial domain into the Fourier coefficients in the frequency domain; the first K low-frequency Fourier coefficients are taken as the Fourier descriptor to realize the compact encoding of the leaf shape and reduce the dimension of the shape feature; the invariant moment features of the leaf contour, such as contour area, perimeter, circularity, etc., are calculated to obtain the scale-invariant shape feature vector to describe the shape characteristics of the leaf.

[0083] The RGB image of the leaf area in the growth image is converted into the HSV color space, which is more consistent with the human eye's perception of color; in the HSV color space, the color histogram of the leaf area is extracted, and the distribution of the number of pixels in different color ranges is counted; the statistics of the color histogram, such as the mean, variance, skewness, etc., are calculated as color features to describe the average color and color distribution of the leaves; the first-order moment and second-order moment of the leaf area are calculated, the first-order moment represents the average value of the pixel intensity in the leaf area, and the second-order moment represents the variance of the pixel intensity, which is used to describe the change characteristics of the leaf color; the leaf area is divided into a color aggregation area and a color non-aggregation area, the color aggregation area represents a connected area with similar colors, and the color non-aggregation area represents an area with large color differences; the number of pixels in the color aggregation area and the color non-aggregation area are counted respectively to obtain the color aggregation feature to describe the aggregation degree of the leaf color; the color change feature and the color aggregation feature are combined to construct a color aggregation vector to comprehensively describe the color characteristics of the leaf.

[0084] The continuous video frames in the growth video are taken as input, and every two adjacent video frames constitute a frame pair; the optical flow algorithm is applied to each frame pair, and the motion vector of each pixel between adjacent video frames is calculated to obtain the motion field at the pixel level; according to the pixel motion vector, the motion field of each part of the plant, such as the movement of leaves, stems and other parts, is determined; the motion field of each part of the plant is statistically analyzed, and the average speed and acceleration are calculated as the growth rate characteristics to describe the growth rate and change trend of each part of the plant; the horizontal gradient and vertical gradient of the motion field are calculated respectively to obtain the motion boundary image to highlight the boundary area of ​​the plant movement; the motion boundary image is gridded to divide the image into several small grids; in each grid, the histogram of the pixel gradient direction is calculated, and the gradient intensity distribution in different directions is statistically analyzed; the gradient direction histograms of all grids are spliced ​​into a vector to obtain the motion boundary histogram feature vector to describe the characteristics of the plant motion boundary.

[0085] The steps describe in detail the implementation process of leaf segmentation and shape feature extraction, leaf color feature extraction, and plant growth rate feature extraction. These steps automatically extract key features of plant growth from growth images and growth videos by applying computer vision and deep learning algorithms, providing rich feature representation for subsequent growth anomaly detection and analysis.

[0086] The leaf segmentation and shape feature extraction steps utilize a region-based convolutional neural network model. Through the combination of region proposal and convolutional neural network, the leaf region is automatically extracted and segmented, and the compact and scale-invariant leaf shape features are obtained through Fourier descriptors and invariant moment features.

[0087] The leaf color feature extraction step converts the image into the HSV color space, extracts the color histogram and statistics, and combines the first-order moment, second-order moment and color aggregation features to comprehensively describe the color distribution and changes of the leaves.

[0088] The plant growth rate feature extraction step uses the optical flow algorithm to calculate the pixel motion vector, obtain the motion field of each part of the plant, and obtain the growth rate characteristics through statistical analysis. At the same time, by calculating the gradient direction histogram of the motion boundary image, the motion boundary histogram feature vector is obtained to describe the characteristics of the plant motion boundary.

[0089] In this embodiment, automatic leaf segmentation and recognition: through the region-based convolutional neural network model, automatic leaf region segmentation and recognition are realized, manual intervention is reduced, and the accuracy and efficiency of segmentation are improved; through Fourier transform and invariant moment feature extraction, compact and scale-invariant leaf shape features are provided. It helps to accurately describe the geometric shape of the leaf under different conditions and enhance the recognition ability of shape features; the image is converted into HSV color space, and the color histogram, statistics and color aggregation features are extracted to achieve a comprehensive analysis of the color features of the leaf. Accurately describing the color distribution and changes of the leaf helps to monitor the health status of the plant and identify potential problems; the motion vector between video frames is analyzed by the optical flow algorithm to provide detailed growth rate characteristics, including the average speed and acceleration of each part of the plant. This makes the monitoring of plant growth rate more accurate, which helps to evaluate growth dynamics and detect abnormal conditions; by calculating the gradient direction histogram of the motion field, the motion boundary histogram feature vector is obtained to highlight the boundary area of ​​plant movement. It helps to identify the dynamic changes of plants and further analyze the growth trend and movement pattern of plants; the comprehensive and detailed representation of plant growth characteristics is provided by integrating the shape, color and movement characteristics of the leaves. These features can provide strong data support for subsequent growth abnormality detection and analysis.

[0090] In an optional embodiment, based on the abnormality identification smart contract, identifying the abnormal growth of the plant seed felt includes:

[0091] Receive growth features, input into the anomaly recognition smart contract, read anomaly feature data from a pre-built anomaly feature library, map the high-dimensional feature vector corresponding to the growth feature into a low-dimensional hash code based on the local sensitive hashing algorithm, search for approximate features through hash collision, and determine candidate anomaly features;

[0092] Among the candidate abnormal features, combining R-tree search, determining the abnormal feature that is closest to the growth feature in the feature space, and determining the abnormal feature comparison result;

[0093] Calculate the similarity score between the growth feature and the abnormal feature comparison results, set a multi-level similarity threshold based on the preset severity of the abnormal type, compare the similarity score with the corresponding threshold, convert the preset abnormal judgment rule into a conditional statement, match the conditional statement on the growth feature, and determine the abnormal type.

[0094] The anomaly identification smart contract receives growth features extracted from growth images and growth videos as input data; the smart contract reads the anomaly feature data from a pre-built anomaly feature library, which contains high-dimensional feature vectors of plant growth features under known abnormal conditions; a local sensitive hashing algorithm is used to map the high-dimensional feature vectors corresponding to the growth features into low-dimensional hash codes to achieve compressed representation of high-dimensional features; through hash collision search, candidate anomaly features similar to the growth feature hash codes are found in the anomaly feature library, and these candidate anomaly features are closer to the growth features in the feature space; among the candidate anomaly features, an R-tree search algorithm is used to accelerate the calculation and comparison of feature distances by constructing an index structure in the feature space; through R-tree search, the anomaly feature that is closest to the growth feature in the feature space is determined as the anomaly feature comparison result.

[0095] Calculate the similarity score between the comparison results of the growth feature and the abnormal feature, the similarity score quantifies the degree of proximity between the two feature vectors; set multiple levels of similarity thresholds according to the preset severity of the abnormal type, and different thresholds correspond to abnormal types of different severity; compare the calculated similarity score with the corresponding threshold to determine the threshold interval to which the similarity score belongs; convert the preset abnormal judgment rule into a conditional statement, and the conditional statement determines the abnormal type according to the comparison result of the similarity score and the threshold; match the conditional statement on the growth feature, and determine the abnormal type corresponding to the growth feature according to the conditions satisfied by the similarity score; the abnormal type may include different levels and types of growth abnormalities such as leaf discoloration, leaf withering, and plant lodging.

[0096] The correspondence between growth characteristics and abnormal types is used as the abnormality identification result and output to the subsequent business process; the abnormality identification result includes the identification information of the growth characteristics, the corresponding abnormality type, and related data such as the similarity score; the smart contract writes the abnormality identification result into the blockchain to ensure the immutability and traceability of the result; the abnormality identification result can trigger the early warning mechanism to notify the relevant parties to take countermeasures, such as adjusting the planting conditions, arranging manual inspections, etc.; the abnormality identification results can also be used as the input of data analysis. Through data mining and statistical analysis, the laws and trends of abnormal occurrence can be discovered, providing a basis for optimizing the planting plan.

[0097] The steps describe in detail the implementation process of abnormal feature search and comparison, abnormal type determination and abnormal identification result output. These steps achieve efficient and accurate abnormal feature identification and type determination by applying technologies such as local sensitive hashing algorithm, R-tree search and similarity comparison.

[0098] The abnormal feature search and comparison step uses the local sensitive hashing algorithm to map the high-dimensional feature vector into a low-dimensional hash code, quickly searches for similar features through hash collision, and combines the R-tree search to determine the nearest neighbor abnormal features to achieve efficient feature comparison.

[0099] The abnormality type determination step calculates the similarity score and compares it with the preset multi-level threshold. According to the conditions satisfied by the similarity score, the abnormality type is determined using the preset judgment rules to achieve grading and classification of growth abnormalities.

[0100] The abnormal identification result output step associates the identified abnormal type with the growth characteristics to form a structured abnormal identification result, and writes the result into the blockchain to ensure the credibility and traceability of the result. The abnormal identification result can trigger early warning and optimize decision-making, providing strong support for plant cultivation management.

[0101] In this embodiment, the local sensitive hash algorithm and the R-tree search algorithm are used to realize the rapid search and comparison of high-dimensional feature vectors, thereby improving the efficiency of abnormal feature recognition. By comparing the similarity score with the preset multi-level threshold value and combining the preset judgment rules, the accurate classification and classification of growth abnormalities can be realized, and various types of abnormalities, such as leaf discoloration, leaf withering, plant lodging, etc., can be effectively identified. The abnormal identification results are written into the blockchain to ensure the immutability and traceability of the data, and improve the credibility of the abnormal identification results. The abnormal identification results can trigger the early warning mechanism to notify the relevant parties to take timely response measures, such as adjusting the planting conditions or arranging manual inspections, and also provide data support for optimizing the planting plan. The output abnormal identification results include the identification information of the growth characteristics, the corresponding abnormal type and the similarity score, etc., providing detailed data for subsequent business processes and data analysis. The abnormal identification results are used as the input of data analysis. Through data mining and statistical analysis, the laws and trends of abnormal occurrence can be discovered, providing a scientific basis for optimizing the planting plan.

[0102] S103. Based on the digital twin smart contract in the blockchain network, a three-dimensional digital twin model of the plant seed felt is generated. Through the three-dimensional digital twin model, the growth state of the plant seed felt based on different growth parameter information is simulated to generate simulated growth results. Based on the simulated growth results, the growth parameters of the plant seed felt are dynamically adjusted.

[0103] In this embodiment, by generating a three-dimensional digital twin model of the plant seed felt, the growth state of the plant under different growth parameters can be accurately simulated to provide realistic growth simulation; based on the simulated growth results, the growth parameters of the plant seed felt can be dynamically adjusted to optimize environmental conditions such as temperature, humidity, and light, and promote the healthy growth of plants; the digital twin smart contract runs in the blockchain network, ensuring the transparency and immutability of the data, and improving the efficiency and reliability of plant seed felt growth management; by simulating and predicting the impact of different growth parameters on plant growth, potential problems can be discovered in advance and the occurrence of growth abnormalities can be prevented; scientific decision-making support based on simulation data is provided to help growers formulate more effective management strategies and improve planting effects and yields.

[0104] In an optional embodiment, generating a three-dimensional digital twin model of a plant seed felt based on a digital twin smart contract in a blockchain network includes:

[0105] Perform multi-site scanning on the plant seed felt to obtain point cloud data containing geometric coordinates and color information, register and splice the point cloud data, generate a rigid body transformation matrix between different sites through an ICP algorithm, unify the point cloud data corresponding to all sites into the same coordinate system, and generate a seed felt point cloud model;

[0106] Based on the seed felt point cloud model, the point cloud data is downsampled using a voxel grid downsampling algorithm, and the downsampled point cloud is denoised based on a statistical filtering algorithm to obtain preprocessed point cloud data. The preprocessed point cloud data is classified point by point using a PointNet++ semantic segmentation model, and each point is divided into different semantic categories, and a semantic label is determined to obtain a semantic segmentation result;

[0107] Based on the semantic segmentation result, the point cloud data is grouped according to the semantic labels to obtain point cloud subsets of different parts, the point cloud subsets of each part are reconstructed by Poisson surface, and a mesh model composed of triangular facets is generated. Based on the semantic information corresponding to the semantic segmentation result, semantic attributes are assigned to the triangular facets, the mesh model is smoothed and simplified, and the mesh models corresponding to different parts are assembled to obtain a three-dimensional mesh model of the plant seed felt.

[0108] The three-dimensional grid model is parameterized and mapped into a parameter domain. Combined with the growth parameter function defined in the parameter domain, the growth parameter function is sampled onto the vertices of the three-dimensional grid model. The growth parameter value at any position is calculated using an interpolation method to obtain a three-dimensional digital twin model.

[0109] A 3D scanner is used to perform multi-site scanning of the plant seed felt, and the surface data of the plant seed felt is collected from different angles and positions. The scanning data of each site contains a large number of 3D points, and each point contains geometric coordinates (X, Y, Z) and color information (RGB value). The point cloud data collected from multiple sites are aligned and spliced, and the iterative closest point algorithm is used to calculate the rigid body transformation matrix between different sites; the rigid body transformation matrix is ​​used to uniformly convert the point cloud data of all sites to the same coordinate system to eliminate the position and direction differences between sites; the aligned and spliced ​​point cloud data are merged to generate a complete point cloud model of the plant seed felt, which represents the 3D geometric shape of the plant seed felt.

[0110] The plant seed felt point cloud model is preprocessed, and the point cloud data is downsampled using the voxel grid downsampling algorithm to reduce the density of the point cloud; the downsampled point cloud data is denoised, and the statistical filtering algorithm is used to remove noise points to improve the quality of the point cloud; the preprocessed point cloud data is input into the PointNet++ semantic segmentation model, and each point is classified point by point; the PointNet++ model divides each point into different semantic categories, such as leaves, stems, flowers, etc., by learning the local and global features of the point cloud; the model outputs the semantic label of each point, indicating which semantic category the point belongs to, and obtains the semantic segmentation result of the point cloud.

[0111] According to the semantic segmentation results, the point cloud data are grouped according to semantic labels to obtain point cloud subsets of different plant parts, such as leaf point cloud, stem point cloud, etc.; the point cloud subset of each part is reconstructed by Poisson surface to generate a mesh model composed of triangular facets to represent the three-dimensional surface of the part; the semantic information provided by the semantic segmentation results is used to assign semantic attributes to the reconstructed triangular facets, and the semantic category to which each facet belongs is marked; the reconstructed mesh model is smoothed to eliminate surface noise and unevenness and improve the visual quality of the model; the mesh model is simplified to reduce the number of triangular facets, reduce the complexity of the model, and improve rendering efficiency; the mesh models of different parts are assembled, and according to the corresponding relationship of semantic labels, the mesh models of leaves, stems and other parts are spliced ​​into a complete three-dimensional mesh model of plant seed felt.

[0112] The three-dimensional grid model of the plant seed felt is parameterized and mapped to a two-dimensional parameter domain space; growth parameter functions are defined in the parameter domain, such as the distribution functions of environmental factors such as temperature, humidity, and light; the growth parameter functions are sampled to the vertices of the three-dimensional grid model, and the growth parameter values ​​in the parameter domain are mapped to the corresponding positions of the model; interpolation methods, such as linear interpolation or high-order interpolation, are used to calculate the growth parameter values ​​at any position of the three-dimensional grid model; and a three-dimensional digital twin model of the plant seed felt is generated, which includes geometric shape, semantic information, and the distribution of growth parameters.

[0113] In this embodiment, multi-site scanning and point cloud registration technology are used to generate a complete three-dimensional geometric shape of a plant seed felt to ensure the high accuracy and integrity of the model; through downsampling and denoising processing, the quality of point cloud data is improved, noise and redundancy are reduced, and data processing efficiency is optimized; the PointNet++ model is used for semantic segmentation of point clouds to accurately identify and classify different parts of plants (such as leaves, stems, etc.) and enhance the semantic information of the model; through parameterized three-dimensional grid models and combined with growth parameter functions, the growth parameter values ​​at any position in three-dimensional space can be accurately calculated to achieve accurate simulation of plant growth status; based on the three-dimensional digital twin model, the plant growth parameters are dynamically adjusted to achieve optimization and precise control of environmental conditions, thereby improving the efficiency and effectiveness of plant planting management; through smoothing and simplification of the grid model, the model complexity is reduced, the rendering efficiency is improved, high-quality visualization effects are ensured, and real-time monitoring and analysis are facilitated; blockchain technology is used to store and manage three-dimensional digital twin model data to ensure the immutability and traceability of the data and improve data security and credibility.

[0114] In an optional embodiment, the method further comprises: the growth rate function, whose formula is as follows:

[0115] ;

[0116] in, t Indicates time, i represents the vertex number, R i ( t ) represents the vertex i In time t The growth rate, k i Represents a vertex i The maximum growth rate, r i Represents a vertex i The growth rate coefficient, A i ( t ) represents the vertex i The auxin concentration distribution function is j Indicates the ordinal number of environmental factors, m represents the total number of environmental factors, α j Indicates the growth rate j The influence of environmental factors, E j ( t ) indicates the j The regulatory factor of an environmental factor at time t, β represents the nutrient influence coefficient, N (t ) represents the nutrient partitioning factor;

[0117] The growth direction function has the following formula:

[0118] ;

[0119] in, D i ( t ) represents the vertex i In time t The growth direction vector, normalize(·) represents the vector unit length adjustment operation, v i Represents a vertex i The current position vector of or represents the random perturbation factor, RV () represents a random unit vector, c represents the stress vector influence factor, SV (·) represents the stress vector, l represents the gravity vector influence factor, GV (·) represents the gravity vector, d j Indicates the growth direction j The influence of environmental factors.

[0120] In the formula, the growth rate function comprehensively considers the influence of multiple factors such as maximum growth rate, growth rate coefficient, auxin concentration, environmental factors and nutrients; the growth direction function comprehensively considers the influence of current position, random disturbance, stress vector, gravity vector and environmental factors.

[0121] The nonlinear term in the growth rate function simulates the nonlinear process of biological growth, reflecting the rapid growth in the early stage of growth and the gradual stabilization in the later stage.

[0122] Both formulas are time functions and can dynamically simulate the growth status and change trends of plants at different time points.

[0123] Both the growth rate function and the growth direction function take into account the influence of multiple environmental factors and nutrient distribution factors, reflecting the importance of external conditions on plant growth.

[0124] The growth direction function contains random disturbance factors and stress vectors, which can more realistically simulate the growth changes of plants in the natural environment.

[0125] According to the formula, by comprehensively considering multiple factors, the growth rate and direction of plants under different conditions can be simulated more accurately, providing high-precision growth predictions; through the time-dependent growth model, growth parameters can be adjusted at different time points, growth conditions can be optimized, and planting efficiency can be improved; the formula contains the influence of multiple environmental factors and nutrients, which can reflect the growth status of plants under different environmental conditions and provide a basis for refined management; the nonlinear and random disturbance terms in the growth rate and growth direction can simulate the complex growth dynamics of plants in the natural environment, and improve the realism and applicability of the model; through accurate growth models and dynamic adjustment capabilities, planting strategies can be optimized, resource waste can be reduced, and plant health and yield can be improved; the accurate data and prediction results provided by the model can support data-driven decision-making and provide strong support for plant growth management and scientific research.

[0126] In an optional embodiment,

[0127] During the preparation stage, choose appropriate sponge materials, considering water absorption, air permeability, flexibility, etc. to ensure the best conditions for plant growth, cut the sponge blocks to the appropriate size, consider the plant growth space and the available area on the ship, mark the planting point position on the sponge according to the preset plant spacing, punch holes at each planting point, the hole diameter is slightly larger than the seed diameter, and the depth is about 2-3 times the seed diameter. Sow the seeds accurately in the holes, 1-2 seeds per hole, cover with soil (or sponge crumbs), compact, cover the sowing surface with a layer of non-woven fabric with good air permeability to prevent seeds from falling off, roll up the sponge blocks after sowing, and tie them tightly with ropes for easy transportation.

[0128] During the transportation stage, the rolled sponge blocks should be packed in boxes to avoid squeezing and deformation; pay attention to temperature and humidity control during transportation to avoid germination or mildew; try to shorten the transportation time to reduce the loss of seed vitality.

[0129] During the planting stage on the ship, a waterproof layer is laid on the planting rack and a drainage outlet is left. The rolled sponge blocks are unfolded and laid flat on the rack. The irrigation system is connected and turned on regularly by the crew to ensure water supply. The sponge absorbs water and expands fully. Fill lights are set up and the lighting duration is adjusted to make it suitable for the growth of leafy vegetables. According to the needs of the plants, nutrient solution is sprayed on the sponge regularly to balance the nutrient supply. Attention is paid to ventilation and heat dissipation, and the temperature and humidity in the cabin are controlled to be constant within the optimal range.

[0130] During the plant growth management stage, observe the growth of seedlings, replant or thin out seedlings in time, ensure plant density, check plant health, remove diseased and insect-infested seedlings in time to prevent their spread, monitor the moisture status of the sponge, adjust irrigation frequency as appropriate, prevent over-humidity or drought, monitor nutrient solution concentration, add fertilizer as appropriate based on plant growth rate, reduce irrigation before harvest, reduce stem and leaf moisture content, and extend shelf life.

[0131] Consider harvesting and replanting. When the leaves grow to the predetermined specifications, harvest them. Cut them evenly to facilitate secondary growth. After harvesting, add water and fertilizer in time and wait for new leaves to sprout. It can be harvested again in about 5-7 days. After one crop of leafy vegetables, replace the sponge matrix as appropriate depending on the aging of the sponge matrix. Clean and disinfect the frame and start the next round of sowing. The harvested vegetables are stored and refrigerated in time to extend the shelf life. Supply them to the crew in batches to ensure freshness.

[0132] Based on this embodiment, by fixing the seeds directly in the sponge, the steps of seedling and transplanting are omitted, and the planting process is simplified. The volume is small and convenient for transportation. The sponge block can be compressed for easy transportation, saving space and cost. It is easy to store and can be supplied in real time during long-distance voyages.

[0133] Figure 2 Schematic diagram of the structure of the plant seed blanket management system based on blockchain in an embodiment of the present invention. Figure 2 As shown, the system comprises:

[0134] The first unit is used to obtain growth parameter information and identification information of the plant seed felt, upload the growth parameter information and the identification information to the blockchain network, generate a growth file of the plant seed felt, and bind the growth file to the electronic tag of the plant seed felt;

[0135] The second unit is used to collect and upload the growth images and growth videos of the plant seed felt in real time through the Internet of Things devices in the blockchain network, analyze the growth images and the growth videos through the visual feature extraction smart contract in the blockchain network, obtain the growth features, and identify the abnormal growth of the plant seed felt based on the abnormality identification smart contract. When the abnormal growth occurs, the information of the plant seed felt manufacturer is obtained from the growth file, and the abnormality alarm is sent to the manufacturer of the plant seed felt through the supply chain management smart contract in the blockchain network to trigger the quality traceability process of the plant seed felt; the batch number of the plant seed felt is obtained from the growth file, and based on the supply chain management smart contract, an early warning prompt is issued to other plant seed felts in the same batch;

[0136] The third unit is used to generate a three-dimensional digital twin model of the plant seed felt based on the digital twin smart contract in the blockchain network, simulate the growth state of the plant seed felt based on different growth parameter information through the three-dimensional digital twin model, generate simulated growth results, and dynamically adjust the growth parameters of the plant seed felt based on the simulated growth results.

[0137] According to a third aspect of the embodiments of the present invention,

[0138] An electronic device is provided, comprising:

[0139] processor;

[0140] a memory for storing processor-executable instructions;

[0141] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0142] According to a fourth aspect of the embodiments of the present invention,

[0143] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0144] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0145] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A plant seed blanket management method based on blockchain, characterized in that: include: Acquire growth parameter information and identification information of the plant seed felt, upload the growth parameter information and the identification information to the blockchain network, generate a growth profile of the plant seed felt, and bind the growth profile to the electronic tag of the plant seed felt; Through the Internet of Things devices in the blockchain network, the growth images and growth videos of the plant seed felt are collected and uploaded in real time. Through the visual feature extraction smart contract in the blockchain network, the growth images and the growth videos are analyzed to obtain the growth features. Based on the abnormality identification smart contract, the abnormal growth of the plant seed felt is identified. When the abnormal growth occurs, the information of the plant seed felt manufacturer is obtained from the growth file. Through the supply chain management smart contract in the blockchain network, an abnormality alarm is sent to the manufacturer of the plant seed felt, triggering the quality traceability process of the plant seed felt. Obtaining the plant seed felt batch number from the growth archive, and providing early warning reminders for other plant seed felts from the same batch based on the supply chain management smart contract; Based on the digital twin smart contract in the blockchain network, a three-dimensional digital twin model of the plant seed felt is generated, and through the three-dimensional digital twin model, the growth state of the plant seed felt based on different growth parameter information is simulated to generate simulated growth results, and based on the simulated growth results, the growth parameters of the plant seed felt are dynamically adjusted; Based on the digital twin smart contract in the blockchain network, a three-dimensional digital twin model of the plant seed felt is generated, including: Perform multi-site scanning on the plant seed felt to obtain point cloud data containing geometric coordinates and color information, register and splice the point cloud data, generate a rigid body transformation matrix between different sites through an ICP algorithm, unify the point cloud data corresponding to all sites into the same coordinate system, and generate a seed felt point cloud model; Based on the seed felt point cloud model, the point cloud data is downsampled using a voxel grid downsampling algorithm, and the downsampled point cloud is denoised based on a statistical filtering algorithm to obtain preprocessed point cloud data. The preprocessed point cloud data is classified point by point using a PointNet++ semantic segmentation model, and each point is divided into different semantic categories, and a semantic label is determined to obtain a semantic segmentation result; Based on the semantic segmentation result, the point cloud data is grouped according to the semantic labels to obtain point cloud subsets of different parts, the point cloud subsets of each part are reconstructed by Poisson surface, and a mesh model composed of triangular facets is generated. Based on the semantic information corresponding to the semantic segmentation result, semantic attributes are assigned to the triangular facets, the mesh model is smoothed and simplified, and the mesh models corresponding to different parts are assembled to obtain a three-dimensional mesh model of the plant seed felt. The three-dimensional grid model is parameterized and mapped into a parameter domain. Combined with the growth parameter function defined in the parameter domain, the growth parameter function is sampled onto the vertices of the three-dimensional grid model. The growth parameter value at any position is calculated using an interpolation method to obtain a three-dimensional digital twin model.

2. The method according to claim 1, characterized in that By extracting the smart contract of visual features in the blockchain network, parsing the growth image and the growth video, and obtaining the growth features include: Based on computer vision algorithms, the growth images and growth videos of plant seed felts are analyzed to obtain growth features, wherein the growth features include scale-invariant shape feature vectors, color aggregation vectors, and motion boundary histogram feature vectors; The growth image is input into the region-based convolutional neural network model, and the leaf region candidate frame is automatically extracted through the region proposal network. The convolutional neural network is used to classify and regress the leaves in the leaf region candidate frame to determine the leaf segmentation result, and the coordinate sequence of the leaf contour is obtained. The coordinate sequence is Fourier transformed, and the first K low-frequency coefficients are taken as Fourier descriptors to perform compact encoding of the leaf shape, and the invariant moment features of the leaf contour are calculated to obtain the scale-invariant shape feature vector. Convert the RGB image of the leaf area in the growth image into the HSV color space, extract the color histogram, calculate the statistics of the color histogram, determine the color features, combine the first-order moment and the second-order moment of the leaf area, determine the color change features, divide the leaf area into a color aggregation area and a color non-aggregation area, count the corresponding pixel numbers respectively, obtain the color aggregation features, and determine the color aggregation vector in combination with the color change features; The continuous video frames corresponding to the growth video are input into the optical flow algorithm, the pixel motion vector between adjacent video frames is calculated, the motion field of each part of the plant is determined, the average speed and acceleration of each part of the plant are calculated, the growth rate characteristics are determined, the horizontal gradient and vertical gradient of the motion field are calculated respectively, the motion boundary image is obtained, and the grid is divided. The gradient direction histogram in each grid is calculated to obtain the motion boundary histogram feature vector.

3. The method according to claim 1, characterized in that: Based on the abnormal identification smart contract, the abnormal growth of plant seed felt is identified, including: Receive growth features, input into the anomaly recognition smart contract, read anomaly feature data from a pre-built anomaly feature library, map the high-dimensional feature vector corresponding to the growth feature into a low-dimensional hash code based on the local sensitive hashing algorithm, search for approximate features through hash collision, and determine candidate anomaly features; Among the candidate abnormal features, combining R-tree search, determining the abnormal feature that is closest to the growth feature in the feature space, and determining the abnormal feature comparison result; Calculate the similarity score between the growth feature and the abnormal feature comparison results, set a multi-level similarity threshold based on the preset severity of the abnormal type, compare the similarity score with the corresponding threshold, convert the preset abnormal judgment rule into a conditional statement, match the conditional statement on the growth feature, and determine the abnormal type.

4. The method according to claim 1, characterized in that Through the three-dimensional digital twin model, the growth state of the plant seed felt based on different growth parameter information is simulated, and the simulated growth results are generated including: Based on predetermined plant growth prior knowledge, determining a growth model, wherein the growth model includes a growth rate function, a growth direction function, and a branching and bifurcation rule; The growth rate function simulates the auxin concentration distribution based on the pre-constructed auxin polar transport model, maps the auxin concentration distribution to each vertex of the three-dimensional digital twin model, constructs a growth rate adjustment factor in combination with environmental factors, combines the environmental factors with the auxin concentration distribution through a function interpolation method, and dynamically adjusts the growth rate of each vertex; The growth direction function is based on the vertices of each part of the plant seed felt, takes the position vector of the vertex as the initial vector, combines the predetermined growth stress strain field and gravity influence factor, generates a growth direction vector through vector synthesis, combines the random disturbance factor, generates a random unit vector through random sampling, and performs weighted synthesis with the growth direction vector to determine the growth direction; The branching and bifurcation rule simulates the distribution of plant hormone concentration and maps the plant hormone concentration distribution to each vertex of the three-dimensional digital twin model. When the hormone concentration of the vertex exceeds the preset branching and bifurcation threshold, a branching and bifurcation event is triggered through random sampling, and the growth rate function and the growth direction function are called to dynamically generate new branches. Based on the pre-constructed nutrient transport and distribution model, the nutrient transport and distribution process in different growth stages is determined, the nutrient transport and distribution results are fed back to the growth rate function, the growth rate of different parts is dynamically adjusted, and the growth state under different nutritional conditions is determined; Initialize the three-dimensional digital twin model according to the preset initial growth conditions, determine the growth parameters of each vertex, calculate the growth direction and growth rate of each vertex in each time step by solving the growth model and the nutrient transport and distribution model according to the preset time step, update the vertex coordinates, dynamically add new vertices in combination with the branching and bifurcation rules, and update the topological structure of the three-dimensional digital twin model; Repeating the iteration until a preset growth time is reached, recording the state data of the three-dimensional digital twin model at each time step, and generating a simulated growth result of the plant seed felt; Based on the random forest algorithm, the mapping relationship between the growth parameters of the seed felt and the growth state is established by training the three-dimensional digital twin model to simulate the growth results under different parameter settings.

5. The method according to claim 4, characterized in that Also includes: The growth rate function has the following formula: ； in, t Indicates time, i represents the vertex number, R i ( t ) represents the vertex i In time t The growth rate, k i Represents a vertex i The maximum growth rate, r i Represents a vertex i The growth rate coefficient, A i ( t ) represents the vertex i The auxin concentration distribution function is j Represents the ordinal number of environmental factors, m represents the total number of environmental factors, α j Indicates the growth rate j The influence of environmental factors, E j ( t ) indicates the j The regulatory factor of an environmental factor at time t, β represents the nutrient influence coefficient, N ( t ) represents the nutrient partitioning factor; The growth direction function has the following formula: ； in, D i ( t ) represents the vertex i In time t The growth direction vector, normalize(·) represents the vector unit length adjustment operation, v i Represents a vertex i The current position vector of η represents the random perturbation factor, RV () represents a random unit vector, γ represents the stress vector influence factor, SV (·) represents the stress vector, λ represents the gravity vector influence factor, GV (·) represents the gravity vector, δ j Indicates the growth direction j The influence of environmental factors.

6. A plant seed blanket management system based on blockchain, used to implement the method described in any one of claims 1 to 5, characterized in that: include: The first unit is used to obtain growth parameter information and identification information of the plant seed felt, upload the growth parameter information and the identification information to the blockchain network, generate a growth file of the plant seed felt, and bind the growth file to the electronic tag of the plant seed felt; The second unit is used to collect and upload the growth image and growth video of the plant seed felt in real time through the Internet of Things device in the blockchain network, analyze the growth image and the growth video through the visual feature extraction smart contract in the blockchain network, obtain the growth characteristics, and identify the growth abnormality of the plant seed felt based on the abnormality identification smart contract. When the growth abnormality occurs, the information of the plant seed felt manufacturer is obtained from the growth file, and an abnormality alarm is sent to the manufacturer of the plant seed felt through the supply chain management smart contract in the blockchain network, triggering the quality traceability process of the plant seed felt; Obtaining the plant seed felt batch number from the growth archive, and providing early warning reminders for other plant seed felts from the same batch based on the supply chain management smart contract; The third unit is used to generate a three-dimensional digital twin model of the plant seed felt based on the digital twin smart contract in the blockchain network, simulate the growth state of the plant seed felt based on different growth parameter information through the three-dimensional digital twin model, generate simulated growth results, and dynamically adjust the growth parameters of the plant seed felt based on the simulated growth results.

7. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method described in any one of claims 1 to 5.

8. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images