Grape tree body growth vigor estimation method, system and equipment based on deep learning and medium

Abstract

The invention discloses a grape tree body growth vigor estimation method, system and device based on deep learning and a medium, and relates to the technical field of intelligent agriculture. The method comprises the following steps: acquiring a grapevine body picture; preprocessing the grapevine body image, inputting the preprocessed image into a feature extraction model for identification, and determining grapevine body structural features; the feature extraction model is constructed based on an HRNet architecture and a spatial linear self-attention mechanism, comprises a plurality of stages for extracting features with different resolutions, and is connected with a multi-scale feature fusion module; drawing a node and branch association diagram according to the grapevine body structure characteristics; and reconstructing a three-dimensional structure model of the grapevine by using a graph theory algorithm and the node and branch association graph, displaying the spatial layout of the grapevine, and completing the growth vigor estimation of the grapevine body. According to the method, the labor intensity of workers can be reduced through automatic tree body growth vigor identification.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent agriculture, and particularly to a method, system, device and medium for estimating the growth trend of grapevine based on deep learning. Background Art

[0002] In modern agricultural production, the accurate estimation of the growth trend of grapevines is of great significance for improving the management efficiency of vineyards, optimizing pruning strategies, enhancing fruit quality, and realizing precision agriculture. Traditional methods for estimating the growth trend of grapevines mainly rely on manual measurement and empirical judgment, which have problems such as low efficiency, strong subjectivity, and insufficient accuracy. With the development of computer vision and deep learning technologies, more and more research has begun to explore automated methods for plant structure estimation.

[0003] Currently, there have been studies on estimating the growth trend of grapevines through image processing and machine learning technologies, but these methods usually require a large amount of manually labeled data and have limited ability to identify the growth trend of grapevines in complex scenarios. In addition, existing methods are difficult to meet the requirements of actual production due to limitations in computing resources and real-time performance when dealing with large-scale vineyards. Summary of the Invention

[0004] The purpose of the present invention is to provide a method, system, device and medium for estimating the growth trend of grapevine based on deep learning, which can reduce the labor intensity through automated identification of the growth trend of grapevines.

[0005] To achieve the above purpose, the present invention provides the following solutions:

[0006] A method for estimating the growth trend of grapevine based on deep learning, comprising:

[0007] Obtaining an image of a grapevine;

[0008] Preprocessing the grapevine image and inputting the preprocessed image into a feature extraction model for recognition to determine the structural features of the grapevine; the feature extraction model is constructed based on the HRNet architecture and a spatial linear self-attention mechanism, includes multiple stages of feature extraction with different resolutions, and is connected with a multi-scale feature fusion module;

[0009] Drawing a node and branch association graph according to the structural features of the grapevine;

[0010] Using graph theory algorithms and the node and branch association graph to reconstruct a three-dimensional structure model of the grapevine to display the spatial layout of the grapevine and complete the estimation of the growth trend of the grapevine.

[0011] Optionally, the preprocessing of the grapevine image specifically includes:

[0012] Crop the grapevine image and use Otsu's binarization to remove the irrelevant background, retaining the key structural parts of the grapevine; the key structural parts include branches and nodes;

[0013] Adjust the brightness and contrast of the image after background removal, and use a filter to remove image noise, obtaining the processed image.

[0014] Optionally, input the preprocessed image into a feature extraction model for recognition to determine the grapevine structural features, specifically including:

[0015] In each resolution branch, extract image features through a convolutional layer;

[0016] Use upsampling and downsampling operations to align the image features of different resolutions, and perform weighted summation on the aligned image features to obtain the fused key feature points; the key feature points include the position coordinates of the nodes and the direction vectors of the branches;

[0017] Determine the grapevine structural features based on the key feature points.

[0018] Optionally, draw a node and branch association graph according to the grapevine structural features, specifically including:

[0019] Perform initial pairing according to the grapevine structural features to determine the initial association graph;

[0020] Use the shortest path algorithm to calculate the shortest path between any two nodes in the initial association graph, and remove the redundant edges according to the obtained shortest path to obtain the final node and branch association graph.

[0021] Optionally, perform initial pairing according to the grapevine structural features to determine the initial association graph, specifically including:

[0022] First, use the SIFT feature matching algorithm to match the grapevine structural features, and then use the K-Means clustering algorithm to group the feature points to determine the clustering result; the clustering result includes the pairs of nodes and branches that are mutually associated in the image;

[0023] According to the clustering result, connect the nodes within each cluster to form the initial association graph; the nodes in the initial association graph represent the structural nodes of the grapevine, and the edges represent the connection relationships between the nodes.

[0024] The present invention also provides a grapevine growth trend estimation system based on deep learning, including:

[0025] An image acquisition unit for acquiring grapevine pictures;

[0026] A key point recognition unit is used to preprocess the grapevine image and input the preprocessed image into a feature extraction model for recognition to determine the structural features of the grapevine; the feature extraction model is constructed based on the HRNet architecture and the spatial linear self-attention mechanism, includes multiple stages of feature extraction with different resolutions, and is connected with a multi-scale feature fusion module;

[0027] A drawing unit is used to draw a node and branch association graph according to the structural features of the grapevine;

[0028] A growth trend estimation unit is used to reconstruct a three-dimensional structural model of the grapevine by using graph theory algorithms and the node and branch association graph, display the spatial layout of the grapevine, and complete the growth trend estimation of the grapevine.

[0029] The present invention also provides an electronic device, including a memory and a processor, the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the above-mentioned method for estimating the growth trend of a grapevine based on deep learning.

[0030] The present invention also provides a computer-readable storage medium, which is characterized in that it stores a computer program, and when the computer program is executed by a processor, it implements the above-mentioned method for estimating the growth trend of a grapevine based on deep learning.

[0031] According to the specific embodiments provided by the present invention, the following technical effects are disclosed:

[0032] The present invention discloses a method, system, device and medium for estimating the growth trend of a grapevine based on deep learning. The method includes obtaining a grapevine image; preprocessing the grapevine image and inputting the preprocessed image into a feature extraction model for recognition to determine the structural features of the grapevine; the feature extraction model is constructed based on the HRNet architecture and the spatial linear self-attention mechanism, includes multiple stages of feature extraction with different resolutions, and is connected with a multi-scale feature fusion module; drawing a node and branch association graph according to the structural features of the grapevine; reconstructing a three-dimensional structural model of the grapevine by using graph theory algorithms and the node and branch association graph, displaying the spatial layout of the grapevine, and completing the growth trend estimation of the grapevine. The present invention can reduce the manual labor intensity through automatic recognition of the tree growth trend. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0034] Figure 1 This is a schematic flowchart of the grapevine growth estimation method based on deep learning of the present invention;

[0035] Figure 2 This is the framework diagram of the HRNet model in this embodiment;

[0036] Figure 3 This is the framework diagram of the feature extraction model in this embodiment. Detailed implementation manners

[0037] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0038] The purpose of the present invention is to provide a grapevine growth estimation method, system, device and medium based on deep learning, which can reduce the manual labor intensity through automatic tree growth recognition.

[0039] To make the above objects, features and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0040] As Figure 1 shown, the present invention provides a grapevine growth estimation method based on deep learning, including:

[0041] Step 100: Obtain grapevine pictures;

[0042] Step 200: Preprocess the grapevine image, and input the preprocessed image into a feature extraction model for recognition to determine the grapevine structure features; the feature extraction model is constructed based on the HRNet architecture and the spatial linear self-attention mechanism, includes multiple stages of feature extraction with different resolutions, and is connected with a multi-scale feature fusion module;

[0043] Step 300: Draw a node and branch association diagram according to the grapevine structure features;

[0044] Step 400: Use graph theory algorithms and the node and branch association diagram to reconstruct the three-dimensional structure model of the grapevine, display the spatial layout of the grapevine, and complete the growth estimation of the grapevine.

[0045] As a specific implementation manner, the key processing flow includes:

[0046] 1. Automated Image Processing:

[0047] Image Acquisition: During the grape growth season, representative grapevines are selected as the analysis objects. High-resolution cameras are used to capture images of the grapevines at different angles and times, ensuring that the images contain sufficient details for subsequent processing and analysis.

[0048] Image Preprocessing: The images are cropped, and Otsu's binarization is used to remove the irrelevant background, retaining the key structural parts of the grapevines, such as branches and nodes. The brightness and contrast of the images are adjusted to enhance the visibility of the grapevine structural features. Filters are applied to remove image noise and improve the image quality, providing a clear input for the feature extraction stage.

[0049] 2. Deep Learning Model Architecture:

[0050] Model Selection: The present invention innovatively adopts HRNet (High-Resolution Network) as the basic architecture. With its parallel high-resolution and low-resolution branches, HRNet can accurately capture both the details and context information of the images simultaneously, showing excellent performance in image recognition and segmentation tasks, especially suitable for the complex task of estimating the growth trend of grapevines. It can effectively address the challenges brought by the diverse grapevine morphologies and complex growth environments, providing strong support for accurately extracting the grapevine structure information. This basic architecture is as Figure 2 shown.

[0051] Spatial Linear Self-Attention Mechanism: Based on the HRNet architecture, the present invention ingeniously introduces a spatial linear self-attention mechanism to further improve the model's performance. This mechanism calculates the attention weights of each pixel point with its surrounding pixel points in a refined manner, dynamically adjusting the feature representation of each pixel point in the feature map, thereby highlighting the pixel points of important structural features in the grapevine images and suppressing unimportant pixel points, significantly improving the accuracy of the model's estimation of the tree growth trend and enabling the model to more sensitively capture the subtle changes in the grapevine structure.

[0052] Multi-Scale Feature Fusion Optimization: At each stage of the model, a carefully designed multi-scale feature fusion module deeply fuses features of different resolutions. In each resolution branch, feature maps are accurately extracted through convolutional layers. Subsequently, upsampling and downsampling operations are skillfully used to accurately align the feature maps of different resolutions, and then they are fused through weighted summation. The fused feature maps are enhanced by the spatial linear self-attention mechanism to further improve the feature expression ability. Through rigorous experimental verification, the weight parameters in the multi-scale feature fusion module are finely adjusted to continuously optimize the feature fusion effect, enabling the model to more comprehensively and accurately understand and analyze the grapevine structure information. The overall framework of the model is as Figure 3 shown.

[0053] 3. Construction of Node and Branch Association Diagram:

[0054] Feature Point Extraction and Matching: The deep learning model accurately extracts the key feature points of each node and branch in the grapevine image, such as the position coordinates of the nodes, the direction vectors of the branches, etc. Then, the efficient SIFT feature matching algorithm is used to accurately match the feature points, and the K-Means clustering algorithm is used to scientifically group the feature points to accurately find the associated node and branch pairs in the image. According to the clustering results, the nodes within each cluster are cleverly connected to form an initial association diagram. The nodes in the diagram accurately represent the structural nodes of the grapevine, and the edges clearly represent the connection relationships between the nodes. Finally, the shortest path algorithm is used to accurately calculate the shortest path between any two nodes in the diagram, effectively removing redundant edges, making the association diagram more concise and accurate in reflecting the actual structure of the grapevine.

[0055] Based on the above technical solutions, the following embodiments are provided.

[0056] Image Acquisition: Select representative grapevines as the analysis objects to ensure that the samples cover different growth stages and conditions. Use a high-resolution camera to take images of the grapevines at different angles and times, including the canopy and trunk structures. Ensure that the images contain sufficient details for subsequent processing and analysis, and pay special attention to the consistency of lighting conditions and backgrounds to reduce the complexity of preprocessing.

[0057] Feature Extraction: Input the preprocessed image into the deep learning model, which has been trained to recognize the structural features of the grapevine. The model automatically recognizes the structural features of the grapevine in the image, including the trunk, branches, nodes, etc., and marks the positions of these features. The extracted features will be used for subsequent structural analysis and reconstruction, including the connection relationships of the nodes and the directions of the branches.

[0058] Structural Analysis: Utilize the extracted features to analyze the structure of the grapevine, including the distribution of branches, the connection relationships of nodes, and the distribution of leaves. Determine the relative positions of each node and branch, as well as their functions and roles in the grapevine structure, providing a basis for pruning. The analysis results will be used to construct the structural model of the grapevine, including the stretching directions and mutual relationships of the branches.

[0059] Structural Reconstruction: According to the results of the structural analysis, use graph theory algorithms to reconstruct the three-dimensional structural model of the grapevine to display the spatial layout of the grapevine. The model can display the three-dimensional shape of the grapevine, including the stretching directions and mutual relationships of the branches, providing an intuitive view for pruning and growth analysis. The reconstructed structure can be used to simulate pruning effects, predict the growth trends of grapevines, and evaluate the effectiveness of pruning strategies.

[0060] Result Output: The reconstructed grapevine structure is output in a visual form, such as a 3D model or a structured data report, for use by vineyard managers and researchers. The output results can be used to guide actual grapevine pruning work or as a reference for grapevine growth research, providing scientific data support.

[0061] System Optimization: According to the actual application effects of the output results, the parameters of the deep learning model are adjusted and optimized to improve the accuracy of recognition and analysis. User feedback is collected, and the algorithms for image preprocessing, feature extraction, and structure reconstruction are continuously improved to adapt to different grapevine varieties and growth conditions. The model is updated regularly to adapt to changes in the grapevine growth environment and new data analysis requirements, ensuring the continuous effectiveness and accuracy of the system.

[0062] Therefore, this solution has the following beneficial effects:

[0063] Improve Automation Level: Through automated image processing and structure analysis, the present invention can significantly improve the automation level of vineyard management, reduce manual intervention, lower labor costs, and improve operation efficiency at the same time.

[0064] Enhance Pruning Quality: The precise structure information provided by the present invention helps to formulate more effective pruning strategies, improve the pruning quality of grapevines, thereby enhancing grape yield and quality, and increasing the economic value of vineyards.

[0065] Scientific Data Support: The present invention provides precise data support for the growth and development research of grapevines, helping scientific researchers better understand the growth characteristics of grapevines and providing a new perspective for agricultural scientific research.

[0066] In this specification, each embodiment is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other.

[0067] In this article, specific examples are used to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the core idea of the present invention; at the same time, for those of ordinary skill in the art, based on the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A grapevine growth estimation method based on deep learning, characterized in that: include: Get the grapevine body picture; Preprocessing the grape vine body image, and inputting the preprocessed image into a feature extraction model for recognition to determine the grape vine body structural features; The feature extraction model is built based on the HRNet architecture and the spatial linear self-attention mechanism, including multiple stages of feature extraction at different resolutions, and connected with a multi-scale feature fusion module; Draw a node and branch association diagram according to the grapevine body structural characteristics; The three-dimensional structural model of the grapevine is reconstructed using the graph theory algorithm and the node and branch association graph, the spatial layout of the grapevine is displayed, and the growth potential of the grapevine is estimated.

2. The grapevine growth estimation method based on deep learning according to claim 1, characterized in that: The preprocessing of the grape tree body image specifically includes: The grapevine body image is cropped, and irrelevant background is removed by using Otsu's binarization, so as to retain key structural parts of the grapevine; the key structural parts include branches and nodes; Adjust the brightness and contrast of the image after background removal, and use filters to remove image noise to obtain the processed image.

3. The grapevine growth estimation method based on deep learning according to claim 1, characterized in that: The pre-processed image is input into the feature extraction model for identification to determine the grapevine structure features, specifically including: In each resolution branch, image features are extracted through convolutional layers; Using upsampling and downsampling operations to align image features of different resolutions, and weighted summing the aligned image features to obtain fused key feature points; the key feature points include the position coordinates of the nodes and the direction vectors of the branches; The structural characteristics of the grapevine are determined according to the key characteristic points.

4. The grapevine growth estimation method based on deep learning according to claim 1, characterized in that: Drawing a node and branch association diagram according to the grape tree structure characteristics specifically includes: Performing initial pairing according to the grapevine body structural characteristics to determine an initial association graph; The shortest path algorithm is used to calculate the shortest path between any two nodes in the initial association graph, and redundant edges are removed according to the obtained shortest path to obtain a final node and branch association graph.

5. The grapevine growth estimation method based on deep learning according to claim 4, characterized in that: The initial pairing is performed according to the grapevine body structure characteristics to determine the initial association graph, specifically including: Firstly, the structural features of the grapevine are matched by using SIFT feature matching algorithm, and then the feature points are grouped by using K-Means clustering algorithm to determine the clustering result; the clustering result includes the node and branch pairs that are related to each other in the image; According to the clustering result, the nodes in each cluster are connected to form an initial association graph; the nodes in the initial association graph represent the structural nodes of the grapevine, and the edges represent the connection relationship between the nodes.

6. A grapevine growth estimation system based on deep learning, characterized in that: include: An image acquisition unit, used for acquiring images of grape vine bodies; A key point recognition unit is used to pre-process the grape vine body image and input the pre-processed image into a feature extraction model for recognition to determine the grape vine body structure features; The feature extraction model is built based on the HRNet architecture and the spatial linear self-attention mechanism, including multiple stages of feature extraction at different resolutions, and is connected to a multi-scale feature fusion module; A drawing unit, used for drawing a node and branch association diagram according to the grape tree body structural characteristics; The growth potential estimation unit is used to reconstruct the three-dimensional structural model of the grapevine using the graph theory algorithm and the node and branch association graph, display the spatial layout of the grapevine, and complete the growth potential estimation of the grapevine body.

7. An electronic device, characterized in that: The electronic device comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the grapevine growth estimation method based on deep learning according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: It stores a computer program, which, when executed by a processor, implements the grapevine growth estimation method based on deep learning as described in any one of claims 1 to 5.

Images