Crop plant three-dimensional dynamic model construction method and device based on digital twinning

Through digital twin technology and deep learning methods, a three-dimensional dynamic model of crop plants is constructed, solving the problems of time-consuming and low sensibility in the existing technology, and a fast and efficient dynamic visualization of crop plants model is achieved.

CN118967971APending Publication Date: 2024-11-15HUNAN CREATOR INFORMATION TECH CO LTD
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202410973729.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-19
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

In the prior art, artificially constructing a dynamic plant model takes a long time and has poor generalization ability, low simulation degree, making it difficult to quickly construct a dynamic visual crop plant model.

Method used

Using a digital twin method, a plant growth model database is constructed, the organ depth information is predicted through a single-view depth estimator, and the unit spherical projection is generated using an adversarial generation network, and the model is scaled and spliced ​​in combination with OpenGL to realize the plant Rapid construction of three-dimensional dynamic models.

Benefits of technology

The three-dimensional dynamic model construction of plants with short time and high realistic performance is realized, which improves the geometric accuracy and generalization ability of the model, and can dynamically display model changes in different growth environments and growth stages.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118967971A_ABST
    Figure CN118967971A_ABST
Patent Text Reader

Abstract

The invention discloses a crop plant three-dimensional dynamic model construction method and device based on digital twinning, and the method comprises the steps: constructing a plant growth model database, and storing the size and relative position relation of each organ of a plant in different growth environments and different growth stages; obtaining the image of each organ of the plant, and generating a depth image of each organ of the plant through a single-view depth estimator; converting the depth image of each organ of the plant into a three-dimensional point cloud data set, extracting three-dimensional surface information of the organ from the three-dimensional point cloud data set, and then converting the three-dimensional surface information into unit spherical projection of the organ; using a generator to obtain a unit fine three-dimensional model of each organ according to the unit spherical projection; setting different growth environments and growth stages, extracting the size and relative position relationship of each organ of the plant from the plant growth model database, and performing scaling and splicing by taking the unit fine three-dimensional model of each organ as a reference to form an overall three-dimensional model of the plant. According to the method, the time consumed for building the whole plant model is short, and the obtained plant has good simulation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of computer image data generation, and specifically relates to a method and device for constructing a three-dimensional dynamic model of crop plants based on digital twins. Background Art

[0002] With the development of botanical research and planting industry, the demand for rapid construction and dynamic visualization of crop plant models is increasing. Such models can not only help scientists understand the growth process of plants, but also conduct real-time monitoring and management in the field of crop planting. In the process of dynamic visualization, the establishment of crop models is very important.

[0003] Traditional methods of 3D reconstruction of plants usually rely on manual drawing. Due to the wide variety of crops, the models under different growth periods are obviously different, which requires a lot of time and manpower to model. The method of using photography technology to shoot plants from multiple angles and using structured light, stereo matching or deep learning to restore the 3D structure of plants requires a large amount of image data and complex algorithms. In addition, due to the complexity of individual plant models, the generated 3D models are usually far from the real model and have weak generalization ability. Summary of the invention

[0004] The existing technology of manually constructing plant dynamic models is time-consuming. Figure 3 The reconstruction of three-dimensional models has technical problems such as poor generalization ability and low simulation degree. The present invention provides a method and equipment for constructing a three-dimensional dynamic model of crop plants based on digital twins, which is time-saving and the resulting plants have good simulation degree.

[0005] In order to achieve the above technical objectives, the present invention adopts the following technical solutions:

[0006] A method for constructing a three-dimensional dynamic model of a crop plant based on digital twins, comprising:

[0007] Constructing a plant growth model database, which stores the size of each organ of the plant in different growth environments and at different growth stages, as well as the relative position relationship between different organs;

[0008] Acquire an image of each organ of the plant, predict the depth information of each organ through a single-view depth estimator, and generate a depth image of each organ of the plant;

[0009] The depth image of each organ of the plant is converted into a three-dimensional point cloud dataset, and then the three-dimensional surface information of the organ is extracted from it, and the three-dimensional surface information of the organ is converted into a unit sphere projection of the organ;

[0010] The generator trained by the adversarial generative network is used to output the unit fine 3D models of various organs of the plant according to the unit sphere projection of various organs of the plant.

[0011] By setting different growth environments and growth stages, the corresponding size of each organ of the plant and the relative position relationship between different organs are extracted from the plant growth model database, and the unit fine three-dimensional models of various organs are used as the benchmark for scaling to obtain a fine three-dimensional model of each organ of the plant; then the fine three-dimensional models of all the organs of the plant are spliced ​​to form an overall three-dimensional model of the plant.

[0012] Furthermore, the growth environment includes meteorological environment and soil environment. The meteorological environment includes longitude and latitude, temperature, rainfall, and sunshine duration; the soil environment includes pH value, texture, and organic matter content.

[0013] Furthermore, the size of each organ of the plant and the relative position relationship between different organs stored in the plant growth model database are obtained through historical measurements of plants, including leaf height, leaf length, maximum leaf width, stem node length, stem-leaf angle, stem curl diameter, and plant height.

[0014] Furthermore, the single-view depth estimator uses a pre-trained ResNet-18 network as the basic network structure, and modifies the output layer to output only a single-channel depth image; on the basis of the Shapenet dataset, plant organ images and corresponding depth images of multiple plant categories are added, and the images and corresponding depth images in the dataset are used as input and output respectively, and the ResNet-18 network with the modified output layer is trained to obtain a single-view depth estimator for generating depth images of various plant organs.

[0015] Furthermore, the depth images of various organs of the plant are converted into a three-dimensional point cloud data set, specifically including: according to the camera parameters, the depth image is first converted into a camera coordinate system, and then converted into a world coordinate system to generate point cloud data:

[0016]

[0017] Among them, (u,v) represents the pixel position coordinates in the depth image, (X,Y,Z) and (X w ,Y w ,Z w ) represent the coordinates of the pixel position (u, v) in the depth image in the camera coordinate system and the point cloud coordinates in the world coordinate system; d(u, v) represents the depth value at the pixel position (u, v) in the depth image, f is the focal length of the camera, (c x ,c y ) is the camera principal point; R is the camera's rotation matrix, and T is the camera's position matrix.

[0018] Furthermore, the three-dimensional surface information of the organ is extracted from the three-dimensional point cloud data set, specifically including:

[0019] Divide the three-dimensional point cloud data set into a number of cubic units, each of which is called a voxel, and each voxel consists of 8 vertices; obtain the isosurface value c for distinguishing organs;

[0020] For each voxel, compare the depth of each vertex in the depth image with the isosurface value c to determine whether each vertex is on the isosurface for distinguishing organs, or whether different vertices of the voxel are on both sides of the isosurface;

[0021] If a voxel has a vertex on the isosurface or its different vertices are distributed on both sides of the isosurface, the voxel is still judged to intersect with the isosurface distinguishing the organs;

[0022] The intersection points of the isosurface and the edges of each voxel are calculated by linear interpolation. The connection order of all intersection points is obtained by querying the triangle correspondence table of the Marching Cubes algorithm, and the triangular mesh of the isosurface is generated, which is the extracted three-dimensional surface information of the organ.

[0023] Furthermore, the three-dimensional surface information of the organ is specifically a triangular mesh of an isosurface, and the three-dimensional surface information of the organ is converted into a unit sphere projection of the organ, that is, each point in the triangular mesh of the organ is represented by coordinates in a spherical coordinate system.

[0024] Furthermore, the generative adversarial network in which the generator is located has a discriminator composed of 5 convolutional layers with ReLU activation functions, and the generative adversarial network is trained using the Shapenet dataset, where the loss function of the training is:

[0025]

[0026] Among them, S1 represents the predicted output of the discriminator based on the unit sphere projection of the input organ, S2 represents the real unit fine three-dimensional model of the organ; x and y are points in S1 and S2 respectively; CD(S1,S2) represents the average distance between S1 and S2.

[0027] Furthermore, the plant organ model is scaled and spliced ​​in the 3D graphics library OpenGL, and the plant organ models are color rendered, texture mapped and illuminated in the 3D graphics library OpenGL.

[0028] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor implements any one of the above methods.

[0029] Beneficial Effects

[0030] In view of the relatively complex characteristics of the overall image of crop plants, the present invention first reconstructs the organ model of the relatively simple plant single organ image in three dimensions, and uses the linearization characteristics of plant growth to construct the overall plant model by combining models according to growth data, and sets different growth environments and different growth stages to dynamically display the changes in the model effect of the plant at each growth stage. In the process of reconstructing the three-dimensional model of each organ, a depth estimator is used to extract the depth image from the plant single organ image, the Marching Cubes algorithm is used to obtain the isosurface of the point cloud data and convert it into a spherical projection model, and a three-dimensional generative adversarial network is introduced to judge the authenticity of the three-dimensional model, so as to establish a more realistic three-dimensional single organ model.

[0031] Compared with the higher complexity of the overall crop plant model, the three-dimensional reconstruction of the single organ view of the plant can ensure the geometric accuracy of the model. The present invention extracts the depth image through a depth estimator based on the ResNet-18 network, and adds spherical projection and model discrimination to further ensure the details and generalization ability of the reconstructed three-dimensional model, so that the training results will not be too dependent on the data set, solving the problem of overfitting.

[0032] In the dynamic visualization process, the embodiment of the present invention uses OpenGL to perform color rendering, texture mapping and lighting processing on the organ models of crop plants, and restores the information such as the organ position, rotation angle, scaling ratio, quantity, etc. of the plant in different growth cycles through the plant growth model database. Since the changes of various organs of crop plants are linear and measurable during their growth, the plant model constructed by combining organs can have better simulation and realize the rapid construction of three-dimensional dynamic visualization of crop plants. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is a flow chart of the method described in the embodiment of the present application;

[0034] Figure 2 Schematic diagram of input and output of training adversarial generative network in the method described in the embodiment of the present application;

[0035] Figure 3 This is a diagram showing the effect of plants generated by the method described in the embodiment of the present application;

[0036] Figure 4 The embodiment of the present application generates models of different growth stages based on plant growth data, wherein (a)(b)(c)(d) correspond to different growth stages respectively. DETAILED DESCRIPTION

[0037] The following is a detailed description of an embodiment of the present invention. This embodiment is based on the technical solution of the present invention, and provides a detailed implementation method and a specific operation process to further explain the technical solution of the present invention.

[0038] This embodiment provides a method for constructing a three-dimensional dynamic model of crop plants based on digital twins. Figure 1 As shown, the following steps are included:

[0039] Step 1: construct a plant growth model database, which stores the size of each organ of the plant in different growth environments and different growth stages, as well as the relative position relationship between different organs.

[0040] The growth environment includes meteorological environment and soil environment. The meteorological environment includes longitude and latitude, temperature (including daily maximum temperature and daily minimum temperature), rainfall, sunshine duration, etc.; the soil environment includes pH value, texture, organic matter content, etc.

[0041] In different historical growth environments and different growth stages, the sizes and relative position relationships of various plant organs are obtained through actual measurements for storage, including leaf height, leaf length, maximum leaf width, internode length, stem-leaf angle, stem roll diameter, plant height and other morphological data, which are then stored in the plant growth model database.

[0042] Step 2: Obtain an image of each organ of the plant, predict the depth information of each organ through a single-view depth estimator, and generate a depth image of each organ of the plant.

[0043] The single-view depth estimator uses a pre-trained ResNet-18 network as the basic network structure and modifies the output layer to output only a single channel depth image.

[0044] Based on the Shapenet dataset (including 55 common object categories and corresponding 51,300 3D models), plant organ images and corresponding depth images of multiple plant categories are added. The images and corresponding depth images in the dataset are used as input and output respectively, and the ResNet-18 network with the output layer modified is trained to obtain a single-view depth estimator for generating depth images of various plant organs.

[0045] In this embodiment, the input and input images for training and reasoning of the ResNet-18 network after the output layer is modified use normalized images. That is, the pixel values ​​of each channel of the plant organ image are divided by 255, and the pixel values ​​are scaled to the range of [0,1] before being input into the ResNet-18 network; the depth image output by the ResNet-18 network is normalized according to the following formula:

[0046]

[0047] Where, Depth v is the original depth value output by the ResNet-18 network, Depth n is the normalized depth value, min_depth and max_depth are the minimum and maximum depth values ​​in the depth image respectively.

[0048] The above training is for ResNet-18 network, using the two-dimensional intersection over union (IoU) and mean square error (MES) of the real image and the output image as network evaluation indicators:

[0049]

[0050] Where: A represents the predicted output depth image, and B represents the real depth image. The higher the IoU value, the better the prediction effect, and the smaller the MES value, the better the effect.

[0051] Use the trained ResNet-18 network model to infer the input single organ image of the crop plant and generate a depth image of each organ of the plant.

[0052] Step 3: Convert the depth image of each organ of the plant into a three-dimensional point cloud dataset, extract the three-dimensional surface information of the organ from it, and convert the three-dimensional surface information of the organ into a unit sphere projection of the organ.

[0053] Step 3.1, convert the depth image of each organ of the plant into a three-dimensional point cloud data set, specifically including: according to the camera parameters, convert the depth image to the camera coordinate system first, and then convert it to the world coordinate system to generate point cloud data:

[0054]

[0055] Among them, (u,v) represents the pixel position coordinates in the depth image, (X,Y,Z) and (X w ,Y w ,Z w ) represent the coordinates of the pixel position (u, v) in the depth image in the camera coordinate system and the point cloud coordinates in the world coordinate system; d(u, v) represents the depth value at the pixel position (u, v) in the depth image, f is the focal length of the camera, (c x ,c y ) is the camera principal point; R is the camera's rotation matrix, and T is the camera's position matrix.

[0056] Step 3.2, extracting the three-dimensional surface information of the organ from the three-dimensional point cloud data set, specifically including:

[0057] (1) The 3D point cloud dataset is divided into several cubic units, each of which is called a voxel. Each voxel consists of 8 vertices. The isosurface value c for distinguishing organs is obtained.

[0058] (2) For each voxel, compare the depth value of each vertex in the depth image with the isosurface value c to determine whether each vertex is on the isosurface that distinguishes the organs, or whether different vertices of the voxel are on both sides of the isosurface:

[0059]

[0060] In the formula, f(x,y,z)=a0+a1x+a2y+a3z+a4xy+a5yz+a6xz+a7xyz represents the plane in the cube, and the coefficient a i (i=0,1,…,7) are the function values ​​of the 8 vertices in the cube. When the simultaneous equations are set to c, the equation of the plane in the cube that is equal to the value of the isosurface is found.

[0061] The depth values ​​of the eight vertices of the voxel are compared with the values ​​of the isosurface. If the depth values ​​of the eight vertices are greater than or less than the isosurface, the voxel is on both sides of the isosurface, otherwise the voxel is on the isosurface. If a voxel has a vertex on the isosurface or its different vertices are distributed on both sides of the isosurface, the voxel is still judged to intersect with the isosurface that distinguishes organs.

[0062] (3) The intersection points of the isosurface and the edges of each voxel are calculated by linear interpolation. The connection order of all intersection points is obtained by querying the triangle correspondence table of the Marching Cubes algorithm, and the triangular mesh of the isosurface is generated, which is the extracted three-dimensional surface information of the organ.

[0063] For each edge on the voxel that intersects with the isosurface, assuming that the two endpoints of the edge are p1 and p2, their scalar values ​​are v1 and v2, and the isosurface value is v, then the position of the intersection can be calculated by linear interpolation: t = (v-v1) / (v2-v1), thereby obtaining the position of the intersection point p: p = p1+t*(p2-p1).

[0064] Step 3.3, convert the three-dimensional surface information of the organ into the unit sphere projection of the organ, that is, each point in the triangular mesh of the organ is represented by the coordinates in the spherical coordinate system, specifically represented by two angle parameters. Normalize the radius in the spherical coordinates to generate the unit sphere projection of the isosurface information, that is, convert the radius of all vertices to between 0 and 1. For example, the coordinates of a point are (x i ,y i ,z i ), then the radius of the point is So we use two angles to represent this point: θi =arccos(z i / r i ),

[0065] Step 4: Use the generator trained by the generative adversarial network to output the unit fine three-dimensional model of each organ according to the unit sphere projection of each organ of the plant.

[0066] The training of the adversarial generation network is as follows Figure 2 As shown, the unit sphere projection of the organ obtained in step 3 is input into the generator, and the generator predicts and outputs a more refined unit sphere projection; the discriminator determines the similarity between the unit sphere projection of the predicted output and the real unit sphere projection of the organ, and constructs a loss function based on the similarity to train the adversarial generative network.

[0067] The generator is located in the adversarial generative network, whose discriminator is composed of 5 convolutional layers with ReLU activation function, and the adversarial generative network is trained using the Shapenet dataset, where the training loss function is:

[0068]

[0069] Among them, S1 represents the predicted output of the discriminator based on the unit sphere projection of the input organ, S2 represents the real unit sphere projection of the organ; x and y are points in S1 and S2 respectively; CD(S1,S2) represents the average distance between S1 and S2.

[0070] Finally, the generator trained by the generative adversarial network is used to output the unit fine three-dimensional model of each organ according to the unit sphere projection of each organ of the plant, thus completing the three-dimensional model reconstruction.

[0071] Step 5, set different growth environments and growth stages, extract the size and relative position relationship of each organ of the plant from the plant growth model database, and scale the unit fine three-dimensional model of each organ to obtain a fine three-dimensional model of each organ of the plant; then splice the fine three-dimensional models of all the organs of the plant to form an overall three-dimensional model of the plant.

[0072] This embodiment performs plant organ model scaling and plant organ splicing in the 3D graphics library OpenGL, and performs color rendering, texture mapping and lighting processing on each plant organ model in the 3D graphics library OpenGL. Specifically, the following steps are included:

[0073] Step 5.1, import the unit fine three-dimensional model of each organ of the crop plant (obtained in step 4) into OpenGL through the loading library.

[0074] Step 5.2, according to the crop plant growth model database, set different growth environments and growth stages, and scale the unit fine three-dimensional model of each organ as a benchmark to obtain a fine three-dimensional model of each organ of the plant, that is, visualize the shape and size of each organ model of the plant.

[0075] Step 5.3, write vertex shaders and fragment shaders, obtain the vertex positions of each organ model through model view projection transformation, pass the vertex positions to the fragment shader, and adjust the parameters for color rendering according to the color changes of crop plants in different growth stages.

[0076] Step 5.4, load the corresponding texture images of crop plants at different growth stages in the texture library, set the mapping method of the crop plant organ texture mapping to the geometric model surface, and perform texture mapping by defining the model coordinates of the texture image on the organ three-dimensional model through a function.

[0077] Step 5.5, set the position of the light source at different simulation times, and use the Phong lighting model to perform lighting processing on the plant model.

[0078] Step 5.6, render the plant single organ model data, and assemble the rendered plant single organ model according to the positional relationship between the plant organs (including the position and relative rotation angle) to generate a plant monomer model, thereby completing the three-dimensional dynamic visualization of crop plants under different conditions.

[0079] Compared with the overall crop plant model, the complexity of the single organ model of the plant is greatly reduced, and some organs of the same plant have linear morphological changes in different growth cycles. Therefore, the present invention generates a three-dimensional model from the single organ images of crop plants through deep learning, and then combines the single organ models according to the plant growth data to generate a complete model. The model can change dynamically with the conditions, which improves the generalization ability of the model.

[0080] The above embodiments are preferred embodiments of the present application. Ordinary technicians in this field can also make various changes or improvements on this basis. Without departing from the overall concept of the present application, these changes or improvements should fall within the scope of protection required by the present application.

Claims

1. A method for constructing a three-dimensional dynamic model of crop plants based on digital twins, characterized in that: include: Constructing a plant growth model database, which stores the size of each organ of the plant in different growth environments and at different growth stages, as well as the relative position relationship between different organs; Acquire an image of each organ of the plant, predict the depth information of each organ through a single-view depth estimator, and generate a depth image of each organ of the plant; The depth image of each organ of the plant is converted into a three-dimensional point cloud dataset, and then the three-dimensional surface information of the organ is extracted from it, and the three-dimensional surface information of the organ is converted into a unit sphere projection of the organ; The generator trained by the adversarial generative network is used to output the unit fine 3D models of various organs of the plant according to the unit sphere projection of various organs of the plant. By setting different growth environments and growth stages, the corresponding size of each organ of the plant and the relative position relationship between different organs are extracted from the plant growth model database, and the unit fine three-dimensional models of various organs are used as the benchmark for scaling to obtain a fine three-dimensional model of each organ of the plant; then the fine three-dimensional models of all the organs of the plant are spliced ​​to form an overall three-dimensional model of the plant.

2. The method for constructing a three-dimensional dynamic model of crop plants based on digital twins according to claim 1, characterized in that: The growth environment includes meteorological environment and soil environment. The meteorological environment includes longitude and latitude, temperature, rainfall, and sunshine duration; the soil environment includes pH value, texture, and organic matter content.

3. The method for constructing a three-dimensional dynamic model of crop plants based on digital twins according to claim 1, characterized in that: The size of each organ of the plant and the relative position relationship between different organs stored in the plant growth model database are obtained through historical measurements of plants, including leaf height, leaf length, maximum leaf width, stem node length, stem-leaf angle, stem roll diameter, and plant height.

4. The method for constructing a three-dimensional dynamic model of crop plants based on digital twins according to claim 1, characterized in that: The single-view depth estimator uses a pre-trained ResNet-18 network as the basic network structure, and modifies the output layer to output only a single-channel depth image; based on the Shapenet dataset, plant organ images and corresponding depth images of multiple plant categories are added, and the images and corresponding depth images in the dataset are used as input and output respectively. The ResNet-18 network with the modified output layer is trained to obtain a single-view depth estimator for generating depth images of various plant organs.

5. The method for constructing a three-dimensional dynamic model of crop plants based on digital twins according to claim 1, characterized in that: The depth images of various organs of the plant are converted into a three-dimensional point cloud data set, specifically including: according to the camera parameters, the depth image is first converted to the camera coordinate system, and then converted to the world coordinate system to generate point cloud data: Among them, (u, v) represents the pixel position coordinates in the depth image, (X, Y, Z) and (X w , Y w , Z w ) represent the coordinates of the pixel position (u, v) in the depth image in the camera coordinate system and the point cloud coordinates in the world coordinate system respectively; d(u, v) represents the depth value at the pixel position (u, v) in the depth image, f is the focal length of the camera, (c x , c y ) is the camera principal point; R is the camera's rotation matrix, and T is the camera's position matrix.

6. The method for constructing a three-dimensional dynamic model of crop plants based on digital twins according to claim 1, characterized in that: Extracting 3D surface information of organs from 3D point cloud datasets, including: Divide the three-dimensional point cloud data set into a number of cubic units, each of which is called a voxel, and each voxel consists of 8 vertices; obtain the isosurface value c for distinguishing organs; For each voxel, compare the depth of each vertex in the depth image with the isosurface value c to determine whether each vertex is on the isosurface for distinguishing organs, or whether different vertices of the voxel are on both sides of the isosurface; If a voxel has a vertex on the isosurface or its different vertices are distributed on both sides of the isosurface, the voxel is still judged to intersect with the isosurface distinguishing the organs; The intersection points of the isosurface and the edges of each voxel are calculated by linear interpolation. The connection order of all intersection points is obtained by querying the triangle correspondence table of the Marching Cubes algorithm, and the triangular mesh of the isosurface is generated, which is the extracted three-dimensional surface information of the organ.

7. The method for constructing a three-dimensional dynamic model of crop plants based on digital twins according to claim 1, characterized in that: The three-dimensional surface information of the organ is specifically a triangular mesh of an isosurface. The three-dimensional surface information of the organ is converted into a unit sphere projection of the organ, that is, each point in the triangular mesh of the organ is represented by coordinates in a spherical coordinate system.

8. The method for constructing a three-dimensional dynamic model of crop plants based on digital twins according to claim 1, characterized in that: The generator is located in the adversarial generative network, whose discriminator is composed of 5 convolutional layers with ReLU activation function, and the adversarial generative network is trained using the Shapenet dataset, where the training loss function is: Among them, S1 represents the predicted output of the discriminator based on the unit sphere projection of the input organ, S2 represents the real unit fine three-dimensional model of the organ; x and y are points in S1 and S2 respectively; CD(S1, S2) represents the average distance between S1 and S2.

9. The method for constructing a three-dimensional dynamic model of crop plants based on digital twins according to claim 1, characterized in that: The plant organ model is scaled and spliced ​​in the 3D graphics library OpenGL, and the color rendering, texture mapping and lighting processing are performed on the plant organ models in the 3D graphics library OpenGL.

10. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the computer program is executed by the processor, the processor is caused to implement the method according to any one of claims 1 to 9.

Citation Information

Cited By

  • Visual intelligent agricultural planting control system

    CN120255625A

  • Tomato dynamic growth monitoring method based on robot inspection and digital twinning

    CN122023933A

  • Eggplant double-trunk three-dimensional model construction method and system

    CN122636903A