Leaf phenotype parameter calculation method based on soybean leaf tiled image generation model

CN118967785AActive Publication Date: 2024-11-15SOUTH CHINA AGRICULTURAL UNIVERSITY

Patent Information

Application Number
CN202411279367.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-12
Publication Date
2024-11-15
Estimated Expiration
2044-09-12

AI Technical Summary

Technical Problem

The traditional method of calculating the aspect ratio of soybean leaves requires contact with leaves, resulting in leaf damage and low calculation efficiency, and requires a lot of manpower and time.

Method used

Using a non-contact method based on the soybean leaf tiling image generation model, by taking soybean leaf images in a natural curved state, image processing was performed using an improved SwimIR model to generate soybean leaf tiling images and calculate their aspect ratio.

Benefits of technology

The efficiency of calculating the aspect ratio of soybean leaves is improved, the risk of damage to soybean plants is reduced, the problem of leaf color style conversion in natural curved state is solved, and efficient and low-damage leaf type parameter calculation is achieved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118967785A_ABST
    Figure CN118967785A_ABST
Patent Text Reader

Abstract

The invention discloses a leaf phenotypic parameter calculation method based on a soybean leaf tiled image generation model, and the method comprises the steps: carrying out the photographing of soybean leaves in a natural bending state in a non-contact manner, and carrying out the tiled processing of a natural state image of the soybean leaves, and finally, the leaf type phenotypic parameters of the soybean leaves are calculated by using the soybean leaf tiled image, and compared with a method for directly measuring and calculating the length-width ratio of the soybean leaves manually, the measurement and calculation efficiency of the length-width ratio of the leaf types of the soybean leaves can be greatly improved through intervention of a deep learning technology for generating the soybean leaf tiled image, and meanwhile, the risk of damaging soybean plants is reduced; according to the method, improvement is carried out on the basis of the SwimIR model, so that the spreading processing of the soybean leaf image in the natural bending state is realized; and under the condition that the color style difference exists between the shot image and the scanned image in the model training data set, the model can also keep the color style of the output soybean leaf tiled image consistent with the color style of the input shot image.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of soybean leaf phenotypic parameter calculation, and in particular to a leaf phenotypic parameter calculation method based on a soybean leaf tiled image generation model. Background Art

[0002] In order to improve the efficiency of soybean breeding and enhance the self-sufficiency of soybean, the support of crop phenotyping technology for soybean breeding is indispensable. The phenotypic parameters of crops are important data support for accelerating the breeding speed of crops and increasing their yields. Among them, the Leaf Area Index (LAI), one of the important factors affecting the photosynthesis efficiency of crops, is closely related to the length-to-width ratio of soybean leaves (i.e., soybean leaf length-to-width ratio) and other leaf phenotypic parameters (i.e., leaf phenotypic parameters, phenotypic parameters), because different soybean leaf length-to-width ratios will affect the state of mutual shading between soybean leaves, thereby affecting the ability of soybean leaves to receive light.

[0003] Therefore, the measurement of the length-to-width ratio of soybean leaves is very important for soybean breeding. Since soybean leaves in a naturally growing state will have different degrees of natural curvature, the traditional method of measuring the length-to-width ratio of soybean leaves generally requires contacting the soybean leaves and flattening them before measuring. Therefore, the soybean leaves need to be manually flattened before measurement before each contact measurement. This contact measurement method requires contact with the soybean leaves, which inevitably increases the possibility of damage to the soybean leaves of the soybean plant. It also requires more manpower and time costs to be invested in the measurement of the length-to-width ratio of the soybean leaf shape, resulting in low measurement efficiency. Summary of the invention

[0004] 1. Technical issues to be resolved

[0005] In view of the deficiencies in the prior art, the present invention provides a method for calculating leaf phenotypic parameters based on a soybean leaf tiled image generation model, which can solve the above technical problems.

[0006] (II) Technical solution

[0007] In order to solve the above technical problems, the present invention provides the following technical solution: a method for calculating leaf phenotypic parameters based on a soybean leaf tiled image generation model, comprising the following steps:

[0008] S1. photographing soybean leaves in a natural bending state in a non-contact manner to obtain a natural state image of the soybean leaves;

[0009] S2, inputting the soybean leaf natural state image into a pre-trained soybean leaf tiled image generation model to obtain a soybean leaf tiled image;

[0010] S3. Calculate leaf shape phenotypic parameters of the soybean leaf using the soybean leaf tiled image, wherein the leaf shape phenotypic parameters include the length-to-width ratio of the soybean leaf.

[0011] Preferably, the soybean leaf tiled image generation model is an improved SwimIR model.

[0012] Preferably, the improved SwimIR model includes a data preprocessing module, a shallow feature extraction module, a deep feature optimization module, an auxiliary style preservation module and a high-quality image reconstruction module.

[0013] Preferably, step S2 includes sub-step S21: the data preprocessing module rotates the soybean leaf natural state image so that the line between the leaf tip and the leaf stalk of the soybean leaf natural state image remains horizontally centered with the leaf tip facing right and the leaf stalk facing left.

[0014] Preferably, step S2 further includes a sub-step S22: inputting the rotated soybean leaf natural state image into a shallow feature extraction module to preliminarily extract shallow feature information of the soybean leaf natural state image.

[0015] Preferably, step S2 further includes sub-step S23: further inputting the output of the shallow feature extraction module into the deep feature optimization module to perform deep optimization processing on the shallow feature information of the soybean leaf natural state image.

[0016] Preferably, step S2 further includes a sub-step S24: further inputting the outputs of the shallow feature extraction module and the deep feature optimization module into the auxiliary style preservation module for adaptive instance normalization processing.

[0017] Preferably, step S2 further includes a sub-step S25: inputting the output of the auxiliary style preserving module into a high-quality image reconstruction module for image reconstruction to output a soybean leaf tiled image.

[0018] Preferably, step S3 includes: inputting the soybean leaf tiled image into the image segmentation model SAM to obtain a mask map of the soybean leaf tiled image; further, obtaining the relative leaf length and relative leaf width of the soybean leaf by traversing the mask map horizontally and vertically; and finally calculating the aspect ratio of the soybean leaf shape using the relative leaf length and relative leaf width.

[0019] Preferably, the leaf phenotypic parameters also include the leaf area of ​​the soybean leaf type; step S3 also includes: calculating the leaf area of ​​the soybean leaf type using the soybean leaf tiled image.

[0020] (III) Beneficial effects

[0021] Compared with the prior art, the present invention provides a method for calculating leaf phenotypic parameters based on a soybean leaf flattened image generation model, which has the following beneficial effects: the present invention photographs soybean leaves in a naturally bent state in a non-contact manner to obtain a natural state image of the soybean leaves, further inputs the natural state image of the soybean leaves into a pre-trained soybean leaf flattened image generation model to obtain a soybean leaf flattened image, and finally uses the soybean leaf flattened image to calculate the aspect ratio of the soybean leaf shape; compared with the traditional method of directly manually measuring the aspect ratio of soybean leaves, the present invention can greatly improve the efficiency of calculating the aspect ratio of the soybean leaf shape by intervening in deep learning technology for generating soybean leaf flattened images, while reducing the risk of damaging soybean plants. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a flow chart of the leaf phenotypic parameter calculation method based on the soybean leaf flattening image generation model of the present invention;

[0023] Figure 2 is a flow chart of step S2 of the present invention;

[0024] Figure 3 is a flow chart of steps S1-S2 of the present invention;

[0025] Figure 4 is a flow chart of step S3 of the present invention;

[0026] Figure 5 Comparison of the photographed image, scanned image, and tiled image output by the model of soybean leaves. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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.

[0028] The present invention provides a method for calculating leaf phenotypic parameters based on a soybean leaf tiled image generation model, comprising the following steps:

[0029] S1. Photographing soybean leaves in a natural bending state in a non-contact manner to obtain a natural state image of the soybean leaves.

[0030] It is understandable that soybean leaves in a naturally growing state will have natural curvature to varying degrees. In step S1, a camera or other device is used to photograph the soybean leaves in a naturally curved state in a non-contact manner.

[0031] S2. Inputting the soybean leaf natural state image into a pre-trained soybean leaf tiled image generation model to obtain a soybean leaf tiled image.

[0032] Preferably, the above-mentioned soybean leaf tiled image generation model is an improved SwimIR (Swim-IR) model; the improved SwimIR model of the present invention includes a data preprocessing module, a shallow feature extraction module, a deep feature optimization module, an auxiliary style preservation module and a high-quality image reconstruction module; in the data preprocessing module, the captured images of soybean leaves in their natural state are unified in the horizontal orientation of the leaf tips and leaf stalks; in the subsequent four modules, the soybean leaf images after the unified horizontal orientation will be tiled and the color style of the captured image (soybean leaf natural state image) will be retained.

[0033] Specifically, step S2 includes the following sub-steps:

[0034] Sub-step S21: The data preprocessing module rotates the soybean leaf natural state image so that the line between the leaf tip and the leaf stalk of the soybean leaf natural state image remains horizontally centered with the leaf tip facing right and the leaf stalk facing left.

[0035] The data preprocessing module is a module that rotates the image of the soybean leaf in its natural state obtained by non-contact photography (i.e., the image of the soybean leaf in its natural state) so that the line between the tip and the stalk of the soybean leaf image remains horizontal and centered, with the tip pointing to the right and the stalk pointing to the left. The data preprocessing module first performs key point detection on the image of the soybean leaf in its natural state using a pre-trained YOLOv8 model. The purpose is to detect the coordinate information of the tip (top) and stalk of the soybean leaf, and record the two coordinate points of the tip and stalk as (x top ,y top )、(x stalk ,y stalk ); further calculate the slope k of the line connecting the leaf tip and the leaf stalk by the following formula (1).

[0036]

[0037] After obtaining the above slope k, the data preprocessing module further determines the relative position of the leaf tip and the leaf stalk point on the horizontal axis through the following formula (2) to determine whether the rotation angle of the soybean leaf natural state image needs to be added by 180° or 0°, and the final rotation angle a of the soybean leaf natural state image is obtained by the following formula (3).

[0038]

[0039]

[0040] The data preprocessing module rotates the above anglea As a key input of "getRotationMatrix2D" in the OpenCV tool library, the rotation matrix R (Rotate Matrix) of the natural state image of the soybean leaf is obtained, and then the rotation matrix R is used as a key input of "warpAffine" in the OpenCV tool library to realize the preprocessing of the natural state image of the soybean leaf to a unified orientation, so as to facilitate the subsequent calculation of leaf phenotypic parameters.

[0041] Sub-step S22: inputting the rotated soybean leaf natural state image into a shallow feature extraction module to preliminarily extract shallow feature information of the soybean leaf natural state image, so as to be used for generating a soybean leaf tiled image.

[0042] Shallow feature extraction module, the input of this module is the natural state image of soybean leaves preprocessed by the above data preprocessing module. The shallow feature extraction module consists of a convolution layer, the number of input channels of this convolution layer is 3, the number of output channels is 96, and the size and step size of its convolution kernel are 4×4 and 2 respectively. The purpose of the shallow feature extraction module is to preliminarily extract the shallow feature information of the input image and use its own output as the input of the next module (deep feature optimization module).

[0043] Sub-step S23: The output of the shallow feature extraction module is further input into the deep feature optimization module to perform deep optimization processing on the shallow feature information of the soybean leaf natural state image, so as to better ensure the quality of the subsequently generated soybean leaf tiled image.

[0044] The deep feature optimization module, namely the deep feature extraction module, takes the output of the previous module (shallow feature extraction module) as input, and is used to continue the optimization process at the deep feature level of the image. The deep feature optimization module consists of 7 submodules in series: 6 RSTB (Residual Swim Transformer Block) blocks and a convolutional layer. Among them, RSTB is composed of 6 STL (Swim Transformer Layer) and a convolutional layer in series. The number of channels of all convolutional layer inputs and outputs in the deep feature optimization module is 96, and the convolution kernel size and step size are 3×3 and 1 respectively.

[0045] Sub-step S24: The outputs of the shallow feature extraction module and the deep feature optimization module are further input into the auxiliary style preservation module for adaptive instance normalization processing.

[0046] Auxiliary style preservation module, this module is essentially adaptive instance normalization (Adaptive InstanceNormalization, AdaIN), which takes the output of the above-mentioned shallow feature extraction module and deep feature optimization module as its input. The two inputs of the auxiliary style preservation module are respectively recorded as x1 and x2. The calculation corresponding to the auxiliary style preservation module is as follows (4), where σ(·) is the standard deviation and μ(·) is the mean. The calculation of the following formula (4) is beneficial to maintaining the color style of the input soybean leaf natural state image, thereby helping to reduce the influence of color difference between the soybean leaf scanned image and the photographed image.

[0047]

[0048] Sub-step S25: The output of the auxiliary style preservation module is further input into the high-quality image reconstruction module for image reconstruction to output a soybean leaf tiled image.

[0049] The high-quality image reconstruction module is used to reconstruct the output of the above module (auxiliary style preservation module) into a 3-channel RGB image to obtain a soybean leaf tiled image. The high-quality image reconstruction module can be specifically composed of a convolution layer with 96-channel input, 3-channel output, a convolution kernel size of 4×4, and a stride of 2.

[0050] In addition, the data set used for training the improved SwimlR model of the present invention is 16,841 natural state photos of soybean leaves (i.e., natural state images of soybean leaves) and their corresponding tiled scanned images (the tiled scanned images are images of soybean leaves after being unfolded and flattened by using a scanner), as well as binary mask images of the photos and scanned images after image segmentation by SAM (i.e., the image segmentation model described below). Figure 5 As shown, the captured image is used as the input of the model and is denoted as I input , and its corresponding mask image is denoted as I I-mask ; The scanned image is used as the ground truth during model training and is denoted as T GT , and its corresponding mask image is denoted as I G_mask ; The output of the improved SwimlR model is denoted as I output The loss functions used in training the improved SwimlR model can be generative adversarial loss, perceptual loss, leaf area color loss, and normalized L1 loss, which consists of a total of 4 loss terms.

[0051] Generative adversarial loss, also known as GAN (Generative Adversarial Networks) loss, needs to be used with the discriminator. The discriminator uses the network structure of the U-Net model in the field of semantic segmentation to build the discriminator used in this invention. Its input is a three-channel RGB image and its output is a one-dimensional vector. The discriminator receives I GT with I output The closer the two output vectors are to each other as input, the better. In other words, the binary cross entropy loss between them is calculated. The definition of the binary cross entropy loss is shown in the following formula (5). By calculating the generative adversarial loss, it is helpful to improve the quality of the tiled soybean leaf image generated by the improved SwimlR model.

[0052]

[0053] In the above formula (5), N represents the total number of samples, and the category to which the i-th sample belongs is y i , generally speaking i The value of is 1 or 0, representing the true category and the false category respectively, and the predicted value of the i-th sample is recorded as g(x i ) and it is generally a probability value.

[0054] Perceptual loss, convolutional neural networks have a powerful ability to extract image features for digital images. Each layer of the convolutional neural network can be regarded as a feature extractor, and each layer can extract features of different abstract levels from the input image. Generally speaking, the farther away from the input layer, the easier it is to extract information at the global level of the image, and vice versa, it is easier to extract information at the detail level of the image. The definition of the perceptual loss function is shown in the following formula (6).

[0055]

[0056] In the above formula (6), Φ q represents the qth layer among the M feature extraction layers, and γ q Refers to the weight coefficient of the qth feature extraction layer, and the weight coefficient is defined as 0.8 or 1.2.

[0057] The color loss of the leaf area is due to the fact that the data set collected for training the improved SwimlR model includes images of soybean leaves in a natural bending state (photographed in Figure 1 input ) and its corresponding tiled scanned image (scanned image I GT ), shooting picture I input With Scanning Figure I GTDue to the differences in light sources, lighting angles, and image acquisition equipment, there is an inevitable color difference between the natural state images of soybean leaves in the training data set and their corresponding scanned images. output with I GT Establishing a loss function will cause the model to fail to learn I during training input The color style of soybean leaves turns more towards I GT The color style of soybean leaves. To this end, the present invention utilizes I input ,I I_mask with I output ,I G_mask Establishing the color loss of the leaf surface area can effectively solve the above color difference problem. The color loss function of the leaf surface area is defined as shown in the following formula (11).

[0058]

[0059]

[0060]

[0061]

[0062]

[0063] Where, as shown in the above formulas (7)-(10), μ input , μ output , σ input , σ output The input image I input , output image I output The mean and standard deviation of the pixels in the soybean leaf surface area, N I_mask 、N G_mask I input ,I output The total number of pixels in the leaf area, m is the total number of pixels in the image. In the above formula (11), n ​​is the number of channels of the image, namely: three color channels: R, G, and B. In addition, I I_mask ,I G_mask The value of the pixel points in the soybean leaf surface area is 1, and the value of the pixel points in the rest of the area is 0.

[0064] Normalized L1 loss, due to I input with I GT Due to the color difference, I GT with I output Establishing L1 loss together with the desired model output image I output The color style is I input Therefore, it is necessary toGT ,I output Reprojection is performed in three channels respectively as shown in the following formula (12), and then I′ obtained after reprojection is used GT , I′ output The L1 loss is established as shown in equation (13).

[0065]

[0066] L norm_L1 =||I′ GT -I′ output ||1 (13)

[0067] In the above formula (12), i represents the three channels of the image R, G, and B, m represents the total number of image pixels, and j is an integer, and 1×10 -6 To prevent the divisor from being 0.

[0068] The above four loss functions are added together to form the overall loss function during model training as shown in the following formula (14).

[0069] L TOTAL =β1L GAN +β2L perceptual +β3L leaf_color +β4L norm_L1 (14)

[0070] In the above formula (14), β is the weight coefficient of each loss term, which is defined as 0.01, 1.5, 20, and 25 respectively.

[0071] S3. Calculate the leaf phenotypic parameters of soybean leaves using the soybean leaf tiled image.

[0072] Among them, leaf phenotypic parameters include the length-to-width ratio of soybean leaves. The length-to-width ratio of soybean leaves is an important influencing indicator of the photosynthetic efficiency of soybean plants and is closely related to the yield of soybean plants.

[0073] The aspect ratio of soybean leaves. Although the straight-line distance between the camera and the soybean leaves is not fixed when the soybean leaf images are taken manually, which makes it impossible to directly obtain the absolutely true data of the length and width of the soybean leaves through image measurement, the calculation of the aspect ratio parameters does not require an absolutely true length and width data. Because the aspect ratio data obtained by directly measuring the relative data obtained by dividing the leaf length and leaf width of the soybean leaf tile image is logically consistent with the aspect ratio data obtained by dividing the absolutely true leaf length and leaf width data, the acquisition of the aspect ratio parameters does not require other reference values ​​to assist in the calculation.

[0074] Preferably, step S3 comprises: inputting the soybean leaf tiled image into an image segmentation model SAM (SegmentAnything Model) to obtain a mask image I of the soybean leaf tiled image. output_mask It can be understood that the pixel values ​​of the leaf surface area and other areas of the mask image are different: the pixel value of the leaf surface area is 1, and the pixel value of other areas is 0. Further, the relative leaf length and relative leaf width of the soybean leaf are obtained by traversing the mask image horizontally and vertically, that is, by traversing I horizontally and vertically respectively. output_mask The X-axis and Y-axis directions of the traversal are used to calculate the number of pixels with a pixel value of 1 to obtain the relative leaf length and relative leaf width data of the soybean leaf, which is recorded as L unreal , H unreal , the relative leaf length corresponds to the longest leaf length obtained by traversing the mask graph, and the relative leaf width corresponds to the longest leaf width obtained by traversing the mask graph. Finally, the relative leaf length and relative leaf width are used to calculate the length-to-width ratio R of the actual soybean leaf shape as shown in the following formula (15).

[0075]

[0076] In summary, it can be understood that the present invention uses the pre-trained YOL0v8 model through the data preprocessing module to perform key point detection on the leaf tip and leaf stalk points in the soybean leaf image, calculates the slope of the line connecting the leaf tip point and the leaf stalk point coordinates, and unifies the soybean leaf image with the leaf tip facing right and the leaf stalk facing left through the calculated slope. The soybean leaf image with standardized leaf orientation is input into the improved SwimlR model, and the tiled image of the input soybean leaf image is finally obtained through the "shallow feature extraction module", "deep feature optimization module", "auxiliary style retention module" and "high-quality image reconstruction module" in the trained model, and the color style of the output tiled soybean leaf image is aligned with the color style of the input soybean leaf image. Further, the output soybean leaf tiled image is passed through the SAM model to obtain its mask map (mask), and then the length and width (counting the number of pixels) of the soybean leaf area in the mask map in the horizontal and vertical directions are calculated respectively, and the ratio between the two is calculated to obtain the phenotypic aspect ratio parameter of the tiled soybean leaf.

[0077] In addition, the leaf phenotypic parameters may also include the leaf area of ​​the soybean leaf type. The leaf area is also an important influencing index of the photosynthesis efficiency of the soybean leaf plant and is closely related to the yield of the soybean plant. Accordingly, step S3 also includes: calculating the leaf area of ​​the soybean leaf type using the soybean leaf tiled image. For the calculation of the actual leaf area of ​​the soybean leaf type, it can be conveniently calculated by firstly statistically analyzing the above I output_mask The number of pixels in the leaf surface area, each pixel is defined as 1, and the total number of pixels in the leaf surface area is regarded as non-real relative image leaf area, recorded as S unreai, assuming that the length of the target soybean leaf shape has been obtained through actual measurement and is recorded as L real , the actual leaf area is S real The soybean leaf shape is approximately regarded as an ellipse, and the actual measured major axis length, minor axis length, and area of ​​the ellipse are denoted as a real 、b real , Then the non-real relative major axis length, minor axis length and area of ​​the ellipse measured in the image are recorded as a unreal 、b unreal , The following equations (16) to (18) are obtained, where equation (18) is an approximate calculation formula for the true leaf area (leaf area of ​​soybean leaf type).

[0078]

[0079]

[0080]

[0081] Compared with the prior art, the present invention provides a method for calculating leaf phenotypic parameters based on a soybean leaf flattened image generation model, which has the following beneficial effects: (1) The present invention photographs soybean leaves in a naturally bent state in a non-contact manner to obtain a soybean leaf natural state image, further inputs the soybean leaf natural state image into a pre-trained soybean leaf flattened image generation model to obtain a soybean leaf flattened image, and finally uses the soybean leaf flattened image to calculate the aspect ratio of the soybean leaf shape; compared with the traditional method of directly manually measuring the aspect ratio of the soybean leaf, the present invention can greatly improve the efficiency of calculating the aspect ratio of the soybean leaf shape by introducing deep learning technology for generating soybean leaf flattened images, while reducing the risk of damaging the soybean plant; (2) At the data set level, since there is inevitably color difference between the natural state captured image of the soybean leaf and its corresponding scanned image, the soybean leaf shape is not necessarily the same as the natural state image of the soybean leaf. The color of the leaf can reflect the content of photosynthesis-related plant pigments (such as chlorophyll and carotenoids) contained in soybean leaves to a certain extent. Therefore, solving the color difference problem between data pairs in the data set is also a problem solved by the present invention. The present invention can effectively solve the above-mentioned color difference problem by establishing the color loss of the leaf surface area during the training process of the model. In addition, the auxiliary style maintenance module is used to help reduce the influence of the color difference between the scanned image and the photographed image of the soybean leaf. The present invention is improved on the basis of the SwimIR model, which not only realizes the tiling processing of the soybean leaf image in the natural bending state, but also enables the model to keep the output soybean leaf tiled image consistent with the color style of the input photographed image when there is a color style difference between the photographed image and the scanned image in the model training data set; (3) In addition, compared with the commonly used convolutional neural network (ConvolutionNeural Compared with the CNN, the deep feature optimization module in the SwimIR model uses the SwimTransformer Layer submodule on a large scale. The latter has a stronger global perception of digital images and the ability to understand the connections between the various blocks of the image. In addition, the model also uses the residual connection method, which enables the model to achieve better training results under a deeper network structure. (4) By using the tiled image of soybean leaves to calculate the non-real relative leaf length and leaf area data of soybean leaves, and only by measuring the real leaf length data, the leaf area parameters of the soybean leaf type can be calculated. Compared with the method of directly measuring the leaf area manually, it can reduce the contact measurement of the soybean leaves, thereby improving the efficiency of leaf area measurement and reducing the damage to the soybean leaves.

[0082] It should be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0083] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. Leaf phenotypic parameter calculation method based on soybean leaf tiled image generation model, It is characterized in that The following steps are involved: S1. photographing soybean leaves in a natural bending state in a non-contact manner to obtain a natural state image of the soybean leaves; S2, inputting the soybean leaf natural state image into a pre-trained soybean leaf tiled image generation model to obtain a soybean leaf tiled image; S3. Calculating leaf phenotypic parameters of the soybean leaf using the soybean leaf tiled image, wherein the leaf phenotypic parameters include the aspect ratio of the soybean leaf.

2. The method for calculating leaf phenotypic parameters based on the soybean leaf flattening image generation model according to claim 1, Features: The soybean leaf tiled image generation model is an improved SwimIR model.

3. The method for calculating leaf phenotypic parameters based on the soybean leaf flattening image generation model according to claim 2, Features: The improved SwimIR model includes a data preprocessing module, a shallow feature extraction module, a deep feature optimization module, an auxiliary style preservation module and a high-quality image reconstruction module.

4. The method for calculating leaf phenotypic parameters based on the soybean leaf flattening image generation model according to claim 3, It is characterized in that The step S2 includes a sub-step S21: the data preprocessing module rotates the soybean leaf natural state image so that the line between the leaf tip and the leaf stalk of the soybean leaf natural state image remains horizontally centered with the leaf tip facing right and the leaf stalk facing left.

5. The method for calculating leaf phenotypic parameters based on the soybean leaf flattening image generation model according to claim 4, It is characterized in that The step S2 further includes a sub-step S22: inputting the rotated soybean leaf natural state image into the shallow feature extraction module to preliminarily extract shallow feature information of the soybean leaf natural state image.

6. The method for calculating leaf phenotypic parameters based on the soybean leaf flattening image generation model according to claim 5, It is characterized in that The step S2 further includes a sub-step S23: further inputting the output of the shallow feature extraction module into the deep feature optimization module to perform deep optimization processing on the shallow feature information of the natural state image of the soybean leaf.

7. The method for calculating leaf phenotypic parameters based on the soybean leaf flattening image generation model according to claim 6, It is characterized in that The step S2 further includes a sub-step S24: inputting the outputs of the shallow feature extraction module and the deep feature optimization module into the auxiliary style preservation module for adaptive instance normalization processing.

8. The method for calculating leaf phenotypic parameters based on the soybean leaf flattening image generation model according to claim 7, It is characterized in that The step S2 further includes a sub-step S25: inputting the output of the auxiliary style preserving module into the high-quality image reconstruction module for image reconstruction to output the soybean leaf tiled image.

9. The method for calculating leaf phenotypic parameters based on the soybean leaf flattening image generation model according to claim 1, It is characterized in that The step S3 comprises: inputting the soybean leaf tiled image into an image segmentation model SAM to obtain a mask image of the soybean leaf tiled image; further, obtaining the relative leaf length and relative leaf width of the soybean leaf by traversing and calculating the mask image horizontally and vertically; and finally calculating the aspect ratio of the soybean leaf shape using the relative leaf length and the relative leaf width.

10. The method for calculating leaf phenotypic parameters based on the soybean leaf flattening image generation model according to claim 1, It is characterized in that The leaf phenotypic parameters also include the leaf area of ​​the soybean leaf type; the step S3 also includes: calculating the leaf area of ​​the soybean leaf type using the soybean leaf tiled image.

Citation Information

Patent Citations

  • Image style migration method and system based on deep learning

    CN114581341A

  • Multi-mode and adversarial learning-based multi-task target detection and identification method and device

    CN114821014A

  • Soybean leaf area measuring and calculating method and system, computer equipment and storage medium

    CN115861409A

  • Two-stage image privacy protection method and system based on dynamic mask and generation and recovery

    CN117131520A

  • Plant leaf form identification method

    CN117576195A

Cited By

  • Kelp leaf surface area determination method and application thereof

    CN119991772A