A method for predicting crop genotypes throughout their growth period based on artificial intelligence

Abstract

This invention discloses an artificial intelligence-based method for predicting crop genotypes using images throughout their growth cycle. The method involves crop image acquisition and processing, constructing a multi-layer perceptron neural network and training a conditionally controlled generative adversarial network to learn the mapping relationship between crop genotypes, phenotypes, and images throughout their growth cycle. This method then generates a crop gene-phenotype prediction model and a phenotype-image prediction model. Finally, these two models predict, based on the predicted crop genotype, a visualization of the crop's entire growth cycle under the control of the gene in an ideal environment. This method, based on artificial intelligence technology, constructs a multi-dimensional multi-layer perceptron network and a conditionally controlled generative adversarial network. This method leverages the mapping patterns inherent in crop gene, phenotype, and image data to achieve realistic and accurate visual prediction of crops.

Description

Technical Field

[0001] The present invention relates to technical fields such as smart agriculture, artificial intelligence, crop informatics and smart crop breeding, and in particular to a method for predicting crop genotypes using images of their entire growth period based on artificial intelligence. Background Art

[0002] Traditional crop breeding techniques are developing slowly, and yield growth rates are insufficient to meet future food demands. With the rapid development of high-throughput omics and artificial intelligence technologies, a vast amount of crop genomics and high-throughput phenotypic data has been accumulated. Currently, the effective use of genomics technologies plays a crucial role in addressing challenges related to rapid breeding and omics research, and in promoting sustainable agricultural development.

[0003] Genomic prediction can predict crop field phenotypes based on genotypes before planting. Existing genomic prediction methods include classic statistical learning models such as RRBLUP and BayesA, as well as deep learning methods such as DNNGP. DNNGP performs PCA dimensionality reduction on whole-genome genotype data and uses the reduced features to achieve genomic prediction. Existing genomic prediction methods have high computational complexity, low genomic prediction accuracy, and are challenging to predict phenotypes throughout the entire growth period.

[0004] Furthermore, converting crop phenotypic information into intuitive, predictable growth images—implementing crop profiling—has become both a hot topic and a challenge in current research. This technology can simulate the growth of crops of different genotypes at different growth stages, providing farmers with scientific planting guidance and management recommendations, significantly improving the intelligence and precision of agricultural production.

[0005] Patent publication number CN115880433A collects environmental data and crop appearance images to construct a crop environmental data prediction network and a crop growth image prediction network, enabling crop simulation visualization. Patent publication number CN118097372A proposes a crop growth visualization prediction method that uses crop appearance images to create a portrait of the future crop. These studies demonstrate that it is feasible to predict crop appearance images from phenotypic information using existing machine vision methods. Summary of the Invention

[0006] Artificial intelligence technology has made significant progress in generating realistic images. However, in the agricultural sector, directly converting crop genotype data into corresponding appearance images remains challenging, particularly because gene expression patterns are highly specific and dynamic across different tissues and developmental stages, and crop phenotypes are shaped by multiple genes, their complex interaction networks, and external environmental factors. This makes it extremely difficult to construct models that learn nonlinear, multi-factor association mechanisms from static genotype data and accurately predict crop appearance images.

[0007] Currently, there are no mature datasets and algorithms that can directly establish an accurate mapping relationship between genotype data and crop appearance images. This is mainly due to the lack of sufficient high-quality sample data to train models and the lack of advanced algorithms that can handle such highly complex and high-dimensional data.

[0008] The present invention discloses a method for predicting crop genotypes using images of the entire growth period based on artificial intelligence, the method comprising:

[0009] Step S1, preparing a plurality of crop seeds of known varieties, planting them in different pots under the same growing environment, and using a camera to regularly capture crop images throughout the entire growth period of the crops to obtain original images of the crops throughout the entire growth period;

[0010] Step S2: training a crop semantic segmentation model through manual annotation, and using the model to process the original crop image obtained in step S1, removing irrelevant content in the image and retaining only the area where the crop is located to obtain a crop image;

[0011] Step S3, performing image processing on the crop image obtained in step S2 to extract phenotypic information of the crop at different growth stages;

[0012] Step S4, obtaining the crop genotype sequence through sequencing technology, organizing the one-to-one correspondence between the genotype sequence and the crop phenotypic information, constructing a multidimensional space multilayer perceptron neural network (MDMLP), training to obtain a genotype-phenotype prediction model, and establishing a mapping relationship between the genotype and phenotype of the crop throughout its growth period;

[0013] Step S5: sorting the crop phenotypes and corresponding image data for each growth stage, constructing a conditionally controlled crop image generator and a discriminator that integrates conditional information based on an improved StyleGAN high-resolution generative adversarial network model, and designing corresponding loss functions to train and obtain phenotype-image prediction models for each growth stage, thereby establishing a mapping relationship between the crop phenotypes and crop images for each growth stage.

[0014] Step S6, inputting the genotype sequence to be predicted into the genotype-phenotype prediction model obtained in step S4 to obtain the predicted crop phenotype;

[0015] In step S7, the crop phenotype predicted in step S6 and the corresponding growth period information are input into the phenotype-image prediction model obtained in step S5 to obtain a predicted crop image.

[0016] Furthermore, the method in step S2 of the present invention includes:

[0017] Step S21, selecting different plants with large growth differences and crop images at different time points as training sets;

[0018] Step S22, performing pixel-level semantic annotation on the crop area in the training set image to obtain annotation information corresponding to the image;

[0019] Step S23: inputting the training set images and corresponding annotation information into a semantic segmentation network for training to obtain a crop semantic segmentation model;

[0020] Step S24, using the crop semantic segmentation model to segment the original crop image acquired in step S1 to obtain a crop binary image;

[0021] Step S25 , using the crop binary image obtained in step S24 to perform mask processing on the original crop image, to obtain a crop image that only retains the area where the crop is located.

[0022] Furthermore, the method in step S4 of the present invention includes:

[0023] Step S41: Sequence the crops to obtain the single nucleotide variant site data SNP of all crop genomes, and randomly screen m SNP sites are selected to ensure that the SNP distribution density of the whole genome after screening is consistent with that before screening, and the genotype sequence for predicting crop phenotypes is obtained;

[0024] Step S42, constructing a multidimensional space multilayer perceptron neural network MDMLP to achieve the fusion of local and global features between the extracted SNP sites;

[0025] In step S43, the neural network designed in step S42 is used to learn the one-to-one corresponding genotype and phenotype data of each period to obtain a genotype-phenotype prediction model for each growth period.

[0026] Furthermore, the method for constructing the MDMLP in step S42 of the present invention includes:

[0027] Step S421, set the patch embedding layer, and use the patch embedding layer to embed the input m The SNP site features are converted into word embeddings of different dimensions including height, channel and width;

[0028] In step S422, an MDblock layer is set to construct a height-mixed MLP layer, a channel-mixed MLP layer, and a token-width-mixed MLP layer for word embeddings of different dimensions, respectively, using a fully connected neural network to further realize local feature extraction of different dimensions. The specific implementation process is shown in formula (1).

[0029] (1)

[0030] In formula (1), y represents the feature vector extracted by the hybrid MLP layer, x represents the encoded word embedding features, W 1 and W 2 represents the weight of the two-layer linear mapping, b 1 and b 2 is the bias of the two-layer linear mapping, σ Represents the activation function Gelu , the captured linear features can be made nonlinear by the activation function. The activation function calculation is shown in formula (2).

[0031] (2)

[0032] In the above formula, X Indicates compliance N (0,1) Normally distributed random numbers, P ( X ≤ x ) represents a random variable X Less than or equal to x The probability of X~N (0,1), so P ( X ≤ x ) refers to the cumulative distribution function of the standard normal distribution;

[0033] Step S423: Setting a global average pooling layer, and using global average pooling to achieve global feature fusion based on features extracted from different dimensions;

[0034] Step S424: Setting a fully connected neural network layer, and using a fully connected neural network based on the globally fused features to further improve the effect of feature extraction;

[0035] In step S425, a dropout regularization method and nonlinear activation functions such as ReLU are added to the network to prevent overfitting, and finally an MDMLP network is constructed.

[0036] Furthermore, the method in step S5 of the present invention includes:

[0037] Step S51, obtaining a high-resolution generative adversarial network based on StyleGAN with the conditional vector introduced;

[0038] Step S52: In the discriminator, a multi-scale discriminator network is constructed, and a downsampling convolution layer is used to extract image features to obtain a feature vector;

[0039] Step S53: Input the phenotypic data as a conditional vector into the discriminator network. The conditional vector has a size of 1×1×S, where S is the number of phenotypes. The first two dimensions of the vector are expanded by replication to obtain a 4×4×S conditional vector. At this time, the expanded conditional vector and the image feature vector obtained after downsampling are consistent in the first two dimensions. The image feature vector has a size of 4×4×C, where C is the number of channels. The expanded conditional vector and the image feature vector are aligned in the first two dimensions and superimposed in the third dimension to obtain a fused feature vector of size 4×4×(S+C). Then, the fused features are extracted through convolution, and finally a fully connected layer is used to obtain the discriminator score.

[0040] Step S54, calculate the cross entropy between the discriminator score and 0 or 1 as the loss function of the generative adversarial network, as shown in formula (3),

[0041] (3)

[0042] In the formula D is the discriminator, G For the generator, E is the mathematical expectation, I To input a real image, y S is a conditional vector of size 1×1×S, z is Gaussian distributed noise, V ( D , G ) represents the difference between the generated rice image and the real rice image, using the cross entropy loss function of the true and false categories, including and Two parts: Represents a fixed generator G Parameter training discriminator D parameters, by maximizing the cross entropy loss V ( D , G ) to update the discriminator D parameter; represents a fixed discriminator D Parameter training generator G Parameters, generator G To be in the discriminator D Maximize cross entropy loss V ( D , G ), minimize the cross entropy loss.

[0043] This invention establishes a complete process from crop planting to image acquisition, and also details all necessary preprocessing steps to ensure data quality and accuracy. By introducing the MDMLP artificial intelligence algorithm, it accurately predicts key phenotypic values throughout the crop's growth cycle, covering all stages of crop development, providing comprehensive data support for agricultural research and production.

[0044] Based on the accurately predicted phenotypic values for each growth stage, this paper further innovatively applies an improved StyleGAN algorithm to render crop images. This method not only retains StyleGAN's advantages in generating high-quality, high-resolution images, but also incorporates crop-specific phenotypic information to generate images that are more realistic. It can intuitively display the growth status of crops at different growth stages, providing an intuitive and reliable visual reference for agricultural decision makers.

[0045] This invention significantly advances the intelligent and automated development of agricultural production. By building an intelligent prediction model and image generation system, agricultural workers can obtain a comprehensive picture of crop growth without the need for field observation, effectively reducing labor intensity and improving work efficiency. This invention also provides important technical support for research in related disciplines such as crop genetics and plant physiology, helping to reveal the molecular mechanisms of crop growth and development and offering new ideas and methods for crop breeding and genetic improvement. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is the overall technical solution diagram of the present invention.

[0047] Figure 2 This is the MDMLP network structure diagram constructed by the present invention.

[0048] Figure 3 This is the structure diagram of the improved StyleGAN network discriminator of the present invention.

[0049] Figure 4 This is an example diagram of the prediction effect of a specific embodiment of the present invention. DETAILED DESCRIPTION

[0050] An exemplary embodiment of the present invention will now be described in detail. This detailed description should not be considered a limitation of the present invention, but rather should be understood as a more detailed description of certain aspects of the present invention. It should be understood that the terms used herein are intended solely to describe particular embodiments and are not intended to limit the present invention. For parameter numerical ranges herein, it should be understood that the upper and lower limits of the ranges, as well as intermediate values therebetween, are specifically disclosed.

[0051] Unless otherwise specified, all techniques and terms used herein have the same meaning as commonly understood by those skilled in the art in the art to which the present invention pertains. Although only one specific exemplary embodiment has been described herein, methods and materials similar or equivalent to those described herein may also be used in the practice or testing of the present invention.

[0052] like Figure 1 The flowchart of the method for predicting crop genotype images throughout the growth period based on artificial intelligence provided by this embodiment is shown, including:

[0053] Step S1: Prepare seeds of 533 rice varieties with known genotype sequences, select four healthy seeds for each variety, and plant them in different pots under the same growth environment. Two weeks after sowing, use a camera to capture rice images every six days for eight times to obtain original images of rice throughout its entire growth period.

[0054] Step S2: training a rice semantic segmentation model through manual annotation, and using the model to process the original rice image obtained in step S1, removing irrelevant content in the image and retaining only the rice area to obtain a rice image;

[0055] Step S3, using a high-throughput crop phenotyping device RAP to process the rice image obtained in step S2, extracting 13 phenotypic information of the rice, including six grayscale histogram features, plant projected area, height, width, ratio of projected area to circumscribed rectangle area, ratio of perimeter to projected area, fractal dimension, and green portion projected area, and normalizing all data;

[0056] Step S4, obtaining the rice genotype sequence through sequencing technology, organizing the one-to-one correspondence between the rice genotype sequence and phenotypic information, constructing a multidimensional space multilayer perceptron neural network (MDMLP), training to obtain a genotype-phenotype prediction model, and establishing a mapping relationship between the genotype and phenotype of rice throughout its entire growth period;

[0057] Step S5: sorting the crop phenotypes and corresponding image data for each growth stage, constructing a conditionally controlled crop image generator and a discriminator that integrates conditional information based on an improved StyleGAN high-resolution generative adversarial network model, and designing corresponding loss functions to train and obtain phenotype-image prediction models for each growth stage, thereby establishing a mapping relationship between the crop phenotypes and crop images for each growth stage.

[0058] Step S6, inputting the genotype sequence to be predicted and the expected rice growth period information into the genotype-phenotype prediction model obtained in step S4 to obtain the predicted rice phenotype for that period;

[0059] Step S7: input the rice phenotype and corresponding growth period information predicted in step S6 into the phenotype-image prediction model obtained in step S5 to obtain a predicted rice image. The prediction effect is shown in the following example: Figure 4 shown.

[0060] Furthermore, the method in step S2 of the present invention includes:

[0061] Step S21, selecting 150 original rice images of plants with large growth differences at different time points as a training set;

[0062] Step S22: Mark the pixels outside the rice area in the training set image as black, and save the marked image, which is the marking information corresponding to the original image;

[0063] Step S23, inputting the above training set images and corresponding annotation information into the Segformer semantic segmentation network for training to obtain a rice semantic segmentation model;

[0064] Step S24, using the rice semantic segmentation model to segment the original rice image acquired in step S1. After segmentation, the model sets the rice plant area to white and other areas to black, thereby obtaining a rice binary image.

[0065] Step S25 , using the rice binary image obtained in step S24 to perform masking on the original rice image, setting the pixels outside the rice area to black, and obtaining an image that only retains the rice area.

[0066] Furthermore, the method in step S4 of the present invention includes:

[0067] Step S41: obtaining single nucleotide variant site (SNP) data for the entire genome of 533 rice varieties through sequencing, ensuring that the SNP distribution density of the entire genome after screening is consistent with the SNP distribution density of the entire genome before screening, and randomly screening 150,000 SNP sites across the entire genome for predictive phenotypic analysis;

[0068] Step S42, constructing a multidimensional multilayer perceptron neural network MDMLP, inputting the 150,000 SNP sites screened in step S41 into the MDMLP to achieve the fusion of local and global features between the extracted SNP sites;

[0069] In step S43, the neural network designed in step S42 is used to learn the one-to-one corresponding genotype and phenotype data of each period to obtain a genotype-phenotype prediction model for each growth period.

[0070] Furthermore, the method for constructing MDMLP in step S42 of the present invention is as follows: Figure 2 As shown, including:

[0071] Step S421: Set the patch embedding layer and use it to convert the input 150,000 SNP site features into word embeddings of different dimensions including height, channel, and width;

[0072] In step S422, an MDblock layer is set to construct a height-mixed MLP layer, a channel-mixed MLP layer, and a token-width-mixed MLP layer for word embeddings of different dimensions, respectively, using a fully connected neural network to further realize local feature extraction of different dimensions. The specific implementation process is shown in formula (4).

[0073] (4)

[0074] In formula (4), y represents the feature vector extracted by the hybrid MLP layer, x represents the encoded word embedding features, W 1 and W 2 represents the weight of the two-layer linear mapping, b 1 and b 2 is the bias of the two-layer linear mapping, σ Represents the activation function Gelu , the captured linear features can be made nonlinear by the activation function. The activation function calculation is shown in formula (5).

[0075] (5)

[0076] In the above formula, X Indicates compliance N (0,1) Normally distributed random numbers, P ( X ≤ x ) represents a random variable X Less than or equal to x The probability of X~N (0,1), so P ( X ≤ x ) refers to the cumulative distribution function of the standard normal distribution;

[0077] Step S423: Setting a global average pooling layer, and using global average pooling to achieve global feature fusion based on features extracted from different dimensions;

[0078] Step S424: Setting a fully connected neural network layer, and using a fully connected neural network based on the globally fused features to further improve the effect of feature extraction;

[0079] In step S425, a dropout regularization method and nonlinear activation functions such as ReLU are added to the network to prevent overfitting, and finally an MDMLP network is constructed.

[0080] Furthermore, the method in step S5 of the present invention includes:

[0081] Step S51, removing data of plant lodging, wilting, and blurred images from the rice images obtained in step S2 and the rice phenotypic information obtained in step S3, organizing the data into one-to-one corresponding phenotype-image pairs, and dividing the data into two groups: before heading and after heading;

[0082] Step S52, 533 rice varieties are randomly divided into a training set and a test set at a ratio of 9:1, 16,321 images and corresponding phenotypic data before heading are used as the training set, 1,830 images and corresponding phenotypic data are used as the test set, 12,352 images and corresponding phenotypic data after heading are used as the training set, and 1,790 images and corresponding phenotypic data are used as the test set;

[0083] Step S53, based on the conditional control high-resolution generative adversarial network of the StyleGAN series, improve its discriminator network structure and corresponding loss function, such as Figure 3 As shown;

[0084] In step S54, the two training sets obtained in step S52 are used to train the improved high-resolution generative adversarial network in step S53 to obtain two phenotype-image prediction models corresponding to the pre-heading period and the post-heading period, respectively.

[0085] Furthermore, the method for improving the high-resolution generative adversarial network ConditionalStyleGAN in step S53 of the present invention includes:

[0086] Step S531: In the discriminator of Conditional StyleGAN, a multi-scale discriminator network is constructed, and a downsampling convolution layer is used to extract image features to obtain a feature vector.

[0087] Step S532: The phenotypic data is input into the discriminator network as a conditional vector with a size of 1×1×13. The first two dimensions of the vector are expanded by replication to obtain a 4×4×13 conditional vector. At this time, the expanded conditional vector and the image feature vector obtained after downsampling are consistent in the first two dimensions. The image feature vector has a size of 4×4×3. The expanded conditional vector and the image feature vector are aligned in the first two dimensions and superimposed in the third dimension to obtain a fused feature vector with a size of 4×4×(13+3). The fused features are then extracted through convolution, and finally a fully connected layer is used to obtain the discriminator score.

[0088] Step S533, calculate the cross entropy between the discriminator score and 0 or 1 as the loss function of the generative adversarial network, as shown in formula (6),

[0089] (6)

[0090] In the formula D is the discriminator, G For the generator, E is the mathematical expectation, I To input a real image, y S is a conditional vector of size 1×1×13, z is Gaussian distributed noise, V ( D , G ) represents the difference between the generated rice image and the real rice image, using the cross entropy loss function of the true and false categories, including and Two parts: Represents a fixed generator G Parameter training discriminator D parameters, by maximizing the cross entropy loss V ( D , G ) to update the discriminator D parameter; represents a fixed discriminator D Parameter training generator G Parameters, generator G To be in the discriminator D Maximize cross entropy loss V ( D , G ), minimize the cross entropy loss.

[0091] Although the present invention has been described with reference to exemplary embodiments, it should be understood that the exemplary embodiments described herein are merely illustrative of the spirit of the present invention. Persons skilled in the art may modify, supplement, or substitute the exemplary embodiments described herein with similar approaches without departing from the spirit of the present invention or exceeding the scope of the appended claims.

Claims

1. A method for predicting crop genotypes using images of the entire growth period based on artificial intelligence, the method comprising: Step S1, preparing a plurality of crop seeds of known varieties, planting them in different pots under the same growing environment, and using a camera to regularly capture crop images throughout the entire growth period of the crops to obtain original images of the crops throughout the entire growth period; Step S2: training a crop semantic segmentation model through manual annotation, and using the model to process the original crop image obtained in step S1, removing irrelevant content in the image and retaining only the area where the crop is located to obtain a crop image; Step S3, performing image processing on the crop image obtained in step S2 to extract phenotypic information of the crop at different growth stages; Step S4, obtaining the crop genotype sequence through sequencing technology, organizing the one-to-one correspondence between the genotype sequence and the crop phenotypic information, constructing a multidimensional space multilayer perceptron neural network (MDMLP), training to obtain a genotype-phenotype prediction model, and establishing a mapping relationship between the genotype and phenotype of the crop throughout its growth period; Step S5: sorting the crop phenotypes and corresponding image data for each growth stage, constructing a conditionally controlled crop image generator and a discriminator that integrates conditional information based on an improved StyleGAN high-resolution generative adversarial network model, and designing corresponding loss functions to train and obtain phenotype-image prediction models for each growth stage, thereby establishing a mapping relationship between the crop phenotypes and crop images for each growth stage. Step S6, inputting the genotype sequence to be predicted into the genotype-phenotype prediction model obtained in step S4 to obtain the predicted crop phenotype; In step S7, the crop phenotype predicted in step S6 and the corresponding growth period information are input into the phenotype-image prediction model obtained in step S5 to obtain a predicted crop image.

2. The method for predicting crop genotypes throughout their growth period based on artificial intelligence according to claim 1, characterized in that: The method in step S2 includes: Step S21, selecting different plants with large growth differences and crop images at different time points as training sets; Step S22, performing pixel-level semantic annotation on the crop area in the training set image to obtain annotation information corresponding to the image; Step S23: inputting the training set images and corresponding annotation information into a semantic segmentation network for training to obtain a crop semantic segmentation model; Step S24, using the crop semantic segmentation model to segment the original crop image acquired in step S1 to obtain a crop binary image; Step S25 , using the crop binary image obtained in step S24 to perform mask processing on the original crop image, to obtain a crop image that only retains the area where the crop is located.

3. The method for predicting crop genotypes throughout their growth period based on artificial intelligence according to claim 1, characterized in that: The method in step S4 includes: Step S41: Sequence the crops to obtain the single nucleotide variant site data SNP of all crop genomes, and randomly screen m SNP sites are selected to ensure that the SNP distribution density of the whole genome after screening is consistent with that before screening, and the genotype sequence for predicting crop phenotypes is obtained; Step S42, constructing a multidimensional space multilayer perceptron neural network MDMLP to achieve the fusion of local and global features between the extracted SNP sites; In step S43, the neural network designed in step S42 is used to learn the one-to-one corresponding genotype and phenotype data of each period to obtain a genotype-phenotype prediction model for each growth period.

4. The method for predicting crop genotypes throughout their growth period based on artificial intelligence according to claim 3, characterized in that: The method for constructing the MDMLP in step S42 includes: Step S421, set the patch embedding layer, and use the patch embedding layer to embed the input m The SNP site features are converted into word embeddings of different dimensions including height, channel and width; In step S422, an MDblock layer is set to construct a height-mixed MLP layer, a channel-mixed MLP layer, and a token-width-mixed MLP layer for word embeddings of different dimensions, respectively, using a fully connected neural network to further realize local feature extraction of different dimensions. The specific implementation process is shown in formula (1). （1） In formula (1), y represents the feature vector extracted by the hybrid MLP layer, x represents the encoded word embedding features, W 1 and W 2 represents the weight of the two-layer linear mapping, b 1 and b 2 is the bias of the two-layer linear mapping, σ Represents the activation function Gelu , the captured linear features can be made nonlinear by the activation function. The activation function calculation is shown in formula (2). （2） In the above formula, X Indicates compliance N (0,1) Normally distributed random numbers, P ( X ≤ x ) represents a random variable X Less than or equal to x The probability of X~N (0,1), so P ( X ≤ x ) refers to the cumulative distribution function of the standard normal distribution; Step S423: Setting a global average pooling layer, and using global average pooling to achieve global feature fusion based on features extracted from different dimensions; Step S424: Setting a fully connected neural network layer, and using a fully connected neural network based on the globally fused features to further improve the effect of feature extraction; In step S425, a dropout regularization method and nonlinear activation functions such as ReLU are added to the network to prevent overfitting, and finally an MDMLP network is constructed.

5. The method for predicting crop genotypes throughout their growth period based on artificial intelligence according to claim 1, characterized in that: The method in step S5 includes: Step S51, obtaining a high-resolution generative adversarial network based on StyleGAN with the conditional vector introduced; Step S52: In the discriminator, a multi-scale discriminator network is constructed, and a downsampling convolution layer is used to extract image features to obtain a feature vector; Step S53: Input the phenotypic data as a conditional vector into the discriminator network. The conditional vector has a size of 1×1×S, where S is the number of phenotypes. The first two dimensions of the vector are expanded by replication to obtain a 4×4×S conditional vector. At this time, the expanded conditional vector and the image feature vector obtained after downsampling are consistent in the first two dimensions. The image feature vector has a size of 4×4×C, where C is the number of channels. The expanded conditional vector and the image feature vector are aligned in the first two dimensions and superimposed in the third dimension to obtain a fused feature vector of size 4×4×(S+C). Then, the fused features are extracted through convolution, and finally a fully connected layer is used to obtain the discriminator score. Step S54, calculate the cross entropy between the discriminator score and 0 or 1 as the loss function of the generative adversarial network, as shown in formula (3), m i n G m a x D V D G = m i n G m a x D E l o g D I y S + E l o g 1 - D G z y S y S （3） In the formula D is the discriminator, G For the generator, E is the mathematical expectation, I To input a real image, y S is a conditional vector of size 1×1×S, z is Gaussian distributed noise, V ( D , G ) represents the difference between the generated image and the real image, and adopts the cross entropy loss function of the two categories of true and false. and Two parts; Represents a fixed generator G Training the discriminator D , by maximizing the cross entropy loss V ( D , G ) to update the discriminator D Parameters; represents a fixed discriminator D Training the Generator G , the generator needs to maximize the cross entropy loss between real and fake images in the discriminator V ( D , G ), minimize the cross entropy loss.

Images