Corn germplasm resource viability cross-germplasm discrimination method, system, equipment and medium

Abstract

The invention discloses a corn germplasm resource viability cross-germplasm discrimination method, system, equipment and medium, and relates to the technical field of corn viability discrimination, in the method, hyperspectral data of an embryo area and hyperspectral data of an endosperm area are obtained; obtaining hyperspectral data of each sample corn seed; each piece of hyperspectral data is preprocessed; feature selection is carried out on each piece of preprocessed hyperspectral data, and a constructed double-branch spectrum Transform corn germplasm cross-germplasm viability detection model is trained in combination with viability label data of each sample corn seed; the trained double-branch spectrum Transform corn germplasm cross-germplasm viability detection model is utilized to discriminate the cross-germplasm corn to be discriminated, the viability prediction result of the cross-germplasm corn to be discriminated is obtained, and the accuracy and reliability of cross-germplasm corn germplasm resource viability discrimination can be remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of corn vitality identification, and in particular to a method, system, equipment and medium for cross-germplasm identification of corn germplasm vitality. Background Art

[0002] Hyperspectral imaging technology can capture continuous-band spectral information about seeds, encompassing their chemical composition, structure, and other characteristics. It is a commonly used method for non-destructive seed viability testing. However, due to differences in genetic background and chemical composition, different varieties of corn seeds exhibit significant spectral variations, making accurate determination of seed viability across different germplasms challenging.

[0003] Currently, most traditional algorithms fail to fully exploit the potential connections and differences between different corn seed varieties when processing data. They also lack the ability to integrate key information such as seed spectral characteristics, resulting in poor accuracy and stability in the discrimination models. Faced with large amounts of seed data from diverse varieties, these algorithms suffer from low computational efficiency and are unable to meet the rapid detection requirements of actual production. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to address the deficiencies of the existing technology and specifically provide a method, system, device and medium for distinguishing the vitality of corn germplasm resources across germplasms, as follows:

[0005] 1) In the first aspect, the present invention provides a method for determining the viability of corn germplasm resources across germplasms. The specific technical solution is as follows:

[0006] Acquire a hyperspectral image of each sample corn seed in the sample corn germplasm resource, and obtain vitality label data of each sample corn seed in the sample corn germplasm resource;

[0007] Threshold segmentation is performed on the hyperspectral image of each sample corn seed to obtain a hyperspectral image of each sample corn seed after background removal, and regional segmentation is performed on the hyperspectral image of each sample corn seed after background removal to determine a hyperspectral image of the embryo region and a hyperspectral image of the endosperm region;

[0008] Obtaining hyperspectral data of the embryo region and hyperspectral data of the endosperm region according to the hyperspectral image of the embryo region and the hyperspectral image of the endosperm region;

[0009] Obtaining hyperspectral data of each sample corn seed from the hyperspectral image of each sample corn seed after background removal;

[0010] Preprocess each hyperspectral data;

[0011] Feature selection was performed on each pre-processed hyperspectral data set, and the constructed dual-branch spectral Transformer corn germplasm cross-germplasm vitality detection model was trained based on the vitality label data of each corn seed sample.

[0012] The trained dual-branch spectral Transformer corn germplasm cross-germplasm viability detection model is used to discriminate the cross-germplasm corn to obtain the viability prediction results of the cross-germplasm corn to be discriminated.

[0013] The beneficial effects of the cross-germplasm discrimination method for the vitality of corn germplasm resources provided by the present invention are as follows:

[0014] In the process of judging the vitality of corn seeds, different parts of the seeds contribute differently to the vitality of the seeds. For example, the endosperm is the main source of nutrients in the seed germination process, and its state is directly related to whether the seeds can germinate normally. The present invention can make full use of the pre-processed hyperspectral data of different parts of the corn seeds to train the dual-branch spectral Transformer corn germplasm cross-germplasm vitality detection model, so that the trained dual-branch spectral Transformer corn germplasm cross-germplasm vitality detection model can accurately capture the characteristic patterns of different varieties of corn seeds under vitality status, effectively solving the problems of low accuracy and poor stability of traditional cross-germplasm algorithms. In practical applications, by comparing the characteristics of the segmented embryo region, endosperm region and the entire seed region, the present invention can more deeply explore the intrinsic information of the vitality of corn germplasm resources, significantly improve the accuracy and reliability of cross-germplasm corn germplasm vitality judgment, and provide strong technical support for seed quality detection in agricultural production.

[0015] On the basis of the above scheme, the method for distinguishing the vitality of corn germplasm resources across germplasms of the present invention can be further improved as follows.

[0016] Furthermore, the hyperspectral image of each sample corn seed after background removal is segmented to determine the hyperspectral image of the embryo region and the hyperspectral image of the endosperm region, including:

[0017] Manually annotating the boundaries of the embryo region and the endosperm region in the hyperspectral images of multiple corn seed samples to obtain embryo annotated data and endosperm annotated data;

[0018] The constructed moving inverted bottleneck UNet germplasm hyperspectral image segmentation model was trained using hyperspectral images of multiple corn seed samples, combined with embryo and endosperm annotation data.

[0019] The trained moving inverted bottleneck UNet germplasm hyperspectral image segmentation model was used to perform regional segmentation on the hyperspectral image of each sample corn seed after background removal to determine the hyperspectral image of the embryo region and the hyperspectral image of the endosperm region.

[0020] The beneficial effect of adopting the above-mentioned further scheme is that the existing hyperspectral detection methods often do not fully consider the characteristic differences of different parts of seeds. On the one hand, they ignore the characteristic differences of different parts of seeds and adopt the analysis method of averaging all pixels in the seed area, which leads to the weakening of important characteristic information; moreover, traditional image segmentation is inaccurate in the segmentation of corn seeds with irregular surfaces, which easily causes the segmented areas to be missed or incorrectly divided, resulting in deviations in the feature analysis based on the segmentation results, and cannot provide a reliable basis for vitality judgment. However, the present application can accurately determine the hyperspectral images of the embryo area and the hyperspectral images of the endosperm area through the mobile inverted bottleneck UNet germplasm hyperspectral image segmentation model, providing a reliable basis for vitality judgment.

[0021] Furthermore, each hyperspectral data is preprocessed, including:

[0022] Each hyperspectral data set is processed with noise reduction, baseline drift correction, light scattering correction and data scaling.

[0023] Furthermore, the dual-branch spectral Transformer corn germplasm cross-germplasm viability detection model includes: a multi-scale CNN module, a Transformer module, a multi-level feature fusion attention module and a linear classification module, wherein the multi-scale CNN module is used to calculate the local features of each preprocessed hyperspectral data, the Transformer module is used to calculate the global features of each preprocessed hyperspectral data, the multi-level feature fusion attention module is used to: map local features and global features into multiple feature maps with consistent dimensions, and assign attention score weights to each feature map through the activation function layer, and fuse all feature maps based on the assigned attention score weights to obtain multi-level global-local fusion features; the linear classification module is used to: perform global average pooling and linear layer operations on the multi-level global-local fusion features in sequence to obtain vitality prediction results.

[0024] Furthermore, a hyperspectral image of each sample corn seed in the sample corn germplasm resource is obtained, and vitality label data of each sample corn seed in the sample corn germplasm resource is obtained, including:

[0025] A hyperspectral imaging device is used to obtain a hyperspectral image of each sample corn seed in the sample corn germplasm resources, and the vitality label data of each sample corn seed in the sample corn germplasm resources is obtained through a germination experiment.

[0026] 2) In a second aspect, the present invention further provides a system for distinguishing the vitality of corn germplasm resources across different germplasms. The specific technical solution is as follows:

[0027] It includes acquisition module, segmentation module, hyperspectral data acquisition module, preprocessing module, model training module and discrimination module;

[0028] The acquisition module is used to: acquire a hyperspectral image of each sample corn seed in the sample corn germplasm resource, and obtain vitality label data of each sample corn seed in the sample corn germplasm resource;

[0029] The segmentation module is used to perform threshold segmentation on the hyperspectral image of each sample corn seed to obtain a hyperspectral image of each sample corn seed after background removal, and perform regional segmentation on the hyperspectral image of each sample corn seed after background removal to determine a hyperspectral image of the embryo region and a hyperspectral image of the endosperm region;

[0030] The hyperspectral data acquisition module is used to obtain hyperspectral data of the embryo region and hyperspectral data of the endosperm region according to the hyperspectral image of the embryo region and the hyperspectral image of the endosperm region;

[0031] The hyperspectral data acquisition module is further used to: obtain hyperspectral data of each sample corn seed from the hyperspectral image of each sample corn seed after removing the background;

[0032] The preprocessing module is used to: preprocess each hyperspectral data;

[0033] The model training module is used to perform feature selection on each pre-processed hyperspectral data set and train the constructed dual-branch spectral Transformer corn germplasm cross-germplasm viability detection model based on the viability label data of each corn seed sample.

[0034] The discrimination module is used to: use the trained dual-branch spectral Transformer corn germplasm cross-germplasm vitality detection model to discriminate the cross-germplasm corn to be discriminated, and obtain the vitality prediction result of the cross-germplasm corn to be discriminated.

[0035] Based on the above solution, the corn germplasm resource vitality cross-germplasm discrimination system of the present invention can also be improved as follows.

[0036] Furthermore, the segmentation module is specifically used to:

[0037] Manually annotating the boundaries of the embryo region and the endosperm region in the hyperspectral images of multiple corn seed samples to obtain embryo annotated data and endosperm annotated data;

[0038] The constructed moving inverted bottleneck UNet germplasm hyperspectral image segmentation model was trained using hyperspectral images of multiple corn seed samples, combined with embryo and endosperm annotation data.

[0039] The trained moving inverted bottleneck UNet germplasm hyperspectral image segmentation model was used to perform regional segmentation on the hyperspectral image of each sample corn seed after background removal to determine the hyperspectral image of the embryo region and the hyperspectral image of the endosperm region.

[0040] Furthermore, each hyperspectral data is preprocessed, including:

[0041] Each hyperspectral data set is processed with noise reduction, baseline drift correction, light scattering correction and data scaling.

[0042] Furthermore, the dual-branch spectral Transformer corn germplasm cross-germplasm viability detection model includes: a multi-scale CNN module, a Transformer module, a multi-level feature fusion attention module and a linear classification module, wherein the multi-scale CNN module is used to calculate the local features of each preprocessed hyperspectral data, the Transformer module is used to calculate the global features of each preprocessed hyperspectral data, the multi-level feature fusion attention module is used to: map local features and global features into multiple feature maps with consistent dimensions, and assign attention score weights to each feature map through the activation function layer, and fuse all feature maps based on the assigned attention score weights to obtain multi-level global-local fusion features; the linear classification module is used to: perform global average pooling and linear layer operations on the multi-level global-local fusion features in sequence to obtain vitality prediction results.

[0043] Furthermore, the acquisition module is specifically used to:

[0044] A hyperspectral imaging device is used to obtain a hyperspectral image of each sample corn seed in the sample corn germplasm resources, and the vitality label data of each sample corn seed in the sample corn germplasm resources is obtained through a germination experiment.

[0045] 3) In a third aspect, the present invention further provides an electronic device, comprising a processor coupled to a memory, wherein the memory stores at least one computer program, and the at least one computer program is loaded and executed by the processor, so that the electronic device can implement any of the above-mentioned methods for cross-germplasm discrimination of corn germplasm vitality.

[0046] 4) In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned methods for cross-germplasm identification of the vitality of corn germplasm resources.

[0047] It should be noted that the beneficial effects achieved by the technical solutions of the second to fourth aspects of the present invention and the corresponding possible implementation methods can be found in the above-mentioned technical effects of the first aspect and its corresponding possible implementation methods, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments of the present invention:

[0049] Figure 1 This is a flow chart of a method for distinguishing the vitality of corn germplasm resources across germplasms according to an embodiment of the present invention;

[0050] Figure 2 Schematic diagram of the network structure of the mobile inverted bottleneck UNet germplasm hyperspectral image segmentation model;

[0051] Figure 3 This is a schematic diagram of the network structure of the dual-branch spectral Transformer model for detecting the viability of corn germplasm across germplasms;

[0052] Figure 4 This is the interface diagram of the detection software corresponding to the cross-germplasm discrimination method for the vitality of corn germplasm resources;

[0053] Figure 5 This is a second flow chart of a method for distinguishing the vitality of corn germplasm resources across germplasms according to an embodiment of the present invention;

[0054] Figure 6 This is a schematic structural diagram of a system for distinguishing the vitality of corn germplasm resources across germplasms according to an embodiment of the present invention;

[0055] Figure 7 The figure is a schematic structural diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0056] The principles and features of the present invention are described below. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0057] The following describes in detail the technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems using specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The following embodiments of the present invention are described in conjunction with the accompanying drawings.

[0058] like Figure 1 As shown, a method for determining the vitality of corn germplasm resources across germplasms according to an embodiment of the present invention includes the following steps:

[0059] S1. Acquire a hyperspectral image of each sample corn seed in the sample corn germplasm resources, and obtain vitality label data of each sample corn seed in the sample corn germplasm resources;

[0060] Among them, a hyperspectral imaging device is used to obtain hyperspectral images, and then a germination experiment is carried out to obtain the vitality label data of all samples. The vitality label data is divided into germination (vitality) and non-germination (no vitality).

[0061] S2. performing threshold segmentation on the hyperspectral image of each sample corn seed to obtain a hyperspectral image of each sample corn seed after background removal, and performing regional segmentation on the hyperspectral image of each sample corn seed after background removal to determine a hyperspectral image of an embryo region and a hyperspectral image of an endosperm region;

[0062] Among them, the hyperspectral image of each sample corn seed is threshold segmented to obtain the hyperspectral image of each sample corn seed after removing the background. The specific implementation process is as follows:

[0063] First, the corn seeds are separated from the background using threshold segmentation processing. Specifically, a single-channel image with the maximum spectral reflectance difference between the seeds and the background is selected from the hyperspectral full-channel image, and an appropriate threshold is set for the single-channel image. Pixels with reflectance values ​​less than the threshold are regarded as background and represented by 0, while pixels with reflectance values ​​greater than the appropriate threshold constitute a corn seed mask. The corn seed mask is applied to the hyperspectral full-channel image to remove the background, eliminate noise and separate the seed area, obtaining a hyperspectral image of the seed area, that is, a hyperspectral image of each sample corn seed after removing the background.

[0064] The hyperspectral image of each corn seed sample after background removal is segmented to determine the hyperspectral image of the embryo region and the hyperspectral image of the endosperm region, including:

[0065] Manually annotating the boundaries of the embryo region and the endosperm region in the hyperspectral images of multiple corn seed samples to obtain embryo annotated data and endosperm annotated data;

[0066] The constructed moving inverted bottleneck UNet germplasm hyperspectral image segmentation model was trained using hyperspectral images of multiple corn seed samples, combined with embryo and endosperm annotation data.

[0067] The trained moving inverted bottleneck UNet germplasm hyperspectral image segmentation model was used to perform regional segmentation on the hyperspectral image of each sample corn seed after background removal to determine the hyperspectral image of the embryo region and the hyperspectral image of the endosperm region.

[0068] like Figure 2As shown in the figure, the moving inverted bottleneck UNet germplasm hyperspectral image segmentation model is obtained by improving UNet, which is obtained by adding a bottleneck module, an initial block and a moving flip bottleneck convolution module to UNet. UNet can be specifically MBUNet. The hyperspectral image of the sample corn seed after background removal is processed by an initial block, and the output of the initial block is used as the input of the first moving flip bottleneck convolution module, and the output of the initial block is downsampled using the convolution layer to obtain a first downsampling result. The first downsampling result is processed by 4 bottleneck modules in sequence to obtain a first processing result. The first processing result is used as the input of the second moving flip bottleneck convolution module, and the convolution layer is used to downsample the first processing result to obtain a second downsampling result. The second downsampling result is processed by 7 bottleneck modules in sequence to obtain a second processing result. The second processing result is used as the input of the third moving flip bottleneck convolution module, and the convolution layer is used to downsample the second processing result. A downsampling operation is performed to obtain a third downsampling result. The third downsampling result is processed by seven bottleneck modules in sequence to obtain a fourth processing result. The fourth processing result and the output of the third moving flip bottleneck convolution module are upsampled using a convolution layer to obtain a first upsampling result. The first upsampling result is processed by two bottleneck modules in sequence to obtain a fifth processing result. The fifth processing result and the output of the second moving flip bottleneck convolution module are upsampled using a convolution layer to obtain a second upsampling result. The second upsampling result is processed by two bottleneck modules in sequence to obtain a sixth processing result. The sixth processing result and the output of the first moving flip bottleneck convolution module are upsampled using a convolution layer to obtain a third upsampling result. The third upsampling result is processed by one bottleneck module in sequence to obtain a seventh processing result. The seventh processing result is operated on by a conversion convolution layer to obtain a hyperspectral image of the embryo region and a hyperspectral image of the endosperm region.

[0069] The initial block performs convolution operations (specifically, this can be achieved through a 3×3 convolution layer) and maximum pooling operations on the input, and concatenates the two operation results to obtain the output of the initial block.

[0070] Among them, the moving flip bottleneck convolution module performs convolution operation (specifically, it can be implemented by a 1×1 convolution layer), separable convolution operation (specifically, it can be implemented by a 1×1 separable convolution layer), squeeze excitation module, and zero layer (specifically, it can be implemented by a 1×1 convolution layer) on the input in sequence, and then adds the processing result to the input of the moving flip bottleneck convolution module element by element to obtain the output of the moving flip bottleneck convolution module.

[0071] Among them, after the bottleneck module performs convolution operation (specifically, it can be achieved through a 1×1 convolution layer) and asymmetric operation on the input, the result is divided into two paths for processing. In the first path, the result is processed by a 3×1 convolution layer and ReLU activation function, and a 1×3 convolution layer. In the second path, the result is maximum pooled, and the output of the first path and the output of the second path are processed in turn through "ReLU activation function and 1×1 convolution layer", "1×1 convolution layer, ReLU activation function and zeroing layer", and the processing result is added element by element to the input of the bottleneck module, and the BN+ReLU activation function is used to obtain the output of the bottleneck module.

[0072] Among them, the bottleneck module first uses point-by-point convolution to reduce the dimension of the input features, and then restores the original feature dimension, thereby significantly reducing the number of parameters; the initial block reduces the dimension of the input image and extracts basic features, converting high-resolution features into low-dimensional feature maps to reduce the computational complexity of the network; the jump connection transfers the feature map of the encoder to the corresponding layer of the decoder, realizing the fusion of high-resolution spatial information and low-resolution semantic information, and enhancing the network's ability to simultaneously capture local details and global information; MBConv is used to reduce the spatial information loss caused by jump connections, thereby improving the overall segmentation performance.

[0073] During the training of the constructed mobile inverted bottleneck UNet germplasm hyperspectral image segmentation model, seed images and segmentation label data (i.e., embryo and endosperm annotation data) from the segmentation dataset were input into the model to obtain prediction results. A loss function was used to calculate the loss between the segmentation results and the true annotations. The parameter gradients were calculated and updated through backpropagation in the optimizer. Hyperparameters were adjusted in a timely manner based on performance to improve the model's generalization ability and segmentation accuracy. Finally, a segmentation result with a small loss value was obtained. A hyperspectral image of the embryo region was obtained through mapping, and a hyperspectral image of the endosperm region was obtained by subtracting the embryo region image.

[0074] S3. obtaining hyperspectral data of the embryo region and hyperspectral data of the endosperm region based on the hyperspectral image of the embryo region and the hyperspectral image of the endosperm region;

[0075] The average reflectance of the hyperspectral image of the embryo region is extracted as the hyperspectral data of the embryo region, and the average reflectance of the hyperspectral image of the endosperm region is extracted as the hyperspectral data of the endosperm region.

[0076] S4. Obtaining hyperspectral data of each sample corn seed from the hyperspectral image of each sample corn seed after background removal.

[0077] The average reflectance of the hyperspectral image of the sample corn seeds is extracted as the hyperspectral data of the sample corn seeds.

[0078] Among them, the hyperspectral data of the embryo region, the hyperspectral data of the endosperm region and the hyperspectral data of the whole-grain germplasm of each corn sample corn seed are obtained, and together with the corresponding vitality label data, they constitute the data set of the dual-branch spectral Transformer corn germplasm cross-germplasm vitality detection model; among them, the hyperspectral image of the seed region constitutes the whole-grain germplasm hyperspectral data, the segmented hyperspectral image of the embryo part constitutes the embryo region hyperspectral data, and the segmented hyperspectral image of the endosperm part constitutes the endosperm region hyperspectral data.

[0079] The hyperspectral data are divided into three hyperspectral data sets X = {X0, X1, X2}∈B: whole grain germplasm, embryo region and endosperm region. R×C , where X0 is the hyperspectral dataset of whole grain germplasm, X1 is the hyperspectral dataset of embryo region germplasm, X2 is the hyperspectral dataset of endosperm region germplasm, B is the number of batches, and C is the number of hyperspectral data bands.

[0080] S5. Preprocess each hyperspectral data set. Specifically, perform noise reduction, baseline drift correction, light scattering correction, and data scaling on each hyperspectral data set. Each hyperspectral data set includes: hyperspectral data of the embryo region of each corn seed sample, hyperspectral data of the endosperm region of each corn seed sample, and hyperspectral data of each corn seed sample.

[0081] Data preprocessing is divided into four categories from the perspectives of noise reduction, baseline drift correction, light scattering correction and data scaling, which respectively target the noise, baseline drift, light scattering factors in hyperspectral images and the risk of overfitting of models at different data scales. Each preprocessing category contains multiple corresponding preprocessing methods. For example, the noise reduction preprocessing category includes moving average smoothing, SG filtering, wavelet transform and other preprocessing methods. Single preprocessing and combined preprocessing methods are used. The combined preprocessing method is to select two or more preprocessing categories and only select one preprocessing method under each preprocessing category.

[0082] S6. Perform feature selection on each pre-processed hyperspectral data set, and train the constructed dual-branch spectral Transformer corn germplasm cross-germplasm vitality detection model based on the vitality label data of each sample corn seed;

[0083] like Figure 3As shown in the figure, the constructed dual-branch spectral Transformer corn germplasm cross-germplasm viability detection model includes: a multi-scale CNN module (also called a multi-scale convolution module), a Transformer module, a multi-level feature fusion attention module and a linear classification module, wherein the multi-scale CNN module is used to calculate the local features of each preprocessed hyperspectral data, the Transformer module is used to calculate the global features of each preprocessed hyperspectral data, the multi-level feature fusion attention module is used to: map local features and global features into multiple feature maps with consistent dimensions, and assign attention score weights to each feature map through the activation function layer, and fuse all feature maps based on the assigned attention score weights to obtain multi-level global-local fusion features; the linear classification module is used to: perform global average pooling and linear layer operations on the multi-level global-local fusion features in sequence to obtain vitality prediction results.

[0084] The multi-scale CNN module is used to calculate the local features of (the pre-processed hyperspectral data of the embryo region of each sample corn seed, the pre-processed hyperspectral data of the endosperm region, and the pre-processed hyperspectral data of each sample corn seed). The Transformer module calculates the global features of (the pre-processed hyperspectral data of the embryo region of each sample corn seed, the pre-processed hyperspectral data of the endosperm region, and the pre-processed hyperspectral data of each sample corn seed). The multi-level feature fusion attention module maps local features (through a two-dimensional convolutional layer) and global features (linear layer) into feature maps with consistent dimensions (local features correspond to three feature maps, and global features correspond to three feature maps). The activation function layer of the multi-level feature fusion attention module assigns fusion feature attention score weights to the six feature maps. Based on the assigned weights, the six feature maps are fused to obtain multi-level global-local fusion features. The linear classification module performs global average pooling and linear layer operations on the multi-level global-local fusion features in sequence to obtain the germplasm vitality detection results. Specifically:

[0085] 1) The multi-scale CNN module contains one convolutional neural network with kernel sizes of 1×1, 1×3, 1×5, and 1×7. Each convolutional neural network contains a one-dimensional convolution layer, a batch normalization layer (BN), and a Gaussian error linear unit (GELU); the Transformer module contains two layer normalization layers (LN), a multi-head self-attention mechanism layer (MHSA), and a multi-layer perceptron (MLP); the multi-level feature fusion attention module contains a two-dimensional convolution layer, a layer normalization layer, and a normalized exponential function (Softmax);

[0086] The multi-scale CNN module is used to calculate local features. It processes the input corn seed hyperspectral image in parallel using convolution kernels of different sizes. Larger convolution kernels can capture global features in the seed image, while smaller convolution kernels focus on local details of the image. This enriches the dimension of feature information, enabling the model to learn more comprehensive and detailed seed features, more effectively mining vitality-related information contained in the seed image, and ultimately obtaining a local feature map.

[0087] 2) The Transformer module is used to calculate global features. Through its self-attention mechanism, it can adapt to the diversity of corn seed data of different varieties, sizes, and shapes. By calculating the correlation between any two pixels in the image, it comprehensively analyzes the spectral characteristics of different seed parts and discovers potential patterns related to vitality. The MLP in the Transformer module adjusts the feature dimensions based on the characteristics of the task, converting the features output by the self-attention module into an expression more suitable for the specific task, ultimately obtaining a global feature map.

[0088] 3) The multi-level feature fusion attention module is used to effectively integrate local features and global features. The local feature map is converted into the first feature map Q1, the second feature map K1 and the third feature map V1 with consistent dimensions through two-dimensional convolution. The global feature map is converted into the fourth feature map Q2, the fifth feature map K2 and the sixth feature map V2 with consistent dimensions through layer normalization. Furthermore, the first feature map Q1 and the fourth feature map Q2 are spliced ​​to obtain the seventh feature map, the second feature map K1 and the fifth feature map K2 are spliced ​​to obtain the eighth feature map, and the third feature map V1 and the sixth feature map V2 are spliced ​​to obtain the ninth feature map. The eighth feature map and the ninth feature map are then multiplied element by element to obtain the representation fusion information matrix, which is calculated as the feature attention score weight matrix by the activation function and linearly transformed with the seventh feature map to finally obtain the multi-level global and local fusion features;

[0089] 4) The linear classification module obtains the germplasm vitality detection results through the global average pooling layer and the linear layer.

[0090] Optionally, the trained dual-branch spectral Transformer corn germplasm cross-germplasm viability detection model is used to discriminate the cross-germplasm corn to be identified, and the vitality prediction result of the cross-germplasm corn to be identified is obtained.

[0091] S7. Use the trained dual-branch spectral Transformer corn germplasm cross-germplasm vitality detection model to discriminate the cross-germplasm corn to obtain the vitality prediction result of the cross-germplasm corn to be discriminated.

[0092] Optionally, in S1, a hyperspectral image of each sample corn seed in the sample corn germplasm resource is obtained, and vitality label data of each sample corn seed in the sample corn germplasm resource is obtained, including:

[0093] A hyperspectral imaging device is used to obtain a hyperspectral image of each sample corn seed in the sample corn germplasm resources, and the vitality label data of each sample corn seed in the sample corn germplasm resources is obtained through a germination experiment.

[0094] The present invention constructs the vitality discrimination model of the endosperm region, embryo region and the whole seed region based on the neural network model, namely the trained dual-branch spectral Transformer corn germplasm cross-germplasm vitality detection model. For the convenience of users, the present invention designs an interactive interface, such as Figure 4 This interface loads a pre-trained viability discrimination model, namely the pre-trained dual-branch spectral Transformer corn germplasm cross-germplasm viability detection model. The user then inputs hyperspectral image data of the corn seeds to be tested. The interface automatically invokes data reading, image segmentation algorithms, spectral data preprocessing algorithms, and the viability discrimination model to determine seed viability, output the results, and allow the user to save the data. The interface also provides visualization capabilities, generating spectral curves for different seed parts, helping users intuitively understand the spectral characteristics of the seeds and providing a reference for further analysis.

[0095] The seed different parts segmentation algorithm of the present invention, namely the moving inverted bottleneck UNet germplasm hyperspectral image segmentation model, has been deeply improved on the basis of the U-Net network architecture. By introducing the attention mechanism, the model can focus more on the key features of different parts of the seeds, enhance the recognition ability of subtle differences, and significantly improve the segmentation accuracy. At the same time, the cross-germplasm discrimination algorithm designed by the present invention improves the discrimination reliability through regional association analysis, makes full use of the characteristic information of different parts of the seeds after segmentation, and constructs a more comprehensive and effective feature vector. Through training based on big data, the algorithm can accurately capture the characteristic patterns of seeds of different varieties under vitality states, effectively solving the problems of low accuracy and poor stability of traditional cross-germplasm algorithms. In practical applications, by comparing the characteristics of the endosperm region and the entire seed region after segmentation, the present invention can more deeply explore the intrinsic information of the vitality of corn germplasm resources, significantly improve the accuracy and reliability of cross-germplasm corn germplasm vitality discrimination, and provide strong technical support for seed quality detection in agricultural production.

[0096] The cross-germplasm corn germplasm resource hyperspectral vitality detection software of the present invention has the following specific implementation process: uploading hyperspectral data in the data upload window, and selecting .hdr files and .dat files in sequence according to the prompt box; then selecting one of the entire seed area, embryo area, and endosperm area in the area division window, and the hyperspectral spectrum of the corresponding area is displayed in the middle graphic window; selecting one of the entire seed area, embryo area, and endosperm area in the data analysis window and clicking below to start analysis, then the middle table window displays the seed serial number and the presence or absence of vitality prediction results, and the result display window displays the average vitality of the current batch of seeds; clicking Save Results in the result save window, then saving the result data in the table as an Excel file to the corresponding path.

[0097] The structure of the cross-germplasm corn germplasm resource hyperspectral vitality detection software of the present invention is as follows:

[0098] Define the UI_windows class to encapsulate the settings and layout related to the user interface, define the UI_setup() instance method to receive the Main_windows parameter, which sets the basic properties such as the main window name and main window size, then sets the icon file in the specified path as the main window icon, and uses the setStyleSheet method to set the background color of the main window button in the normal, mouse hover and mouse press states, and sets the main window icon size; create five QGroupBox objects under the central window to set the size, position layout and style of the relevant button components and display them in groups, create a QGraphicsView graphics view object to set the size and position layout for displaying two-dimensional spectral images, create a QTableWidget table control object to set the size and position layout for displaying table data; the window multiple control signals are connected to the slot function through the connectSlotsByName function to execute the response signal, then the file selection function can be triggered by pressing button 1, the data generation function can be triggered by pressing buttons 2-4, the corresponding seed analysis area can be selected by pressing buttons 5-7, the data analysis function can be triggered by pressing button 8, and the result saving function can be triggered by pressing button 9;

[0099] Define the retranslateUi function to update the control content on the interface, use the setTitle method to assign group box titles to the five QGroupBox objects, and the titles are upload data, regional division, analysis data, result saving and result display respectively. Use the setText method to set the text of the button under the upload data group box, that is, button 1, to hyperspectral data, set the text of the three buttons under the regional division group box, that is, buttons 2-4, to all seed areas, embryo areas and endosperm areas respectively, set the text of the three buttons under the analysis data group box, that is, buttons 5-7, to all seed areas, embryo areas and endosperm areas respectively, and set the text of the button 8 below to start analysis, set the text of the button under the result saving group box, that is, button 9, to save results, and set the text of the label control under the result display group box to average vitality;

[0100] A custom upload_check function is used to detect whether the hyperspectral data button in the upload data group box is selected by the user. If selected, the open_filedialog() function is executed to open a message prompt box to prompt the user to upload data and store the uploaded hyperspectral image; then the Data_make() function is called to obtain three hyperspectral datasets of the embryo region, endosperm region, and whole grain germplasm.

[0101] In the upload_check function, the custom open_filedialog() function is used to open two file selection dialog boxes, save the selected file path in the instance attribute of the class, and display "Hyperspectral data submission completed" in a message prompt box. Specifically, the QFileDialog class is first used to create a file selection dialog box, and then the getOpenFileName method is used to display the first file selection dialog box. Specifically, the dialog box title is set to "Submit hdr file" to prompt the user to upload the file format, set the file filter to limit the file type that the user can select to files with the .hdr suffix, and finally store the path of the .hdr file selected by the user in the instance attribute of the class for subsequent calls; then the getOpenFileName method is used to display the second file selection dialog box, specifically, the dialog box title is set to "Submit dat file" to prompt the user to upload Transfer the file format, set the file filter to limit the file types that the user can select to files with the .dat suffix, and finally store the path of the .dat file selected by the user in the instance attribute 2 of the class for subsequent calls; then create an instance of QMessageBox msg_box to create a message box, and set the icon type and text content of the message box to "Data submission completed", and set an "OK" button in the message box. The user clicks the button to close the message box. Finally, use the exec_ method to display the message box and enter a modal loop, allowing the user to interact with the message box. That is, the program will execute subsequent code only after clicking the "OK" button, otherwise the program will pause.

[0102] In the upload_check function, a custom Data_make() function is used to use the sepctral module to read the .hdr header file stored in the instance attribute 1 of the class and the .dat data file stored in the instance attribute 2 of the class. The wavelength data in the .csv format under the data path is opened, and a specific band is selected for the hyperspectral image to form a single-band image. Then, the single-band image is thresholded to remove the background and obtain the image of the seed area. Specifically, the binarization threshold is manually set, and the pixels with pixel values ​​greater than the threshold in the single-band image are retained and the pixels with values ​​less than the threshold are set to 0 to obtain the binarized single-band image. The connected area component is used to segment the pixel area with pixel values ​​greater than 0 in the binarized single-band image. An image contains multiple seed areas, so multiple segmented seed area masks are formed, and they are grouped and sorted according to the X and Y axis coordinate values. The sorting sequence matches each seed number. After sorting, multiple seed area lists are formed. A blank array of the same shape as the original hyperspectral image is created and the hyperspectral data of the corresponding area are extracted into the blank array in turn to form each seed area hyperspectral image. Next, we perform image segmentation on the hyperspectral image of each seed region to obtain three hyperspectral datasets: the embryo region, the endosperm region, and the whole-grain germplasm region. The whole-grain germplasm hyperspectral dataset is obtained as follows: call the Ref_read() function with the file name parameter whole_seed, and use a loop statement to read the whole-grain germplasm reflectance into the data processing path, naming it whole_seed.csv. The method for obtaining the hyperspectral dataset of the endosperm region is as follows: the trained MBUNet model is called to segment the embryo and endosperm parts of the seed region hyperspectral image, and the hyperspectral images of the embryo and endosperm parts are extracted from the original hyperspectral image according to the segmentation mask of the corresponding part output by the model. Specifically, a mask of the same size as the original hyperspectral image is created for each part. The mask takes the value of 1 at the pixel position belonging to the category and the value of 0 at other positions. Then, the mask is multiplied element by element with the original hyperspectral image to obtain the hyperspectral image of the part; finally, the Ref_read() function is called with the given file name parameters as embryo_seed and endosperm_seed, and the reflectance of the embryo region and endosperm region are read in the data processing path using a loop statement, and named as: embryo_seed.csv and endosperm_seed.csv respectively.

[0103] Finally, the Data_make() function creates an instance of QMessageBox, msg_box, to create a message box, and sets the icon type and text content of the message box to "All seed area data sets have been produced", and sets an "OK" button in the message box. The user clicks this button to close the message box. Finally, the exec_ method is used to display the message box and enter a modal loop, allowing the user to interact with the message box. That is, the program will execute subsequent code only after clicking the "OK" button, otherwise the program will pause.

[0104] In the custom Data_make() function of the upload_check function, the Ref_read() function accepts a file name parameter representing the reflectance data, and uses a loop statement to read the reflectance of each seed. Specifically, the empty list Ref_list is initialized to store the reflectance average of each band in the seed area, and the outer for loop is used to traverse the coordinate list of each seed area. Then, each channel of the hyperspectral image under the coordinates of the seed area is traversed to create an empty list ref_sums. The inner loop traverses each seed area coordinate and calculates the pixel value and adds it to ref_sums. The calculated average value of ref_sums is the average reflectance of the seed in the band. The ref_sums obtained in each loop are put into the Ref_list list. Finally, the loop obtains the average reflectance data of each seed in the hyperspectral image in all bands. The wavelength data and average reflectance data are given the headers "wavelength" and "reflectance" respectively and saved in the data processing path with the specified name in .csv format.

[0105] A custom region_check function is used to detect whether the three buttons of all seed regions, embryo regions, and endosperm regions under the region division group box are selected by the user. If selected, the data_processing function is executed accordingly.

[0106] Among them, the custom data_processing function is used to load data and draw hyperspectral images. Specifically, the seed.csv file under the data processing path is loaded, the wavelength column and reflectance column are read, the matplotlib library is called to draw the reflectance curve graph, the drawing size, font, coordinate axis, line, etc. are reasonably set to present the drawing effect beautifully, the width and height of the QGraphicsView graphics view window are obtained as the target width and height of the image scaling, the image is adjusted to adapt to the size of the QGraphicsView graphics view and displayed in the view.

[0107] Among them, the custom analysis_check function is used to detect whether the three buttons of all seed areas, embryo areas and endosperm areas under the analysis data group box are selected by the user. If one of the buttons is selected, the reflectivity of the selected seed area is recorded as the current reflectivity. When button 8 is pressed, the current reflectivity data parameter is passed in to execute the data_analysis function.

[0108] The data_analysis function accepts and preprocesses the incoming reflectance data, then uses the torch.load function to load the model. This model is then called and passed the preprocessed reflectance column data to obtain the viability prediction results for the maize germplasm resource. Three columns are added to the QTableWidget table control window: the first column is the test sequence number, the second column is the seed numbering, which starts at sequence 1, and the third column is the viability determination result. The average viability of the current seed batch is calculated as the number of viable seeds divided by the total number of seeds. The result is printed under the Result Display group box.

[0109] Among them, the custom result_check function is used to detect whether the result save button under the area division group box is selected by the user. If it is selected, the result_save function is executed accordingly.

[0110] Among them, the result_save function obtains the data in the QTableWidget table control window row by row and column by column into DataFrame, and saves the DataFrame as an Excel file in the corresponding data result path.

[0111] The present invention is further described by the following examples:

[0112] like Figure 5 As shown, the method for distinguishing the vitality of corn germplasm resources across germplasms in this embodiment includes: data collection, image segmentation, data preprocessing, model training and prediction, specifically:

[0113] A hyperspectral dataset was obtained, including 600 hyperspectral images and their reflectance data. A germination test was performed on each kernel of corn to obtain the corresponding vitality label set. The hyperspectral images were obtained by preparing a hyperspectral device equipped with a sample plate with grooves for placing samples, and the size and shape of the grooves were designed to be suitable for corn seeds. Each corn seed was placed with the embryo side facing up in the corresponding groove of the sample plate to facilitate subsequent batch segmentation of the embryo and endosperm areas of the image. The hyperspectral device was started to scan the sample, and a hyperspectral image was obtained after each scan, ultimately obtaining a hyperspectral image set.

[0114] The optimal hyperspectral image segmentation model and model parameters are determined by the following steps:

[0115] S101. Manually annotate the germplasm embryo region of 50 hyperspectral images in the hyperspectral dataset to form an image segmentation label set, which has three labels: the entire germplasm region, the embryo region, and the endosperm region; and form three image segmentation hyperspectral datasets of the embryo region, the endosperm region, and the whole grain germplasm region and their corresponding region labels.

[0116] S102. Create training and testing datasets based on the image segmentation labels of S101 and three image segmentation hyperspectral datasets.

[0117] S103. Based on the structure and features of the training dataset produced in step S102, image segmentation operations are performed on the hyperspectral image, including image threshold segmentation to separate corn seeds from the background and constructing a moving inverted bottleneck UNet (MBUNet) germplasm hyperspectral image segmentation model.

[0118] S104. Experiment and evaluate MBUNet and the commonly used Enet, UNet, UNet++, DeeplabV3+, and SegFormer image segmentation models respectively to obtain the best model MBUNet and its optimal model parameters, and complete S1 to obtain the embryo region segmentation of the remaining 550 hyperspectral images in the hyperspectral dataset, as follows: first, use threshold segmentation processing to separate corn seeds from the background, and then construct a moving inverted bottleneck UNet (MBUNet) germplasm hyperspectral image segmentation model to perform embryo and endosperm region segmentation on the seed region hyperspectral images to obtain three hyperspectral image datasets of embryo region, endosperm region and whole grain germplasm region, and calculate the average spectrum of the three regions to form three hyperspectral datasets of embryo region, endosperm region and whole grain germplasm region.

[0119] S105. Based on the three hyperspectral data sets of the embryo region, endosperm region, and whole grain germplasm obtained in S104, preprocess the hyperspectral data in the data sets respectively.

[0120] Among them, data preprocessing is divided into four preprocessing categories from the four perspectives of noise reduction, baseline drift correction, light scattering correction and data scaling, which are respectively aimed at the noise, baseline drift, light scattering factors in the hyperspectral spectrum and the risk of overfitting of models at different data scales; each preprocessing category contains multiple corresponding preprocessing methods. For example, the noise reduction preprocessing category includes moving average smoothing, SG filtering, wavelet transform and other preprocessing methods. Single preprocessing and combined preprocessing methods are adopted. Among them, the combined preprocessing method is to select two or more preprocessing categories and only select one preprocessing method under each preprocessing category.

[0121] S106. Based on the dataset after pre-processing and band selection in step S105 and its corresponding vitality label set, a training dataset and a test dataset are generated.

[0122] S107. Based on the structure and characteristics of the training data set produced in step S106, a dual-branch spectral Transformer corn germplasm cross-germplasm vitality detection model is constructed.

[0123] S108. Test and evaluate the dual-branch spectral Transformer model for detecting cross-germplasm viability of corn germplasm.

[0124] The vitality label set in S101 is a label of vitality or non-vitality assigned to each sample.

[0125] In S107, the dual-branch spectral Transformer corn germplasm cross-germplasm viability detection model includes a multi-scale CNN module, a Transformer module, a multi-level feature fusion attention module and a linear classification module. The multi-scale CNN module contains one convolutional neural network with convolution kernel sizes of 1×1, 1×3, 1×5 and 1×7, each of which contains a one-dimensional convolution layer, a batch normalization layer (BN) and a Gaussian error linear unit (GELU); the Transformer module contains two layer normalization layers, a multi-head self-attention mechanism layer and a multi-layer perceptron; the multi-level feature fusion attention module contains a two-dimensional convolution layer, a layer normalization layer and a normalized exponential function (Softmax). The hyperspectral dataset X∈R B×C Input 1×3, 1×5 and 1×7 convolutional neural networks respectively to obtain three different feature maps: X l1 ∈R B×C , X l2 ∈R B×C and X l3 ∈R B×C , the three feature maps are fused with X and input into the 1×1 convolutional neural network to obtain the local feature map X1∈R B×C , where C is the feature map dimension; input X into the Transformer module to obtain the global feature map: Among them, C2 is the global feature map dimension; {X1, X2} is input into the multi-level feature fusion attention module to obtain the multi-level global local fusion feature F∈R B×C ; F input linear classification module to obtain germplasm vitality detection result S∈R B×2 , where 2 is the corresponding label for vitality or not.

[0126] In S103, the moving inverted bottleneck UNet (MBUNet) germplasm hyperspectral image segmentation model includes the UNet main architecture, bottleneck module, initial block, skip connection and moving inverted bottleneck convolution (MBConv); the specific steps are:

[0127] S1031, the bottleneck module first uses point-by-point convolution to reduce the dimension of the input features, and then restores the original feature dimension, thereby significantly reducing the number of parameters.

[0128] S1032, the initial block performs dimensionality reduction and basic feature extraction on the input image, converting high-resolution features into low-dimensional feature maps to reduce the computational complexity of the network.

[0129] S1033, skip connection transfers the encoder's feature map to the corresponding layer of the decoder, achieving the fusion of high-resolution spatial information and low-resolution semantic information, and enhancing the network's ability to capture both local details and global information;

[0130] S1034 and MBConv are used to reduce the spatial information loss caused by skip connections, thereby improving the overall performance.

[0131] Optionally, in the above technical solution, MBUNet is experimented and evaluated in S104 to obtain the optimal model parameters, and the embryo region segmentation of the remaining 550 hyperspectral images in the hyperspectral dataset obtained in S101 is completed to form two hyperspectral datasets: embryo region, endosperm region, and whole grain germplasm, as follows:

[0132] S1041. Conduct a test on the mobile inverted bottleneck UNet (MBUNet) germplasm hyperspectral image segmentation model constructed in S103 to verify the detection performance of the model.

[0133] Furthermore, the mobile inverted bottleneck UNet (MBUNet) germplasm hyperspectral image segmentation model was experimented and evaluated as follows:

[0134] S1042. Experiments are conducted on the datasets in S101 to S102, and the mean intersection over union (mIoU) and test accuracy are used to comprehensively evaluate the model's hyperspectral image segmentation performance.

[0135] S1043. Experiments and evaluations were conducted using MBUNet and existing commonly used image segmentation algorithms: ENet, UNet, UNet++, DeeplabV3+, and SegFormer. Experiments were also conducted on the datasets from S101 to S102. Comparative analysis was performed using mean intersection over union (mIoU) and test accuracy from S1042. This provided a basis for a comprehensive evaluation of the performance of the Moving Inverted Bottleneck UNet (MBUNet) germplasm hyperspectral image segmentation model. The test results are shown in Table 1.

[0136] Table 1:

[0137] Model mIoU (%) Accuracy(%) ENet 93.93±0.08 99.53±0.01 UNet 94.25±0.02 99.56±0.01 UNet++ 94.14±0.07 99.55±0.01 DeeplabV3+ 94.11±0.02 99.55±0.02 SegFormer 94.08±0.05 99.54±0.01 MBUNet 94.31±0.05 99.57±0.01

[0138] Optionally, in S105, preprocessing and feature selection are performed on the hyperspectral data in the dataset, specifically:

[0139] S1051. Perform multivariate scattering correction, SG filtering, normalization, and second-order derivative processing on the hyperspectral data channel by channel to obtain preprocessed hyperspectral data.

[0140] S1052: Select all characteristic bands of preprocessed hyperspectral data for modeling.

[0141] Optionally, the dual-branch spectral Transformer corn germplasm cross-germplasm viability detection model in S107 includes a multi-scale CNN module, a Transformer module, a multi-level feature fusion attention module, and a linear classification module. The multi-scale CNN module includes one convolutional neural network with convolution kernel sizes of 1×1, 1×3, 1×5, and 1×7, each of which includes a one-dimensional convolution layer, a batch normalization layer (BN), and a Gaussian error linear unit (GELU); the Transformer module includes two layer normalization layers (LN), a multi-head self-attention mechanism layer (MHSA), and a multi-layer perceptron (MLP); the multi-level feature fusion attention module includes a two-dimensional convolution layer, a layer normalization layer, and a normalized exponential function (Softmax); the specific steps are:

[0142] S1071. The process of calculating local features using the multi-scale CNN module is represented by the following formula:

[0143] X1=GELU(BN(Conv1d(X+X l1 +X l2 +X l3 ))

[0144] X l1 =GELU(BN(Conv1d 1×3 (X)

[0145] X l2 =GELU(BN(Conv1d 1×5 (X)

[0146] X l3 =GELU(BN(Conv1d 1×7 (X)

[0147]

[0148] Among them, x and θ are the input features of the BN layer and the GELU activation function respectively, μ and σ are the mean and variance of x, γ and β are the learnable parameters in the training process, GELU is used to perform Gaussian error linear calculation, BN is used to perform batch normalization operation, Conv1d is used to perform one-dimensional convolution operation with a convolution kernel size of 1×1, and Conv1d is used to perform one-dimensional convolution operation with a convolution kernel size of 1×1. 1×3 To perform a one-dimensional convolution operation with a convolution kernel size of 1×3, Conv1d 1×5 To perform a one-dimensional convolution operation with a convolution kernel size of 1×5, Conv1d 1×7 To perform a one-dimensional convolution operation with a convolution kernel size of 1×7, Tanh performs a hyperbolic tangent operation.

[0149] S1072. The process of calculating global features using the Transformer module is represented by the following formula:

[0150] X2=MLP(LN((y)))+y

[0151] y=X+MHSA(LN(X))

[0152] MHSA=Concat(SA i (y)),i∈{1,2,...,h}

[0153]

[0154] MLP(y)=Linear(GELU(Linear(y)))

[0155] Linear(y)=W×y+B

[0156] Among them, y is the intermediate output feature, h is the number of heads of the multi-head self-attention mechanism, Q, K and V are weight matrices, d k is the dimension of K, LN is the execution layer normalization operation, SA is the self-attention mechanism, Concat is the splicing operation, Linear is the linear layer, W and B are the weight matrix and parameters respectively.

[0157] S1073, use two-dimensional convolution and LN to process X1 and X2 respectively, and convert them into feature maps Z1 and Z2 with consistent dimensions respectively. The specific process is represented by the following formula:

[0158] Z1=Conv2d(X1)∈R B×K

[0159] Z2=Conv2d(X2)∈R B×K

[0160] Among them, K is the feature map dimension.

[0161] S1074, map Z1 and Z2 to Q1, K1 and V1 vectors containing local information and Q2, K2 and V2 vectors containing global information respectively, and then obtain the weight matrix Q representing the fusion information by concatenating (Q1, Q2), (K1, K2) and (V1, V2) Fus , K Fus and V Fus , the specific process is represented by the following formula:

[0162] Q i ,K i ,V i =Linear(Z i ),i∈{1,2}

[0163] Q Fus =Concat(Q1,Q2)

[0164] K Fus =Concat(K1,K2)

[0165] V Fus =Concat(V1,V2)

[0166] Among them, Concat is a concatenation operation.

[0167] S1075, Q Fus With K Fus Multiply to obtain the fusion feature attention score, the score weight is assigned by the activation function layer and multiplied by V Fus Get the multi-level global and local fusion features F∈R B×K , the specific process is represented by the following formula:

[0168] F=Softmax(Q Fus ,K Fus )V Fus

[0169] Among them, Softmax is the activation function operation.

[0170] S1076, input the multi-level global and local fusion features F into the linear classification layer to obtain the germplasm vitality detection result S∈R B×2 , which is specifically achieved through the following formula:

[0171] S=GAP(Linear(F))

[0172] Among them, GAP is the global average pooling operation, and Linear is the linear layer operation.

[0173] Optionally, in S108, a dual-branch spectral Transformer corn germplasm cross-germplasm viability detection model is experimented and evaluated; specifically, as follows:

[0174] S1081. Conducting a test experiment on the constructed dual-branch spectral Transformer corn germplasm cross-germplasm viability detection model to verify the detection performance of the model. In other words, conducting experiments and evaluating the dual-branch spectral Transformer corn germplasm cross-germplasm viability detection model is as follows:

[0175] S1082. Experiments were conducted on a constructed dataset of embryo, endosperm, and whole-grain germplasm regions, using accuracy, F1-score, and area under the receiver operating characteristic (ROC) curve (AUC) to comprehensively evaluate the performance of the model for viability detection.

[0176] S1083: Experiments were conducted using the dual-branch spectral Transformer model for detecting viability across maize germplasm regions, including embryo, endosperm, and whole-kernel regions. Comparative analysis was performed using the accuracy, F1-score, and area under the receiver operating characteristic (ROC) curve (AUC) from S1082 to provide a basis for comprehensive performance evaluation of the dual-branch spectral Transformer model for detecting viability across maize germplasm regions. The test results are shown in Tables 2 through 4. Table 2 shows the accuracy comparison results, Table 3 shows the AUC comparison results, and Table 4 shows the F1-score comparison results.

[0177] Table 2:

[0178]

[0179]

[0180] Table 3:

[0181]

[0182] Table 4:

[0183]

[0184]

[0185] In the above embodiments, although the steps are numbered S1, S2, etc., these are only specific embodiments given by the present invention. Those skilled in the art may adjust the execution order of S1, S2, etc. according to actual conditions, which is also within the scope of protection of the present invention. It can be understood that in some embodiments, some or all of the above embodiments may be included.

[0186] like Figure 6As shown, a system 200 for distinguishing the vitality of corn germplasm resources across germplasms according to an embodiment of the present invention includes an acquisition module 201, a segmentation module 202, a hyperspectral data acquisition module 203, a preprocessing module 204, a model training module 205 and a discrimination module 206;

[0187] The acquisition module 201 is used to: acquire a hyperspectral image of each sample corn seed in the sample corn germplasm resource, and obtain vitality label data of each sample corn seed in the sample corn germplasm resource;

[0188] The segmentation module 202 is used to perform threshold segmentation on the hyperspectral image of each sample corn seed to obtain a hyperspectral image of each sample corn seed after background removal, and perform regional segmentation on the hyperspectral image of each sample corn seed after background removal to determine a hyperspectral image of the embryo region and a hyperspectral image of the endosperm region;

[0189] The hyperspectral data acquisition module 203 is used to obtain hyperspectral data of the embryo region and hyperspectral data of the endosperm region according to the hyperspectral image of the embryo region and the hyperspectral image of the endosperm region;

[0190] The hyperspectral data acquisition module 204 is further configured to: obtain hyperspectral data of each sample corn seed from the hyperspectral image of each sample corn seed after background removal;

[0191] The pre-processing module 205 is used to: pre-process each hyperspectral data;

[0192] The model training module is used to perform feature selection on each pre-processed hyperspectral data set and train the constructed dual-branch spectral Transformer corn germplasm cross-germplasm viability detection model based on the viability label data of each corn seed sample.

[0193] The discrimination module 206 is used to: use the trained dual-branch spectral Transformer corn germplasm cross-germplasm vitality detection model to discriminate the cross-germplasm corn to be discriminated, and obtain the vitality prediction result of the cross-germplasm corn to be discriminated.

[0194] Optionally, in the above technical solution, the segmentation module 202 is specifically configured to:

[0195] Manually annotating the boundaries of the embryo region and the endosperm region in the hyperspectral images of multiple corn seed samples to obtain embryo annotated data and endosperm annotated data;

[0196] The constructed moving inverted bottleneck UNet germplasm hyperspectral image segmentation model was trained using hyperspectral images of multiple corn seed samples, combined with embryo and endosperm annotation data.

[0197] The trained moving inverted bottleneck UNet germplasm hyperspectral image segmentation model was used to perform regional segmentation on the hyperspectral image of each sample corn seed after background removal to determine the hyperspectral image of the embryo region and the hyperspectral image of the endosperm region.

[0198] Optionally, in the above technical solution, the pre-processing module 204 is specifically used to perform noise reduction processing, baseline drift correction processing, light scattering correction processing and data scaling processing on each hyperspectral data.

[0199] Optionally, in the above technical solution, the dual-branch spectral Transformer corn germplasm cross-germplasm vitality detection model includes: a multi-scale CNN module, a Transformer module, a multi-level feature fusion attention module and a linear classification module, wherein the multi-scale CNN module is used to calculate the local features of each preprocessed hyperspectral data, the Transformer module is used to calculate the global features of each preprocessed hyperspectral data, the multi-level feature fusion attention module is used to: map local features and global features into multiple feature maps with consistent dimensions, and assign attention score weights to each feature map through an activation function layer, and fuse all feature maps based on the assigned attention score weights to obtain multi-level global-local fusion features; the linear classification module is used to: perform global average pooling and linear layer operations on the multi-level global-local fusion features in sequence to obtain vitality prediction results.

[0200] Optionally, in the above technical solution, the acquisition module 201 is specifically configured to:

[0201] A hyperspectral imaging device is used to obtain a hyperspectral image of each sample corn seed in the sample corn germplasm resources, and the vitality label data of each sample corn seed in the sample corn germplasm resources is obtained through a germination experiment.

[0202] It should be noted that the beneficial effects of the corn germplasm resource vitality cross-germplasm discrimination system 200 provided in the above embodiment are the same as the beneficial effects of the corn germplasm resource vitality cross-germplasm discrimination method described above, and will not be repeated here. In addition, when the system provided in the above embodiment realizes its functions, it only uses the division of the above-mentioned functional modules as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the system can be divided into different functional modules according to actual conditions to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiment belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0203] Among them, the corn germplasm resource vitality cross-germplasm discrimination system of the present invention can be a computer program (including program code) running in a computer device. For example, the corn germplasm resource vitality cross-germplasm discrimination system of the present invention is an application software that can be used to execute the corresponding steps in the corn germplasm resource vitality cross-germplasm discrimination method of the present invention.

[0204] In some embodiments, the corn germplasm vitality cross-germplasm discrimination system of the present invention can be implemented by a combination of software and hardware. As an example, the corn germplasm vitality cross-germplasm discrimination system of the present invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the corn germplasm vitality cross-germplasm discrimination method of the present invention. For example, the processor in the form of a hardware decoding processor can adopt one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs) or other electronic components.

[0205] The modules described in the embodiments of the present invention may be implemented in software or hardware, and the name of a module does not necessarily limit the module itself.

[0206] An electronic device according to an embodiment of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, any one of the above-mentioned methods for distinguishing the vitality of corn germplasm resources across germplasms is implemented. That is, an electronic device according to an embodiment of the present invention may include but is not limited to: a processor and a memory; the memory is used to store the computer program; the processor is used to execute the method for distinguishing the vitality of corn germplasm resources across germplasms shown in any one of the embodiments of the present invention by calling the computer program.

[0207] In an alternative embodiment, an electronic device is provided, such as Figure 7 As shown, Figure 7The electronic device 4000 shown includes: a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, for example, via a bus 4002. Optionally, the electronic device 4000 may further include a transceiver 4004, which can be used for data exchange between the electronic device and other electronic devices, such as data transmission and / or data reception. It should be noted that in actual applications, the number of transceivers 4004 is not limited to one, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present invention.

[0208] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the present disclosure. Processor 4001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, and the like.

[0209] Bus 4002 may include a path for transmitting information between the above components. Bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. Bus 4002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 In the figure, only one thick line is used to represent the bus 4002, but this does not mean that there is only one bus or one type of bus.

[0210] The memory 4003 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to these.

[0211] The memory 4003 is used to store application code (computer program) for executing the solution of the present invention, and is controlled by the processor 4001. The processor 4001 is used to execute the application code stored in the memory 4003 to implement the content shown in the above method embodiment.

[0212] Among them, the electronic device can also be a terminal device, and the terminal device can be any device that can install applications, including at least one of a smartphone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, a smart TV, and a smart car device.

[0213] It should be noted that Figure 7 The electronic device shown is only an example and should not limit the functions and scope of use of the embodiments of the present invention.

[0214] A computer-readable storage medium according to an embodiment of the present invention stores a computer program, which, when executed by a processor, implements any of the above-mentioned methods for cross-germplasm discrimination of corn germplasm vitality.

[0215] Alternatively, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, or the like.

[0216] In an exemplary embodiment, a computer program product or computer program is also provided, comprising computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform any of the above-described methods for determining the viability of corn germplasm resources across germplasms.

[0217] Computer program code for performing the operations of the present invention may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on the remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0218] It should be understood that the flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the methods and computer program products according to various embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the module, program segment, or a part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or can be implemented using a combination of dedicated hardware and computer instructions.

[0219] The computer-readable storage medium provided in the embodiment of the present invention may be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EEPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or component.

[0220] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device executes the method shown in the above embodiment.

[0221] The above description is merely a preferred embodiment of the present invention and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of disclosure involved in the present invention is not limited to the technical solutions formed by the specific combination of the above-mentioned technical features, but also includes other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the above-mentioned disclosed concepts. For example, a technical solution formed by replacing the above-mentioned features with (but not limited to) technical features with similar functions disclosed in the present invention.

[0222] It should be noted that the terms "first," "second," and the like in the specification and claims of this application are used to distinguish similar objects and to define a specific order or precedence. Where appropriate, the order used for similar objects may be interchanged, such that the embodiments of the present application described herein can be implemented in an order other than the order shown or described.

[0223] Those skilled in the art will appreciate that the present invention may be implemented as a system, method, or computer program product. Therefore, the present invention may be implemented in the following forms: entirely in hardware, entirely in software (including firmware, resident software, microcode, etc.), or in a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, the present invention may be implemented in the form of a computer program product embodied in one or more computer-readable media containing computer-readable program code.

[0224] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A method for distinguishing the vitality of corn germplasm resources across germplasms, characterized in that: include: Acquire a hyperspectral image of each sample corn seed in the sample corn germplasm resource, and obtain vitality label data of each sample corn seed in the sample corn germplasm resource; Threshold segmentation is performed on the hyperspectral image of each sample corn seed to obtain a hyperspectral image of each sample corn seed after background removal, and regional segmentation is performed on the hyperspectral image of each sample corn seed after background removal to determine a hyperspectral image of the embryo region and a hyperspectral image of the endosperm region; Obtaining hyperspectral data of the embryo region and hyperspectral data of the endosperm region according to the hyperspectral image of the embryo region and the hyperspectral image of the endosperm region; Obtaining hyperspectral data of each sample corn seed from the hyperspectral image of each sample corn seed after background removal; Preprocess each hyperspectral data; Feature selection was performed on each pre-processed hyperspectral data set, and the constructed dual-branch spectral Transformer corn germplasm cross-germplasm vitality detection model was trained based on the vitality label data of each corn seed sample. The trained dual-branch spectral Transformer corn germplasm cross-germplasm viability detection model is used to discriminate the cross-germplasm corn to be discriminated, and the vitality prediction result of the cross-germplasm corn to be discriminated is obtained.

2. A method for distinguishing the vitality of corn germplasm resources across germplasms according to claim 1, characterized in that: Perform regional segmentation on the hyperspectral image of each corn seed sample after removing the background, and determine the hyperspectral image of the embryo region and the hyperspectral image of the endosperm region, including: Manually annotating the boundaries of the embryo region and the endosperm region in the hyperspectral images of multiple corn seed samples to obtain embryo annotated data and endosperm annotated data; The constructed moving inverted bottleneck UNet germplasm hyperspectral image segmentation model was trained using hyperspectral images of multiple corn seed samples, combined with embryo and endosperm annotation data. The trained moving inverted bottleneck UNet germplasm hyperspectral image segmentation model was used to perform regional segmentation on the hyperspectral image of each sample corn seed after background removal, and the hyperspectral images of the embryo region and the endosperm region were determined.

3. The method for distinguishing the vitality of corn germplasm resources across germplasms according to claim 1, characterized in that: Each hyperspectral data is preprocessed, including: Each hyperspectral data set is processed with noise reduction, baseline drift correction, light scattering correction and data scaling.

4. A method for distinguishing the vitality of corn germplasm resources across germplasms according to any one of claims 1 to 3, characterized in that: The dual-branch spectral Transformer corn germplasm cross-germplasm viability detection model includes: a multi-scale CNN module, a Transformer module, a multi-level feature fusion attention module and a linear classification module, wherein the multi-scale CNN module is used to calculate the local features of each preprocessed hyperspectral data, the Transformer module is used to calculate the global features of each preprocessed hyperspectral data, the multi-level feature fusion attention module is used to: map the local features and global features into multiple feature maps with consistent dimensions, and assign an attention score weight to each feature map through an activation function layer, and fuse all feature maps based on the assigned attention score weight to obtain multi-level global-local fusion features; the linear classification module is used to: perform global average pooling and linear layer operations on the multi-level global-local fusion features in sequence to obtain vitality prediction results.

5. A method for distinguishing the vitality of corn germplasm resources across germplasms according to any one of claims 1 to 3, characterized in that: Acquire a hyperspectral image of each sample corn seed in the sample corn germplasm resource, and obtain vitality label data of each sample corn seed in the sample corn germplasm resource, including: A hyperspectral imaging device is used to obtain a hyperspectral image of each sample corn seed in the sample corn germplasm resource, and vitality label data of each sample corn seed in the sample corn germplasm resource is obtained through a germination experiment.

6. A system for distinguishing the vitality of corn germplasm resources across germplasms, characterized by: It includes acquisition module, segmentation module, hyperspectral data acquisition module, preprocessing module, model training module and discrimination module; The acquisition module is used to: acquire a hyperspectral image of each sample corn seed in the sample corn germplasm resource, and obtain vitality label data of each sample corn seed in the sample corn germplasm resource; The segmentation module is used to perform threshold segmentation on the hyperspectral image of each sample corn seed to obtain a hyperspectral image of each sample corn seed after background removal, and perform regional segmentation on the hyperspectral image of each sample corn seed after background removal to determine a hyperspectral image of an embryo region and a hyperspectral image of an endosperm region; The hyperspectral data acquisition module is used to obtain hyperspectral data of the embryo region and hyperspectral data of the endosperm region based on the hyperspectral image of the embryo region and the hyperspectral image of the endosperm region; The hyperspectral data acquisition module is further used to: obtain hyperspectral data of each sample corn seed from the hyperspectral image of each sample corn seed after removing the background; The preprocessing module is used to: preprocess each hyperspectral data; The model training module is used to perform feature selection on each pre-processed hyperspectral data, and train the constructed dual-branch spectral Transformer corn germplasm cross-germplasm vitality detection model in combination with the vitality label data of each sample corn seed; The discrimination module is used to: use the trained dual-branch spectral Transformer corn germplasm cross-germplasm vitality detection model to discriminate the cross-germplasm corn to be discriminated, and obtain the vitality prediction result of the cross-germplasm corn to be discriminated.

7. A corn germplasm resource vitality cross-germplasm discrimination system according to claim 6, characterized in that: The segmentation module is specifically used for: Manually annotating the boundaries of the embryo region and the endosperm region in the hyperspectral images of multiple corn seed samples to obtain embryo annotated data and endosperm annotated data; The constructed moving inverted bottleneck UNet germplasm hyperspectral image segmentation model was trained using hyperspectral images of multiple corn seed samples, combined with embryo and endosperm annotation data. The trained moving inverted bottleneck UNet germplasm hyperspectral image segmentation model was used to perform regional segmentation on the hyperspectral image of each sample corn seed after background removal, and the hyperspectral images of the embryo region and the endosperm region were determined.

8. A corn germplasm resource vitality cross-germplasm discrimination system according to claim 6, characterized in that: Each hyperspectral data is preprocessed, including: Each hyperspectral data set is processed with noise reduction, baseline drift correction, light scattering correction and data scaling.

9. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for cross-germplasm discrimination of the vitality of corn germplasm resources according to any one of claims 1 to 5 is implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for cross-germplasm discrimination of the vitality of corn germplasm resources according to any one of claims 1 to 5 is implemented.

Images