A multi-modal biomass yield prediction method and device for wheat breeding, medium and product

Abstract

The application discloses a kind of multi-modal biomass yield estimation methods, device, medium and product for wheat breeding, it is related to crop breeding field, method includes obtaining target image and target data, image segmentation is carried out to target image and the segmented image and target data are input to trained multi-modal feature wheat biomass estimation model, and the yield of wheat biomass estimation is obtained;Trained multi-modal feature wheat biomass estimation model includes image processing submodel, vegetation index and environmental data feature extraction submodel and feature vector fusion submodel.By using deep learning method, wheat breeding related information can be automatically analyzed, and the efficiency of data acquisition and analysis is improved;Using multi-modal information, the problem of incomplete information amount of single data source is reduced, and the yield of wheat biomass is estimated comprehensively and efficiently.

Description

Technical Field

[0001] This invention relates to the field of crop breeding, and in particular to a method, apparatus, medium and product for predicting multimodal biomass yield in wheat breeding. Background Technology

[0002] In the field of wheat breeding, traditional manual surveys and data collection methods are labor-intensive and inefficient, and it is difficult to obtain comprehensive and accurate information in a short period of time. Traditional methods mainly rely on human experience and manual operations, including field observation, data recording, and decision-making based on manual analysis. Due to the above-mentioned problems, traditional wheat breeding methods face the following challenges: (1) Manual methods lead to low efficiency in the wheat breeding process, which limits the speed of new variety development; (2) Traditional methods are difficult to capture multimodal information in the field, such as plant growth status, comprehensively and accurately; (3) The lack of advanced data analysis tools limits the in-depth mining and effective utilization of large-scale farmland data.

[0003] Currently, some automated systems, including sensors and data acquisition devices, are used for field data collection. However, these systems have limitations in processing multimodal data (such as images and meteorological data), and traditional data analysis methods are mainly based on statistics and single-modal information, which cannot fully tap the potential information of multimodal data. Summary of the Invention

[0004] The purpose of this invention is to provide a method, device, medium, and product for predicting multimodal biomass yield in wheat breeding, which can efficiently and comprehensively collect and analyze wheat breeding-related information and predict wheat biomass yield.

[0005] To achieve the above objectives, the present invention provides the following solution:

[0006] A multimodal biomass yield prediction method for wheat breeding, the method comprising:

[0007] Obtain the target dataset; the target dataset includes target images and target data; the target images are obtained by preprocessing original wheat field images; the target data includes vegetation index data and environmental data corresponding to the wheat fields; the original wheat field images were taken when the wheat was in the tillering stage;

[0008] The target image is segmented to obtain a wheat field segmentation image;

[0009] The segmented wheat field image and the target data are input into a trained multimodal feature wheat biomass prediction model to obtain the estimated wheat biomass yield. The trained multimodal feature wheat biomass prediction model includes an image processing sub-model, a vegetation index and environmental data feature extraction sub-model, and a feature vector fusion sub-model. The image processing sub-model is used to extract features from the segmented wheat field image. The vegetation index and environmental data feature extraction sub-model is used to extract features from the vegetation index data and the environmental data. The feature vector fusion sub-model is used to fuse and output the feature vectors output by the image processing sub-model and the vegetation index and environmental data feature extraction sub-model.

[0010] Optionally, the preprocessing of the original wheat field images specifically includes:

[0011] The original wheat field image is scaled up, and the scaled wheat field image is cropped to obtain a cropped wheat field image.

[0012] The cropped wheat field image is subjected to random horizontal and vertical mirroring, random cropping, rotation, and noise addition operations in sequence to obtain the target image.

[0013] Optionally, the step of performing image segmentation on the target image to obtain wheat field segmentation images specifically includes:

[0014] The target image is input into a trained wheat field bounding box recognition model to obtain the bounding boxes of the wheat fields; the wheat field bounding box recognition model is built based on the convolutional neural network YOLOv8.

[0015] The bounding boxes of the wheat fields are input into the trained image segmentation model to obtain segmented images of the wheat fields.

[0016] Optionally, the image processing sub-model includes a shallow feature extraction network module, a deep feature extraction network module, a multi-scale feature extraction module, and a feature filtering and integration module connected in sequence.

[0017] The shallow feature extraction network module includes a first convolutional layer, a first max pooling layer, a second convolutional layer, and a first average pooling layer connected in sequence.

[0018] The deep feature extraction network module includes a third convolutional layer, a fourth convolutional layer, a fifth convolutional layer, and a second max pooling layer connected in sequence.

[0019] The multi-scale feature extraction module includes a first RFB layer and a second RFB layer connected in sequence.

[0020] The feature filtering and integration module includes a filtering layer, a second average pooling layer, and a linear integration layer connected in sequence.

[0021] Optionally, the vegetation index and environmental data feature extraction sub-model includes a first feature extraction module, a second feature extraction module, and a feature fusion module; the first feature extraction module and the second feature extraction module are connected in parallel and then connected in series with the feature fusion module;

[0022] The first feature extraction module includes an LSTM layer and a Dropout layer connected in sequence;

[0023] The second feature extraction module includes a first multi-scale convolutional layer, a second multi-scale convolutional layer, a third max pooling layer, a third multi-scale convolutional layer, a fourth multi-scale convolutional layer, and a global pooling layer connected in sequence.

[0024] The feature fusion module includes a first fully connected layer.

[0025] Optionally, the feature vector fusion sub-model includes a second fully connected layer and a Softmax layer connected in sequence.

[0026] Optionally, the vegetation index data includes terrestrial chlorophyll index, chlorophyll content index, and leaf area index;

[0027] The environmental data includes temperature and precipitation.

[0028] To achieve the above objectives, the present invention also provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the multimodal biomass yield prediction method for wheat breeding.

[0029] To achieve the above objectives, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the multimodal biomass yield prediction method for wheat breeding.

[0030] To achieve the above objectives, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the multimodal biomass yield prediction method for wheat breeding.

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

[0032] This invention provides a method, device, medium, and product for multimodal biomass yield prediction in wheat breeding. By employing deep learning methods, it can automatically analyze wheat breeding-related information, improving the efficiency of data collection and analysis. By mining potential patterns in large-scale farmland data, it provides a scientific basis for breeding decisions. By utilizing multimodal information, it reduces the problem of incomplete information from a single data source, enabling comprehensive and efficient prediction of wheat biomass yield. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This is a schematic diagram of the multimodal biomass yield prediction method for wheat breeding provided in Embodiment 1 of the present invention;

[0035] Figure 2 Aerial images of wheat during its growth period taken by drone;

[0036] Figure 3 A schematic diagram showing the bounding box recognition results and wheat field segmentation image results for wheat fields;

[0037] Figure 4 A schematic diagram of the module structure of a multimodal wheat biomass prediction model;

[0038] Figure 5 This is a schematic diagram of the preprocessed field area image;

[0039] Figure 6 A schematic diagram of an electronic spreadsheet created according to the crop planting sequence designed for the experimental field;

[0040] Figure 7 A schematic diagram of the output results of the multimodal wheat biomass prediction model;

[0041] Figure 8 This is a diagram of the internal structure of a computer device. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] The purpose of this invention is to provide a method, device, medium, and product for predicting multimodal biomass yield in wheat breeding, aiming to efficiently and comprehensively collect and analyze wheat breeding-related information and predict wheat biomass yield.

[0044] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0045] Example 1

[0046] like Figure 1 As shown, the multimodal biomass yield prediction method for wheat breeding in this embodiment includes the following steps:

[0047] S1: Obtain the target dataset; the target dataset includes target images and target data; the target images are obtained by preprocessing original wheat field images; the target data includes vegetation index data and environmental data corresponding to the wheat fields; the original wheat field images were taken when the wheat was in the tillering stage. Specifically, wheat goes through a series of developmental stages, which can generally be divided into the following main stages: emergence stage, main leaf stage, tillering stage, overwintering stage, greening stage, tillering stage, jointing stage, booting stage, heading stage, flowering stage, grain filling stage, and maturity stage. This invention, through observation of UAV aerial images of wheat throughout its growth period, found that the outline of the field area formed by wheat in the tillering stage is the most obvious, that is, the crop isolation is relatively obvious, see [reference]. Figure 2 Therefore, the original wheat field images used were drone images of wheat in its early growth stage.

[0048] Further, in step S1, the preprocessing of the original wheat field image specifically includes:

[0049] S11: Scale the original wheat field image and crop the scaled wheat field image to obtain a cropped wheat field image.

[0050] S12: Perform random horizontal and vertical mirroring, random cropping, rotation, and noise addition operations on the cropped wheat field image in sequence to obtain the target image.

[0051] The process begins with capturing 140 images at a resolution of 5472×3468. Due to the large pixel size of the raw data, direct deep learning is not feasible, requiring further preprocessing. Traditional UAV image preprocessing involves stitching together a complete flight path image, followed by cropping. However, traditional methods often result in images with complete field area information, and the initial stitching process places significant hardware requirements. Therefore, this invention employs an innovative cropping method. The raw wheat field images captured by the UAV are first batch-scaled to a resolution of 1396×912. These scaled images are then further cropped to obtain more complete field information. Next, the scaled images are cropped to a resolution of 512×512, preparing them for feeding into the deep learning network. Finally, 750 images at a resolution of 512×512 are obtained, representing the 750 target images. The wheat images can be captured using UAVs, satellite remote sensing, or other methods, without limitation.

[0052] Furthermore, data augmentation techniques were employed to perform multiple processing steps on the 750 cropped wheat field images, including random horizontal and vertical mirroring, random cropping, rotation, and noise addition. These processes expanded the number of images to 5000. This expansion aimed to increase the diversity of the dataset and improve the model's robustness to better adapt to various scenarios and changes.

[0053] S2: Perform image segmentation on the target image to obtain a wheat field segmentation image; specifically including:

[0054] S21: The target image is input into the trained wheat field bounding box recognition model to obtain the bounding boxes of the wheat fields. The wheat field bounding box recognition model is built based on the YOLOv8 convolutional neural network. Specifically, the target image with a resolution of 512×512 is labeled using labelImg. Finally, the obtained labels and the original target image are set as the training set, test set, and validation set in a ratio of 7:2:1, respectively. Then, the custom-trained YOLOv8 model (the trained wheat field bounding box recognition model) is used to obtain the bounding boxes of the wheat fields. The bounding box recognition results of the wheat fields are as follows: Figure 3 As shown.

[0055] S22: Input the bounding boxes of the wheat fields into the trained image segmentation model to obtain segmented wheat field images. The results of the wheat field segmentation images are as follows: Figure 3As shown in the figure. The trained image segmentation model is built on the image segmentation model SAM. After obtaining the bounding box of the wheat field, SAM can segment it or add a mask to it.

[0056] S3: Input the segmented wheat field image and the target data into the trained multimodal feature wheat biomass prediction model to obtain the estimated wheat biomass yield; see [link / reference]. Figure 4 The trained multimodal wheat biomass prediction model includes an image processing sub-model, a vegetation index and environmental data feature extraction sub-model, and a feature vector fusion sub-model. The image processing sub-model is used to extract features from the segmented wheat field image. The vegetation index and environmental data feature extraction sub-model is used to extract features from the vegetation index data and the environmental data. The feature vector fusion sub-model is used to fuse and output the feature vectors output by the image processing sub-model and the vegetation index and environmental data feature extraction sub-model.

[0057] Specifically, such as Figure 4 As shown, the image processing sub-model includes a shallow feature extraction network module, a deep feature extraction network module, a multi-scale feature extraction module, and a feature filtering and integration module connected in sequence.

[0058] The shallow feature extraction network module comprises a first convolutional layer, a first max-pooling layer, a second convolutional layer, and a first average pooling layer connected in sequence. (See also...) Figure 4 , Figure 4 The first and second convolutional layers are both represented by conv2d+BN+ReLU, the first max pooling layer is represented by Maxpool, and the first average pooling layer is represented by AvgPool. The goal of the shallow feature extraction network module is to capture basic features in the input data, such as edges and textures, so that higher-level networks can learn more abstract and complex representations based on these features.

[0059] The deep feature extraction network module includes a third convolutional layer, a fourth convolutional layer, a fifth convolutional layer, and a second max-pooling layer connected in sequence. See also... Figure 4 , Figure 4 The third, fourth, and fifth convolutional layers are all represented using conv2d+BN+ReLU, and the second max pooling layer is represented using Maxpool. By adding more convolutional layers, features are extracted at a deeper level, and the number of feature channels is doubled, increasing the network complexity to learn more abstract representations of the input data. Pooling layers follow the convolutional layers, helping to reduce the spatial dimensionality of the data and improve computational efficiency. The pooling layers then aggregate the information from the entire feature map into a single value, forming the final deep feature representation.

[0060] The multi-scale feature extraction module includes a first RFB layer and a second RFB layer connected in sequence. See also Figure 4 , Figure 4 In this model, both the first and second RFB layers are denoted by RFB. The RFB module is used to extract multi-scale features. By using variable convolutional kernel sizes and incorporating multi-scale information, the RFB module enhances the network's ability to perceive different target sizes. The introduction of convolutional kernels of varying sizes allows for better capture of multi-scale target information. Furthermore, by concatenating features from different scales, the module further enhances target perception. Finally, additional convolutional layers nested within the RFB module process the concatenated features to generate an enhanced feature map.

[0061] The feature filtering and integration module includes a filtering layer, a second average pooling layer, and a linear integration layer connected in sequence. See also Figure 4 , Figure 4 The middle filtering layer is denoted by Cloofomer, the second average pooling layer is denoted by AvgPool, and the linear integration layer is denoted by Linear+LeakyReLU.

[0062] Further, see Figure 4 The vegetation index and environmental data feature extraction sub-model includes a first feature extraction module, a second feature extraction module, and a feature fusion module; the first feature extraction module and the second feature extraction module are connected in parallel and then connected in series with the feature fusion module.

[0063] The first feature extraction module includes an LSTM layer and a Dropout layer connected in sequence.

[0064] The second feature extraction module includes a first multi-scale convolutional layer, a second multi-scale convolutional layer, a third max pooling layer, a third multi-scale convolutional layer, a fourth multi-scale convolutional layer, and a global pooling layer connected in sequence. For example... Figure 4 As shown, the first, second, third, and fourth multi-scale convolutional layers are all represented by MACM, the third max pooling layer is represented by Maxpool, and the global pooling layer is represented by Globe Pool.

[0065] The feature fusion module includes a first fully connected layer, wherein the first fully connected layer in Figure 4 In Chinese, it is represented by Concat.

[0066] Further, see Figure 4The feature vector fusion sub-model includes a second fully connected layer and a Softmax layer connected in sequence. Through the aforementioned image processing sub-model and vegetation index and environmental data feature extraction sub-model, 1×512 image feature vectors, 1×512 vegetation index feature vectors, and 1×512 environmental feature vectors are obtained, respectively. The feature vector fusion sub-model concatenates these three vectors into a 1×1536 feature vector. Finally, through the second fully connected layer and the Softmax layer, the estimated biomass yield of wheat is obtained.

[0067] Optionally, the vegetation index data includes the MERIS terrestrial chlorophyll index (MTCI), the Datt chlorophyll content index (Datt), and the leaf area index (LAI). The environmental data includes temperature and precipitation. Furthermore, other feature parameters can be selected as vegetation index data and environmental data, or other types of data can be used as inputs for multimodal features; this is not limited and all are within the scope of protection of this invention.

[0068] like Figures 5-7 As shown, when using it, the user only needs to prepare: an orthophoto image of the initial field (such as...). Figure 2 (as shown), preprocessed field area images (such as...) Figure 5 (as shown) and an electronic spreadsheet created according to the crop planting sequence designed for the experimental field (such as... Figure 6 As shown), when a user submits a preprocessed image of a field area, the method of this invention will ultimately yield the following information: the serial number of the field area, the specific location of the field area in the orthophoto map, and the estimated wheat biomass yield of the field area. (See attached image.) Figure 7 .

[0069] Furthermore, replacing the wheat crop of this invention with other crops, such as corn or rice, is also within the scope of protection of this invention. This invention does not limit the types of crops.

[0070] Technical effects of the present invention:

[0071] 1) It can automatically analyze wheat breeding-related information, improving the efficiency of data collection and analysis and reducing the manual burden.

[0072] 2) By utilizing multimodal information, the problem of incomplete information from a single data source is avoided, and the impact of the environment on crop growth and development is taken into account.

[0073] 3) By using deep learning methods, we can uncover potential patterns in large-scale farmland data, providing a scientific basis for breeding decisions and enabling more comprehensive and efficient collection and analysis of wheat breeding-related information.

[0074] Example 2

[0075] A computer device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the computer program to implement the steps of the multimodal biomass yield prediction method for wheat breeding.

[0076] Example 3

[0077] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the multimodal biomass yield prediction method for wheat breeding.

[0078] Example 4

[0079] A computer program product includes a computer program that, when executed by a processor, implements the steps of the multimodal biomass yield prediction method for wheat breeding.

[0080] Example 5

[0081] A computer device, which may be a database, may have an internal structure diagram as shown below. Figure 8 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores pending transactions. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements the multimodal biomass yield prediction method for wheat breeding in Example 1.

[0082] It should be noted that the object information (including but not limited to object device information, object personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this invention are all information and data authorized by the object or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0083] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided by this invention may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided by this invention may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0084] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0085] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A multimodal biomass yield prediction method for wheat breeding, characterized in that, The method includes: Obtain the target dataset; the target dataset includes target images and target data; the target images are obtained by preprocessing original wheat field images; the target data includes vegetation index data and environmental data corresponding to the wheat fields; the original wheat field images were taken when the wheat was in the tillering stage; The target image is segmented to obtain a wheat field segmentation image; The segmented wheat field image and the target data are input into a trained multimodal feature wheat biomass prediction model to obtain the estimated wheat biomass yield. The trained multimodal feature wheat biomass prediction model includes an image processing sub-model, a vegetation index and environmental data feature extraction sub-model, and a feature vector fusion sub-model. The image processing sub-model is used to extract features from the segmented wheat field image. The vegetation index and environmental data feature extraction sub-model is used to extract features from the vegetation index data and the environmental data. The feature vector fusion sub-model is used to fuse and output the feature vectors output by the image processing sub-model and the vegetation index and environmental data feature extraction sub-model. The image processing sub-model includes a shallow feature extraction network module, a deep feature extraction network module, a multi-scale feature extraction module, and a feature filtering and integration module connected in sequence. The shallow feature extraction network module includes a first convolutional layer, a first max pooling layer, a second convolutional layer, and a first average pooling layer connected in sequence. The deep feature extraction network module includes a third convolutional layer, a fourth convolutional layer, a fifth convolutional layer, and a second max pooling layer connected in sequence. The multi-scale feature extraction module includes a first RFB layer and a second RFB layer connected in sequence. The feature filtering and integration module includes a filtering layer, a second average pooling layer, and a linear integration layer connected in sequence. The vegetation index and environmental data feature extraction sub-model includes a first feature extraction module, a second feature extraction module, and a feature fusion module; the first feature extraction module and the second feature extraction module are connected in parallel and then connected in series with the feature fusion module. The first feature extraction module includes an LSTM layer and a Dropout layer connected in sequence; The second feature extraction module includes a first multi-scale convolutional layer, a second multi-scale convolutional layer, a third max pooling layer, a third multi-scale convolutional layer, a fourth multi-scale convolutional layer, and a global pooling layer connected in sequence. The feature fusion module includes a first fully connected layer.

2. The multimodal biomass yield prediction method for wheat breeding according to claim 1, characterized in that, The preprocessing of the original wheat field images specifically includes: The original wheat field image is scaled up, and the scaled wheat field image is cropped to obtain a cropped wheat field image. The cropped wheat field image is subjected to random horizontal and vertical mirroring, random cropping, rotation, and noise addition operations in sequence to obtain the target image.

3. The multimodal biomass yield prediction method for wheat breeding according to claim 1, characterized in that, The step of segmenting the target image to obtain a wheat field segmentation image specifically includes: The target image is input into a trained wheat field bounding box recognition model to obtain the bounding boxes of the wheat fields; the wheat field bounding box recognition model is built based on the convolutional neural network YOLOv8. The bounding boxes of the wheat fields are input into the trained image segmentation model to obtain segmented images of the wheat fields.

4. The multimodal biomass yield prediction method for wheat breeding according to claim 1, characterized in that, The feature vector fusion sub-model includes a second fully connected layer and a Softmax layer connected in sequence.

5. The multimodal biomass yield prediction method for wheat breeding according to claim 1, characterized in that, The vegetation index data includes terrestrial chlorophyll index, chlorophyll content index, and leaf area index. The environmental data includes temperature and precipitation.

6. A computer device, comprising: The memory and processor are computer programs stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the multimodal biomass yield prediction method for wheat breeding as described in any one of claims 1-5.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the multimodal biomass yield estimation method for wheat breeding as described in any one of claims 1-5.

8. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the multimodal biomass yield estimation method for wheat breeding as described in any one of claims 1-5.

Images