A method, system and device for predicting biochemical parameters of crop seeds based on hyperspectral

Abstract

The present invention discloses a method, system and device for predicting biochemical parameters of crop seeds based on hyperspectral, including: using a data acquisition device to capture multi-band imaging information of seeds in real time using a hyperspectral camera, then using a convolutional neural network to extract image features, and combining a graph convolutional neural network and a long short-term memory network to construct a multi-scale spatiotemporal graph neural network model to predict biochemical parameters such as protein content, moisture content, and starch content of seeds. Finally, the prediction results are visualized so that users can check the status of seeds at any time. The design of the present invention realizes accurate monitoring, intelligent management and efficient decision-making of seed biochemical parameters, improves the intelligence level of agricultural production, helps optimize the crop growth environment, and improves crop yield and quality.

Description

Technical Field

[0001] The present invention relates to the technical field of big data intelligent analysis and seed quality detection, and more specifically, to a method, system and device for predicting crop seed biochemical parameters based on hyperspectral data and spatiotemporal graph network model. Background Art

[0002] Rapid and accurate detection of seed quality is a key link in ensuring seed quality. It can ensure smooth and orderly food production from the source, reduce agricultural production risks, and promote agricultural scientific and technological progress. Most traditional seed quality detection methods require destructive analysis of seeds, and the detection time is long and the process is complicated, which makes it difficult to meet the requirements of modern agriculture for non-destructive and rapid production of seeds. With the trend of intelligent crop breeding technology (biotechnology + artificial intelligence + big data technology) and the scale of seed data to be tested, it is particularly necessary and urgent to carry out research on intelligent, high-throughput, dynamic and non-destructive seed quality rapid and accurate detection technology. Common prediction models include linear regression, logistic regression and other statistical methods, which are not effective in dealing with complex nonlinear relationships. Neural networks can capture the complex structure and implicit correlations in the data, and thus make more accurate predictions.

[0003] Hyperspectral data can be continuously sampled within an extremely narrow spectral range, containing hundreds to thousands of bands, each of which corresponds to a very narrow spectral range. This high spectral resolution enables the model to capture more detailed and richer spectral information, thereby providing more accurate spatiotemporal analysis. The spatiotemporal graph network model is capable of processing multi-dimensional and highly complex data, taking into account both spatial and temporal dimensions, and can fully capture the spatiotemporal characteristics of the data. Therefore, how to construct a spatiotemporal graph network model and integrate hyperspectral data into the learning and reasoning process of the model to accurately predict seed biochemical parameters is a key technical problem that needs to be solved urgently. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a method, system and device for predicting crop seed biochemical parameters based on hyperspectral, with the purpose of capturing the hyperspectral data of seeds and combining them with a multi-scale spatiotemporal graph neural network model to predict the protein content, moisture content, starch content and other biochemical parameters of seeds.

[0005] The first aspect of the present invention provides a method for predicting crop seed biochemical parameters based on hyperspectral, comprising:

[0006] Acquire imaging information of crop seeds in different bands through a hyperspectral camera, and input the imaging information into a convolutional neural network to extract image features and generate a sequence vector;

[0007] Based on the graph neural network and LSTM network, a spatiotemporal graph network structure is established to construct a seed biochemical parameter prediction model, the sequence vector is used as the model input, and the graph neural network is used to construct a spatial embedding vector sequence to obtain spatial dependency information and structural information;

[0008] Input the spatial embedding vector sequence into the LSTM network to capture the dynamic characteristics of the image changing with the spectrum, and obtain the output vector of each band;

[0009] The output vectors are weighted summed using a fully connected layer, and an activation function is used to process the vectors to obtain the prediction results of the biochemical parameters of the crop seeds.

[0010] In this solution, the imaging information is input into a convolutional neural network to extract image features and generate a sequence vector, specifically:

[0011] Use hyperspectral cameras to obtain different types of crop seeds in different bands Imaging information , inputting the imaging information into a convolutional neural network;

[0012] Through multi-layer convolution and pooling operations, local features and hierarchical representations in imaging information are automatically extracted to identify key features of crop seeds in different bands, including shape features, texture features, and color features, and the key features are encoded into sequence vectors. .

[0013] In this solution, the sequence vector is used as the model input, and a graph neural network is used to construct a spatial embedding vector sequence, specifically:

[0014] The sequence vector is input into the graph neural network for spatial modeling. Each key feature is used as a node, and the relationship between nodes is used as the edge structure. The graph neural network is used to obtain the spatial dependency information and structural information of the key features based on the relationship between different key features.

[0015] In the graph convolution layer of the graph neural network, the state of each node is updated through information aggregation, and the spatial structure embedding matrix is ​​constructed through the graph neural network to obtain the spatial embedding vector sequence.

[0016] In this solution, the spatial embedding vector sequence is input into the LSTM network to capture the dynamic characteristics of the image as the spectrum changes, specifically:

[0017] The spatial embedding vector sequence output by the graph neural network is input into the LSTM network. The relationship between key features in the spatial embedding vector sequence is obtained through the forget gate and memory cell state update, capturing the dynamic characteristics of imaging information changing with the spectrum.

[0018] The spatial embedding vector sequence is calculated through the LSTM network to obtain the output vector of each band ,in These are the different bands of the hyperspectral camera.

[0019] In this solution, the crop seed biochemical parameters include protein content, starch content and water content;

[0020] The output vector The data is passed to the fully connected layer, which performs weighted summation of the vectors and sigmoid activation function to obtain the prediction results of protein content, starch content and moisture content.

[0021] The second aspect of the present invention provides a crop seed biochemical parameter prediction system based on hyperspectral, the system comprising: a data acquisition module, a seed biochemical parameter prediction module and a data visualization module;

[0022] The data acquisition module uses a spectral camera to obtain imaging information of different bands of different crop seeds and stores it in a memory;

[0023] The seed biochemical parameter prediction module extracts key features of imaging information through a convolutional neural network, processes feature sequences using a graph neural network and a long short-term memory network, and finally maps the features to specific biochemical parameters through a fully connected layer to obtain prediction results of protein content, moisture content, and starch content;

[0024] The data visualization module sends the obtained prediction results to the memory and visualizes the prediction results.

[0025] The third aspect of the present invention provides an intelligent monitoring device for crop seed biochemical parameters, comprising: a main control board, a memory, a feed place, a hyperspectral camera, a halogen lamp, a display, a transformer and an external power supply, wherein the halogen lamp is evenly distributed around the hyperspectral camera;

[0026] The feeding place is used for placing and taking out crop seeds, and the crop seeds are placed in by pressing the feeding place;

[0027] The hyperspectral camera is used to obtain imaging information of different bands of crop seeds;

[0028] The halogen lamp is used to irradiate the crop seeds and provide an internal lighting environment for the detection device;

[0029] The display is used to display the prediction results of crop seed biochemical parameters.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] The present invention constructs a hyperspectral and spatiotemporal graph network model for seed biochemical parameter prediction, realizing accurate prediction of biochemical parameters of seed growth. In addition, an efficient and fast seed biochemical parameter prediction system is designed to realize the automatic collection and analysis of seed biochemical related phenotypic data, provide a data basis for seed vitality prediction, and realize intelligent seed quality monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the embodiments or exemplary embodiments of the present invention, the drawings required for use in the embodiments or exemplary descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained according to the drawings without paying creative work.

[0033] Figure 1 A flow chart of a method for predicting biochemical parameters of crop seeds based on hyperspectral is shown;

[0034] Figure 2 A schematic diagram of a seed biochemical parameter prediction model is shown;

[0035] Figure 3 The framework diagram of the crop seed biochemical parameter prediction system based on hyperspectral is shown;

[0036] Figure 4 A schematic diagram of an intelligent monitoring device for biochemical parameters of crop seeds is shown. DETAILED DESCRIPTION

[0037] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0038] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0039] Figure 1 A flow chart of the hyperspectral-based crop seed biochemical parameter prediction method is shown.

[0040] like Figure 1 As shown, this embodiment provides a method for predicting crop seed biochemical parameters based on hyperspectral, including:

[0041] S102, acquiring imaging information of crop seeds in different bands through a hyperspectral camera, and inputting the imaging information into a convolutional neural network to extract image features and generate a sequence vector;

[0042] S104, constructing a spatiotemporal graph network structure based on a graph neural network and an LSTM network, building a seed biochemical parameter prediction model, taking the sequence vector as the model input, and using the graph neural network to construct a spatial embedding vector sequence to obtain spatial dependency information and structural information;

[0043] S106, inputting the spatial embedding vector sequence into the LSTM network to capture the dynamic characteristics of the image changing with the spectrum, and obtaining the output vector of each band;

[0044] S108, using a fully connected layer to perform weighted summation on the output vectors, and using an activation function to process to obtain prediction results of crop seed biochemical parameters.

[0045] It should be noted that if Figure 2 As shown in the figure, a hyperspectral camera is used to detect different types of crop seeds in different bands. Take photos and divide them into different spectral bands to obtain imaging information , input the imaging information into a convolutional neural network; automatically extract local features and hierarchical representations in the imaging information through multi-layer convolution and pooling operations, identify key features of crop seeds in different bands, including shape features, texture features and color features, and encode the key features into sequence vectors .

[0046] The sequence vector Input into the graph neural network for spatial modeling, the graph neural network is represented as: , Represents a collection of nodes. Represents a set of edge structures, takes each key feature as a node, and the association between nodes as an edge structure. The graph neural network is used to obtain the spatial dependency information and structural information of the key features according to the association between different key features. In the graph convolution layer of the graph neural network, the state of each node is updated by information aggregation. Information aggregation is defined as:

[0047] ;

[0048] in Representation Node In the Aggregate information of the layer, represents the weight matrix, represents a nonlinear activation function, represents the neighborhood, Represents the neighborhood node. Next, update the status of each node. The node update definition is as follows:

[0049] ;

[0050] in represents another activation function.

[0051] The sequence vector is passed through the graph neural network to construct a spatial structure embedding matrix to obtain a spatial embedding vector sequence ,The spatial embedding vector sequence output by the graph neural network is input into the LSTM network to capture the dynamic characteristics of the image as the spectrum changes. LSTM is used for each time step A hidden state will be generated ,This state contains the information summary of the previous input sequence, and obtains the relationship between the key features in the spatial embedding vector sequence through the forget gate and the memory cell state update, capturing the dynamic characteristics of the imaging information as the spectrum changes; forget gate and cell status Update is defined as:

[0052] ;

[0053] in represents the weight matrix, represents the bias vector, Represents the Sigmoid activation function, whose output range is between 0 and 1. Represents the time step The key features of represents element-wise multiplication, represents the input gate, Indicates candidate status.

[0054] The spatial embedding vector sequence is calculated through the LSTM network to obtain the output vector of each band ,in are different bands of the hyperspectral camera. The output vector The data is passed to the fully connected layer, which performs weighted summation of the vectors and sigmoid activation function to obtain the prediction results of protein content, starch content and moisture content.

[0055] Figure 3 The framework diagram of the crop seed biochemical parameter prediction system based on hyperspectral is shown.

[0056] This embodiment provides a crop seed biochemical parameter prediction system based on hyperspectral, the system comprising: a data acquisition module, a seed biochemical parameter prediction module and a data visualization module;

[0057] The data acquisition module uses a spectral camera to obtain imaging information of different bands of different crop seeds and stores it in a memory;

[0058] The seed biochemical parameter prediction module extracts key features of imaging information through a convolutional neural network, processes feature sequences using a graph neural network and a long short-term memory network, and finally maps the features to specific biochemical parameters through a fully connected layer to obtain prediction results of protein content, moisture content, and starch content;

[0059] The data visualization module sends the obtained prediction results to the memory and visualizes the prediction results.

[0060] The crop seed data collection module stores the seed imaging information of different bands in the memory, and then passes it to the CNN network through the memory, automatically extracting local features and hierarchical representations in the image through multi-layer convolution and pooling operations. After the imaging information enters the convolutional neural network, it learns and identifies the key features of the seeds in different bands and encodes them into sequence vectors. . Then the sequence vector Input into the GCN network, The spatial embedding matrix is ​​constructed through the GCN network to obtain the spatial embedding vector sequence , and then obtain the spatial dependency information and structural information between features. Then the vector sequence Input to the LSTM network, and get the output vector of each band through LSTM . The output vector The data is passed to the next fully connected layer, and the fully connected layer FCL obtains the prediction results of protein content, starch content, moisture content, etc. by weighted summation of the vectors and sigmoid activation function processing. In order to visualize the prediction results, the prediction results are sent to the memory and displayed through the data visualization module.

[0061] Figure 4 A schematic diagram of an intelligent monitoring device for biochemical parameters of crop seeds is shown.

[0062] This embodiment provides an intelligent monitoring device for crop seed biochemical parameters, including: a main control board, a memory, a feed place, a hyperspectral camera, a halogen lamp, a display, a transformer and an external power supply;

[0063] The feeding place is used for placing and taking out crop seeds, and the crop seeds are placed in by pressing the feeding place;

[0064] The hyperspectral camera is used to obtain imaging information of different bands of crop seeds;

[0065] The halogen lamps are evenly distributed around the hyperspectral camera to illuminate the crop seeds and provide an internal lighting environment for the detection device;

[0066] The display is used to display the prediction results of crop seed biochemical parameters.

[0067] The user places the crop seeds at the feed point and starts the equipment immediately. The main control board not only has precise control capabilities, but also can quickly respond to the user's operating instructions to control the halogen lamp to reach the preset optimal brightness conditions in a short time. Through the precise control of the main control board, the evenly distributed halogen lamp array is activated to create an ideal lighting environment for the hyperspectral camera. When the transposed internal brightness meets the preset conditions, the hyperspectral camera automatically starts and uses its highly sensitive sensor and advanced multi-band imaging technology to capture the hyperspectral information of crop seeds. After capturing the data, the hyperspectral camera transfers it back to the memory for storage. The memory uses high-speed reading and writing technology to ensure that the data can be recorded and saved quickly and accurately for subsequent analysis and processing. The stored data is predicted using the pre-trained spatiotemporal graph network to obtain the predicted results of the biochemical parameters of the crop seeds. The data collection is completed, and the predicted results of the biochemical parameters of the crop seeds are visualized using the display.

[0068] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0069] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0070] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0071] Those skilled in the art can understand that all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiments; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), disks or optical disks, and other media that can store program codes.

[0072] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0073] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for predicting crop seed biochemical parameters based on hyperspectral, characterized in that: The following steps are involved: Acquire imaging information of crop seeds in different bands through a hyperspectral camera, and input the imaging information into a convolutional neural network to extract image features and generate a sequence vector; Based on the graph neural network and LSTM network, a spatiotemporal graph network structure is established to construct a seed biochemical parameter prediction model, the sequence vector is used as the model input, and the graph neural network is used to construct a spatial embedding vector sequence to obtain spatial dependency information and structural information; Input the spatial embedding vector sequence into the LSTM network to capture the dynamic characteristics of the image changing with the spectrum, and obtain the output vector of each band; The output vectors are weighted summed using a fully connected layer, and an activation function is used to process the vectors to obtain the prediction results of the biochemical parameters of the crop seeds.

2. The method for predicting crop seed biochemical parameters based on hyperspectral according to claim 1, characterized in that: The imaging information is input into a convolutional neural network to extract image features and generate a sequence vector. Specifically, a hyperspectral camera is used to obtain different types of crop seeds in different bands. Imaging information , inputting the imaging information into a convolutional neural network; Through multi-layer convolution and pooling operations, local features and hierarchical representations in imaging information are automatically extracted to identify key features of crop seeds in different bands, including shape features, texture features, and color features, and the key features are encoded into sequence vectors. .

3. The method for predicting crop seed biochemical parameters based on hyperspectral according to claim 1, characterized in that: The sequence vector is used as the model input, and the graph neural network is used to construct a spatial embedding vector sequence, specifically: The sequence vector is input into the graph neural network for spatial modeling. Each key feature is used as a node, and the relationship between nodes is used as the edge structure. The graph neural network is used to obtain the spatial dependency information and structural information of the key features based on the relationship between different key features. In the graph convolution layer of the graph neural network, the state of each node is updated through information aggregation, and the spatial structure embedding matrix is ​​constructed through the graph neural network to obtain the spatial embedding vector sequence.

4. The method for predicting crop seed biochemical parameters based on hyperspectral according to claim 1, characterized in that: The spatial embedding vector sequence is input into the LSTM network to capture the dynamic characteristics of the image as the spectrum changes, specifically: The spatial embedding vector sequence output by the graph neural network is input into the LSTM network. The relationship between key features in the spatial embedding vector sequence is obtained through the forget gate and memory cell state update, capturing the dynamic characteristics of imaging information changing with the spectrum. The spatial embedding vector sequence is calculated through the LSTM network to obtain the output vector of each band ,in These are the different bands of the hyperspectral camera.

5. The method for predicting crop seed biochemical parameters based on hyperspectral according to claim 1, characterized in that: The crop seed biochemical parameters include protein content, starch content and moisture content; The output vector The data is passed to the fully connected layer, which performs weighted summation of the vectors and sigmoid activation function to obtain the prediction results of protein content, starch content, and moisture content.

6. A crop seed biochemical parameter prediction system based on hyperspectral, characterized in that: The method for predicting crop seed biochemical parameters based on hyperspectral according to any one of claims 1 to 5 is implemented, the system comprising: a data acquisition module, a seed biochemical parameter prediction module and a data visualization module; The data acquisition module uses a spectral camera to obtain imaging information of different bands of different crop seeds and stores it in a memory; The seed biochemical parameter prediction module extracts key features of imaging information through a convolutional neural network, processes feature sequences using a graph neural network and a long short-term memory network, and finally maps the features to specific biochemical parameters through a fully connected layer to obtain prediction results of protein content, moisture content, and starch content; The data visualization module sends the obtained prediction results to the memory and visualizes the prediction results.

7. An intelligent monitoring device for crop seed biochemical parameters, applied to the crop seed biochemical parameter prediction method based on hyperspectral according to any one of claims 1 to 5, characterized in that: include: Main control board, memory, feeder, hyperspectral camera, halogen lamp, display, transformer and external power supply. The halogen lamp is evenly distributed around the hyperspectral camera. The feeding place is used for placing and taking out crop seeds, and the crop seeds are placed in by pressing the feeding place; The hyperspectral camera is used to obtain imaging information of different bands of crop seeds; The halogen lamp is used to irradiate the crop seeds and provide an internal lighting environment for the detection device; The display is used to display the prediction results of crop seed biochemical parameters.

Images