Wheat disease identification method and device

Abstract

The invention provides a wheat disease identification method and device, and is applied to the technical field of crop disease identification. The method comprises the steps of obtaining a training sample set; determining mass-to-charge ratio signal intensity data of volatile organic compounds of the wheat leaf samples in the training sample set, determining a maximum characteristic value of each mass-to-charge ratio channel in the mass-to-charge ratio signal intensity data, mapping the maximum characteristic values of all the mass-to-charge ratio channels into pixel points, and obtaining a two-dimensional pixel matrix of the wheat leaf samples; extracting parent ions and pixel points where fragment ions corresponding to the parent ions are located from the two-dimensional pixel matrix, and rearranging the positions of the extracted pixel points in the two-dimensional pixel matrix to obtain a recombined feature map; and inputting the recombinant feature map into a wheat disease recognition model for network training, wherein the wheat disease recognition model is used for recognizing the disease type and infection degree of wheat according to the mass-to-charge ratio signal features in the recombinant feature map.

Description

Technical Field

[0001] The present invention relates to the technical field of crop disease identification, and in particular to a wheat disease identification method and device. Background Art

[0002] Traditional disease detection technologies, including polymerase chain reaction, enzyme-linked immunosorbent assay, and fluorescence in situ hybridization, rely heavily on the appearance of disease symptoms, leading to technical bottlenecks in early and accurate detection. Plant diseases often develop rapidly and spread rapidly, leaving a very short window for effective prevention and control. These technologies, with inherent drawbacks such as long detection cycles and delayed responses, are no longer able to meet the urgent need for rapid disease warning and early intervention in modern agriculture. Summary of the Invention

[0003] The present invention provides a wheat disease identification method and device to solve the problems of the prior art such as long detection cycle and delayed response, which are difficult to adapt to the urgent needs of modern agriculture for rapid disease warning and early intervention.

[0004] The invention provides a wheat disease identification method, comprising: obtaining a training sample set, wherein the training sample set includes wheat leaf samples in different health states, and each wheat leaf sample is annotated with a disease type and a disease degree; determining mass-to-charge ratio signal intensity data of volatile organic compounds of the wheat leaf samples in the training sample set, and determining the maximum eigenvalue of each mass-to-charge ratio channel in the mass-to-charge ratio signal intensity data, mapping the maximum eigenvalues of all mass-to-charge ratio channels into pixel points to obtain a two-dimensional pixel matrix of the wheat leaf samples, wherein one wheat leaf sample corresponds to one two-dimensional pixel matrix; extracting pixel points where a parent ion and its corresponding fragment ion are located from the two-dimensional pixel matrix, and rearranging the positions of the extracted pixel points in the two-dimensional pixel matrix to obtain a recombined feature map; and inputting the recombined feature map into a wheat disease identification model for network training, wherein the wheat disease identification model is used to identify the disease type and the disease degree of wheat according to the mass-to-charge ratio signal characteristics in the recombined feature map.

[0005] According to a wheat disease identification method provided by the present invention, the pixel points where the parent ion and its corresponding fragment ion are located are extracted from the two-dimensional pixel matrix, and the positions of the extracted pixel points in the two-dimensional pixel matrix are rearranged to obtain a recombined feature map, including: when the prior knowledge of the target disease is determined, the pixel points of the corresponding mass-to-charge ratio channel are extracted for feature recombining according to the correlation between the parent ion and its fragment ions in the prior knowledge of the target disease; when the prior knowledge of the target disease is not determined, the pixel points where the parent ion and its corresponding fragment ion are located are feature recombined according to the universal feature combination of volatile organic compounds released by wheat leaf samples.

[0006] According to a wheat disease identification method provided by the present invention, the prior knowledge of the target disease is derived from analysis literature of gas chromatography-mass spectrometry technology.

[0007] According to a wheat disease identification method provided by the present invention, the wheat disease identification model adopts a lightweight convolutional neural network, and the lightweight convolutional neural network includes: an input layer, three convolution layers, three maximum pooling layers, a flattening layer, two fully connected layers and an output layer.

[0008] According to a wheat disease identification method provided by the present invention, determining the mass-to-charge ratio signal intensity data of volatile organic compounds of wheat leaf samples in the training sample set includes: The mass-to-charge ratio signal intensity data of volatile organic compounds in wheat leaf samples were detected by proton transfer reaction-mass spectrometry.

[0009] The present invention also provides a wheat disease identification device, comprising the following modules: an acquisition module and a processing module; the acquisition module is used to acquire a training sample set, the training sample set including wheat leaf samples in different health states, each wheat leaf sample being marked with a disease type and a disease degree; the processing module is used to determine the mass-to-charge ratio signal intensity data of volatile organic compounds of the wheat leaf samples in the training sample set, and determine the maximum eigenvalue of each mass-to-charge ratio channel in the mass-to-charge ratio signal intensity data, and map the maximum eigenvalues of all mass-to-charge ratio channels into pixel points to obtain a two-dimensional pixel matrix of the wheat leaf samples, wherein one wheat leaf sample corresponds to one two-dimensional pixel matrix; the pixel points where the parent ion and its corresponding fragment ion are located are extracted from the two-dimensional pixel matrix, and the positions of the extracted pixel points in the two-dimensional pixel matrix are rearranged to obtain a recombined feature map; the recombined feature map is input into a wheat disease identification model for network training, and the wheat disease identification model is used to identify the disease type and the disease degree of wheat according to the mass-to-charge ratio signal characteristics in the recombined feature map.

[0010] According to a wheat disease identification device provided by the present invention, the processing module is used to extract the pixel points of the corresponding mass-to-charge ratio channel for feature reconstruction based on the correlation between the parent ion and its fragment ions in the prior knowledge of the target disease when the prior knowledge of the target disease is determined; when the prior knowledge of the target disease is not determined, the pixel points where the parent ion and its corresponding fragment ion are located are subjected to feature reconstruction based on the universal feature combination of volatile organic compounds released by the wheat leaf sample.

[0011] According to a wheat disease identification device provided by the present invention, the target disease prior knowledge is derived from analysis literature of gas chromatography-mass spectrometry technology.

[0012] According to a wheat disease identification device provided by the present invention, the wheat disease identification model adopts a lightweight convolutional neural network, and the lightweight convolutional neural network includes: an input layer, three convolution layers, three maximum pooling layers, a flattening layer, two fully connected layers and an output layer.

[0013] According to the wheat disease identification device provided by the present invention, the processing module is used to detect the mass-to-charge ratio signal intensity data of volatile organic compounds in wheat leaf samples using proton transfer reaction mass spectrometry.

[0014] The present invention also provides an electronic device, comprising 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, any one of the wheat disease identification methods described above is implemented.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the wheat disease identification methods described above.

[0016] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-mentioned wheat disease identification methods.

[0017] The wheat disease identification method and device provided by the present invention, on the one hand, effectively suppresses background noise and signal fluctuation interference by extracting the maximum eigenvalue of each mass-to-charge ratio channel and mapping it into a two-dimensional pixel matrix, highlighting key disease-related features and improving the sensitivity of the wheat disease identification model to weak disease signals. On the other hand, by extracting and rearranging the pixels of the parent ion and its fragment ions, the scattered chemical signals are integrated into a structured feature map, significantly enhancing the ability to express disease characteristics. In this way, high-precision early identification of wheat diseases can be achieved, providing key technical support for rapid early warning and early intervention of diseases in modern agriculture. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 This is one of the flow charts of the wheat disease identification method provided by the present invention.

[0020] Figure 2 This is the second flow chart of the wheat disease identification method provided by the present invention.

[0021] Figure 3 It is a structural schematic diagram of the wheat disease identification device provided by the present invention.

[0022] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0023] To make the objectives, technical solutions, and advantages of this application more clear, the technical solutions of this application will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments of this application, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of this application.

[0024] It should be noted that in the embodiments of this application, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be interpreted as being more preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0025] It should be noted that, in this article, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the statement "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element. In addition, it should be noted that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in the opposite order according to the functions involved. For example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0026] In order to facilitate a clear description of the technical solutions of the embodiments of the present application, in the embodiments of the present application, words such as "first" and "second" are used to distinguish between identical or similar items with basically the same functions and effects. Those skilled in the art can understand that words such as "first" and "second" do not limit the quantity and execution order.

[0027] The embodiments of the present application describe some exemplary embodiments for the purpose of explanation. It should be understood that the present application can be implemented in other ways that are not specifically shown in the drawings.

[0028] like Figure 1 As shown, the embodiment of the present application provides a wheat disease identification method, which can be applied to a wheat disease identification device. The wheat disease identification method may include S101-S104: S101. A wheat disease recognition device obtains a training sample set, where the training sample set includes wheat leaf samples in different health states, and each wheat leaf sample is labeled with a disease type and infection degree.

[0029] Specifically, to build an accurate wheat disease identification model, it is necessary to scientifically collect and label wheat leaf samples in different health states. The core of this process stems from the knowledge accumulated by plant pathologists through long-term research, namely that different wheat diseases present unique visual symptoms on leaves. For example, in the early stages of wheat powdery mildew, small white mold spots appear on the leaf surface. As the disease worsens, the mold spots gradually expand and connect into patches, forming a thick layer of powdery mildew. In the early stages of stripe rust, chlorotic spots appear on the leaves, followed by the growth of bright yellow summer spores. In severe cases, the spores cover the entire leaf, causing the leaves to wither and turn yellow. Based on these characteristics, experts can make scientific judgments about the disease.

[0030] In the actual operation, sample collection must follow a rigorous process. First, representative wheat plants are selected from wheat fields at different growth stages and in different plots. For healthy wheat leaves, select samples with complete leaves, bright green color, and no lesions; for diseased leaves, collect samples at different infection stages according to the degree of disease development. Taking powdery mildew as an example, in the early infection stage, carefully look for leaves with only a few small white mold spots; in the severe infection stage, collect samples that are covered with white mold, curled leaves, or even yellowed. When collecting, record the collection location, time, wheat variety, and other information of the sample. After collection, the sample is sent to a professional laboratory, and the pathologist will mark the disease type and degree of infection of each sample based on the morphology, distribution range, color, and other symptoms of the leaf lesions, combined with professional knowledge and experience. During the labeling process, tools such as microscopes can be used to observe the microstructure of the lesion tissue to further confirm the disease type and ensure the accuracy of the labeling.

[0031] This approach, on the one hand, comprehensively covers samples from various disease types and their different infection stages, providing a rich data foundation for the constructed training sample set. This enables the subsequently constructed wheat disease recognition model to learn more comprehensive and diverse disease characteristics, improving the model's recognition accuracy and generalization capabilities. Furthermore, relying on the expertise of pathologists for labeling ensures the scientific and reliable nature of sample annotation, avoiding model training biases caused by inaccurate labeling.

[0032] S102. The wheat disease identification apparatus determines mass-to-charge ratio signal intensity data of volatile organic compounds of wheat leaf samples in the training sample set, determines the maximum eigenvalue of each mass-to-charge ratio channel in the mass-to-charge ratio signal intensity data, and maps the maximum eigenvalues of all mass-to-charge ratio channels to pixels to obtain a two-dimensional pixel matrix of the wheat leaf samples. One wheat leaf sample corresponds to one two-dimensional pixel matrix.

[0033] Studies have found that plants release specific volatile organic compounds (VOCs) when they are stressed by pathogens. These VOCs serve as important signal molecules between plants or between plants and the environment, and can reflect the physiological state of plants before the disease shows obvious symptoms.

[0034] Alternatively, the wheat disease identification device can use proton transfer reaction mass spectrometry to detect the mass-to-charge ratio signal intensity data of volatile organic compounds in wheat leaf samples. Specifically, wheat can release VOCs of different compositions and concentrations when in a healthy state and when infected with different diseases. Proton transfer reaction mass spectrometry (PTR-MS) technology can use hydronium ions ( ) as reagent ions, undergoing proton transfer reaction with the VOCs molecules in the wheat leaf sample, ionizing the VOCs molecules and detecting them by the mass spectrometer to obtain mass-to-charge ratio signal intensity data.

[0035] like Figure 2 As shown in the figure, after obtaining the mass-to-charge ratio signal intensity data, the wheat disease identification device can perform data preprocessing. Because the signal is unstable during the initial detection phase, this data needs to be eliminated, retaining only the mass-to-charge ratio signal intensity data during the stable phase. The maximum eigenvalue of each mass-to-charge ratio channel is then extracted as the valid eigenvalue. Finally, the maximum eigenvalues of all mass-to-charge ratio channels are mapped to pixels in the RGB three-channel pseudo-color image and arranged in a fixed order in a two-dimensional image grid of size H×W. For example, if the image size is set to 100×100, the maximum eigenvalues of different mass-to-charge ratio channels are mapped to the corresponding pixel positions, forming a two-dimensional pixel matrix of the wheat leaf sample.

[0036] The specific processing process is as follows: First, place the wheat leaf samples that have been marked with the disease type and degree of infection into 50mL sealed gas collection bottles one by one. For example, healthy leaves, leaves infected with early and severe powdery mildew are placed in different gas collection bottles, and left to stand at room temperature for 30-40 minutes to allow the leaves to fully release VOCs. The gas collection bottle is equipped with an inlet and outlet for easy connection to the PTR-MS injection system. After the enrichment is completed, connect the air inlet of the PTR-MS system to the air outlet of the gas collection bottle, start the detection program, and continue measuring for 2 minutes. The key parameters are set during the detection. The mass-to-charge ratio detection range is 20-150, which can cover the mass-to-charge ratio range of common VOCs; the drift tube voltage is 450V, and the single mass-to-charge ratio signal scanning time is 100ms to ensure the sensitivity and accuracy of the detection.

[0037] Based on the above scheme, from the perspective of detection efficiency, compared with the traditional method that relies on manual observation of leaf morphology, the PTR-MS detection and data processing process can complete a large number of sample analyses in a relatively short period of time, greatly improving detection efficiency. In terms of accuracy, VOCs, as an early physiological response indicator of plant diseases, can detect disease information before obvious visible spots appear on the leaves. Combined with the high-sensitivity detection characteristics of PTR-MS, it can achieve early and accurate diagnosis of wheat diseases. In addition, converting the data into a two-dimensional pixel matrix in RGB format can be adapted to existing mature image analysis models without the need to redevelop complex algorithms, which lowers the technical threshold. At the same time, it also improves the stability and reliability of model training and disease identification, which helps to quickly deploy wheat disease identification systems in agricultural production, take timely prevention and control measures, and reduce economic losses caused by diseases.

[0038] S103. The wheat disease identification device extracts pixel points where the parent ion and its corresponding fragment ion are located from the two-dimensional pixel matrix, and rearranges the positions of the extracted pixel points in the two-dimensional pixel matrix to obtain a recombined feature map.

[0039] Optionally, the wheat disease identification device extracts the pixel points where the parent ion and its corresponding fragment ion are located from the two-dimensional pixel matrix, and rearranges the positions of the extracted pixel points in the two-dimensional pixel matrix to obtain a recombined feature map, including: when the prior knowledge of the target disease is determined, according to the correlation between the parent ion and its fragment ion in the prior knowledge of the target disease, the pixel points of the corresponding mass-to-charge ratio channel are extracted for feature recombining; when the prior knowledge of the target disease is not determined, according to the universal feature combination of volatile organic compounds released by the wheat leaf sample, the pixel points where the parent ion and its corresponding fragment ion are located are feature recombined.

[0040] Specifically, when using PTR-MS to detect VOCs in wheat leaves for disease identification, the chemical ionization process in PTR-MS can cause long-chain VOC molecules to break apart, producing fragment ions. These fragment ions overlap with the parent ion signal, making individual VOC identification difficult and affecting the accuracy of disease identification. Furthermore, in practical applications, prior knowledge of relevant VOCs for certain target diseases may be lacking, which also poses a challenge to accurate identification.

[0041] However, on the one hand, it is known that when compounds of different chemical structures undergo structural fragmentation during the PTR-MS ionization process, the released fragment ions often show a regular distribution, and there is a correlation between the compound's parent ion and its major fragment ions, which provides a basis for feature reconstruction based on prior knowledge. On the other hand, the common categories of VOCs released by wheat, such as aldehydes, alcohols, and terpenes, have a certain degree of universality, which can be used to construct a universal feature combination strategy to address the lack of prior knowledge of VOCs related to target diseases.

[0042] The specific processing process is as follows: First, determine whether there is prior knowledge of VOCs related to the target disease. Figure 2 If present, the corresponding mass-to-charge ratio channel signals are extracted from the preliminarily constructed two-dimensional pixel matrix based on the correlation between the parent ion and its major fragment ions for feature reconstruction. For example, if powdery mildew-specific VOCs are known, their positions in the image are rearranged according to preset rules based on their parent ions and corresponding major fragment ions. If there is no prior knowledge of relevant VOCs for the target disease, a universal feature combination strategy is constructed based on common wheat VOC categories (such as alkenes, cycloalkanes, alkylbenzenes, alcohols, aldehydes, terpenes, etc.).

[0043] For cases where prior knowledge of the target disease is available, the mass-to-charge ratio of the corresponding parent ion in PTR-MS analysis is calculated by adding one to the relative molecular mass of the known compound ([M+H]+). Because PTR-MS utilizes chemical ionization, target compounds are susceptible to structural fragmentation during the ionization process. Compounds of different chemical structures often exhibit a regular distribution of fragment ions released during fragmentation. Table 1 summarizes the typical fragment ion distributions for six common VOC types, which can serve as a basis for feature reconstruction.

[0044] In the absence of prior knowledge about VOCs associated with target diseases, a generalized feature combination strategy was used for recombinant analysis. This strategy targets common VOCs emitted by wheat, including alkylbenzenes (with parent ions 105, 121, 129, and 135), aldehydes (with parent ions 107 and 143), alkenes (with parent ions 109 and 111), alcohols (with parent ions 117 and 127), cycloalkanes (with parent ion 119), and monoterpenes (with parent ion 137). Typical fragment ions corresponding to each category are shown in Table 1. This strategy allows for the construction of a well-structured, generalized feature combination scheme with broad applicability.

[0045] Table 1

[0046] Optionally, the above-mentioned prior knowledge of target diseases is derived from analytical literature on gas chromatography-mass spectrometry technology.

[0047] Based on the above scheme, firstly, by integrating the feature recombination of prior knowledge, the expression intensity of disease-related signals in the model is effectively enhanced, and the misjudgment caused by fragment interference is significantly reduced. Secondly, it supports two combinations of specificity of known disease-related VOCs and universality of unknown disease-related VOCs, which greatly improves the adaptability of the model. When the relevant VOCs knowledge released by the target disease is known, the combination of specific features based on prior knowledge can improve the exclusive recognition ability; when there is a lack of relevant knowledge, a universal combination scheme based on common VOCs structural categories can improve the scenario adaptability of the model. In addition, this processing process realizes the structured expression of VOCs data, provides more interpretable input features for model training, helps to improve the generalization ability of the model and the accuracy of disease identification, and provides strong support for the early and accurate identification of wheat diseases.

[0048] S104. The wheat disease recognition device inputs the recombined feature map into a wheat disease recognition model for network training.

[0049] The wheat disease recognition model is used to identify the type and degree of wheat disease based on the mass-to-charge ratio signal characteristics in the recombinant feature map.

[0050] Specifically, if Figure 2As shown in the figure, the recombined feature image is input into a wheat disease recognition model for training, a mapping relationship between mass-to-charge ratio signal features and wheat disease severity is established, and an intelligent discrimination model for wheat health status is constructed. Once the wheat disease recognition model is established, the wheat leaves to be tested are placed in a gas collection bottle to enrich the VOCs released. The mass-to-charge ratio signal intensity data is obtained using a PTR-MS system. Feature extraction, recombination, and image conversion are completed according to the established recognition process. Finally, the trained wheat disease recognition model outputs the corresponding disease type and infection severity of the sample, achieving rapid identification and intelligent classification of wheat diseases.

[0051] Optionally, the wheat disease recognition model adopts a lightweight convolutional neural network, which includes: an input layer, three convolution layers, three maximum pooling layers, a flattening layer, two fully connected layers and an output layer.

[0052] Specifically, the convolutional layer uses a 3×3 convolution kernel for local feature extraction, with a stride of 1 and a "same" padding method to maintain the feature map size. To enhance the model's generalization and prevent overfitting, a dropout layer is introduced after each max pooling layer, randomly dropping 20% of the neuronal connections. This network structure effectively extracts spatial correlation information from the two-dimensional mass-to-charge ratio feature map and improves the discrimination performance for samples of different disease levels.

[0053] In the embodiments of the present application, on the one hand, by extracting the maximum eigenvalue of each mass-to-charge ratio channel and mapping it into a two-dimensional pixel matrix, background noise and signal fluctuation interference are effectively suppressed, key features related to the disease are highlighted, and the sensitivity of the wheat disease recognition model to weak disease signals is improved; on the other hand, by extracting the pixel points of the parent ion and its fragment ions and rearranging them, the scattered chemical signals are integrated into a structured feature map, which significantly enhances the expression ability of disease characteristics. In this way, high-precision identification of wheat diseases in the early stage can be achieved, providing key technical support for rapid early warning and early intervention of diseases in modern agriculture.

[0054] The above mainly introduces the solution provided by the embodiment of the present application from the perspective of method. In order to realize the above functions, it includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should easily appreciate that, in combination with the units and algorithm steps of each example described in the embodiments disclosed herein, the embodiments of the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in a hardware or computer software driven hardware manner depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0055] The wheat disease identification method provided in the embodiments of the present application can be executed by a wheat disease identification device or a control module for wheat disease identification in the wheat disease identification device. The wheat disease identification device provided in the embodiments of the present application is described by taking the wheat disease identification device executing the wheat disease identification method as an example.

[0056] It should be noted that the embodiment of the present application can divide the wheat disease identification device into functional modules according to the above method example. For example, each functional module can be divided according to each function, or two or more functions can be integrated into one processing module. The above integrated modules can be implemented in the form of hardware or in the form of software functional modules. Optionally, the division of modules in the embodiment of the present application is schematic and is only a logical functional division. In actual implementation, other division methods can be used.

[0057] like Figure 3 As shown, an embodiment of the present application provides a wheat disease identification device 300. The wheat disease identification device 300 includes: an acquisition module 301 and a processing module 302. The acquisition module 301 can be used to obtain a training sample set, wherein the training sample set includes wheat leaf samples in different health states, each wheat leaf sample is annotated with the disease type and disease severity; the processing module 302 is used to determine the mass-to-charge ratio signal intensity data of volatile organic compounds of the wheat leaf samples in the training sample set, and determine the maximum eigenvalue of each mass-to-charge ratio channel in the mass-to-charge ratio signal intensity data, and map the maximum eigenvalues of all mass-to-charge ratio channels to pixel points to obtain a two-dimensional pixel matrix of the wheat leaf samples, where one wheat leaf sample corresponds to one two-dimensional pixel matrix; extract the pixel points where the parent ion and its corresponding fragment ion are located from the two-dimensional pixel matrix, and rearrange the positions of the extracted pixel points in the two-dimensional pixel matrix to obtain a recombined feature map; the recombined feature map is input into a wheat disease identification model for network training, and the wheat disease identification model is used to identify the disease type and disease severity of wheat based on the mass-to-charge ratio signal characteristics in the recombined feature map.

[0058] Optionally, the processing module 302 is used to extract the pixel points of the corresponding mass-to-charge ratio channel for feature reconstruction based on the correlation between the parent ion and its fragment ions in the prior knowledge of the target disease when the prior knowledge of the target disease is determined; when the prior knowledge of the target disease is not determined, the pixel points where the parent ion and its corresponding fragment ion are located are feature reconstructed based on the universal feature combination of volatile organic compounds released by the wheat leaf sample.

[0059] Optionally, the prior knowledge of the target disease is derived from analytical literature on gas chromatography-mass spectrometry technology.

[0060] Optionally, the wheat disease recognition model adopts a lightweight convolutional neural network, which includes: an input layer, three convolution layers, three maximum pooling layers, a flattening layer, two fully connected layers and an output layer.

[0061] Optionally, the processing module 302 is configured to detect the mass-to-charge ratio signal intensity data of volatile organic compounds in the wheat leaf sample using proton transfer reaction mass spectrometry.

[0062] In the embodiments of the present application, on the one hand, by extracting the maximum eigenvalue of each mass-to-charge ratio channel and mapping it into a two-dimensional pixel matrix, background noise and signal fluctuation interference are effectively suppressed, key features related to the disease are highlighted, and the sensitivity of the wheat disease recognition model to weak disease signals is improved; on the other hand, by extracting the pixel points of the parent ion and its fragment ions and rearranging them, the scattered chemical signals are integrated into a structured feature map, which significantly enhances the expression ability of disease characteristics. In this way, high-precision identification of wheat diseases in the early stage can be achieved, providing key technical support for rapid early warning and early intervention of diseases in modern agriculture.

[0063] Figure 4 An example of a physical structure diagram of an electronic device is shown below. Figure 4 As shown, the electronic device may include: a processor (processor) 410 , a communication interface (Communications Interface) 420 , a memory (memory) 430 and a communication bus 440 , wherein the processor 410 , the communication interface 420 , and the memory 430 communicate with each other via the communication bus 440 . The processor 410 can call the logic instructions in the memory 430 to execute the wheat disease identification method, which includes: obtaining a training sample set, wherein the training sample set includes wheat leaf samples in different health states, and each wheat leaf sample is marked with a disease type and a degree of infection; determining the mass-to-charge ratio signal intensity data of volatile organic compounds of the wheat leaf samples in the training sample set, and determining the maximum eigenvalue of each mass-to-charge ratio channel in the mass-to-charge ratio signal intensity data, mapping the maximum eigenvalues of all mass-to-charge ratio channels into pixel points to obtain a two-dimensional pixel matrix of the wheat leaf samples, where one wheat leaf sample corresponds to one two-dimensional pixel matrix; extracting the pixel points where the parent ion and its corresponding fragment ion are located from the two-dimensional pixel matrix, and rearranging the positions of the extracted pixel points in the two-dimensional pixel matrix to obtain a recombined feature map; inputting the recombined feature map into a wheat disease identification model for network training, and the wheat disease identification model is used to identify the disease type and degree of infection of wheat based on the mass-to-charge ratio signal characteristics in the recombined feature map.

[0064] Furthermore, the logic instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product, stored in a storage medium, includes instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0065] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the wheat disease identification method provided by the above methods, the method including: obtaining a training sample set, the training sample set including wheat leaf samples in different health states, each wheat leaf sample marked with a disease type and a degree of infection; determining the mass-to-charge ratio signal intensity data of volatile organic compounds of the wheat leaf samples in the training sample set, and determining the maximum eigenvalue of each mass-to-charge ratio channel in the mass-to-charge ratio signal intensity data, mapping the maximum eigenvalues of all mass-to-charge ratio channels into pixel points to obtain a two-dimensional pixel matrix of the wheat leaf samples, one wheat leaf sample corresponding to one two-dimensional pixel matrix; extracting the pixel points where the parent ion and its corresponding fragment ion are located from the two-dimensional pixel matrix, and rearranging the positions of the extracted pixel points in the two-dimensional pixel matrix to obtain a recombined feature map; inputting the recombined feature map into a wheat disease recognition model for network training, the wheat disease recognition model is used to identify the disease type and degree of infection of wheat according to the mass-to-charge ratio signal characteristics in the recombined feature map.

[0066] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the wheat disease identification method provided by the above-mentioned methods, the method comprising: obtaining a training sample set, the training sample set comprising wheat leaf samples in different health states, each wheat leaf sample being marked with a disease type and a degree of infection; determining the mass-to-charge ratio signal intensity data of volatile organic compounds of the wheat leaf samples in the training sample set, and determining the maximum eigenvalue of each mass-to-charge ratio channel in the mass-to-charge ratio signal intensity data, mapping the maximum eigenvalues of all mass-to-charge ratio channels into pixel points to obtain a two-dimensional pixel matrix of the wheat leaf samples, wherein one wheat leaf sample corresponds to one two-dimensional pixel matrix; extracting the pixel points where the parent ion and its corresponding fragment ion are located from the two-dimensional pixel matrix, and rearranging the positions of the extracted pixel points in the two-dimensional pixel matrix to obtain a recombined feature map; inputting the recombined feature map into a wheat disease recognition model for network training, the wheat disease recognition model being used to identify the disease type and degree of infection of wheat based on the mass-to-charge ratio signal characteristics in the recombined feature map.

[0067] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0068] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A wheat disease identification method, characterized in that: include: Obtaining a training sample set, the training sample set comprising wheat leaf samples in different health states, each wheat leaf sample being labeled with a disease type and a disease severity; Determining mass-to-charge ratio signal intensity data of volatile organic compounds of wheat leaf samples in the training sample set, and determining the maximum eigenvalue of each mass-to-charge ratio channel in the mass-to-charge ratio signal intensity data, mapping the maximum eigenvalues of all mass-to-charge ratio channels into pixel points to obtain a two-dimensional pixel matrix of the wheat leaf samples, where one wheat leaf sample corresponds to one two-dimensional pixel matrix; Extracting pixel points where the parent ion and its corresponding fragment ion are located from the two-dimensional pixel matrix, and rearranging the positions of the extracted pixel points in the two-dimensional pixel matrix to obtain a recombined feature map; The recombinant characteristic graph is input into a wheat disease recognition model for network training. The wheat disease recognition model is used to identify the disease type and infection degree of wheat according to the mass-to-charge ratio signal characteristics in the recombinant characteristic graph.

2. The wheat disease identification method according to claim 1, characterized in that: The step of extracting pixel points where the parent ion and its corresponding fragment ion are located from the two-dimensional pixel matrix and rearranging the positions of the extracted pixel points in the two-dimensional pixel matrix to obtain a recombined feature map includes: When prior knowledge of the target disease is determined, pixels of the corresponding mass-to-charge ratio channel are extracted for feature reconstruction based on the correlation between the parent ion and its fragment ions in the prior knowledge of the target disease; Without prior knowledge of the target disease, the pixel points where the parent ion and its corresponding fragment ion are located are feature recombined according to the universal feature combination of volatile organic compounds released by wheat leaf samples.

3. The wheat disease identification method according to claim 2, characterized in that: The prior knowledge of the target disease is derived from analytical literature on gas chromatography-mass spectrometry technology.

4. The wheat disease identification method according to claim 1, characterized in that: The wheat disease recognition model adopts a lightweight convolutional neural network, which includes: an input layer, three convolution layers, three maximum pooling layers, a flattening layer, two fully connected layers and an output layer.

5. The wheat disease identification method according to claim 1, characterized in that: Determining the mass-to-charge ratio signal intensity data of volatile organic compounds of the wheat leaf samples in the training sample set includes: The mass-to-charge ratio signal intensity data of volatile organic compounds in wheat leaf samples were detected by proton transfer reaction-mass spectrometry.

6. A wheat disease identification device, characterized in that: include: Acquisition module and processing module; The acquisition module is used to acquire a training sample set, wherein the training sample set includes wheat leaf samples in different health states, and each wheat leaf sample is annotated with a disease type and disease severity; The processing module is used to determine the mass-to-charge ratio signal intensity data of volatile organic compounds of wheat leaf samples in the training sample set, and determine the maximum eigenvalue of each mass-to-charge ratio channel in the mass-to-charge ratio signal intensity data, map the maximum eigenvalues of all mass-to-charge ratio channels into pixel points, and obtain a two-dimensional pixel matrix of the wheat leaf samples, where one wheat leaf sample corresponds to one two-dimensional pixel matrix; extract the pixel points where the parent ion and its corresponding fragment ion are located from the two-dimensional pixel matrix, and rearrange the positions of the extracted pixel points in the two-dimensional pixel matrix to obtain a recombined feature map; input the recombined feature map into a wheat disease recognition model for network training, and the wheat disease recognition model is used to identify the disease type and infection degree of wheat according to the mass-to-charge ratio signal characteristics in the recombined feature map.

7. The wheat disease identification device according to claim 6, characterized in that: The processing module is used to extract the pixel points of the corresponding mass-to-charge ratio channel for feature reconstruction based on the correlation between the parent ion and its fragment ions in the prior knowledge of the target disease when the prior knowledge of the target disease is determined; Without prior knowledge of the target disease, the pixel points where the parent ion and its corresponding fragment ion are located are feature recombined according to the universal feature combination of volatile organic compounds released by wheat leaf samples.

8. An electronic device comprising 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 wheat disease identification method according to any one of claims 1 to 5 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the wheat disease identification method according to any one of claims 1 to 5 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the wheat disease identification method according to any one of claims 1 to 5 is implemented.

Images