Hyperspectral early nondestructive detection method and device for bacterial wilt

Abstract

The present application provides a hyperspectral early nondestructive detection method and device for bacterial wilt, the method comprising: collecting a target spectrum of a target plant; extracting target feature parameters corresponding to the optimal solution feature combination from the target spectrum according to the optimal solution feature combination; inputting the target feature parameters into a target model and outputting a detection result; wherein the optimal solution feature combination is obtained based on a genetic algorithm model; the target model is obtained by training a binary classification model based on training samples processed by the optimal solution feature combination; the detection result is used to indicate whether the target plant suffers from bacterial wilt. The hyperspectral early nondestructive detection method and device for bacterial wilt provided by the present application are used for early diagnosis of bacterial wilt and are helpful for the prevention and treatment of bacterial wilt.

Description

Technical Field

[0001] The present application relates to the field of detection of bacterial wilt of tomato plants, and in particular to a hyperspectral early non-destructive detection method and device for bacterial wilt. Background Art

[0002] Tomato is one of the most consumed vegetables in the world and one of the world's major cash crops. Bacterial wilt is a major disease of tomatoes. It develops rapidly and spreads quickly. In severe cases, it can cause large-scale plant death, resulting in reduced or even total crop failure, which seriously affects tomato yield and quality. Early, rapid and accurate detection of bacterial wilt is extremely important for disease prevention and control and crop care.

[0003] However, since the plants remain green during the onset of bacterial wilt, only the top leaves of the plants lighten in color and wilt slightly in the early stage, and recover on their own at night, it is extremely difficult to judge with the naked eye. Among the existing detection methods, expert detection relies on experts and only identifies symptoms when they appear, which is time-consuming, labor-intensive, and subjective. Chemical molecular detection results are accurate, but cannot be performed in real time in the field, and are time-consuming and costly, making it difficult to meet the needs of early and rapid detection of bacterial wilt. Summary of the invention

[0004] The purpose of this application is to provide a hyperspectral early non-destructive detection method and device for bacterial wilt, which is used for the early detection of bacterial wilt.

[0005] The present application provides a hyperspectral early nondestructive detection method for bacterial wilt, comprising:

[0006] Collect a target spectrum of a target plant; extract target feature parameters corresponding to the optimal solution feature combination from the target spectrum according to the optimal solution feature combination; input the target feature parameters into a target model, and output a detection result; wherein the optimal solution feature combination is obtained based on a genetic algorithm model; the target model is a binary classification model obtained by training training samples processed based on the optimal solution feature combination; and the detection result is used to indicate whether the target plant suffers from bacterial wilt.

[0007] Optionally, collecting the target spectrum of the target plant includes: acquiring the target spectrum collected from a target part of the target plant based on a sunlight simulation light source; wherein the target part is a part of the stem of the target plant close to the root.

[0008] Optionally, before extracting the target vegetation index from the target spectrum according to the optimal solution feature combination, the method also includes: collecting a first spectrum of a target part of at least one tomato plant, and preprocessing the spectrum corresponding to each tomato plant using a convolution smoothing method to generate a second spectrum corresponding to each tomato plant; extracting the spectral principal components of the second spectrum corresponding to each tomato plant through principal component analysis PCA, and extracting the vegetation index based on the second spectrum corresponding to each tomato plant using a band operation method; generating a data set based on the spectral principal components and vegetation index corresponding to each tomato plant; wherein the vegetation index includes at least one of the following: water stress index, chlorophyll index, and red edge index.

[0009] Optionally, after generating a data set based on the spectral principal components and vegetation indices corresponding to each tomato plant, the method further includes: inputting the data set into a genetic algorithm model, and selecting a bacterial wilt feature combination corresponding to the individual with the largest fitness from the population finally obtained as the optimal solution feature combination; wherein the bacterial wilt feature combination corresponding to each individual is randomly assigned; and the bacterial wilt feature combination includes: at least one item in a feature set consisting of the spectral principal components and vegetation indices corresponding to the tomato plants.

[0010] Optionally, after inputting the data set into a genetic algorithm model and selecting a bacterial wilt feature combination corresponding to the individual with the highest fitness from the finally obtained population as the optimal solution feature combination, the method further comprises: normalizing the feature combination corresponding to each tomato plant in the data set according to the optimal solution feature combination, and marking the feature combination corresponding to each tomato plant according to whether each tomato plant suffers from bacterial wilt; inputting the marked feature combination into the binary classification model, and training the binary classification model using a ten-fold cross validation method to obtain the target model; wherein the binary classification model is a support vector machine (SVM) binary classification model.

[0011] The present application also provides a hyperspectral early nondestructive detection device for bacterial wilt, comprising:

[0012] A spectrum acquisition module is used to collect a target spectrum of a target plant; a parameter extraction module is used to extract target feature parameters corresponding to the optimal solution feature combination from the target spectrum according to the optimal solution feature combination; a disease diagnosis module is used to input the target feature parameters into a target model and output a detection result; wherein the optimal solution feature combination is obtained based on a genetic algorithm model; the target model is obtained by training a binary classification model based on training samples processed by the optimal solution feature combination; and the detection result is used to indicate whether the target plant suffers from bacterial wilt.

[0013] Optionally, the spectrum acquisition module is specifically used to obtain a target spectrum collected from a target part of the target plant based on a sunlight simulation light source; wherein the target part is a part of the stem of the target plant close to the root.

[0014] Optionally, the device also includes: a calculation module and a generation module; the spectrum acquisition module is also used to collect a first spectrum of a target part of at least one tomato plant, and pre-process the spectrum corresponding to each tomato plant using a convolution smoothing method to generate a second spectrum corresponding to each tomato plant; the calculation module is used to extract the spectral principal components of the second spectrum corresponding to each tomato plant through principal component analysis PCA, and use a band operation method to extract a vegetation index based on the second spectrum corresponding to each tomato plant; the generation module is used to generate a data set according to the spectral principal components and vegetation index corresponding to each tomato plant; wherein the vegetation index includes at least one of the following: water stress index, chlorophyll index, and red edge index.

[0015] Optionally, the device also includes: a feature selection module; the feature selection module is used to input the data set into a genetic algorithm model, and screen out the bacterial wilt feature combination corresponding to the individual with the largest fitness from the population finally obtained, as the optimal solution feature combination; wherein the bacterial wilt feature combination corresponding to each individual is randomly assigned; the bacterial wilt feature combination includes: at least one item in the feature set consisting of the spectral principal component and vegetation index corresponding to the tomato plant.

[0016] Optionally, the device also includes: an initialization module and a model training module; the initialization module is used to normalize the feature combination corresponding to each tomato plant in the data set according to the optimal solution feature combination, and mark the feature combination corresponding to each tomato plant according to whether each tomato plant suffers from bacterial wilt; the model training module is used to input the marked feature combination into the binary classification model, and train the binary classification model using a ten-fold cross-validation method to obtain the target model; wherein the binary classification model is a support vector machine SVM binary classification model.

[0017] The present application also provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of any of the above-mentioned hyperspectral early non-destructive detection methods for bacterial wilt.

[0018] The present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of any of the above-mentioned hyperspectral early nondestructive detection methods for bacterial wilt are implemented.

[0019] The present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the above-mentioned hyperspectral early nondestructive detection methods for bacterial wilt.

[0020] The hyperspectral early nondestructive detection method and device for bacterial wilt provided by the present application, in order to determine whether the target plant suffers from bacterial wilt, needs to collect the target spectrum of the target plant, and according to the optimal solution feature combination obtained based on the genetic algorithm model, extract the target feature parameters corresponding to the optimal solution feature combination from the collected target spectrum. Afterwards, the extracted target feature parameters are input into the target model, and the detection result is output, according to which it can be determined whether the target plant suffers from bacterial wilt. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0022] Figure 1 This is one of the flow diagrams of the hyperspectral early nondestructive detection method for bacterial wilt provided in this application;

[0023] Figure 2 This is a comparison of the average spectra of the stems of healthy plants and plants with bacterial wilt provided by the present application;

[0024] Figure 3 This is the second flow chart of the hyperspectral early nondestructive detection method for bacterial wilt provided in this application;

[0025] Figure 4 It is a fitness comparison chart of bacterial wilt characteristic combinations provided in this application;

[0026] Figure 5 It is a ROC curve diagram of bacterial wilt detection accuracy provided by this application;

[0027] Figure 6 It is a schematic diagram of the structure of the hyperspectral early nondestructive detection device for bacterial wilt provided in the present application;

[0028] Figure 7 It is a schematic diagram of the structure of the electronic device provided by this application. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0030] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0031] Since the plants remain green during the onset of bacterial wilt, only the top leaves of the plants lighten in color and wilt slightly in the early stage, and they recover on their own at night, which is extremely difficult to judge with the naked eye. Among the existing detection methods, expert detection relies on experts and only identifies symptoms when they appear, which is time-consuming, labor-intensive, and subjective. The results of chemical molecular detection are accurate, but they cannot be operated in the field in real time, and they are time-consuming and costly, which makes it difficult to meet the needs of early and rapid detection of bacterial wilt.

[0032] Therefore, in order to solve the above-mentioned technical problems existing in the related technologies, the embodiments of the present application provide a hyperspectral early nondestructive detection method for bacterial wilt. Hyperspectral remote sensing monitoring technology can obtain continuous spectral information of ground objects. By using the diagnostic spectral information, it can realize the quantitative parameter inversion and growth stress information extraction of ground objects, and has obvious advantages in crop nutrient diagnosis and disease and insect pest degree identification. At the same time, remote sensing monitoring has the characteristics of safe, convenient and nondestructive information acquisition, which can greatly improve detection efficiency and reduce costs.

[0033] Currently, there is little research on remote sensing non-destructive detection of bacterial wilt, and there is a gap in early high-precision and rapid detection methods based on hyperspectral technology.

[0034] The hyperspectral early nondestructive detection method for bacterial wilt provided in the embodiment of the present application is described in detail below with reference to the accompanying drawings through specific embodiments and their application scenarios.

[0035] like Figure 1 As shown, the embodiment of the present application provides a hyperspectral early nondestructive detection method for bacterial wilt, which may include the following steps 101 to 103:

[0036] Step 101: Collect the target spectrum of the target plant.

[0037] Illustratively, the target plant may be a tomato plant, or any other Solanaceae plant that can be infected with bacterial wilt.

[0038] It should be noted that since tomato plants in areas affected by bacterial wilt are usually protected by bird-proof nets and greenhouses, the spectrum detection technology of conventional natural light sources will be limited. Therefore, in the embodiment of the present application, a HyScan portable intelligent spectrometer can be used to collect the spectrum of tomato plants.

[0039] Exemplarily, in this embodiment, a spectrum acquisition device with its own sunlight simulation light source is used to perform standard whiteboard calibration to collect the spectrum of tomato plants, that is, the above-mentioned target spectrum. The spectral range of the above-mentioned target spectrum is: 398.7nm to 1000.46nm, and the spectral resolution is better than 10nm. The hyperspectral early non-destructive detection method for bacterial wilt provided in the embodiment of the present application can determine whether the target plant has bacterial wilt based on the above-mentioned target spectrum.

[0040] Exemplarily, the target spectrum may be a spectrum of a stem of a target plant near the root, and the step 101 may include the following steps 101a:

[0041] Step 101a: obtaining a target spectrum of a target part of the target plant based on a sunlight simulation light source.

[0042] Wherein, the target part is a part of the stem of the target plant close to the root.

[0043] For example, in order to obtain the target spectrum of the target part of the target plant, the shading device can be used to select the stem near the root of the target plant and measure the spectrum of the stem. Specifically, three places can be selected near the root of the target plant, and spectra can be collected respectively. After collecting the spectrum 5 times each time, the average value is taken to obtain the target spectrum of the target plant.

[0044] Step 102: According to the optimal solution feature combination, extract the target feature parameters corresponding to the optimal solution feature combination from the target spectrum.

[0045] Wherein, the optimal solution feature combination is obtained based on a genetic algorithm model.

[0046] It can be understood that the above-mentioned target spectrum includes some spectra that have weak correlation with determining whether the target plant is suffering from bacterial wilt, and based on the collected target spectrum, relevant vegetation indices such as water stress, red edge, and chlorophyll can be obtained. In order to improve the detection efficiency, it is necessary to screen out the feature combination that has the greatest effect on determining whether the target plant is suffering from bacterial wilt, and constitute the above-mentioned optimal solution feature combination.

[0047] For example, after acquiring the target spectrum of the target plant, the target characteristic parameters corresponding to the optimal characteristic combination can be extracted from the collected target spectrum according to the pre-obtained optimal characteristic combination. Then, using the above target characteristic parameters, it is possible to quickly detect whether the target plant has a disease.

[0048] Step 103: input the target feature parameters into the target model and output the detection result.

[0049] Among them, the target model is a binary classification model obtained by training the training samples after the optimal solution feature combination processing; the detection result is used to indicate whether the target plant suffers from bacterial wilt.

[0050] Exemplarily, after obtaining the above target feature parameters, the target feature parameters can be input into a target model, which is a binary classification model trained based on the training samples processed by the optimal solution feature combination. According to the detection results output by the target model, it can be determined whether the target plant suffers from bacterial wilt.

[0051] For example, Figure 2 The following is a comparison of the average spectra of the stems of healthy plants and plants with bacterial wilt. Figure 2 From the spectral comparison shown, it can be seen that there are differences between the spectrum of the stems of healthy plants and the spectrum of the stems of diseased plants. Based on this, it is possible to determine whether the plant has bacterial wilt disease through the spectrum of the plant stem.

[0052] Thus, in order to determine whether the target plant suffers from bacterial wilt, it is necessary to collect the target spectrum of the target plant, and extract the target feature parameters corresponding to the optimal solution feature combination from the collected target spectrum according to the optimal solution feature combination obtained based on the genetic algorithm model. Afterwards, the extracted target feature parameters are input into the target model, and the detection result is output. According to the detection result, it can be determined whether the target plant suffers from bacterial wilt.

[0053] Optionally, in an embodiment of the present application, the above-mentioned optimal solution feature combination can be obtained through a genetic algorithm, which can utilize the extracted vegetation index and spectral principal components in combination with a genetic algorithm feature selection model established by the genetic algorithm.

[0054] For example, before the above step 102, the hyperspectral early nondestructive detection method for bacterial wilt provided in the embodiment of the present application may further include the following steps 104 to 106:

[0055] Step 104 , collecting a first spectrum of a target part of at least one tomato plant, and preprocessing the spectrum corresponding to each tomato plant using a convolution smoothing method to generate a second spectrum corresponding to each tomato plant.

[0056] Step 105: extract the principal component of the spectrum of the second spectrum corresponding to each tomato plant by principal component analysis PCA, and extract the vegetation index based on the second spectrum corresponding to each tomato plant by using a band operation method.

[0057] Step 106: Generate a data set according to the spectral principal components and vegetation index corresponding to each tomato plant.

[0058] Wherein, the vegetation index includes at least one of the following: water stress index, chlorophyll index, and red edge index.

[0059] For example, in order to obtain a data set containing the main components of the spectra and vegetation indexes of multiple tomato plants, it is necessary to collect the first spectra of the target parts of the multiple tomato plants and pre-process the first spectra to reduce the interference of random noise. Specifically, the convolution smoothing method is used to pre-process the spectrum corresponding to each tomato plant to generate a second spectrum corresponding to each tomato plant.

[0060] For example, after obtaining the second spectrum corresponding to each tomato plant, the second spectrum can be reduced in dimension by principal component analysis (PCA) to obtain the principal component containing 99.98% of the information of the second spectrum. Specifically, the characteristic dimension of the second spectrum can be changed from 303 to 25. At the same time, the band operation method can also be used to extract vegetation indices such as water stress index, chlorophyll index, and red edge index from the second spectrum.

[0061] Exemplarily, as shown in the following table, the vegetation index extracted in the embodiment of the present application is shown, where R represents reflectivity:

[0062]

[0063] For example, the principal components and vegetation index can be used as features for determining whether a plant is suffering from bacterial wilt. Afterwards, a genetic algorithm can be used to select the feature combination that has the greatest effect on determining whether a plant is diseased from the above features as the optimal solution feature combination.

[0064] For example, Figure 3As shown, it is a schematic diagram of the process of the hyperspectral early non-destructive detection method for bacterial wilt provided in the embodiment of the present application. After collecting the spectrum of the tomato plant using the HyScan portable intelligent spectrometer, the collected spectral data (i.e., the first spectrum) is preprocessed. After that, the preprocessed spectral data (i.e., the second spectrum) is subjected to principal component analysis and band operation to obtain the spectral principal components and vegetation index of the spectral data.

[0065] Furthermore, in an embodiment of the present application, after obtaining the above data set, the optimal solution feature combination can be screened out from the features corresponding to each of the above tomato plants through a genetic algorithm model.

[0066] For example, after the above step 106, the hyperspectral early nondestructive detection method for bacterial wilt provided in the embodiment of the present application may further include the following step 107:

[0067] Step 107: input the data set into a genetic algorithm model, and select the bacterial wilt feature combination corresponding to the individual with the largest fitness from the population finally obtained as the optimal solution feature combination.

[0068] The bacterial wilt feature combination corresponding to each individual is randomly assigned; the bacterial wilt feature combination includes: at least one item in the feature set consisting of the spectral principal component and vegetation index corresponding to the tomato plant.

[0069] It can be understood that the above-extracted vegetation indices such as WI, NDVI, GM1 and spectral principal components are combined with genetic algorithms to establish a genetic algorithm feature selection model. Genetic algorithm is a random global search optimization method that randomly initializes the population and generates a group of individuals that are more suitable for the environment by simulating operations such as replication, crossover and mutation in natural selection and inheritance, so that the population evolves to better and better areas in the search space. After continuous iteration of the population, it finally converges to the optimal solution, thereby obtaining the optimal solution of the feature combination method, that is, the above-mentioned optimal solution feature combination.

[0070] Exemplarily, in the random search optimization process of the genetic algorithm model, the bacterial wilt feature combination corresponding to each individual can be initialized by binary coding, and the initialization process can randomly assign the feature combination corresponding to each individual, and select the bacterial wilt feature combination corresponding to the individual with the largest fitness from the final population as the above-mentioned optimal solution feature combination.

[0071] For example, in the random search optimization process of the genetic algorithm, the feature set corresponding to each individual includes: feature 1, feature 2, feature 3 and feature 4. Taking the initialization of the target individual as an example, after randomly assigning the feature combination corresponding to the target individual through binary coding, the binary number of the feature combination corresponding to the target individual is 0101, that is, the feature combination corresponding to the target individual is feature 2 and feature 4.

[0072] Exemplarily, after initializing the feature combination corresponding to each individual through the above binary coding, it is also necessary to decode the binary number to obtain the feature combination corresponding to each individual.

[0073] Afterwards, the fitness value f(x) of the bacterial wilt disease feature combination corresponding to each individual is calculated, and the individuals in the population are selected based on the fitness value using a roulette wheel method. The larger the fitness value, the greater the probability of selecting the individual, and thus a new offspring population based on the fitness value of the previous generation population is obtained.

[0074] Exemplarily, the new offspring population obtained above is equally divided into the father and mother generations, and after crossover and mutation operations, the next generation population is obtained, that is, the population is evolved to better and better areas in the search space.

[0075] Exemplarily, the above population evolution process is repeated. During the iteration process, the individuals in the population will tend to an optimal solution that includes multiple feature dimensions. After reaching a certain number of iterations, the algorithm terminates and outputs the individual with the largest fitness value in the final population. The corresponding bacterial wilt feature combination is the above optimal solution feature combination.

[0076] For example, Figure 4 As shown, it is a fitness comparison diagram of bacterial wilt feature combination provided in the embodiment of the present application, which is composed of Figure 4 It can be seen that when the feature combination contains features of 10 dimensions, the fitness value is the highest, and the feature combination is specifically: vegetation index: WI: NDVI, NWI, NPCI, PRI2, Ctr1, GM1; principal components: 9, 10, 13, 18. Among them, principal components 9, 10, 13, and 18 are the 9th, 10th, 13th, and 18th features of the above-mentioned spectral principal components (containing 25 dimensions), respectively.

[0077] Furthermore, in an embodiment of the present application, after obtaining the above-mentioned optimal solution feature combination, after processing the samples based on the optimal solution feature combination, a binary classification model can be trained to obtain the above-mentioned target model.

[0078] For example, after the above step 107, the hyperspectral early nondestructive detection method for bacterial wilt provided in the embodiment of the present application may further include the following steps 108 and 109:

[0079] Step 108: normalize the feature combination corresponding to each tomato plant in the data set according to the optimal solution feature combination, and mark the feature combination corresponding to each tomato plant according to whether each tomato plant suffers from bacterial wilt.

[0080] Step 109: input the marked feature combination into the binary classification model, and train the binary classification model using a ten-fold cross validation method to obtain the target model.

[0081] Wherein, the binary classification model is a support vector machine (SVM) binary classification model.

[0082] Exemplarily, after obtaining the above-mentioned optimal solution feature combination, the feature combination corresponding to each tomato plant in the above-mentioned data set is normalized based on the optimal solution feature combination, that is, the features contained in the optimal solution combination are selected as the judgment features for judging whether the plant suffers from bacterial wilt disease.

[0083] Exemplarily, after processing the feature combination corresponding to each tomato plant, the feature combination corresponding to each tomato plant is marked according to whether each tomato plant suffers from bacterial wilt, and then input into the binary classification model for training.

[0084] Exemplarily, the binary classification model provided in the embodiment of the present application may be a support vector machine (SVM) binary classification model.

[0085] Specifically, the binary classification model can be trained using a 10-fold cross-validation method, that is, the data set is divided into ten parts, 9 of which are used as training data and 1 as test data in turn, and experiments are performed to finally obtain the optimal parameters of the SVM binary classification model.

[0086] It can be understood that the labeled training data is input into the SVM binary classification model, and the grid search algorithm is adopted to arrange and combine different parameters within a certain range of C and gamma parameters according to the specified step size; the ten-fold cross validation is repeated 10 times for each group of parameter combinations, that is, under different combinations of C and gamma parameters, the training set samples are divided into 10 groups, one group is used as the verification data sample, and the remaining 9 groups are used as training data. Each group of data is used as the verification data sample in turn, and this process is repeated 10 times; a group of C and gamma parameters with the best performance indicators are selected as the final optimal parameter values ​​to construct a classification model, that is, the above-mentioned target model.

[0087] Illustratively, after obtaining the above-mentioned optimal solution feature combination and the above-mentioned target model, the bacterial wilt disease can be diagnosed based on the spectrum of the target part of the target plant.

[0088] For example, Figure 5 As shown in FIG. 1 , the receiver operating characteristic curve (ROC) of the bacterial wilt detection accuracy provided in the embodiment of the present application is shown. The experimental results show that the accuracy (Accuracy) and recall rate (Recall) of the bacterial wilt hyperspectral early non-destructive detection method provided in the embodiment of the present application are 91.67% and 90.91%, respectively. Figure 5 As shown, the curve is steeper and the area under the ROC curve and the coordinate axis (Area UnderCurve, AUC) is 0.94. The experimental results show that the bacterial wilt disease detection model has good performance and meets the requirements of non-destructive, fast and efficient estimation.

[0089] The hyperspectral early nondestructive detection method for bacterial wilt provided in the embodiment of the present application uses crop stem spectra rather than leaf spectra to carry out early disease diagnosis based on the pathogenesis of bacterial wilt, and combines feature optimization and SVM binary classification methods for detection modeling, which solves the problem that bacterial wilt acts on the vascular bundle in the early stage of onset and the symptoms on the leaves are mild and difficult to judge, and provides a theoretical basis and technical support for early, nondestructive, rapid and high-precision detection research of bacterial wilt.

[0090] It should be noted that the hyperspectral early nondestructive detection method for bacterial wilt provided in the embodiment of the present application can be executed by a hyperspectral early nondestructive detection device for bacterial wilt, or a control module in the hyperspectral early nondestructive detection device for bacterial wilt that is used to execute the hyperspectral early nondestructive detection method for bacterial wilt. In the embodiment of the present application, the hyperspectral early nondestructive detection method for bacterial wilt performed by the hyperspectral early nondestructive detection device for bacterial wilt is taken as an example to illustrate the hyperspectral early nondestructive detection device for bacterial wilt provided in the embodiment of the present application.

[0091] It should be noted that, in the embodiments of the present application, the hyperspectral early nondestructive detection method for bacterial wilt shown in the drawings of the above-mentioned methods is illustrated by taking one of the drawings in the embodiments of the present application as an example. In specific implementation, the hyperspectral early nondestructive detection method for bacterial wilt shown in the drawings of the above-mentioned methods can also be implemented in combination with any other drawings that can be combined as shown in the above-mentioned embodiments, which will not be repeated here.

[0092] The hyperspectral early nondestructive detection device for bacterial wilt provided in the present application is described below. The hyperspectral early nondestructive detection method for bacterial wilt described below can be referenced to each other with the hyperspectral early nondestructive detection method for bacterial wilt described above.

[0093] Figure 6A schematic diagram of the structure of a hyperspectral early nondestructive detection device for bacterial wilt provided in one embodiment of the present application is shown in FIG. Figure 6 As shown, specifically including:

[0094] The spectrum acquisition module 601 is used to collect the target spectrum of the target plant; the parameter extraction module 602 is used to extract the target feature parameters corresponding to the optimal solution feature combination from the target spectrum according to the optimal solution feature combination; the disease diagnosis module 603 is used to input the target feature parameters into the target model and output the detection result; wherein the optimal solution feature combination is obtained based on the genetic algorithm model; the target model is a binary classification model obtained by training the training samples after processing the optimal solution feature combination; the detection result is used to indicate whether the target plant suffers from bacterial wilt.

[0095] Optionally, the spectrum acquisition module 601 is specifically used to obtain a target spectrum collected from a target part of the target plant based on a sunlight simulation light source; wherein the target part is a part of the stem of the target plant close to the root.

[0096] Optionally, the device also includes: a calculation module and a generation module; the spectrum acquisition module 601 is also used to collect a first spectrum of a target part of at least one tomato plant, and pre-process the spectrum corresponding to each tomato plant using a convolution smoothing method to generate a second spectrum corresponding to each tomato plant; the calculation module is used to extract the spectral principal components of the second spectrum corresponding to each tomato plant through principal component analysis PCA, and use a band operation method to extract a vegetation index based on the second spectrum corresponding to each tomato plant; the generation module is used to generate a data set according to the spectral principal components and vegetation index corresponding to each tomato plant; wherein the vegetation index includes at least one of the following: water stress index, chlorophyll index, and red edge index.

[0097] Optionally, the device also includes: a feature selection module; inputting the data set into a genetic algorithm model, and selecting a bacterial wilt feature combination corresponding to the individual with the highest fitness from the population finally obtained as the optimal solution feature combination; wherein the bacterial wilt feature combination corresponding to each individual is randomly assigned; the bacterial wilt feature combination includes: at least one item in a feature set consisting of a spectral principal component and a vegetation index corresponding to a tomato plant.

[0098] Optionally, the device also includes: an initialization module and a model training module; the initialization module is used to normalize the feature combination corresponding to each tomato plant in the data set according to the optimal solution feature combination, and mark the feature combination corresponding to each tomato plant according to whether each tomato plant suffers from bacterial wilt; the model training module is used to input the marked feature combination into the binary classification model, and train the binary classification model using a ten-fold cross-validation method to obtain the target model; wherein the binary classification model is a support vector machine SVM binary classification model.

[0099] The hyperspectral early nondestructive detection device for bacterial wilt provided in the present application uses crop stem spectra rather than leaf spectra to carry out early disease diagnosis based on the pathogenesis of bacterial wilt, and combines feature optimization and SVM binary classification methods for detection modeling, thereby solving the problem that bacterial wilt acts on the vascular bundle in the early stage of the disease and the symptoms on the leaves are mild and difficult to judge, providing a theoretical basis and technical support for early, nondestructive, rapid and high-precision detection research of bacterial wilt.

[0100] Figure 7 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 7 As shown, the electronic device may include: a processor 710, a communication interface 720, a memory 730 and a communication bus 740, wherein the processor 710, the communication interface 720 and the memory 730 communicate with each other through the communication bus 740. The processor 710 may call the logic instructions in the memory 730 to execute the hyperspectral early non-destructive detection method for bacterial wilt, which includes: collecting the target spectrum of the target plant; extracting the target feature parameters corresponding to the optimal solution feature combination from the target spectrum according to the optimal solution feature combination; inputting the target feature parameters into the target model and outputting the detection result; wherein the optimal solution feature combination is obtained based on the genetic algorithm model; the target model is obtained by training the binary classification model based on the training sample processed by the optimal solution feature combination; the detection result is used to indicate whether the target plant suffers from bacterial wilt.

[0101] In addition, the logic instructions in the above-mentioned memory 730 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art, and the computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk.

[0102] On the other hand, the present application also provides a computer program product, which includes a computer program stored on a computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the hyperspectral early non-destructive detection method for bacterial wilt provided by the above-mentioned methods, and the method includes: collecting a target spectrum of a target plant; according to an optimal solution feature combination, extracting target feature parameters corresponding to the optimal solution feature combination from the target spectrum; inputting the target feature parameters into a target model and outputting a detection result; wherein the optimal solution feature combination is obtained based on a genetic algorithm model; the target model is obtained by training a binary classification model based on training samples processed by the optimal solution feature combination; and the detection result is used to indicate whether the target plant is suffering from bacterial wilt.

[0103] On the other hand, the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the above-mentioned hyperspectral early nondestructive detection method for bacterial wilt, the method comprising: collecting a target spectrum of a target plant; extracting target feature parameters corresponding to the optimal solution feature combination from the target spectrum according to the optimal solution feature combination; inputting the target feature parameters into a target model, and outputting a detection result; wherein the optimal solution feature combination is obtained based on a genetic algorithm model; the target model is obtained by training a binary classification model based on training samples processed by the optimal solution feature combination; and the detection result is used to indicate whether the target plant is suffering from bacterial wilt.

[0104] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0105] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0106] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application 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 embodiments of the present application.

Claims

1. A hyperspectral early nondestructive detection method for bacterial wilt, characterized in that: include: Collecting target spectra of target plants; According to the optimal solution feature combination, extracting target feature parameters corresponding to the optimal solution feature combination from the target spectrum; Input the target feature parameters into the target model and output the detection results; Wherein, the optimal solution feature combination is obtained based on a genetic algorithm model; the target model is obtained by training a binary classification model based on training samples processed by the optimal solution feature combination; and the detection result is used to indicate whether the target plant suffers from bacterial wilt; The collecting of the target spectrum of the target plant comprises: Acquire a target spectrum of a target part of the target plant based on a sunlight simulation light source; Wherein, the target part is a part of the stem of the target plant close to the root; Before extracting the target vegetation index from the target spectrum according to the optimal solution feature combination, the method further includes: Collecting a first spectrum of a target part of at least one tomato plant, and preprocessing the spectrum corresponding to each tomato plant using a convolution smoothing method to generate a second spectrum corresponding to each tomato plant; The principal component of the spectrum of the second spectrum corresponding to each tomato plant is extracted by principal component analysis PCA, and the vegetation index is extracted based on the second spectrum corresponding to each tomato plant by using a band operation method; Generate a data set based on the spectral principal components and vegetation index corresponding to each tomato plant; Wherein, the vegetation index includes at least one of the following: water stress index, chlorophyll index, red edge index; After generating the data set according to the spectral principal component and vegetation index corresponding to each tomato plant, the method further comprises: Inputting the data set into a genetic algorithm model, and selecting a bacterial wilt feature combination corresponding to an individual with the largest fitness from the population finally obtained as the optimal solution feature combination; The bacterial wilt feature combination corresponding to each individual is randomly assigned; the bacterial wilt feature combination includes: at least one item in the feature set consisting of the spectral principal component and vegetation index corresponding to the tomato plant.

2. The method according to claim 1, characterized in that After inputting the data set into the genetic algorithm model and selecting the bacterial wilt feature combination corresponding to the individual with the largest fitness from the finally obtained population as the optimal solution feature combination, the method further comprises: According to the optimal solution feature combination, the feature combination corresponding to each tomato plant in the data set is normalized, and according to whether each tomato plant suffers from bacterial wilt, the feature combination corresponding to each tomato plant is marked; Inputting the labeled feature combination into the binary classification model, and training the binary classification model using a ten-fold cross validation method to obtain the target model; Wherein, the binary classification model is a support vector machine (SVM) binary classification model.

3. A hyperspectral early nondestructive detection device for bacterial wilt, characterized in that: The device comprises: A spectrum acquisition module, used for acquiring target spectrum of target plants; A parameter extraction module, used for extracting target characteristic parameters corresponding to the optimal solution characteristic combination from the target spectrum according to the optimal solution characteristic combination; A disease diagnosis module, used for inputting the target characteristic parameters into a target model and outputting a detection result; Wherein, the optimal solution feature combination is obtained based on a genetic algorithm model; the target model is obtained by training a binary classification model based on training samples processed by the optimal solution feature combination; and the detection result is used to indicate whether the target plant suffers from bacterial wilt; The spectrum acquisition module is specifically used to acquire the target spectrum of the target part of the target plant based on the sunlight simulation light source; Wherein, the target part is a part of the stem of the target plant close to the root; The device also includes: a calculation module and a generation module; The spectrum acquisition module is further used to acquire a first spectrum of a target part of at least one tomato plant, and to pre-process the spectrum corresponding to each tomato plant using a convolution smoothing method to generate a second spectrum corresponding to each tomato plant; The calculation module is used to extract the spectral principal component of the second spectrum corresponding to each tomato plant by principal component analysis PCA, and to extract the vegetation index based on the second spectrum corresponding to each tomato plant by using a band operation method; The generating module is used to generate a data set according to the spectral principal components and vegetation index corresponding to each tomato plant; Wherein, the vegetation index includes at least one of the following: water stress index, chlorophyll index, red edge index; The device also includes: a feature selection module; The feature selection module is used to input the data set into the genetic algorithm model, and select the bacterial wilt feature combination corresponding to the individual with the largest fitness from the population finally obtained as the optimal solution feature combination; wherein the bacterial wilt feature combination corresponding to each individual is randomly assigned; the bacterial wilt feature combination includes: at least one item in the feature set composed of the spectral principal component and vegetation index corresponding to the tomato plant.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the hyperspectral early nondestructive detection method for bacterial wilt as claimed in claim 1 or 2 are implemented.

5. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the hyperspectral early nondestructive detection method for bacterial wilt described in claim 1 or 2 are implemented.

Images