A tobacco disease identification and prevention method, system and storage medium

Abstract

The present invention discloses a method, system and storage medium for identifying and preventing tobacco diseases, which relate to the field of tobacco pest control. The method for identifying and preventing tobacco viruses includes the following steps: obtaining hyperspectral image information of tobacco plants, preprocessing spectral data information and image data information; extracting average spectral curves and disease spot image features, establishing a tobacco disease recognition model based on the average spectral curves and disease spot image features, and generating recognition results through the spectral recognition module and image recognition module in the tobacco disease recognition model; obtaining the recognition results of the spectral recognition module and the image recognition module, and combining them according to weight information to generate tobacco disease recognition results, and generating disease levels according to the disease recognition results; performing big data processing according to the disease recognition results to generate disease prevention methods, and displaying the disease recognition results, disease levels and disease prevention methods in a preset manner.

Description

Technical Field

[0001] The present invention relates to the field of tobacco pest control, and more specifically, to a tobacco pest identification and control method, system and storage medium. Background Art

[0002] Tobacco diseases are currently a major problem that plagues tobacco growers. my country is a major tobacco producer, and tobacco is one of the main cash crops in my country. If tobacco diseases are not controlled and prevented, it will seriously affect the income of growers and the country's fiscal revenue. The quality of tobacco leaves depends on whether tobacco grows healthily. During the growth process, tobacco is affected by factors such as climate, soil, and technical level, and is extremely susceptible to various tobacco diseases. Inexperienced plant protection personnel and farmers are unable to correctly judge the disease, and even make wrong pesticide application plans, resulting in the inability to curb the disease in time, causing a significant reduction in production and income, affecting the quality of tobacco leaves. Traditional tobacco disease identification requires professional technicians and technical equipment, which cannot meet the needs of growers and cannot be used and promoted on a large scale.

[0003] In order to accurately identify tobacco diseases and scientifically prevent and control them, a system needs to be developed for matching. The system obtains the hyperspectral image information of tobacco plants, extracts the average spectral curve and the image features of lesions, establishes a tobacco disease recognition model based on the average spectral curve and the image features of lesions, and generates recognition results through the spectral recognition module and image recognition module in the tobacco disease recognition model; obtains the recognition results of the spectral recognition module and the image recognition module, and combines them according to the weight information to generate tobacco disease recognition results, and generates disease levels according to the disease recognition results; and generates disease prevention and control methods based on big data processing according to the disease recognition results. In the process of implementing the system, how to establish a tobacco disease recognition model and how to generate a prevention and control plan based on the recognition results are urgent problems that need to be solved. Summary of the invention

[0004] In order to solve at least one of the above technical problems, the present invention proposes a method, system and storage medium for identifying and preventing tobacco diseases.

[0005] The first aspect of the present invention provides a method for identifying and preventing tobacco diseases, comprising:

[0006] Acquire hyperspectral image information of tobacco plants, and pre-process the spectral data information and image data information;

[0007] Generate an average spectral curve and lesion image features, establish a tobacco disease recognition model based on the average spectral curve and lesion image features, and generate a recognition result through a spectral recognition module and an image recognition module in the tobacco disease recognition model;

[0008] Obtaining recognition results of the spectrum recognition module and the image recognition module, combining them according to weight information to generate a tobacco disease recognition result, and generating a disease grade according to the disease recognition result;

[0009] Big data processing is performed based on the disease identification results to generate disease prevention and control methods, and the disease identification results, disease levels and disease prevention and control methods are displayed in a preset manner.

[0010] In this solution, the hyperspectral image information of the tobacco plant is obtained, and the spectral data information and the image data information are preprocessed, specifically:

[0011] Acquire hyperspectral image information of tobacco plants, and acquire spectral data information and image data information from the hyperspectral image information;

[0012] Extracting characteristic wavelengths from the spectral data by preprocessing;

[0013] The image target area is extracted from the image data information through preprocessing, the spectral data information of the image target area is exported, and the spectral data information of the image target area is subjected to noise elimination.

[0014] Acquire spectral reflectance values ​​according to spectral data information of the target area of ​​the image, and generate an average spectral curve according to the spectral reflectance values;

[0015] Among them, the average spectrum curve is generated according to the spectral reflectance value, and the average spectrum expression is:

[0016]

[0017] in, represents the desired average spectrum, p i represents the spectrum obtained by the i-th pixel, and n represents the number of pixels in the target area of ​​the image.

[0018] In this solution, the average spectral curve and the lesion image features are extracted, and a tobacco disease recognition model is established based on the average spectral curve and the lesion image features, wherein the spectral recognition module is specifically:

[0019] Obtaining an average spectral curve, and smoothing the average spectral curve to obtain curve characteristics;

[0020] Determine characteristic peaks through the curve characteristics, and decompose the average spectrum curve into continuous characteristic points according to the curve characteristics and the characteristic peaks;

[0021] The characteristic points are introduced into a spectrum recognition module of a tobacco disease recognition model, and the spectrum recognition module generates a difference coefficient and a difference number through the characteristic point information and an average spectrum curve of healthy tobacco;

[0022] A spectrum recognition result is generated according to the difference coefficient and the number of differences, and the tobacco plant disease type is determined according to the spectrum recognition result.

[0023] In this solution, the average spectral curve and the lesion image features are extracted, and a tobacco disease recognition model is established based on the average spectral curve and the lesion image features, wherein the image recognition module is specifically:

[0024] Acquiring a hyperspectral image of a tobacco plant, and performing mask processing on the hyperspectral image;

[0025] Performing principal component analysis on the masked hyperspectral image to generate a principal component image;

[0026] Select the segmentation threshold, mark the main component image into multiple independent regions, and calculate the gray value of each point in each region;

[0027] In the independent area, if the gray value is greater than the segmentation threshold, the point is changed to a diseased spot area, and a binary image of the diseased tobacco plant is extracted;

[0028] Extract the coordinate information of the background area of ​​the hyperspectral image, mark the background area in the binary image according to the coordinate information, and replace the gray value of the background area with a value different from that of the diseased area and the healthy area;

[0029] Extracting the lesion area, generating lesion image features, and importing the lesion image features into a tobacco disease recognition model;

[0030] The tobacco disease recognition model is used to generate an image recognition result, and tobacco disease types are distinguished according to the image recognition result.

[0031] In this solution, the recognition results of the spectrum recognition module and the image recognition module are obtained, and the tobacco disease recognition results are generated by combining them according to the weight information, and the disease level is generated according to the disease recognition results, specifically:

[0032] Obtaining spectrum recognition results and image recognition results, and performing initial weight assignment on the spectrum recognition results and image recognition results;

[0033] Generate feedback information by verifying the accuracy of tobacco disease recognition model in identifying diseases;

[0034] Adjust the initial weights according to the feedback information, generate an optimal weight combination, and determine the weight information;

[0035] Combining the spectrum recognition result and the image recognition result according to the weight information to generate a disease recognition result;

[0036] A disease index is generated based on the disease identification result, and a disease grade is generated based on the disease index.

[0037] In this solution, the disease identification results are processed by big data to generate disease prevention and control methods, and the disease identification results, disease levels and disease prevention and control methods are displayed in a preset manner, specifically:

[0038] Obtaining disease identification results, performing keyword extraction based on the disease identification results, and generating disease feature information;

[0039] Obtain similar disease prevention and control information and historical disease prevention and control information based on the disease characteristic information through big data processing;

[0040] Calculate the matching degree of disease control plans according to the disease level matching similar disease control information and historical disease control information;

[0041] Determining the matching priority of the similar disease prevention and control information and the historical disease prevention and control information according to the matching degree;

[0042] Analyze similar disease control information with high priority and historical disease control information, determine the required pesticides and dosage information, and generate disease control plans based on the required pesticides and dosage information;

[0043] The disease identification results, disease levels and disease prevention and control plans are displayed in a preset manner.

[0044] The second aspect of the present invention further provides a tobacco disease identification and prevention system, the system comprising: a memory, a processor, the memory comprising a tobacco disease identification and prevention method program, the tobacco disease identification and prevention method program when executed by the processor to implement the following steps:

[0045] Acquire hyperspectral image information of tobacco plants, and pre-process the spectral data information and image data information;

[0046] Generate an average spectral curve and lesion image features, establish a tobacco disease recognition model based on the average spectral curve and lesion image features, and generate a recognition result through a spectral recognition module and an image recognition module in the tobacco disease recognition model;

[0047] Obtaining recognition results of the spectrum recognition module and the image recognition module, combining them according to weight information to generate a tobacco disease recognition result, and generating a disease grade according to the disease recognition result;

[0048] Big data processing is performed based on the disease identification results to generate disease prevention and control methods, and the disease identification results, disease levels and disease prevention and control methods are displayed in a preset manner.

[0049] In this solution, the hyperspectral image information of the tobacco plant is obtained, and the spectral data information and the image data information are preprocessed, specifically:

[0050] Acquire hyperspectral image information of tobacco plants, and acquire spectral data information and image data information from the hyperspectral image information;

[0051] Extracting characteristic wavelengths from the spectral data by preprocessing;

[0052] The image target area is extracted from the image data information through preprocessing, the spectral data information of the image target area is exported, and the spectral data information of the image target area is subjected to noise elimination.

[0053] Acquire spectral reflectance values ​​according to spectral data information of the target area of ​​the image, and generate an average spectral curve according to the spectral reflectance values;

[0054] Among them, the average spectrum curve is generated according to the spectral reflectance value, and the average spectrum expression is:

[0055]

[0056] in, represents the desired average spectrum, p i represents the spectrum obtained by the i-th pixel, and n represents the number of pixels in the target area of ​​the image.

[0057] In this solution, the average spectral curve and the lesion image features are extracted, and a tobacco disease recognition model is established based on the average spectral curve and the lesion image features, wherein the spectral recognition module is specifically:

[0058] Obtaining an average spectral curve, and smoothing the average spectral curve to obtain curve characteristics;

[0059] Determine characteristic peaks through the curve characteristics, and decompose the average spectrum curve into continuous characteristic points according to the curve characteristics and the characteristic peaks;

[0060] The characteristic points are introduced into a spectrum recognition module of a tobacco disease recognition model, and the spectrum recognition module generates a difference coefficient and a difference number through the characteristic point information and an average spectrum curve of healthy tobacco;

[0061] A spectrum recognition result is generated according to the difference coefficient and the number of differences, and the tobacco plant disease type is determined according to the spectrum recognition result.

[0062] In this solution, the average spectral curve and the lesion image features are extracted, and a tobacco disease recognition model is established based on the average spectral curve and the lesion image features, wherein the image recognition module is specifically:

[0063] Acquiring a hyperspectral image of a tobacco plant, and performing mask processing on the hyperspectral image;

[0064] Performing principal component analysis on the masked hyperspectral image to generate a principal component image;

[0065] Select the segmentation threshold, mark the main component image into multiple independent regions, and calculate the gray value of each point in each region;

[0066] In the independent area, if the gray value is greater than the segmentation threshold, the point is changed to a diseased spot area, and a binary image of the diseased tobacco plant is extracted;

[0067] Extract the coordinate information of the background area of ​​the hyperspectral image, mark the background area in the binary image according to the coordinate information, and replace the gray value of the background area with a value different from that of the diseased area and the healthy area;

[0068] Extracting the lesion area, generating lesion image features, and importing the lesion image features into a tobacco disease recognition model;

[0069] The tobacco disease recognition model is used to generate an image recognition result, and tobacco disease types are distinguished according to the image recognition result.

[0070] In this solution, the recognition results of the spectrum recognition module and the image recognition module are obtained, and the tobacco disease recognition results are generated by combining them according to the weight information, and the disease level is generated according to the disease recognition results, specifically:

[0071] Obtaining spectrum recognition results and image recognition results, and performing initial weight assignment on the spectrum recognition results and image recognition results;

[0072] Generate feedback information by verifying the accuracy of tobacco disease recognition model in identifying diseases;

[0073] Adjust the initial weights according to the feedback information, generate an optimal weight combination, and determine the weight information;

[0074] Combining the spectrum recognition result and the image recognition result according to the weight information to generate a disease recognition result;

[0075] A disease index is generated based on the disease identification result, and a disease grade is generated based on the disease index.

[0076] In this solution, the disease identification results are processed by big data to generate disease prevention and control methods, and the disease identification results, disease levels and disease prevention and control methods are displayed in a preset manner, specifically:

[0077] Obtaining disease identification results, performing keyword extraction based on the disease identification results, and generating disease feature information;

[0078] Obtain similar disease prevention and control information and historical disease prevention and control information based on the disease characteristic information through big data processing;

[0079] Calculate the matching degree of disease control plans according to the disease level matching similar disease control information and historical disease control information;

[0080] Determining the matching priority of the similar disease prevention and control information and the historical disease prevention and control information according to the matching degree;

[0081] Analyze similar disease control information with high priority and historical disease control information, determine the required pesticides and dosage information, and generate disease control plans based on the required pesticides and dosage information;

[0082] The disease identification results, disease levels and disease prevention and control plans are displayed in a preset manner.

[0083] The third aspect of the present invention also provides a computer-readable storage medium, which includes a program for a method for identifying and preventing tobacco diseases. When the program for a method for identifying and preventing tobacco diseases is executed by a processor, the steps of the method for identifying and preventing tobacco diseases as described in any one of the above items are implemented.

[0084] The invention discloses a method, system and storage medium for identifying and preventing tobacco diseases, and relates to the field of tobacco pest control. The tobacco virus identification and prevention method includes the following steps: obtaining high-spectral image information of tobacco plants, preprocessing spectral data information and image data information; extracting average spectral curves and lesion image features, establishing a tobacco disease identification model according to the average spectral curves and lesion image features, and generating identification results through the spectral identification module and image identification module in the tobacco disease identification model; obtaining the identification results of the spectral identification module and the image identification module, and combining them according to weight information to generate tobacco disease identification results, and generating disease levels according to the disease identification results; performing big data processing according to the disease identification results to generate disease prevention and control methods, and displaying the disease identification results, disease levels and disease prevention and control methods according to a preset manner. The invention realizes rapid and ready identification of tobacco diseases by establishing a tobacco disease identification model, timely discovers the incidence of tobacco, and adopts scientific prevention and control means to apply pesticides in accordance with the amount, thereby reducing the adverse effects of diseases on tobacco and avoiding the pollution of the environment caused by the abuse of pesticides. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 A flow chart showing a method for identifying and preventing tobacco diseases according to the present invention;

[0086] Figure 2The flowchart of the identification method of the spectrum identification module in the tobacco disease identification model of the present invention is shown;

[0087] Figure 3 A flow chart of the recognition method of the image recognition module in the tobacco disease recognition model of the present invention is shown;

[0088] Figure 4 A block diagram of a tobacco disease identification and prevention system of the present invention is shown. DETAILED DESCRIPTION

[0089] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0090] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0091] Figure 1 A flow chart showing a method for identifying and preventing tobacco diseases according to the present invention;

[0092] like Figure 1 As shown, the first aspect of the present invention provides a method for identifying and preventing tobacco diseases, comprising:

[0093] S102, obtaining hyperspectral image information of tobacco plants, and preprocessing the spectral data information and the image data information;

[0094] S104, generating an average spectral curve and lesion image features, establishing a tobacco disease recognition model according to the average spectral curve and lesion image features, and generating a recognition result through a spectral recognition module and an image recognition module in the tobacco disease recognition model;

[0095] S106, obtaining recognition results of the spectrum recognition module and the image recognition module, combining them according to weight information to generate a tobacco disease recognition result, and generating a disease grade according to the disease recognition result;

[0096] S108, performing big data processing on the disease identification result to generate a disease prevention and control method, and displaying the disease identification result, disease level and disease prevention and control method in a preset manner.

[0097] It should be noted that in order to obtain high-quality hyperspectral images, it is necessary to build a hyperspectral image system structure. The basic components of the hyperspectral image system include light source equipment, wavelength scattering equipment, detectors and supporting control software. Halogen lamps, transmission optical fibers and collimating lenses are selected as lighting units. CCD cameras, spectrometers and fixed-focus lenses form imaging units. After collecting data through the hyperspectral imaging system, ENVI software is also needed to extract and process spectral data, and extract corresponding features through processing of image data for modeling and analysis.

[0098] It should be noted that the method of obtaining the hyperspectral image information of the tobacco plant and preprocessing the spectral data information and the image data information is specifically as follows:

[0099] Acquire hyperspectral image information of tobacco plants, and acquire spectral data information and image data information from the hyperspectral image information;

[0100] Extracting characteristic wavelengths from the spectral data by preprocessing;

[0101] The image target area is extracted from the image data information through preprocessing, the spectral data information of the image target area is exported, and the spectral data information of the image target area is subjected to noise elimination.

[0102] Acquire spectral reflectance values ​​according to spectral data information of the target area of ​​the image, and generate an average spectral curve according to the spectral reflectance values;

[0103] Among them, the average spectrum curve is generated according to the spectral reflectance value, and the average spectrum expression is:

[0104]

[0105] in, represents the desired average spectrum, p i represents the spectrum obtained by the i-th pixel, and n represents the number of pixels in the target area of ​​the image.

[0106] Figure 2 The flowchart of the identification method of the spectrum identification module in the tobacco disease identification model of the present invention is shown.

[0107] According to an embodiment of the present invention, the average spectral curve and the lesion image features are extracted, and a tobacco disease recognition model is established according to the average spectral curve and the lesion image features, wherein the spectral recognition module is specifically:

[0108] S202, obtaining an average spectrum curve, and smoothing the average spectrum curve to obtain curve characteristics;

[0109] S204, determining characteristic peaks according to the curve characteristics, and decomposing the average spectrum curve into continuous characteristic points according to the curve characteristics and the characteristic peaks;

[0110] S206, importing the characteristic points into a spectrum recognition module of a tobacco disease recognition model, wherein the spectrum recognition module generates a difference coefficient and a difference number based on the characteristic point information and an average spectrum curve of healthy tobacco;

[0111] S208, generating a spectrum recognition result according to the difference coefficient and the number of differences, and determining the tobacco plant disease type according to the spectrum recognition result.

[0112] It should be noted that the hyperspectral image information of tobacco samples is collected by the hyperspectral image system, and the image information between the wavelength of 350-1100nm is collected. After the image collection is completed, black and white correction is performed. After preprocessing and feature extraction of the hyperspectral data, the processed data is modeled based on the neural network to obtain the hyperspectral data of healthy tobacco and the growth curve and fitting equation curve of each diseased tobacco. The acquired data are grouped to obtain several training data sets, and several data sets are imported into the spectrum recognition module of the tobacco disease recognition model to generate output results. The spectrum recognition module of the tobacco disease recognition model is adjusted by error back propagation to complete the training of the spectrum recognition module in the tobacco disease recognition model. The data is input into the input layer, and after standardization, the data is set with weights and then transmitted to the hidden layer. The input weights are summed, converted, and transmitted to the third layer in the hidden layer to obtain the output results. The prediction set and the verification set are set, and the correct recognition rate of each prediction set and the verification set is calculated. When the correct recognition rate of the prediction set and the verification set is higher, the model recognition ability is better, and vice versa, the model recognition ability is worse.

[0113] Figure 3 The flowchart of the recognition method of the image recognition module in the tobacco disease recognition model of the present invention is shown.

[0114] According to an embodiment of the present invention, the average spectral curve and the lesion image features are extracted, and a tobacco disease recognition model is established according to the average spectral curve and the lesion image features, wherein the image recognition module is specifically:

[0115] S302, acquiring a hyperspectral image of a tobacco plant, and performing mask processing on the hyperspectral image;

[0116] S304, performing principal component analysis on the hyperspectral image after mask processing to generate a principal component image;

[0117] S306, selecting a segmentation threshold, marking the main component image into multiple independent regions, and calculating the gray value of each point in each region;

[0118] S308, in the independent area, if the gray value is greater than the segmentation threshold, the point is changed to a diseased spot area, and a binary image of the diseased tobacco plant is extracted;

[0119] S310, extracting coordinate information of the background area of ​​the hyperspectral image, marking the background area in the binary image according to the coordinate information, and replacing the grayscale value of the background area with a value different from that of the diseased area and the healthy area;

[0120] S312, extracting the lesion area, generating lesion image features, and importing the lesion image features into a tobacco disease recognition model;

[0121] S314, generating an image recognition result by using the tobacco disease recognition model, and distinguishing tobacco disease types according to the image recognition result.

[0122] It should be noted that the image recognition module in the tobacco recognition model is established based on the convolutional neural network, and the training process of the image recognition module in the tobacco recognition model is specifically as follows: a sufficient number of tobacco disease image data are obtained or a relevant database is accessed, and the obtained image data is processed into a training set and input into the image recognition module in the tobacco recognition model; each convolution layer of the image recognition module in the tobacco recognition model uses the initial convolution kernel and initial bias matrix of each convolution layer to perform convolution calculation and maximum pooling on the image data input into the convolutional neural network model to obtain a first feature image of the training image in the training set; the first feature image of the training image obtained is pooled again to obtain a second feature image of the training image; and according to the training set, the first feature image of the training image is pooled again to obtain a second feature image of the training image. The second feature map of the training image determines the feature vector of each training image, and the obtained feature vector is processed by an initial bias matrix and an initial weight matrix to obtain a classification vector of the training image in the training set. According to the classification vector of the training image in the training set and the initial category of each training image, a category error is calculated, and a convolution kernel of an image recognition module in a tobacco recognition model is adjusted according to the obtained category error. According to multiple training images and the adjusted convolution kernel parameters, the relevant parameters of the image recognition module in the tobacco recognition model are continuously adjusted. After multiple iterations are performed until the error reaches an ideal value, the training of the image recognition module in the tobacco recognition model is stopped, that is, the training of the image recognition module in the tobacco recognition model is completed.

[0123] It should be noted that the identification results of the spectrum identification module and the image identification module are obtained, and the tobacco disease identification results are generated by combining them according to the weight information, and the disease level is generated according to the disease identification results, specifically:

[0124] Obtaining spectrum recognition results and image recognition results, and performing initial weight assignment on the spectrum recognition results and image recognition results;

[0125] Generate feedback information by verifying the accuracy of tobacco disease recognition model in identifying diseases;

[0126] Adjust the initial weights according to the feedback information, generate an optimal weight combination, and determine the weight information;

[0127] Combining the spectrum recognition result and the image recognition result according to the weight information to generate a disease recognition result;

[0128] A disease index is generated based on the disease identification result, and a disease grade is generated based on the disease index.

[0129] The tobacco disease recognition result is generated by combining the spectrum recognition result and the image recognition result according to the weight information. The specific calculation is:

[0130]

[0131] Among them, k represents the tobacco disease recognition result, λ represents the tobacco disease recognition model parameter, α represents the weight information, g represents the spectrum recognition result, and t represents the image recognition result.

[0132] It should be noted that the disease prevention and control method is generated by processing big data based on the disease identification result, and the disease identification result, disease level and disease prevention and control method are displayed in a preset manner, specifically:

[0133] Obtaining disease identification results, performing keyword extraction based on the disease identification results, and generating disease feature information;

[0134] Obtain similar disease prevention and control information and historical disease prevention and control information based on the disease characteristic information through big data processing;

[0135] Calculate the matching degree of disease control plans according to the disease level matching similar disease control information and historical disease control information;

[0136] Determining the matching priority of the similar disease prevention and control information and the historical disease prevention and control information according to the matching degree;

[0137] Analyze similar disease control information with high priority and historical disease control information, determine the required pesticides and dosage information, and generate disease control plans based on the required pesticides and dosage information;

[0138] The disease identification results, disease levels and disease prevention and control plans are displayed in a preset manner.

[0139] According to an embodiment of the present invention, it also includes:

[0140] Collecting environmental change information and analyzing the impact of the environmental change information on tobacco plant diseases;

[0141] Generate a matching sequence model using the environmental change information and the tobacco plant disease status information;

[0142] Segment and extract environmental change information and tobacco plant disease characteristics using the matching sequence model to establish an environmental information database;

[0143] According to the environmental information database, diseases of tobacco plants are monitored and warned, and improvement suggestions and methods are generated for environmental conditions, and pest prevention and treatment plans are generated.

[0144] Figure 4 A block diagram of a tobacco disease identification and prevention system of the present invention is shown.

[0145] The second aspect of the present invention further provides a tobacco disease identification and prevention system 4, the system comprising: a memory 41, a processor 42, the memory comprising a tobacco disease identification and prevention method program, the tobacco disease identification and prevention method program when executed by the processor to implement the following steps:

[0146] Acquire hyperspectral image information of tobacco plants, and pre-process the spectral data information and image data information;

[0147] Generate an average spectral curve and lesion image features, establish a tobacco disease recognition model based on the average spectral curve and lesion image features, and generate a recognition result through a spectral recognition module and an image recognition module in the tobacco disease recognition model;

[0148] Obtaining recognition results of the spectrum recognition module and the image recognition module, combining them according to weight information to generate a tobacco disease recognition result, and generating a disease grade according to the disease recognition result;

[0149] Big data processing is performed based on the disease identification results to generate disease prevention and control methods, and the disease identification results, disease levels and disease prevention and control methods are displayed in a preset manner.

[0150] It should be noted that in order to obtain high-quality hyperspectral images, it is necessary to build a hyperspectral image system structure. The basic components of the hyperspectral image system include light source equipment, wavelength scattering equipment, detectors and supporting control software. Halogen lamps, transmission optical fibers and collimating lenses are selected as lighting units. CCD cameras, spectrometers and fixed-focus lenses form imaging units. After collecting data through the hyperspectral imaging system, ENVI software is also needed to extract and process spectral data, and extract corresponding features through processing of image data for modeling and analysis.

[0151] It should be noted that the method of obtaining the hyperspectral image information of the tobacco plant and preprocessing the spectral data information and the image data information is specifically as follows:

[0152] Acquire hyperspectral image information of tobacco plants, and acquire spectral data information and image data information from the hyperspectral image information;

[0153] Extracting characteristic wavelengths from the spectral data by preprocessing;

[0154] The image target area is extracted from the image data information through preprocessing, the spectral data information of the image target area is exported, and the spectral data information of the image target area is subjected to noise elimination.

[0155] Acquire spectral reflectance values ​​according to spectral data information of the target area of ​​the image, and generate an average spectral curve according to the spectral reflectance values;

[0156] Among them, the average spectrum curve is generated according to the spectral reflectance value, and the average spectrum expression is:

[0157]

[0158] in, represents the desired average spectrum, p i represents the spectrum obtained by the i-th pixel, and n represents the number of pixels in the target area of ​​the image.

[0159] According to an embodiment of the present invention, the average spectral curve and the lesion image features are extracted, and a tobacco disease recognition model is established according to the average spectral curve and the lesion image features, wherein the spectral recognition module is specifically:

[0160] Obtaining an average spectral curve, and smoothing the average spectral curve to obtain curve characteristics;

[0161] Determine characteristic peaks through the curve characteristics, and decompose the average spectrum curve into continuous characteristic points according to the curve characteristics and the characteristic peaks;

[0162] The characteristic points are introduced into a spectrum recognition module of a tobacco disease recognition model, and the spectrum recognition module generates a difference coefficient and a difference number through the characteristic point information and an average spectrum curve of healthy tobacco;

[0163] A spectrum recognition result is generated according to the difference coefficient and the number of differences, and the tobacco plant disease type is determined according to the spectrum recognition result.

[0164] It should be noted that the hyperspectral image information of tobacco samples is collected by the hyperspectral image system, and the image information between the wavelength of 350-1100nm is collected. After the image collection is completed, black and white correction is performed. After preprocessing and feature extraction of the hyperspectral data, the processed data is modeled based on the neural network to obtain the hyperspectral data of healthy tobacco and the growth curve and fitting equation curve of each diseased tobacco. The acquired data are grouped to obtain several training data sets, and several data sets are imported into the spectrum recognition module of the tobacco disease recognition model to generate output results. The spectrum recognition module of the tobacco disease recognition model is adjusted by error back propagation to complete the training of the spectrum recognition module in the tobacco disease recognition model. The data is input into the input layer, and after standardization, the data is set with weights and then transmitted to the hidden layer. The input weights are summed, converted, and transmitted to the third layer in the hidden layer to obtain the output results. The prediction set and the verification set are set, and the correct recognition rate of each prediction set and the verification set is calculated. When the correct recognition rate of the prediction set and the verification set is higher, the model recognition ability is better, and vice versa, the model recognition ability is worse.

[0165] According to an embodiment of the present invention, the average spectral curve and the lesion image features are extracted, and a tobacco disease recognition model is established according to the average spectral curve and the lesion image features, wherein the image recognition module is specifically:

[0166] Acquiring a hyperspectral image of a tobacco plant, and performing mask processing on the hyperspectral image;

[0167] Performing principal component analysis on the masked hyperspectral image to generate a principal component image;

[0168] Select the segmentation threshold, mark the main component image into multiple independent regions, and calculate the gray value of each point in each region;

[0169] In the independent area, if the gray value is greater than the segmentation threshold, the point is changed to a diseased spot area, and a binary image of the diseased tobacco plant is extracted;

[0170] Extract the coordinate information of the background area of ​​the hyperspectral image, mark the background area in the binary image according to the coordinate information, and replace the gray value of the background area with a value different from that of the diseased area and the healthy area;

[0171] Extracting the lesion area, generating lesion image features, and importing the lesion image features into a tobacco disease recognition model;

[0172] The tobacco disease recognition model is used to generate an image recognition result, and tobacco disease types are distinguished according to the image recognition result.

[0173] It should be noted that the image recognition module in the tobacco recognition model is established based on the convolutional neural network, and the training process of the image recognition module in the tobacco recognition model is specifically as follows: a sufficient number of tobacco disease image data are obtained or a relevant database is accessed, and the obtained image data is processed into a training set and input into the image recognition module in the tobacco recognition model; each convolution layer of the image recognition module in the tobacco recognition model uses the initial convolution kernel and initial bias matrix of each convolution layer to perform convolution calculation and maximum pooling on the image data input into the convolutional neural network model to obtain a first feature image of the training image in the training set; the first feature image of the training image obtained is pooled again to obtain a second feature image of the training image; and according to the training set, the first feature image of the training image is pooled again to obtain a second feature image of the training image. The second feature map of the training image determines the feature vector of each training image, and the obtained feature vector is processed by an initial bias matrix and an initial weight matrix to obtain a classification vector of the training image in the training set. According to the classification vector of the training image in the training set and the initial category of each training image, a category error is calculated, and a convolution kernel of an image recognition module in a tobacco recognition model is adjusted according to the obtained category error. According to multiple training images and the adjusted convolution kernel parameters, the relevant parameters of the image recognition module in the tobacco recognition model are continuously adjusted. After multiple iterations are performed until the error reaches an ideal value, the training of the image recognition module in the tobacco recognition model is stopped, that is, the training of the image recognition module in the tobacco recognition model is completed.

[0174] It should be noted that the identification results of the spectrum identification module and the image identification module are obtained, and the tobacco disease identification results are generated by combining them according to the weight information, and the disease level is generated according to the disease identification results, specifically:

[0175] Obtaining spectrum recognition results and image recognition results, and performing initial weight assignment on the spectrum recognition results and image recognition results;

[0176] Generate feedback information by verifying the accuracy of tobacco disease recognition model in identifying diseases;

[0177] Adjust the initial weights according to the feedback information, generate an optimal weight combination, and determine the weight information;

[0178] Combining the spectrum recognition result and the image recognition result according to the weight information to generate a disease recognition result;

[0179] A disease index is generated based on the disease identification result, and a disease grade is generated based on the disease index.

[0180] The tobacco disease recognition result is generated by combining the spectrum recognition result and the image recognition result according to the weight information. The specific calculation is:

[0181]

[0182] Among them, k represents the tobacco disease recognition result, λ represents the tobacco disease recognition model parameter, α represents the weight information, g represents the spectrum recognition result, and t represents the image recognition result.

[0183] It should be noted that the disease prevention and control method is generated by processing big data based on the disease identification result, and the disease identification result, disease level and disease prevention and control method are displayed in a preset manner, specifically:

[0184] Obtaining disease identification results, performing keyword extraction based on the disease identification results, and generating disease feature information;

[0185] Obtain similar disease prevention and control information and historical disease prevention and control information based on the disease characteristic information through big data processing;

[0186] Calculate the matching degree of disease control plans according to the disease level matching similar disease control information and historical disease control information;

[0187] Determining the matching priority of the similar disease prevention and control information and the historical disease prevention and control information according to the matching degree;

[0188] Analyze similar disease control information with high priority and historical disease control information, determine the required pesticides and dosage information, and generate disease control plans based on the required pesticides and dosage information;

[0189] The disease identification results, disease levels and disease prevention and control plans are displayed in a preset manner.

[0190] According to an embodiment of the present invention, it also includes:

[0191] Collecting environmental change information and analyzing the impact of the environmental change information on tobacco plant diseases;

[0192] Generate a matching sequence model using the environmental change information and the tobacco plant disease status information;

[0193] Segment and extract environmental change information and tobacco plant disease characteristics using the matching sequence model to establish an environmental information database;

[0194] According to the environmental information database, diseases of tobacco plants are monitored and warned, and improvement suggestions and methods are generated for environmental conditions, and pest prevention and treatment plans are generated.

[0195] The third aspect of the present invention also provides a computer-readable storage medium, which includes a program for a method for identifying and preventing tobacco diseases. When the program for a method for identifying and preventing tobacco diseases is executed by a processor, the steps of the method for identifying and preventing tobacco diseases as described in any one of the above items are implemented.

[0196] The invention discloses a method, system and storage medium for identifying and preventing tobacco diseases, and relates to the field of tobacco pest control. The tobacco virus identification and prevention method includes the following steps: obtaining high-spectral image information of tobacco plants, preprocessing spectral data information and image data information; extracting average spectral curves and lesion image features, establishing a tobacco disease identification model according to the average spectral curves and lesion image features, and generating identification results through the spectral identification module and image identification module in the tobacco disease identification model; obtaining the identification results of the spectral identification module and the image identification module, and combining them according to weight information to generate tobacco disease identification results, and generating disease levels according to the disease identification results; performing big data processing according to the disease identification results to generate disease prevention and control methods, and displaying the disease identification results, disease levels and disease prevention and control methods according to a preset manner. The invention realizes rapid and ready identification of tobacco diseases by establishing a tobacco disease identification model, timely discovers the incidence of tobacco, and adopts scientific prevention and control means to apply pesticides in accordance with the amount, thereby reducing the adverse effects of diseases on tobacco and avoiding the pollution of the environment caused by the abuse of pesticides.

[0197] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0198] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0199] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0200] Those skilled in the art can understand that: all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiments; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), disks or optical disks, and other media that can store program codes.

[0201] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0202] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for identifying and preventing tobacco diseases, It is characterized in that include: Acquire hyperspectral image information of tobacco plants, and pre-process the spectral data information and image data information; Generate an average spectral curve and lesion image features, establish a tobacco disease recognition model based on the average spectral curve and lesion image features, and generate a recognition result through a spectral recognition module and an image recognition module in the tobacco disease recognition model; Obtaining recognition results of the spectrum recognition module and the image recognition module, combining them according to weight information to generate a tobacco disease recognition result, and generating a disease grade according to the disease recognition result; Performing big data processing on the disease identification results to generate disease prevention and control methods, and displaying the disease identification results, disease levels and disease prevention and control methods in a preset manner; The average spectral curve and the lesion image features are generated, and a tobacco disease recognition model is established according to the average spectral curve and the lesion image features, and a recognition result is generated by a spectral recognition module and an image recognition module in the tobacco disease recognition model, wherein the spectral recognition module is specifically: Acquire spectral reflectance values ​​according to spectral data information of the target area of ​​the image, and generate an average spectral curve according to the spectral reflectance values; Among them, the average spectrum curve is generated according to the spectral reflectance value, and the average spectrum expression is: ， in, represents the desired average spectrum, Indicates The spectrum obtained by pixels, Indicates the number of pixels in the target area of ​​the image; Smoothing the average spectrum curve to obtain curve characteristics; Determine characteristic peaks through the curve characteristics, and decompose the average spectrum curve into continuous characteristic points according to the curve characteristics and the characteristic peaks; The characteristic points are introduced into a spectrum recognition module of a tobacco disease recognition model, and the spectrum recognition module generates a difference coefficient and a difference number through the characteristic point information and an average spectrum curve of healthy tobacco; generating a spectrum recognition result according to the difference coefficient and the number of differences, and determining the tobacco plant disease type according to the spectrum recognition result; The average spectral curve and the lesion image features are generated, and a tobacco disease recognition model is established according to the average spectral curve and the lesion image features, and a recognition result is generated by a spectral recognition module and an image recognition module in the tobacco disease recognition model, wherein the image recognition module is specifically: Acquiring a hyperspectral image of a tobacco plant, and performing mask processing on the hyperspectral image; Perform principal component analysis on the masked hyperspectral image to generate a principal component image; Select the segmentation threshold, mark the main component image into multiple independent regions, and calculate the gray value of each point in each independent region; In the independent area, if the gray value is greater than the segmentation threshold, the point is a diseased spot area, and a binary image of the diseased tobacco plant is extracted; Extract the coordinate information of the background area of ​​the hyperspectral image, mark the background area in the binary image according to the coordinate information, and replace the gray value of the background area with a value different from that of the diseased area and the healthy area; Extracting the lesion area, generating lesion image features, and importing the lesion image features into a tobacco disease recognition model; The tobacco disease recognition model is used to generate an image recognition result, and tobacco disease types are distinguished according to the image recognition result.

2. A method for identifying and preventing tobacco diseases according to claim 1, It is characterized in that The method of obtaining the hyperspectral image information of the tobacco plant and preprocessing the spectral data information and the image data information is specifically as follows: Acquire hyperspectral image information of tobacco plants, and acquire spectral data information and image data information from the hyperspectral image information; Extracting characteristic wavelengths from the spectral data information by preprocessing; The image target area is extracted from the image data information through preprocessing, the spectral data information of the image target area is exported, and the spectral data information of the image target area is subjected to noise elimination.

3. A method for identifying and preventing tobacco diseases according to claim 1, It is characterized in that The identification results of the spectrum identification module and the image identification module are obtained, and combined according to the weight information to generate tobacco disease identification results, and the disease level is generated according to the disease identification results, specifically: Obtaining spectrum recognition results and image recognition results, and performing initial weight assignment on the spectrum recognition results and image recognition results; Generate feedback information by verifying the accuracy of tobacco disease recognition model in identifying diseases; Adjust the initial weights according to the feedback information, generate an optimal weight combination, and determine the weight information; Combining the spectrum recognition result and the image recognition result according to the weight information to generate a disease recognition result; A disease index is generated based on the disease identification result, and a disease grade is generated based on the disease index.

4. A method for identifying and preventing tobacco diseases according to claim 1, It is characterized in that The big data processing is performed based on the disease identification results to generate disease prevention and control methods, and the disease identification results, disease levels and disease prevention and control methods are displayed in a preset manner, specifically: Obtaining disease identification results, performing keyword extraction based on the disease identification results, and generating disease feature information; Obtain similar disease prevention and control information and historical disease prevention and control information based on the disease characteristic information through big data processing; Calculate the matching degree of disease control methods by matching similar disease control information and historical disease control information according to disease levels; Determining the matching priority of the similar disease prevention and control information and the historical disease prevention and control information according to the matching degree; Analyze high-priority similar disease control information and historical disease control information, determine the required pesticides and dosage information, and generate disease control methods based on the required pesticides and dosage information; The disease identification results, disease levels and disease prevention methods are displayed in a preset manner.

5. A tobacco disease identification and prevention system, It is characterized in that The system includes: a memory and a processor, wherein the memory includes a tobacco disease identification and prevention method program, and when the tobacco disease identification and prevention method program is executed by the processor, the following steps are implemented: Acquire hyperspectral image information of tobacco plants, and pre-process the spectral data information and image data information; Generate an average spectral curve and lesion image features, establish a tobacco disease recognition model based on the average spectral curve and lesion image features, and generate a recognition result through a spectral recognition module and an image recognition module in the tobacco disease recognition model; Obtaining recognition results of the spectrum recognition module and the image recognition module, combining them according to weight information to generate a tobacco disease recognition result, and generating a disease grade according to the disease recognition result; Performing big data processing on the disease identification results to generate disease prevention and control methods, and displaying the disease identification results, disease levels and disease prevention and control methods in a preset manner; The average spectral curve and the lesion image features are generated, and a tobacco disease recognition model is established according to the average spectral curve and the lesion image features, and a recognition result is generated by a spectral recognition module and an image recognition module in the tobacco disease recognition model, wherein the spectral recognition module is specifically: Acquire spectral reflectance values ​​according to spectral data information of the target area of ​​the image, and generate an average spectral curve according to the spectral reflectance values; Among them, the average spectrum curve is generated according to the spectral reflectance value, and the average spectrum expression is: ， in, represents the desired average spectrum, Indicates The spectrum obtained by pixels, Indicates the number of pixels in the target area of ​​the image; Smoothing the average spectrum curve to obtain curve characteristics; Determine characteristic peaks through the curve characteristics, and decompose the average spectrum curve into continuous characteristic points according to the curve characteristics and the characteristic peaks; The characteristic points are introduced into a spectrum recognition module of a tobacco disease recognition model, and the spectrum recognition module generates a difference coefficient and a difference number through the characteristic point information and an average spectrum curve of healthy tobacco; generating a spectrum recognition result according to the difference coefficient and the number of differences, and determining the tobacco plant disease type according to the spectrum recognition result; The average spectral curve and the lesion image features are generated, and a tobacco disease recognition model is established according to the average spectral curve and the lesion image features, and a recognition result is generated by a spectral recognition module and an image recognition module in the tobacco disease recognition model, wherein the image recognition module is specifically: Acquiring a hyperspectral image of a tobacco plant, and performing mask processing on the hyperspectral image; Perform principal component analysis on the masked hyperspectral image to generate a principal component image; Select the segmentation threshold, mark the main component image into multiple independent regions, and calculate the gray value of each point in each independent region; In the independent area, if the gray value is greater than the segmentation threshold, the point is a diseased spot area, and a binary image of the diseased tobacco plant is extracted; Extract the coordinate information of the background area of ​​the hyperspectral image, mark the background area in the binary image according to the coordinate information, and replace the gray value of the background area with a value different from that of the diseased area and the healthy area; Extracting the lesion area, generating lesion image features, and importing the lesion image features into a tobacco disease recognition model; The tobacco disease recognition model is used to generate an image recognition result, and tobacco disease types are distinguished according to the image recognition result.

6. A computer-readable storage medium, Features: The computer-readable storage medium includes a tobacco disease identification and prevention method program, and when the tobacco disease identification and prevention method program is executed by the processor, the steps of the tobacco disease identification and prevention method as described in any one of claims 1 to 4 are implemented.

Images