Tobacco mosaic disease early-stage asymptomatic diagnosis method and automatic multispectral diagnosis equipment

Abstract

According to the early-stage asymptomatic diagnosis method for the tobacco mosaic disease and the automatic multispectral diagnosis equipment provided by the invention, the three spectral reflection wavebands are obtained through screening, so that the weak spectral change of the tobacco plant in the early stage of infection of the tobacco mosaic disease can be sensitively detected; the method can be used for distinguishing whether the tobacco plants are infected with viruses or not before the infected tobacco plants have scab symptoms, early-stage symptomless monitoring of the tobacco mosaic disease is really realized, the diagnosis precision exceeds 90%, early discovery and early prevention and treatment of the tobacco mosaic disease are really realized in actual production application, and missing of the optimal prevention and treatment opportunity is effectively prevented; on the basis of the three spectral reflection bands obtained through screening, a normalized tobacco mosaic disease early-stage symptomless diagnosis index is established, whether the tobacco plants are infected with the tobacco mosaic viruses or not is judged, and qualitative distinguishing of the tobacco plants is achieved; in addition, on the basis of three wave bands of 527 nm, 544 nm and 705 nm, corresponding automatic multispectral diagnosis equipment is formed, and real-time, rapid, lossless and large-area tobacco mosaic disease early monitoring is achieved.

Description

Technical Field

[0001] The invention relates to the field of plant lesion detection, and in particular to an early asymptomatic diagnosis method for tobacco mosaic disease and an automated multi-spectral diagnosis device. Background Art

[0002] Tobacco mosaic disease (TMV) is one of the most serious diseases in tobacco cultivation. Tobacco plants infected in the early stage may suffer a yield loss of more than 50%, and in severe cases, it may even lead to total crop failure, greatly affecting the yield and quality of tobacco leaves. Although the early application of chemical control agents can effectively control tobacco mosaic disease, the current information on tobacco mosaic disease infection is mainly obtained through manual sampling, which is time-consuming and labor-intensive, and difficult to monitor in real time, quickly, non-destructively and over a large area. It is also easy to cause cross-infection of healthy plants, resulting in missing the best time for disease prevention and control.

[0003] With the development of machine vision technology, image processing methods based on deep learning have been used in tobacco disease identification. Patent CN202010555276.1 proposes a tobacco disease identification method, system, platform and storage medium, which realizes the extraction of fine-grained features of disease images and solves the problem of small data scale. Although machine vision methods can significantly improve the accuracy and efficiency of disease identification, they are only applicable to single image recognition taken on the ground, and it is difficult to achieve efficient monitoring of large-area diseases.

[0004] Compared with machine vision methods, spectral imaging technology can quickly and accurately identify subtle changes in the health status of plants by capturing the reflectance characteristics of crops in different spectral bands, making early disease monitoring possible. There are already multiple patents that have realized the application of hyperspectral imaging technology in disease monitoring, such as patent CN201910513049.X proposed a method for identifying tea tree anthracnose lesions based on imaging hyperspectral technology, and patent CN202311138666.9 proposed a method for identifying asymptomatic infection of strawberry anthracnose based on hyperspectral features; in the tobacco field, patent CN202111177984.7 mainly focuses on the recognition of lesion image features, and can only distinguish tobacco diseases that have obvious symptoms. Patent CN202310064307.7 proposes a method, system, electronic device and storage medium for identifying early tobacco diseases, but it is based on handheld near-infrared spectral data and cannot be used for large-scale monitoring. In addition, given that hyperspectral equipment is expensive and acquires large amounts of data, high-performance computing resources are required for processing, posing a challenge to the real-time, rapid, large-scale, non-destructive, early asymptomatic monitoring of tobacco mosaic disease. Summary of the invention

[0005] The purpose of the present invention is to provide an early asymptomatic diagnosis method for tobacco mosaic disease and an automated multi-spectral diagnosis device, which collects high-spectral image data of tobacco plants inoculated with tobacco mosaic virus in a test field to generate a number of sample data; pre-processes the sample data to obtain spectral reflection data; based on the spectral reflection data, screens and obtains characteristic wavebands for early diagnosis of tobacco mosaic disease, thereby obtaining early asymptomatic diagnosis wavebands for tobacco mosaic disease; based on the early asymptomatic diagnosis wavebands for tobacco mosaic disease, obtains an early asymptomatic diagnosis index for tobacco mosaic disease, thereby judging whether the tobacco plants are infected with tobacco mosaic virus; by screening and obtaining the three spectral reflection wavebands, the weak spectral changes exhibited by tobacco plants in the early stage of infection with tobacco mosaic virus can be sensitively detected, and the infected tobacco plants can be distinguished before the lesion symptoms appear. Whether the tobacco plants are infected with the virus, truly realize the early asymptomatic monitoring of tobacco mosaic disease, with a diagnostic accuracy of more than 90%, and truly realize the early detection and early prevention of tobacco mosaic disease in actual production applications, effectively preventing the best prevention and control opportunities from being missed; based on the three spectral reflectance bands obtained through screening, a normalized early asymptomatic diagnosis index for tobacco mosaic disease is established to determine whether the tobacco plants are infected with tobacco mosaic virus, and to achieve qualitative distinction of tobacco plants; in addition, based on the three bands of 527nm, 544nm, and 705nm, a corresponding automated multispectral diagnostic device is formed to return the tobacco mosaic disease infection status of tobacco plants in real time. Compared with the original hyperspectral equipment, it has lower cost and is easier to mount on drones, realizing real-time, rapid, non-destructive and large-area early monitoring of tobacco mosaic disease.

[0006] The present invention is achieved through the following technical solutions:

[0007] An early asymptomatic diagnosis method for tobacco mosaic disease, comprising:

[0008] Collecting hyperspectral image data of tobacco plants inoculated with tobacco mosaic virus in a test field, generating corresponding sample data based on the hyperspectral image data; preprocessing the sample data to obtain corresponding spectral reflectance data;

[0009] Based on the spectral reflectance data, a characteristic waveband for early diagnosis of tobacco mosaic disease is screened; and based on the characteristic waveband for early diagnosis of tobacco mosaic disease, an early asymptomatic diagnosis waveband for tobacco mosaic disease is obtained;

[0010] Based on the tobacco mosaic disease early asymptomatic diagnosis band, a tobacco mosaic disease early asymptomatic diagnosis index is obtained; and based on the tobacco mosaic disease early asymptomatic diagnosis index, it is determined whether the tobacco plant is infected with tobacco mosaic virus.

[0011] Optionally, collecting hyperspectral image data of tobacco plants inoculated with tobacco mosaic virus in the test field includes:

[0012] Inoculating tobacco mosaic virus into tobacco plants on the test field, spreading corundum and spraying virus buffer on a number of new leaves of the tobacco plants, and rubbing the new leaves;

[0013] On the 2nd day and the 11th day after the tobacco mosaic virus is inoculated, the first hyperspectral image data and the second hyperspectral image data of the tobacco plants in the test field are collected by using a drone, respectively.

[0014] Optionally, based on the hyperspectral image data, corresponding sample data are generated, including:

[0015] After whiteboard correction is performed on the first hyperspectral image data and the second hyperspectral image data, tobacco plants in the first hyperspectral image data and the second hyperspectral image data are manually marked to form corresponding manually marked areas; and high noise areas and background areas of the first hyperspectral image data and the second hyperspectral image data are deleted respectively;

[0016] The band data of each artificially marked area under the first hyperspectral image data and the second hyperspectral image data are extracted to calculate the average band of each artificially marked area, and the average band of each artificially marked area is used as a sample data.

[0017] Optionally, preprocessing the plurality of sample data to obtain corresponding spectral reflectance data includes:

[0018] The reflectance data of 178 bands contained in each sample data are transformed through standard normal variable transformation to obtain the spectral reflectance data corresponding to each sample data.

[0019] Optionally, based on the spectral reflectance data, characteristic wavebands for early diagnosis of tobacco mosaic disease are screened and obtained, including:

[0020] The spectral reflectance data is processed by a competitive adaptive reweighting algorithm to filter out redundant and irrelevant bands of the spectral reflectance data, thereby screening out characteristic bands for early diagnosis of tobacco mosaic disease.

[0021] Optionally, based on the tobacco mosaic disease early diagnosis characteristic band, an early asymptomatic diagnosis band of tobacco mosaic disease is obtained, including:

[0022] Determine a common band between tobacco mosaic disease early diagnosis characteristic bands corresponding to the first hyperspectral image data and the second hyperspectral image data;

[0023] Based on the common band, the spectral reflectance data corresponding to the first hyperspectral image data and the second hyperspectral image data are processed by a second competitive adaptive reweighting algorithm to obtain the reserved characteristic bands corresponding to the first hyperspectral image data and the second hyperspectral image data;

[0024] Then, the intersection of the reserved characteristic bands corresponding to the hyperspectral image data and the second hyperspectral image data is used as the early asymptomatic diagnosis band of tobacco mosaic disease.

[0025] Optionally, based on the tobacco mosaic disease early asymptomatic diagnosis band, an early asymptomatic diagnosis index of tobacco mosaic disease is obtained, including:

[0026] The asymptomatic early diagnosis band of tobacco mosaic disease is selected to obtain data corresponding to the three key bands of 527nm, 544nm and 705nm;

[0027] The first hyperspectral image data and the second hyperspectral image data are merged, and a Logistic regression model is constructed based on the data corresponding to the three key bands, so as to obtain an early asymptomatic diagnosis index of tobacco mosaic disease.

[0028] Optionally, judging whether a tobacco plant is infected with tobacco mosaic virus based on the early asymptomatic diagnosis index of tobacco mosaic virus comprises:

[0029] A threshold comparison is performed on the tobacco mosaic disease early asymptomatic diagnosis index. If the tobacco mosaic disease early asymptomatic diagnosis index is greater than a preset threshold, it is determined that the tobacco plant is infected with tobacco mosaic virus; otherwise, it is determined that the tobacco plant is not infected with tobacco mosaic virus.

[0030] An automated multi-spectral diagnostic device for realizing the above-mentioned early asymptomatic diagnosis method of tobacco mosaic disease, comprising:

[0031] A multi-lens single camera, including three bandpass filters corresponding to 527nm, 544nm, and 705nm, is used to collect reflected light in the three key bands of 527nm, 544nm, and 705nm;

[0032] The data processing chip is used to process the reflected band data of the reflected light in the three key bands to obtain an early asymptomatic diagnosis index of tobacco mosaic disease; and based on the obtained early asymptomatic diagnosis index of tobacco mosaic disease, determine whether the tobacco plant is infected with tobacco mosaic virus.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] The method for early asymptomatic diagnosis of tobacco mosaic disease and the automated multispectral diagnostic equipment provided by the present application collect hyperspectral image data of tobacco plants inoculated with tobacco mosaic virus in a test field to generate a number of sample data; pre-process the sample data to obtain spectral reflectance data; based on the spectral reflectance data, screen and obtain characteristic bands for early diagnosis of tobacco mosaic disease, thereby obtaining early asymptomatic diagnosis bands for tobacco mosaic disease; based on the early asymptomatic diagnosis bands for tobacco mosaic disease, obtain an early asymptomatic diagnosis index for tobacco mosaic disease, thereby judging whether the tobacco plants are infected with tobacco mosaic virus. The three spectral reflectance bands obtained by screening can sensitively detect the weak spectral changes shown by tobacco plants in the early stage of infection with tobacco mosaic disease, and can distinguish tobacco plants from infected plants before the onset of lesion symptoms. Whether the tobacco plant is infected with the virus, truly realize the early asymptomatic monitoring of tobacco mosaic disease, with a diagnostic accuracy of more than 90%, and truly realize the early detection and early prevention of tobacco mosaic disease in actual production applications, effectively preventing the best prevention and control opportunities from being missed; based on the three spectral reflectance bands obtained by screening, a normalized early asymptomatic diagnosis index for tobacco mosaic disease is established to determine whether the tobacco plant is infected with tobacco mosaic virus, and to achieve qualitative distinction of tobacco plants; in addition, based on the three bands of 527nm, 544nm, and 705nm, a corresponding automated multispectral diagnostic device is formed to return the tobacco mosaic disease infection status of tobacco plants in real time. Compared with the original hyperspectral equipment, it has lower cost and is easier to mount on drones, realizing real-time, rapid, non-destructive and large-area early monitoring of tobacco mosaic disease. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0036] Figure 1 A schematic diagram of a process for diagnosing tobacco mosaic disease in the early stage asymptomatically provided by the present invention.

[0037] Figure 2 Independent test result graphs corresponding to 11 common vegetation indices were introduced for the early asymptomatic diagnosis model of TMV.

[0038] Figure 3 The original RGB image and the visualization image fused from the three retained bands of 527nm, 544nm, and 705nm.

[0039] Figure 4 A result map for the TMV early asymptomatic diagnostic index based on the three retained key bands was established.

[0040] Figure 5 This is a structural block diagram of an automated multi-spectral diagnostic device provided by the present invention. DETAILED DESCRIPTION

[0041] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are described in detail below in conjunction with the accompanying drawings. It is to be understood that the specific embodiments described herein are only used to explain the present application, rather than to limit the present application. It should also be noted that, for ease of description, only some structures related to the present application are shown in the accompanying drawings, rather than all structures. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0042] The terms "including" and "having" and any variations thereof in this application are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices.

[0043] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0044] See also Figure 1 As shown, an embodiment of the present application provides an early asymptomatic diagnosis method for tobacco mosaic disease. The early asymptomatic diagnosis method for tobacco mosaic disease comprises:

[0045] Collecting hyperspectral image data of tobacco plants inoculated with tobacco mosaic virus in the test field, generating corresponding sample data based on the hyperspectral image data; preprocessing the sample data to obtain corresponding spectral reflectance data;

[0046] Based on the spectral reflectance data, the characteristic waveband for early diagnosis of tobacco mosaic disease is screened out; and based on the characteristic waveband for early diagnosis of tobacco mosaic disease, the early asymptomatic diagnosis waveband of tobacco mosaic disease is obtained;

[0047] Based on the tobacco mosaic disease early asymptomatic diagnosis band, a tobacco mosaic disease early asymptomatic diagnosis index is obtained; and based on the tobacco mosaic disease early asymptomatic diagnosis index, it is determined whether the tobacco plant is infected with tobacco mosaic virus.

[0048] The beneficial effects of the above embodiments are as follows: the method for diagnosing early asymptomatic tobacco mosaic disease collects high-spectral image data of tobacco plants inoculated with tobacco mosaic virus in a test field to generate a number of sample data; pre-processes the sample data to obtain spectral reflectance data; based on the spectral reflectance data, screens and obtains characteristic bands for early diagnosis of tobacco mosaic disease, thereby obtaining early asymptomatic diagnosis bands for tobacco mosaic disease; based on the early asymptomatic diagnosis bands for tobacco mosaic disease, obtains an early asymptomatic diagnosis index for tobacco mosaic disease, thereby judging whether the tobacco plants are infected with tobacco mosaic virus; the three spectral reflectance bands obtained by screening can sensitively detect the weak spectral changes exhibited by tobacco plants in the early stage of infection with tobacco mosaic disease, and can distinguish whether the tobacco plants are infected with tobacco mosaic virus before the infected tobacco plants show symptoms of lesions. Whether tobacco plants are infected with the virus, truly realizing early asymptomatic monitoring of tobacco mosaic disease, with a diagnostic accuracy of more than 90%, truly realizing early detection and early prevention of tobacco mosaic disease in actual production applications, and effectively preventing the best prevention and control opportunities from being missed; based on the three spectral reflectance bands obtained through screening, a normalized early asymptomatic diagnosis index for tobacco mosaic disease is established to determine whether tobacco plants are infected with tobacco mosaic virus, thereby achieving qualitative distinction of tobacco plants; in addition, based on the three bands of 527nm, 544nm, and 705nm, a corresponding automated multispectral diagnostic device is formed to return the tobacco mosaic disease infection status of tobacco plants in real time. Compared with the original hyperspectral equipment, it has lower cost and is easier to mount on drones, thus realizing real-time, rapid, non-destructive and large-area early monitoring of tobacco mosaic disease.

[0049] In another embodiment, collecting hyperspectral image data of tobacco plants inoculated with tobacco mosaic virus in a test field includes:

[0050] Inoculating tobacco mosaic virus into tobacco plants on the test field, spreading corundum and spraying virus buffer on a number of new leaves of the tobacco plants, and rubbing the new leaves;

[0051] On the 2nd day and the 11th day after the inoculation of tobacco mosaic virus, the first hyperspectral image data and the second hyperspectral image data of the tobacco plants in the test field were collected by using a drone.

[0052] In actual operation, the tobacco mosaic virus was artificially inoculated on the "Yunyan 87" tobacco plants in the cluster stage in the test field, and corundum was spread on the 3-4 new leaves of the tobacco plants, and then the virus buffer was sprayed. Finally, the back of the tobacco leaves was held with one hand, and the leaf surface covered with corundum was gently rubbed twice with one hand, which can promote the spread of tobacco mosaic virus inside the tobacco plants. On the second day (asymptomatic infection period) and the 11th day (bright vein period) after inoculation of tobacco mosaic virus, the first hyperspectral image data and the second hyperspectral image data of the tobacco plants in the test field were collected by drones, wherein the hyperspectral acquisition equipment carried by the drone can be but not limited to SPECIMA FX10, the spectral resolution is 3.5nm, the acquisition time is 11:30 am, the weather conditions are required to be clear and breezy, and the flight altitude of the drone can be but not limited to 30m, so that all tobacco plants in the test field can be collected in real time over a large area of ​​hyperspectral, and the use of drones can achieve flexible and non-destructive monitoring of tobacco plants in the test field.

[0053] In another embodiment, based on the hyperspectral image data, corresponding sample data are generated, including:

[0054] After whiteboard correction is performed on the first hyperspectral image data and the second hyperspectral image data, tobacco plants in the first hyperspectral image data and the second hyperspectral image data are manually marked to form corresponding manually marked areas; and high noise areas and background areas of the first hyperspectral image data and the second hyperspectral image data are deleted respectively;

[0055] The band data of each artificially marked area under the first hyperspectral image data and the second hyperspectral image data are extracted to calculate the average band of each artificially marked area, and the average band of each artificially marked area is used as a sample data.

[0056] After the drone collects the first and second hyperspectral image data of tobacco plants at different days after artificial inoculation of tobacco mosaic virus, the two hyperspectral image data are first whiteboard corrected, and then the tobacco plants in the hyperspectral image data are manually marked using the Python Labelem module to form corresponding artificially marked areas. Each artificially marked area corresponds to a tobacco plant, and the high noise area and background area in the hyperspectral image data are deleted at the same time, which can reduce the noise interference and background flooding effect in the hyperspectral image data. Then, the band data in each artificially marked area is extracted by writing a Python program, and the average band in each artificially marked area is calculated as a sample data. In this way, if the first hyperspectral image data and the second hyperspectral image data are defined as Dataset1 and Dataset2 respectively, Dataset1 can contain 510 diseased sample data and 331 healthy sample data, and Dataset2 can contain 517 diseased sample data and 402 healthy sample data, so as to comprehensively and effectively characterize the hyperspectral band characteristics of each tobacco plant.

[0057] In another embodiment, the plurality of sample data are preprocessed to obtain corresponding spectral reflectance data, including:

[0058] The reflectance data of 178 bands contained in each sample data are transformed through standard normal variable transformation to obtain the spectral reflectance data corresponding to each sample data.

[0059] Each sample data contains reflectance data of 178 bands. The reflectance data is transformed by Standard Normal Variate (SNV), which can reduce the influence of light scattering and baseline drift. The formula of Standard Normal Variate is as follows:

[0060]

[0061] In the above formula (1), n is the number of wavelengths, x i is the wavelength reflectance; after the standard normal variable transformation, the average value of each sample data of the first hyperspectral image data and the second hyperspectral image data is 0, and the standard deviation is 1.

[0062] In another embodiment, based on the spectral reflectance data, characteristic wavebands for early diagnosis of tobacco mosaic disease are screened and obtained, including:

[0063] The spectral reflectance data is processed by a competitive adaptive reweighting algorithm to filter out redundant and irrelevant bands of the spectral reflectance data, thereby screening out characteristic bands for early diagnosis of tobacco mosaic disease.

[0064] In order to ensure the accuracy of screening characteristic bands for early diagnosis of tobacco mosaic disease, the competitive adaptive reweighted sampling (CARS) algorithm was used to process the spectral reflectance data, and the redundant and irrelevant bands of the spectral reflectance data were filtered out to screen the characteristic bands for early diagnosis of tobacco mosaic disease. Specifically, 80% of the samples were randomly selected in an equal probability manner, and a partial least squares (PLS) model was established to calculate the regression coefficient of each band, and the importance of each band was characterized by the normalized absolute value weight. The exponential decay function was further used to remove the wavelengths with relatively small absolute value weights of the regression coefficient to obtain the candidate characteristic band subset. The performance of the candidate characteristic band subset was evaluated based on the prediction accuracy of 10-fold cross validation (10-CV); after N iterations (for example, 50 times), the candidate band subset with the best prediction accuracy was selected as the final retained band, so as to screen the characteristic bands for early diagnosis of tobacco mosaic disease.

[0065] In another embodiment, based on the tobacco mosaic disease early diagnosis characteristic band, the tobacco mosaic disease early asymptomatic diagnosis band is obtained, including:

[0066] Determine a common band between tobacco mosaic disease early diagnosis characteristic bands corresponding to the first hyperspectral image data and the second hyperspectral image data;

[0067] Based on the shared band, the spectral reflectance data corresponding to the first hyperspectral image data and the second hyperspectral image data are processed by a second competitive adaptive reweighting algorithm, so as to obtain the reserved characteristic bands corresponding to the first hyperspectral image data and the second hyperspectral image data;

[0068] Then, the intersection of the reserved characteristic bands corresponding to the hyperspectral image data and the second hyperspectral image data is used as the early asymptomatic diagnosis band of tobacco mosaic disease.

[0069] For example, when using the competitive adaptive reweighted sampling (CARS) algorithm to process spectral reflectance data, 62 bands are retained for Dataset1 and 79 bands are retained for Dataset2, of which 39 bands are shared between the two. Based on the 39 shared bands, a second feature band screening is performed on Dataset1 and Dataset2 respectively based on the competitive adaptive reweighted sampling algorithm, and finally the intersection of the second retained feature bands of Dataset1 and Dataset2 is used as the early asymptomatic diagnosis band of tobacco mosaic disease, which can ensure the accuracy and reliability of the early asymptomatic diagnosis band of tobacco mosaic disease.

[0070] In another embodiment, based on the tobacco mosaic disease early asymptomatic diagnosis band, the tobacco mosaic disease early asymptomatic diagnosis index is obtained, including:

[0071] The early asymptomatic diagnosis band of tobacco mosaic disease was selected, and the data corresponding to the three key bands of 527nm, 544nm and 705nm were obtained;

[0072] The first hyperspectral image data and the second hyperspectral image data are combined, and a Logistic regression model is constructed based on the data corresponding to the three key bands, so as to obtain an early asymptomatic diagnosis index of tobacco mosaic disease.

[0073] In actual operation, the early asymptomatic diagnosis band of tobacco mosaic disease is selected to obtain the reflectance corresponding to the three key bands of 527nm, 544nm, and 705nm. Then Dataset1 and Dataset2 are merged to construct a Logistic regression model to obtain the early asymptomatic diagnosis index of tobacco mosaic disease TMVI. The calculation formula of the early asymptomatic diagnosis index of tobacco mosaic disease TMVI is as follows:

[0074] TMVI=18.18R 527 -18.57R 544 +14.59R 705 +7.42 (2)

[0075] In the above formula (2), R 527 , R 544 , R 705 They respectively represent the reflectivity corresponding to the three key bands of 527nm, 544nm and 705nm.

[0076] In another embodiment, judging whether a tobacco plant is infected with tobacco mosaic virus based on the early asymptomatic diagnosis index of tobacco mosaic virus comprises:

[0077] A threshold comparison is performed on the early asymptomatic diagnosis index of tobacco mosaic disease. If the early asymptomatic diagnosis index of tobacco mosaic disease is greater than a preset threshold, it is determined that the tobacco plant is infected with tobacco mosaic virus; otherwise, it is determined that the tobacco plant is not infected with tobacco mosaic virus.

[0078] In actual operation, if the early asymptomatic diagnosis index of tobacco mosaic disease is greater than 0, it is judged that the tobacco plant is infected with tobacco mosaic virus; otherwise, it is judged that the tobacco plant is not infected with tobacco mosaic virus.

[0079] After screening by CARS, 62 bands were screened based on Dataset 1, which was defined as the reserved subset A; 79 bands were screened based on Dataset 2, which was defined as the reserved subset B. To verify the external prediction performance of the model, for the reserved subset A, Dataset 1 was used as the training set and Dataset 2 was used as the test set; for the reserved subset B, Dataset 2 was used as the training set and Dataset 1 was used as the test set. Support Vector Classification (SVC), Partial Least Squares Discriminant Analysis (PLS-DA), K-Nearest Neighbor (KNN), Random Forest (RF), and Distributed Gradient Boosting (XGBoost) were introduced to build the classification model. The prediction results of the two sets of reserved subsets based on different classifications on independent tests are shown in Table 1. When no band selection was performed, the average accuracy of the independent test was 67.9%; when CARS performed band screening, the average prediction accuracy of the model was slightly improved, and the accuracy was increased to 69.83%. The early asymptomatic diagnosis of TMV has not yet reached the requirements for field application.

[0080] Table 1 Independent test classification accuracy of two sets of reserved subsets

[0081]

[0082] Furthermore, the three characteristic bands (527nm, 544nm, and 705nm) retained in the second round of band screening were used to construct an early asymptomatic diagnosis model for TMV. In order to compare the diagnostic performance of the retained features, 11 common vegetation indices (as shown in Table 2) were introduced at the same time. The independent test results of each model are shown in Figure 2As shown. When Dataset 1 is used as the training set and Dataset 2 is used as the independent test set, the independent test accuracy of the model using the three characteristic bands retained by the patented method is the highest, with an average accuracy of 86.72%. The independent test accuracy of the four vegetation indices MSAVI, PVI, DVI and EVI is second, at 79.11%, 78.24%, 76.82% and 76.67% respectively. The independent test accuracy of the other seven vegetation indices is less than 50%. When Dataset2 is used as the training set and Dataset1 is used as the independent test set, the independent test accuracy of the model using the three characteristic bands retained by the patented method is still the highest, reaching 87.44%, while the 11 vegetation indices perform poorly, with an accuracy of only about 50%. The independent test accuracy using three characteristic bands has an average accuracy of 87.08 in the two datasets, which is significantly better than the reference vegetation index. Among the five classification models, the SVC had the highest classification accuracy of 90.75% and 91.32%, and its TMV early asymptomatic diagnosis effect met the field needs.

[0083] Table 2 Eleven common vegetation indices

[0084]

[0085]

[0086] also, Figure 3 The original RGB image and the visualization image fused from the three retained bands of 527nm, 544nm, and 705nm are displayed. In the original RGB image, there is no obvious difference between TMV-susceptible tobacco plants and healthy tobacco plants, but in the visualization image based on the three retained bands, TMV-susceptible tobacco plants and healthy tobacco plants show a significant color difference, with the diseased tobacco plants appearing white and the healthy tobacco plants appearing green.

[0087] Finally, the TMV early asymptomatic diagnostic index was established based on the three retained key bands, such as Figure 4 As shown in Figure 1, the normalized index can effectively achieve the qualitative goal of determining whether a tobacco plant is infected with tobacco mosaic virus. When the index is greater than 0, it can be determined that the tobacco plant is infected with tobacco mosaic virus, and the determination accuracy of the two data sets exceeds 82%.

[0088] See also Figure 5 As shown, an embodiment of the present application provides an automated multi-spectral diagnostic device. The automated multi-spectral diagnostic device can implement the above-mentioned early asymptomatic diagnosis method of tobacco mosaic disease, which includes:

[0089] A multi-lens single camera, including three bandpass filters corresponding to 527nm, 544nm, and 705nm, is used to collect reflected light in the three key bands of 527nm, 544nm, and 705nm;

[0090] The data processing chip is used to process the reflected band data of the reflected light in the three key bands to obtain an early asymptomatic diagnosis index of tobacco mosaic disease; and based on the obtained early asymptomatic diagnosis index of tobacco mosaic disease, determine whether the tobacco plant is infected with tobacco mosaic virus.

[0091] The role and effect of this automated multi-spectral diagnostic equipment are the same as the asymptomatic early diagnosis method of tobacco mosaic disease introduced above, and will not be repeated here.

[0092] In general, the method for early asymptomatic diagnosis of tobacco mosaic disease collects hyperspectral image data of tobacco plants inoculated with tobacco mosaic virus in a test field to generate a number of sample data; preprocesses the sample data to obtain spectral reflectance data; based on the spectral reflectance data, screens and obtains characteristic bands for early diagnosis of tobacco mosaic disease, thereby obtaining early asymptomatic diagnosis bands for tobacco mosaic disease; based on the early asymptomatic diagnosis bands for tobacco mosaic disease, obtains an early asymptomatic diagnosis index for tobacco mosaic disease, thereby judging whether the tobacco plants are infected with tobacco mosaic virus. The three spectral reflectance bands obtained by screening can sensitively detect the weak spectral changes exhibited by tobacco plants in the early stage of infection with tobacco mosaic virus, and can distinguish whether the tobacco plants are infected before the infected tobacco plants show symptoms of lesions. Virus, truly realize the early asymptomatic monitoring of tobacco mosaic disease, with a diagnostic accuracy of more than 90%, and truly realize the early detection and early prevention of tobacco mosaic disease in actual production applications, effectively preventing the loss of the best prevention and control opportunity; based on the three spectral reflectance bands obtained by screening, a normalized early asymptomatic diagnosis index of tobacco mosaic disease is established to determine whether tobacco plants are infected with tobacco mosaic virus, and to achieve qualitative distinction of tobacco plants; in addition, based on the three bands of 527nm, 544nm, and 705nm, a corresponding automated multispectral diagnostic device is formed to return the tobacco mosaic disease infection status of tobacco plants in real time. Compared with the original hyperspectral equipment, it has lower cost and is easier to mount on drones, realizing real-time, rapid, non-destructive and large-area early monitoring of tobacco mosaic disease.

[0093] The above is only a specific implementation of the present invention, and any other improvements made based on the concept of the present invention are considered to be within the protection scope of the present invention.

Claims

1. A method for early asymptomatic diagnosis of tobacco mosaic disease, characterized in that: include: Collecting hyperspectral image data of tobacco plants inoculated with tobacco mosaic virus in a test field, and generating corresponding sample data based on the hyperspectral image data; Preprocessing the sample data to obtain corresponding spectral reflectance data; Based on the spectral reflectance data, a characteristic waveband for early diagnosis of tobacco mosaic disease is screened; and based on the characteristic waveband for early diagnosis of tobacco mosaic disease, an early asymptomatic diagnosis waveband for tobacco mosaic disease is obtained; Based on the tobacco mosaic disease early asymptomatic diagnosis band, a tobacco mosaic disease early asymptomatic diagnosis index is obtained; and based on the tobacco mosaic disease early asymptomatic diagnosis index, it is determined whether the tobacco plant is infected with tobacco mosaic virus.

2. The method for diagnosing tobacco mosaic disease in the early stage asymptomatically as described in claim 1, characterized in that: Collect hyperspectral image data of tobacco plants inoculated with tobacco mosaic virus in the experimental field, including: Tobacco mosaic virus is inoculated on tobacco plants in the test field, and then corundum is spread on several new leaves of the tobacco plants and virus buffer is sprayed, and the new leaves are rubbed; on the 2nd day and the 11th day after the inoculation of tobacco mosaic virus, the first hyperspectral image data and the second hyperspectral image data of the tobacco plants in the test field are respectively collected by using a drone.

3. The method for diagnosing tobacco mosaic disease in the early stage asymptomatically as described in claim 2, characterized in that: Based on the hyperspectral image data, a number of corresponding sample data are generated, including: After whiteboard correction is performed on the first hyperspectral image data and the second hyperspectral image data respectively, tobacco plants in the first hyperspectral image data and the second hyperspectral image data are manually marked to form corresponding manually marked areas; high noise areas and background areas of the first hyperspectral image data and the second hyperspectral image data are deleted respectively; band data of each manually marked area under the first hyperspectral image data and the second hyperspectral image data are extracted to calculate the average band of each manually marked area, and the average band of each manually marked area is used as a sample data.

4. The method for diagnosing tobacco mosaic disease in the early stage asymptomatically as described in claim 3, characterized in that: Preprocessing the sample data to obtain corresponding spectral reflectance data includes: The reflectance data of 178 bands contained in each sample data are transformed through standard normal variable transformation to obtain the spectral reflectance data corresponding to each sample data.

5. The method for diagnosing tobacco mosaic disease in the early stage asymptomatically as described in claim 4, characterized in that: Based on the spectral reflectance data, characteristic bands for early diagnosis of tobacco mosaic disease are screened, including: The spectral reflectance data is processed by a competitive adaptive reweighting algorithm to filter out redundant and irrelevant bands of the spectral reflectance data, thereby screening out characteristic bands for early diagnosis of tobacco mosaic disease.

6. The method for diagnosing tobacco mosaic disease in the early stage asymptomatically as described in claim 5, characterized in that: Based on the tobacco mosaic disease early diagnosis characteristic band, the tobacco mosaic disease early asymptomatic diagnosis band is obtained, including: Determine a common band between tobacco mosaic disease early diagnosis characteristic bands corresponding to the first hyperspectral image data and the second hyperspectral image data; Based on the common bands, the spectral reflectance data corresponding to the first hyperspectral image data and the second hyperspectral image data are processed by a second competitive adaptive reweighting algorithm, so as to obtain the reserved characteristic bands corresponding to the first hyperspectral image data and the second hyperspectral image data; Then, the intersection of the reserved characteristic bands corresponding to the hyperspectral image data and the second hyperspectral image data is used as the early asymptomatic diagnosis band of tobacco mosaic disease.

7. The method for diagnosing tobacco mosaic disease in the early stage asymptomatically as described in claim 6, characterized in that: Based on the tobacco mosaic disease early asymptomatic diagnosis band, the tobacco mosaic disease early asymptomatic diagnosis index is obtained, including: The asymptomatic early diagnosis band of tobacco mosaic disease is selected to obtain data corresponding to the three key bands of 527nm, 544nm and 705nm; The first hyperspectral image data and the second hyperspectral image data are merged, and a Logistic regression model is constructed based on the data corresponding to the three key bands, so as to obtain an early asymptomatic diagnosis index of tobacco mosaic disease.

8. The method for diagnosing tobacco mosaic disease in the early stage asymptomatically as described in claim 7, characterized in that: Based on the tobacco mosaic virus early asymptomatic diagnosis index, judging whether the tobacco plant is infected with tobacco mosaic virus includes: A threshold comparison is performed on the tobacco mosaic disease early asymptomatic diagnosis index. If the tobacco mosaic disease early asymptomatic diagnosis index is greater than a preset threshold, it is determined that the tobacco plant is infected with tobacco mosaic virus; otherwise, it is determined that the tobacco plant is not infected with tobacco mosaic virus.

9. An automated multi-spectral diagnostic device for implementing the method for diagnosing early asymptomatic tobacco mosaic disease according to any one of claims 7 to 8, characterized in that: include: A multi-lens single camera, including three bandpass filters corresponding to 527nm, 544nm, and 705nm, is used to collect reflected light in the three key bands of 527nm, 544nm, and 705nm; The data processing chip is used to process the reflected band data of the reflected light in the three key bands to obtain an early asymptomatic diagnosis index of tobacco mosaic disease; and based on the obtained early asymptomatic diagnosis index of tobacco mosaic disease, determine whether the tobacco plant is infected with tobacco mosaic virus.

Images