CN-121994725-A - Early asymptomatic diagnosis method and equipment for tobacco black shank

CN121994725ACN 121994725 ACN121994725 ACN 121994725ACN-121994725-A

Abstract

The invention discloses a method and equipment for diagnosing early-stage asymptomatic tobacco black shank, which belong to the technical field of tobacco black shank diagnosis, and specifically relate to 437 nm, 530, nm, 649, nm and 710 and nm which are main sensitive wave bands of early-stage asymptomatic tobacco black shank after wave band screening, and a Logistic regression model is built based on the four wave bands after Dataset and Dataset 2 are combined to obtain early-stage asymptomatic tobacco black shank diagnostic indexes. Therefore, the invention shows extremely high sensitivity to weak physiological changes at the early stage of black shank infection based on the wave band, and can realize effective differentiation of healthy plants and infected plants by virtue of leaf spectrum details even when macroscopic rot symptoms do not appear on root systems and stem bases.

Inventors

WU SHAOLONG
TANG QIANJUN
CHEN YUAN
XIAO YANSONG
LI JIAYING
TENG KAI
CAI HAILIN

Assignees

中国烟草总公司湖南省公司

Dates

Publication Date: 20260508
Application Date: 20251204

Claims (8)

1. The early asymptomatic diagnosis method for tobacco black shank is characterized by comprising the following specific steps: S1, training data acquisition, namely acquiring spectrum data and a standardized sample of an early asymptomatic stage of tobacco black shank, supporting subsequent modeling and analysis, and carrying out training data acquisition work by adopting a mode of combining manual inoculation test with hyperspectral acquisition and data preprocessing so as to obtain two data sets comprising a disease-sensing sample and a health sample; S2, selecting a tobacco black shank characteristic wave band, namely, based on spectral data after standard normal variable transformation pretreatment, adopting a competitive self-adaptive re-weighting sampling algorithm to carry out screening work, firstly constructing a partial least square regression model by 80% of samples, evaluating the wave band importance by the absolute value weight of a regression coefficient, gradually removing a low-weight wave band by combining an exponential decay function, evaluating the performance of a candidate wave band through 10-fold cross validation, and then respectively screening 60 and 77 characteristic wave bands from two data sets after 70 iterations, and taking the intersection of the two characteristic wave bands as final characteristic input for early asymptomatic recognition of the tobacco black shank; s3, determining a sensitive wave band of the early asymptomatic stage of the tobacco black shank based on a characteristic wave band screening result, combining two data sets, and constructing a Logistic regression model based on the sensitive wave band to finally obtain the early asymptomatic diagnostic index of the tobacco black shank.
2. The method for early asymptomatic diagnosis of tobacco black shank according to claim 1, wherein the specific steps of S1 are as follows: s11, hyperspectral data acquisition: s111, in an open-air test field, selecting a tobacco plant of Yunyan 116 with consistent growth and in a seedling stage to carry out a tobacco black shank patient artificial inoculation test; s112, after inoculation, enabling plants to continuously grow and manage under natural field conditions, and then respectively collecting hyperspectral data of the unmanned aerial vehicle on the 2 nd day and the 4 th day after inoculation, and simultaneously respectively collecting hyperspectral images of the unmanned aerial vehicle on the 2 nd day and the 4 th day after inoculation; S113, after image acquisition, firstly performing whiteboard correction to eliminate the ambient light difference, then manually calibrating a smoke plant area by utilizing a Labelme module of Python to remove background and high noise parts, and then extracting reflectivity data of all wave bands of each marking area by programming and calculating the average value of the reflectivity data as spectral characteristics of a single sample, wherein the data acquired on the 2 nd day after inoculation is defined as Dataset and comprises 495 infected samples and 320 healthy samples, the data on the 4 th day is defined as Dataset and comprises 508 infected samples and 410 healthy samples; S12, data preprocessing: the spectrum data is subjected to standard normal variable transformation processing, and the calculation formula is as follows: Wherein, the N is the number of wavelengths, The average value of the samples is 0 and the standard deviation is 1 after the conversion of standard normal variable, thereby ensuring the comparability among different wave bands.
3. The method for early diagnosis of tobacco black shank according to claim 2, wherein the inoculation mode adopts a stem injury inoculation method, namely firstly cutting a shallow wound at a position which is about 1-2 cm a from the ground surface at the stem base of tobacco seedlings by using scissors sterilized by alcohol, and then inserting a pre-prepared Phytophthora nicotianae inoculum into the wound, and covering with fine wet soil to maintain local humidity so as to promote the colonization of the stem tissues by pathogens.
4. The method for early symptomless diagnosis of tobacco black shank according to claim 2, wherein the two acquisitions in S112 are each arranged to be performed between 11:00-12:00 pm of a clear breeze to reduce the effect of illumination variation.
5. The method for early asymptomatic diagnosis of tobacco black shank according to claim 3, wherein the flying height of the unmanned aerial vehicle in S112 is set to 30m, so that the image resolution of the basal region of the leaf and stem can be ensured to meet the analysis requirement, and the high-precision spectrum acquisition of early asymptomatic plants can be realized.
6. The method for early asymptomatic diagnosis of tobacco black shank according to claim 2, wherein the specific steps of S2 are as follows: S21, filtering redundant or irrelevant wave bands based on spectral data after standard normal variable transformation pretreatment, and extracting key characteristic wave bands for early diagnosis of tobacco black shank by adopting a competitive self-adaptive re-weighting sampling algorithm, wherein the specific process comprises the steps of randomly extracting 80% samples, establishing a partial least square regression model, calculating regression coefficients of all wave bands, evaluating the wave band importance by absolute value weight of the regression coefficients, gradually removing wave bands with lower weight by using an exponential decay function to obtain a candidate wave band subset, evaluating the prediction performance of the candidate wave band set by 10-fold cross validation, and selecting the wave band set with optimal prediction precision as a final characteristic wave band after 70 iterations; s22, screening results based on a competitive self-adaptive re-weighted sampling algorithm show that Dataset 1 reserves 60 wave bands, dataset 2 reserves 77 wave bands, 38 wave bands are reserved, then a second round of competitive self-adaptive re-weighted sampling algorithm optimization screening is carried out on the 38 wave bands, and finally an intersection of the two screened wave bands is taken as final characteristic input of early asymptomatic identification of tobacco black shank for subsequent modeling analysis.
7. The method for early asymptomatic diagnosis of tobacco black shank according to claim 6, wherein the specific steps of S3 are as follows: S31, after band screening, determining 437 nm, 530 nm, 649 nm and 710 nm as main sensitive bands of early asymptomatic stages of tobacco black shank; s32, combining Dataset and Dataset 2, and constructing a Logistic regression model based on the four wave bands to obtain an early asymptomatic diagnosis index of tobacco black shank.
8. The early-stage asymptomatic diagnosis equipment for the tobacco black shank is characterized by comprising four narrow-band DT filters, wherein the narrow-band DT filters are used for acquiring reflected light of corresponding wave bands and integrating a data processing chip to execute wave band reflectivity calculation, pretreatment and automatic operation of early-stage asymptomatic diagnosis indexes for the tobacco black shank in real time so as to rapidly output a tobacco black shank infection state result.

Description

Early asymptomatic diagnosis method and equipment for tobacco black shank Technical Field The invention relates to the technical field of tobacco black shank diagnosis, in particular to a method and equipment for diagnosing early-stage asymptomatic tobacco black shank. Background Tobacco black shank (Tobacco Black SHANK DISEASE, TBS) is a very destructive soil-borne disease caused by plant pathogenic oomycetes Phytophthora nicotianae. Is widely distributed in the main tobacco planting area of the whole world, and is one of key factors influencing the yield and quality of tobacco leaves. The disease mainly infects root systems and stem bases of tobacco plants, causes root rot, blackening of lower stems, yellow plant leaves and then withered, and finally causes wilting of the whole plant. The method has the advantages of short disease course, rapid transmission, strong recurrence, and serious burst in high-temperature and high-humidity environment, so that the yield loss of the tobacco field is as high as 30% -80%, and the whole field can be damaged when serious. At present, the prevention and control strategy for tobacco black shank mainly depends on the treatment of agents in soil and the planting of disease-resistant varieties, but the measures are usually adopted after the symptoms are obvious, so that the optimal prevention and control time can be missed. Meanwhile, at present, most of tobacco diseases are monitored through manual field inspection and symptom observation, so that the method is time-consuming and laborious, has strong subjectivity and poor instantaneity, and also has the risk of cross contamination possibly caused by sampling. Therefore, how to realize early, non-invasive and rapid detection before the visible symptoms of the tobacco plant are not yet generated is a key scientific problem to be broken through in the field of tobacco disease monitoring and control. Different from the traditional machine vision method, the hyperspectral imaging technology can capture the reflection change of plants in different spectral bands, and further reflect the subtle change of the plants in the physiological-biochemical level in the early stage of pathogen infection, so that a new window is opened for early diagnosis of diseases. For example, patent CN113435252a proposes a tobacco pest early warning and monitoring system based on hyperspectral and remote sensing data, which realizes efficient early recognition by selecting sensitive wave bands and establishing a model, patent CN112697723a develops a tobacco yield prediction method and system based on unmanned aerial vehicle hyperspectral images, a spectrum-unmanned aerial vehicle monitoring process has reference significance for early asymptomatic recognition of tobacco black shank, and in addition, the tobacco disease recognition and control system proposed by patent CN113962258A provides a technical idea for large-area, lossless and rapid monitoring. However, current hyperspectral devices are expensive, data processing is large and complex, and these factors limit the wide application of hyperspectral technology in large-area field monitoring. Therefore, future research is urgent to combine the hyperspectral remote sensing platform of the unmanned aerial vehicle with a machine learning algorithm so as to realize large-scale, rapid and non-contact monitoring of the tobacco field on the premise of ensuring the diagnosis precision. By extracting the key spectrum wave bands of the early stage of the tobacco black shank and constructing a prediction model, the disease monitoring is possibly changed from 'symptom identification' to 'potential lesion identification', so that a solid technical support is provided for the accurate prevention and control of the tobacco black shank. Disclosure of Invention In order to solve the technical problems, the invention provides the following technical scheme: A method for diagnosing early-stage asymptomatic tobacco black shank comprises the following specific steps: S1, training data acquisition, namely acquiring spectrum data and a standardized sample of an early asymptomatic stage of tobacco black shank, supporting subsequent modeling and analysis, and carrying out training data acquisition work by adopting a mode of combining manual inoculation test with hyperspectral acquisition and data preprocessing so as to obtain two data sets comprising a disease-sensing sample and a health sample; S2, selecting a tobacco black shank characteristic wave band, namely, based on spectral data after standard normal variable transformation pretreatment, adopting a competitive self-adaptive re-weighting sampling algorithm to carry out screening work, firstly constructing a partial least square regression model by 80% of samples, evaluating the wave band importance by the absolute value weight of a regression coefficient, gradually removing a low-weight wave band by combining an exponential decay function, evaluating the performanc