Tea tree pest degree prediction method based on electronic nose technology

Abstract

The invention discloses a tea tree pest degree prediction method based on an electronic nose technology, and is applied to the technical field of pest detection. Comprising the following steps: pretreating a plurality of tea trees to enable the tea trees to have different types of insect pests; an electronic nose is used for detecting volatile gas of all tea trees within a certain period of time after the tea trees are subjected to insect damage, a response curve is obtained, and tea tree insect damage feature data are extracted; constructing and training a tea tree pest classifier; constructing a pest degree prediction network for each type of pest and training the pest degree prediction network based on the tea tree pest feature data and the corresponding pest degree; and collecting insect pest feature data of a to-be-detected tea tree through an electronic nose, inputting the to-be-detected insect pest feature data into the tea tree insect pest classifier to obtain a classification result, and inputting the classification result into the corresponding insect pest degree prediction network to obtain an insect pest degree prediction result. The method can quickly identify insect pest types and predict insect pest degrees, and provides reference for insect pest control of tea trees.

Description

Technical Field

[0001] The invention relates to the technical field of insect pest detection, and more particularly to a method for predicting the degree of tea tree insect pest based on electronic nose technology. Background Art

[0002] Pests are the main disaster for tea trees. They are characterized by many types and great impact, and they cause the greatest harm to tea trees. Therefore, pest control requires predicting the time and degree of pest damage in order to take measures to avoid greater losses caused by pests. If pests and diseases can be discovered early and corresponding measures can be taken quickly to control them, the yield loss can be reduced. However, most of the existing tea tree pest detection schemes are manually judged by professionals. This method requires a high level of professionalism for the judges and has low judgment efficiency, which cannot meet current needs; another part of the scheme uses image recognition technology to identify leaves to determine the type and degree of pests of tea trees. Although this scheme solves the problems of manual judgment, the image recognition scheme has high requirements for image quality, and often cannot accurately identify the type of pests, nor can it predict the degree of pests. The types and degrees of pest infection are different, and the gases volatilized by plants are also different. The electronic nose is an electronic system that uses the response signal of a gas sensor array to identify odors. It can continuously and in real time monitor the odor conditions at a specific location over a period of hours, days or even months. Therefore, how to propose a method for predicting the degree of tea tree pests based on electronic nose technology is an urgent problem that technicians in this field need to solve. Summary of the invention

[0003] In view of this, the present invention provides a method for predicting the degree of tea tree pests based on electronic nose technology, establishes a tea tree pest classifier and a pest degree prediction network corresponding to different types, uses the data collected by the electronic nose as the basis for judging the pest type and predicting the degree, and realizes rapid detection of pest types and prediction of the degree.

[0004] In order to achieve the above object, the present invention provides the following technical solutions:

[0005] A method for predicting the degree of tea tree pests based on electronic nose technology comprises the following steps:

[0006] S1, pre-treating multiple groups of tea trees so that the tea trees have different types of insect pests;

[0007] S2, using an electronic nose to detect the volatile gases of all tea trees within a certain period of time after being infested by insects, and obtaining the response curves of the electronic noses of tea trees with different insect pest types under different degrees of insect pests;

[0008] S3, processing the response curve of the electronic nose to extract characteristic data of tea tree pests;

[0009] S4, constructing and training a tea tree pest classifier based on tea tree pest feature data and its corresponding pest types;

[0010] S5. For each type of insect pest, construct and train an insect pest severity prediction network based on tea tree insect pest characteristic data and its corresponding insect pest severity;

[0011] S6. The volatile gas of the tea tree to be tested is collected by an electronic nose and processed to obtain characteristic data of the pest to be tested, and the characteristic data of the pest to be tested is input into a tea tree pest classifier to obtain a classification result, and based on the classification result, the characteristic data of the pest to be tested is input into a corresponding pest degree prediction network to obtain a pest degree prediction result.

[0012] Optionally, S1 specifically includes: selecting multiple groups of tea trees in a healthy state, and selecting different types of pests for different groups of tea trees for pest treatment, so that one group of tea trees corresponds to one type of pest, while retaining a group of tea trees in a healthy state.

[0013] Optionally, S2 is specifically as follows: after the tea trees are infested with insect pests, electronic nose detection is performed on the tea trees at all preset time nodes of insect pest degree changes, the sensor is preheated before each detection, the instrument is started and allowed to stand for 1 hour to reach working conditions, a fresh-keeping bag is used to cover the tea tree sample to be tested, air is pumped into it to the maximum volume with an air pump, it is sealed with a clip, and allowed to stand for 30 minutes, and the electronic nose detection value is saved during the detection process.

[0014] Optionally, S3 specifically includes: fitting the response curve of the electronic nose through a binomial function, extracting eigenvalues, and decentralizing the eigenvalue set:

[0015]

[0016] Where, X c is the decentralized eigenvalue set, X is the eigenvalue set, is the feature mean vector, m is the number of samples, i∈[1,m], x i Represents the eigenvalue vector of the i-th sample;

[0017] Calculate the covariance matrix C:

[0018]

[0019] Calculate the eigenvalues ​​and eigenvectors of the covariance matrix, sort the eigenvectors in descending order according to the corresponding eigenvalues, and select the first k eigenvectors as the projection matrix W; calculate the reduced eigenmatrix X based on the projection matrix W R , get the tea tree pest characteristic data:

[0020] XR =X C W

[0021] The feature matrix after dimension reduction is composed of tea tree pest feature data.

[0022] Furthermore, S4 is specifically: using a support vector machine as a tea tree pest classifier, taking tea tree pest feature data as input samples, and taking corresponding pest types as output samples to train the tea tree pest classifier, and obtaining a tea tree pest classifier after the training is completed.

[0023] Furthermore, the tea tree pest classifier is specifically:

[0024]

[0025] In the formula, x is the input sample, y(x) is the classification result, sign represents the classification function, k∈[1,m], m is the number of samples, α k is the Lagrange coefficient, K(x,x k ) is the kernel function and b is a constant.

[0026] Optionally, S4 specifically includes: using a support vector machine as a tea tree pest classifier, taking tea tree pest feature data as input samples, and taking corresponding pest types as output samples to train the tea tree pest classifier, and obtaining a tea tree pest classifier after the training is completed.

[0027] Optionally, the tea tree pest classifier is:

[0028]

[0029] In the formula, x is the input sample, y(x) is the classification result, sign represents the classification function, k∈[1,m], m is the number of samples, α k is the Lagrange coefficient, K(x,x k ) is the kernel function and b is a constant.

[0030] Optionally, S5 specifically comprises: for each type of insect pest, constructing an insect pest degree prediction network and calculating the insect pest degree index of the tea tree at all detection time nodes, taking the insect pest degree index of a time node as the output sample, and the tea tree insect pest characteristic data of the previous time node as the input sample to train the insect pest degree prediction network.

[0031] It can be seen from the above technical solution that compared with the prior art, the present invention provides a method for predicting the degree of tea tree pests based on electronic nose technology, which has the following beneficial effects: the present invention establishes a tea tree pest classifier and a pest degree prediction network corresponding to different types, and uses the data collected by the electronic nose as the basis for judging the type of pests and predicting the degree of pests. The pest degree of different types of pests is predicted respectively, and the pest types can be quickly identified and the pest degree is predicted, providing a reference for the pest control of tea trees. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0033] Figure 1 The present invention is a flow chart of the method for predicting the degree of tea tree pests. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] The embodiment of the present invention discloses a method for predicting the degree of tea tree pests based on electronic nose technology. Figure 1 As shown, the following steps are included:

[0036] S1, pre-treating multiple groups of tea trees so that the tea trees have different types of insect pests;

[0037] S2, using an electronic nose to detect the volatile gases of all tea trees within a certain period of time after being infested by insects, and obtaining the response curves of the electronic noses of tea trees with different insect pest types under different degrees of insect pests;

[0038] S3, processing the response curve of the electronic nose to extract characteristic data of tea tree pests;

[0039] S4, constructing and training a tea tree pest classifier based on tea tree pest feature data and its corresponding pest types;

[0040] S5. For each type of insect pest, construct and train an insect pest severity prediction network based on tea tree insect pest characteristic data and its corresponding insect pest severity;

[0041] S6. The volatile gas of the tea tree to be tested is collected by an electronic nose and processed to obtain characteristic data of the pest to be tested, and the characteristic data of the pest to be tested is input into a tea tree pest classifier to obtain a classification result, and based on the classification result, the characteristic data of the pest to be tested is input into a corresponding pest degree prediction network to obtain a pest degree prediction result.

[0042] Furthermore, S1 specifically includes: selecting multiple groups of tea trees in a healthy state, and selecting different types of pests for different groups of tea trees for pest treatment, so that one group of tea trees corresponds to one type of pest, while retaining a group of tea trees in a healthy state.

[0043] In the embodiment of the present invention, three kinds of pests, namely, Okonomiyaki leafhopper, Green Lygus and Tea leafhopper, are used for treatment, and these three kinds of pests are used in subsequent identification and degree prediction.

[0044] In the embodiment of the present invention, when the number of Xiaoguansongcun leafhoppers is less than 10 heads / 100 leaves, there is no obvious harm and it is classified as Level 1 harm; 10-30 heads / 100 leaves are classified as Level 2 harm; 30-60 heads / 100 leaves are classified as Level 3 harm; more than 60 heads / 100 leaves have serious harm such as leaf blight or even necrosis and it is classified as Level 4 harm; when the number of green stink bugs is less than 1.0 head / 100 shoots, the harm is relatively light and it is classified as Level 1 harm; 1.0-3.5 heads / 100 shoots are classified as Level 2 harm; 3.5-5.5 heads / 100 shoots are classified as Level 3 harm; more than 5.5 heads / 100 shoots have serious harm such as large holes and defects in the leaves and it is classified as Level 4 harm; the new tea leafhopper will form holes on the tender shoots in the early stage of harm, and serious harm will cause the twisted parts of the leaves to turn yellow-brown. The harm level refers to Xiaoguansongcun leafhoppers; in S2, gas detection of electronic nose is performed on different types of insect pests at different levels of harm.

[0045] Furthermore, S2 is specifically as follows: after the tea trees are infested with insect pests, electronic nose detection is performed on the tea trees at all preset time nodes of insect pest degree changes, the sensor is preheated before each detection, the instrument is started and left to stand for 1 hour to reach working conditions, a fresh-keeping bag is used to cover the tea tree sample to be tested, air is pumped into it to the maximum volume with an air pump, it is sealed with a clip, and left to stand for 30 minutes, and the electronic nose detection value is saved during the detection process.

[0046] In the embodiment of the present invention, specifically, the headspace sampling method is used to perform gas detection using the PEN3 portable bionic electronic nose system. The sensor needs to be preheated before each use. After starting the instrument, it is left to stand for 1 hour to reach the working condition. The tea branches are all 20-30 cm high, the headspace time is 30 minutes, and a fresh-keeping bag with a volume of 5000 mL is selected; before using the electronic nose for detection, the tea tree sample to be tested is covered with a fresh-keeping bag, and air is pumped into it to the maximum volume with an air pump, and it is sealed with a clip and left to stand for 30 minutes to stabilize the air in the fresh-keeping bag. During the detection process, the sensor cleaning time is set to 90 seconds, the automatic zeroing time is 10 seconds, the sample preparation time is 5 seconds, the sample measurement interval time is 1 second, the internal flow rate is 600 mL / min, the detection time is 120 seconds, and the injection flow rate is 600 mL / min. The WinMuster software that comes with the electronic nose automatically samples, records and saves the electronic nose detection value per second. Before and after each measurement, the sensor is cleaned and standardized.

[0047] Furthermore, S3 is specifically as follows: fitting the response curve of the electronic nose through a binomial function, extracting eigenvalues, and decentralizing the eigenvalue set:

[0048]

[0049] Where, X c is the decentralized eigenvalue set, X is the eigenvalue set, is the feature mean vector, m is the number of samples, i∈[1,m], x i Represents the eigenvalue vector of the i-th sample;

[0050] Calculate the covariance matrix C:

[0051]

[0052] Calculate the eigenvalues ​​and eigenvectors of the covariance matrix, sort the eigenvectors in descending order according to the corresponding eigenvalues, and select the first k eigenvectors as the projection matrix W; calculate the reduced eigenmatrix X based on the projection matrix W R , get the tea tree pest characteristic data:

[0053] X R =X C W

[0054] The feature matrix after dimension reduction is composed of tea tree pest feature data.

[0055] Furthermore, S4 is specifically: using a support vector machine as a tea tree pest classifier, taking tea tree pest feature data as input samples, and taking corresponding pest types as output samples to train the tea tree pest classifier, and obtaining a tea tree pest classifier after the training is completed.

[0056] Furthermore, the tea tree pest classifier is specifically:

[0057]

[0058] In the formula, x is the input sample, y(x) is the classification result, sign represents the classification function, k∈[1,m], m is the number of samples, α k is the Lagrange coefficient, K(x,x k ) is the kernel function and b is a constant.

[0059] In the embodiment of the present invention, the kernel function may be one of a radial basis kernel function, a polynomial kernel function, and a linear kernel function. In this embodiment, a radial basis kernel function is used:

[0060]

[0061] Where σ is the bandwidth of the kernel function.

[0062] Furthermore, S5 is specifically as follows: for each type of insect pest, a pest severity prediction network is constructed and the pest severity index of the tea tree at all detection time nodes is calculated, the pest severity index of a time node is used as an output sample, and the tea tree pest feature data of the previous time node is used as an input sample to train the pest severity prediction network.

[0063] In an embodiment of the present invention, the pest degree prediction network adopts a BP neural network, and its training process is: initialize sample weights and biases, input a sample set and an expected output set, output prediction results, calculate the error value between the prediction results and the expected output, adjust the sample weights and biases, re-output the prediction results, repeatedly calculate the error value and adjust the sample weights and biases until the error value is less than a preset threshold.

[0064] In an embodiment of the present invention, the pest degree index can be calculated based on the degree of pest damage on the surface of the tea leaves, and the degree of pest damage is compared and calculated with the leaves of the tea tree in a healthy state.

[0065] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0066] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting the degree of tea tree pests based on electronic nose technology, characterized in that: The following steps are involved: S1, pre-treating multiple groups of tea trees so that the tea trees have different types of insect pests; S2, using an electronic nose to detect the volatile gases of all tea trees within a certain period of time after being infested by insects, and obtaining the response curves of the electronic noses of tea trees with different insect pest types under different degrees of insect pests; S3, processing the response curve of the electronic nose to extract characteristic data of tea tree pests; S4, constructing and training a tea tree pest classifier based on tea tree pest feature data and its corresponding pest types; S5. For each type of insect pest, construct and train an insect pest severity prediction network based on tea tree insect pest characteristic data and its corresponding insect pest severity; S6. The volatile gas of the tea tree to be tested is collected by an electronic nose and processed to obtain characteristic data of the pest to be tested, and the characteristic data of the pest to be tested is input into a tea tree pest classifier to obtain a classification result, and based on the classification result, the characteristic data of the pest to be tested is input into a corresponding pest degree prediction network to obtain a pest degree prediction result.

2. The method for predicting the degree of tea tree pests based on electronic nose technology according to claim 1, characterized in that: S1 specifically includes: selecting multiple groups of tea trees in a healthy state, selecting different types of pests for different groups of tea trees for pest treatment, so that one group of tea trees corresponds to one type of pest, and retaining a group of tea trees in a healthy state.

3. The method for predicting the degree of tea tree pests based on electronic nose technology according to claim 1, characterized in that: S2 is specifically as follows: after the tea trees are infested with insect pests, electronic nose detection is performed on the tea trees at all preset time nodes of insect pest degree changes, the sensor is preheated before each detection, the instrument is started and left to stand for 1 hour to reach working conditions, the tea tree sample to be tested is covered with a fresh-keeping bag, air is pumped into it to the maximum volume with an air pump, it is sealed with a clip, and left to stand for 30 minutes, and the electronic nose detection value is saved during the detection process.

4. The method for predicting the degree of tea tree pests based on electronic nose technology according to claim 1, characterized in that: S3 is specifically as follows: fitting the response curve of the electronic nose through a binomial function, extracting eigenvalues, and decentralizing the eigenvalue set: Where, X c is the decentralized eigenvalue set, X is the eigenvalue set, is the feature mean vector, m is the number of samples, i∈[1,m], x i Represents the eigenvalue vector of the i-th sample; Calculate the covariance matrix C: Calculate the eigenvalues ​​and eigenvectors of the covariance matrix, sort the eigenvectors in descending order according to the corresponding eigenvalues, and select the first k eigenvectors as the projection matrix W; calculate the reduced eigenmatrix X based on the projection matrix W R , get the tea tree pest characteristic data: X R =X C W The feature matrix after dimension reduction is composed of tea tree pest feature data.

5. The method for predicting the degree of tea tree pests based on electronic nose technology according to claim 1, characterized in that: S4 is specifically as follows: using support vector machine as tea tree pest classifier, taking tea tree pest feature data as input samples, taking corresponding pest types as output samples to train the tea tree pest classifier, and obtaining the tea tree pest classifier after the training is completed.

6. The method for predicting the degree of tea tree pests based on electronic nose technology according to claim 5, characterized in that: The tea tree pest classifier is as follows: In the formula, x is the input sample, y(x) is the classification result, sign represents the classification function, k∈[1,m], m is the number of samples, α k is the Lagrange coefficient, K(x, x k ) is the kernel function and b is a constant.

7. The method for predicting the degree of tea tree pests based on electronic nose technology according to claim 1, characterized in that: S5 is specifically as follows: for each type of insect pest, a pest severity prediction network is constructed and the pest severity index of the tea tree at all detection time nodes is calculated, the pest severity index of a time node is used as an output sample, and the tea tree pest feature data of the previous time node is used as an input sample to train the pest severity prediction network.

Images