CN103514459A - Method and system for identifying crop diseases and pests based on Android mobile phone platform - Google Patents

Abstract

The invention relates to a method for identifying crop diseases and insect pests based on an Android mobile phone platform, comprising: taking images of diseases and insect pests through a camera and storing them in an SD card of an Android mobile phone; preprocessing the images of disease and insect pests; Extraction; perform feature training on the feature set, use the SVM statistical vector machine method to train the sample set data, and obtain a pest diagnosis model; call the pest diagnosis model to perform SVM statistical vector machine classification, obtain disease picture classification and diagnosis results, and feed back the control methods to Mobile phone users. The invention also discloses a system for identifying crop diseases and insect pests based on an Android mobile phone platform. The present invention performs image preprocessing and feature extraction on the disease images, and uses the statistical vector machine learning method SVM to classify the disease images to establish a disease diagnosis model to achieve the disease image recognition target, and only needs the mobile phone user to align and take pictures, and the recognition efficiency is high. high.

Description

A method and system for identifying crop diseases and insect pests based on an Android mobile phone platform

technical field

The invention relates to the field of client image recognition, in particular to a method and system for identifying crop diseases and insect pests based on an Android mobile phone platform.

Background technique

The traditional way of diagnosis of diseases and insect pests adopts the method of manual observation, which has the disadvantages of subjectivity, limitation and ambiguity. With the development of computer image processing and artificial intelligence technology, people began to use computers instead of humans to diagnose crop diseases and insect pests, and proposed to realize the identification of diseases and insect pests on computers. The development of new technologies in the field of mobile computing has endowed mobile phones with broader application prospects. Mobile phones have become an important platform for personal network and computing services in the future. Android, as one of the most popular smart operating systems, has broken through the concept and mode of traditional mobile phones. Mobile computing The ability is more prominent, and the application is becoming more and more extensive and diversified.

At present, there is a design for identifying pests and diseases on the Android platform: first, the Android client obtains pictures of pests and diseases. In this process, a picture can be taken by the camera that comes with the system platform, or a picture can be obtained through technologies such as Bluetooth; Secondly, the obtained pictures of diseases and insect pests are transmitted to the server through the network, and the server side first trains the pictures of disease and insect pests in the sample library to obtain the characteristic parameters of different types of disease images, that is, to generate a classifier for identifying diseases and insect pests, and the server receives the pictures of disease and insect pests, Process the image on the server-side computer, remove noise, extract image features, and compare the extracted feature parameters with the feature parameters obtained by training the sample library, so as to obtain the analysis and conclusion of the disease picture; finally, the The conclusion of the disease picture is sent and fed back to the Android mobile phone client through the network to inform the user of the result of the disease picture and the prevention and control method. The above-mentioned identification method needs to be connected to the Internet, which is very inconvenient for users; in addition, it needs to be received, identified, and processed by a computer at the server side of the network, and the identification efficiency is low.

Contents of the invention

The primary purpose of the present invention is to provide a method for identifying crop diseases and insect pests based on an Android mobile phone platform, which is convenient to use, easy to operate, and has high identification efficiency.

To achieve the above object, the present invention adopts the following technical solutions: a method for identifying crop diseases and insect pests based on the Android mobile phone platform, the method comprises the steps in the following order:

(1) Mobile phone users take images of pests and diseases through the camera that comes with the Android phone, and store them in the SD card of the Android phone;

(2) Preprocessing the images of pests and diseases;

(3) Extract features from the preprocessed images of pests and diseases;

(4) Perform feature training on the feature set, use the SVM statistical vector machine method to train the sample set data, and obtain the pest diagnosis model;

(5) Call the pest diagnosis model for SVM statistical vector machine classification, obtain the classification and diagnosis results of disease pictures, and feedback the control methods to mobile phone users.

The preprocessing of the images of diseases and insect pests includes grayscale transformation, median filter, threshold segmentation, contour detection and lesion extraction.

Three aspects of feature extraction are carried out on the preprocessed pest image, namely: texture feature, color feature and shape feature, and the color feature, texture feature and shape feature of the disease and pest image are extracted as the recognition feature vector;

For color features, three color features of the first-order moment, second-order moment and third-order moment under the blue channel of the color image are extracted respectively;

For texture features, construct seven texture feature parameters, namely, energy, entropy, contrast and homogeneity in gray-level co-occurrence matrix features, and contrast, angular second-order moment, and entropy in gray-level difference statistical features as identification feature vectors;

For the shape features, the circularity, rectangularity, eccentricity, sphericity ratio, compactness, breadth, and inscribed circle radius parameters are constructed as shape recognition feature vectors.

Carry out feature training on the feature set, use the SVM statistical vector machine method to train the sample set data, and obtain the disease image feature data model. It is mapped to the high-dimensional feature space H, and the function of the original space is used in this space to realize the inner product operation, and the nonlinear problem is converted into a linear problem in another space to obtain the attribution of a sample.

The radial basis kernel function is as follows:

K(x,y)=exp{-|xy| ² /2σ ² }

The kernel function K(x,y) is a monotone function of the Euclidean distance between any point x in space and a certain center y, where y is the center of the kernel function, and σ is the width parameter of the function, which controls the radial effect of the function scope;

After generating the data model file, save the data model file as a .model file and store it in the raw folder of the client program as a pest diagnosis model;

Through the feature vector extraction and comparison of the preprocessed pictures of diseases and insect pests to be identified, the .model disease diagnosis model is called to perform SVM statistical vector machine classification, and the classification and diagnosis results of disease pictures are obtained, and the control methods are fed back to mobile phone users.

The gray-scale transformation means that the collected images of diseases and insect pests are all color images, firstly, the images of diseases and insect pests need to be converted into corresponding gray-scale images, and to convert the color images into gray-scale images, it is necessary to decompose and extract the red (R ), green (G), and blue (B) three image channels, take the R, G, and B color components of the pixel, and use the following formula to calculate the gray value:

Gray (gray value)=R*0.3＋G*0.59＋B*0.11

By performing the above operations on each pixel of a pest image, a gray scale transformed image of the pest image can be obtained.

The smoothing process refers to the use of a nonlinear median filter method to enhance the image, and its basic principle is to perform a neighborhood operation between each pixel in the image and its surrounding pixels; The edge position is determined, and the median filter method is selected to process the image.

The threshold segmentation means that the target of image segmentation is to separate the lesion in the image of diseases and insect pests from the background leaves, so as to obtain an image containing only lesion, to eliminate noise, to obtain more accurate features of lesion, so that the follow-up Perform feature extraction, select a threshold on the gray histogram, and perform segmentation, but the threshold segmentation performance depends on the selection of the threshold;

Using OTSU adaptive threshold segmentation algorithm:

For an image f(x,y), assuming that the image size is M×N, the threshold used to segment the foreground (target) and background of the image is T, and the number of pixels whose gray value is less than the threshold T in the image is recorded as N ₁ , the number of pixels greater than the threshold T is recorded as N ₂ ; if the proportion of foreground pixels in the image is recorded as ω ₁ , the proportion of background in the image is ω ₂ , the average gray level of foreground pixels is μ ₁ , and the average gray level of background pixels is degree is μ ₂ , and the total average gray level of the image is μ, and the variance between classes is denoted as g, then:

${\mathrm{<span\; class="notranslate">\; \&omega;</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; N</span>}}$

${\mathrm{<span\; class="notranslate">\; \&omega;</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; N</span>}}$

N ₁ +N ₂ =M×N

μ=μ ₁ ×ω ₁ +μ ₂ ×ω ₂ 1)

g=ω ₁ ×(μ-μ ₁ ) ² +ω ₂ ×(μ-μ ₂ ) ² 2)

Substituting formula 1) into formula 2), we get:

g=ω ₁ ×ω ₂ ×(μ ₁ -μ ₂ ) ² 3)

In this way, the maximum inter-class variance is obtained, and the gray value corresponding to the maximum variance is the threshold value to be found.

The outline extraction means that the lesion of the diseased leaf contains rich morphological information, and some shape features of the lesion are contained in the outline of the lesion, and the parameters of the shape feature are calculated accordingly, so it is necessary to further extract the outline of the lesion , using the Canny algorithm to detect the lesion outline, the specific method is to use a Gaussian filter to smooth the lesion image, and use the first-order partial derivative finite difference to calculate the gradient magnitude and direction of the lesion image, and on this basis, the gradient magnitude Maximum suppression, and finally a dual-threshold algorithm to detect and connect edges.

The lesion extraction refers to superimposing the contour image and the original image to perform an AND operation to obtain a lesion image with the leaf background removed, and the lesion parts are clearly separated.

The invention also discloses a system for identifying crop diseases and insect pests based on an Android mobile phone platform, including:

The disease image acquisition module starts the camera that the Android mobile phone carries to shoot the disease and insect pest image, and stores it in the SD card of the Android mobile phone;

The image preprocessing module performs grayscale transformation, median filtering, threshold segmentation, contour detection, and disease spot extraction on the image of diseases and insect pests;

The image feature extraction module is used to extract texture features, color features and shape features from the preprocessed images of pests and diseases;

The image pattern recognition module extracts and compares the feature vectors of the preprocessed pictures of diseases and insect pests to be identified, calls the disease and insect pest diagnosis model to perform SVM statistical vector machine classification, obtains the classification and diagnosis results of disease pictures, and feeds back the control methods to mobile phone users.

As can be seen from the above technical scheme, the present invention has studied the characteristics of image processing on the Android system platform, by carrying out image preprocessing and feature extraction on diseased images, and using the statistical vector machine learning method SVM to classify the diseased images and establish a disease diagnosis model to achieve The disease image recognition target only needs the mobile phone user to point and take a picture, without the need for networking, and without the need for a computer on the network server side to receive, identify, and process, and the identification efficiency is high. The interface of the present invention adopts humanized design, is easy to operate, adopts modular programming, and has good expandability.

Description of drawings

Fig. 1 is method flowchart of the present invention;

Fig. 2 is a system structure block diagram of the present invention.

Detailed ways

A method for identifying crop diseases and insect pests based on the Android mobile phone platform, the method includes the steps in the following order: (1) the mobile phone user takes pictures of plant diseases and insect pests through the camera of the Android mobile phone, and stores them in the SD card of the Android mobile phone, which The first step is mainly to jump to the system camera by calling Intent; (2) Preprocess the pest image; (3) Extract features from the preprocessed pest image; (4) Perform feature training on the feature set, using SVM statistical vector (5) Call the pest diagnosis model to perform SVM statistical vector machine classification, obtain disease picture classification and diagnosis results, and feedback the control methods to mobile phone users. As shown in Figure 1.

As shown in Figure 1, the preprocessing of pest images includes grayscale transformation, median filter, threshold segmentation, contour detection, and lesion extraction. The gray-scale transformation means that the collected images of diseases and insect pests are all color images, firstly, the images of diseases and insect pests need to be converted into corresponding gray-scale images, and to convert the color images into gray-scale images, it is necessary to decompose and extract the red (R ), green (G), and blue (B) three image channels, take the R, G, and B color components of the pixel, and use the following formula to calculate the gray value:

Gray (gray value)=R*0.3＋G*0.59＋B*0.11

By performing the above operations on each pixel of a pest image, a gray scale transformed image of the pest image can be obtained.

The smoothing process refers to using a nonlinear median filter method to enhance the image, and its basic principle is to perform a neighborhood operation between each pixel in the image and its surrounding pixels. Spatial domain filtering is divided into linear and nonlinear. Linear filters generally have low-pass characteristics, which will blur the edges of the image, while nonlinear filters can better ensure that the edges of the image are clear. Since the extraction of lesion shape features requires a relatively definite edge position, the median filter method is used to process the image.

The threshold segmentation means that the target of image segmentation is to separate the lesion in the image of diseases and insect pests from the background leaves, so as to obtain an image containing only lesion, which can eliminate noise and obtain more accurate features of lesion, so that the subsequent detection of disease spots for feature extraction. Usually, thresholding is performed on the histogram of the image. Therefore, the threshold is often selected on the gray histogram for segmentation, but the threshold segmentation performance depends on the selection of the threshold. Using OTSU adaptive threshold segmentation algorithm:

For an image f(x,y), assuming that the image size is M×N, the threshold used to segment the foreground (target) and background of the image is T, and the number of pixels whose gray value is less than the threshold T in the image is recorded as N ₁ , the number of pixels greater than the threshold T is recorded as N ₂ ; if the proportion of foreground pixels in the image is recorded as ω ₁ , the proportion of background in the image is ω ₂ , the average gray level of foreground pixels is μ ₁ , and the average gray level of background pixels is degree is μ ₂ , and the total average gray level of the image is μ, and the variance between classes is denoted as g, then:

${\mathrm{<span\; class="notranslate">\; \&omega;</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; N</span>}}$

${\mathrm{<span\; class="notranslate">\; \&omega;</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; N</span>}}$

N ₁ +N ₂ =M×N

μ=μ ₁ ×ω ₁ +μ ₂ ×ω ₂ 1)

g=ω ₁ ×(μ-μ ₁ ) ² +ω ₂ ×(μ-μ ₂ ) ² 2)

Substituting formula 1) into formula 2), we get:

g=ω ₁ ×ω ₂ ×(μ ₁ -μ ₂ ) ² 3)

In this way, the maximum inter-class variance is obtained, and the gray value corresponding to the maximum variance is the threshold value to be found.

The outline extraction means that the lesion of the diseased leaf contains rich morphological information, and some shape features of the lesion are contained in the outline of the lesion, and the parameters of the shape feature are calculated accordingly, so it is necessary to further extract the outline of the lesion . Canny edge algorithm is a commonly used edge detection method. Practice has proved that Canny detection has good results. Therefore, the Canny algorithm is used to detect the lesion contour. The specific method is to use a Gaussian filter to smooth the lesion image, and use the first-order partial derivative Calculate the gradient magnitude and direction of the lesion image, on this basis, carry out non-maximum suppression on the gradient magnitude, and finally use the double threshold algorithm to detect and connect the edges.

The lesion extraction refers to superimposing the contour image and the original image to perform an AND operation to obtain a lesion image with the background of the leaf removed, and the lesion part is clearly separated. So far, the entire preprocessing process is completed, and the obtained image is passed through the preprocessing process. The processing eliminates the noise, which is convenient for the subsequent lesion feature extraction operation and calculation.

As shown in Figure 1, three aspects of feature extraction are performed on the preprocessed pest image, namely: texture feature, color feature and shape feature, and the color feature, texture feature and shape feature of the pest image are extracted as the recognition feature vector ; For color features, three color features of the first-order moment, second-order moment, and third-order moment under the blue channel of the color image are extracted respectively; for texture features, seven texture feature parameters are constructed, that is, the energy in the gray-level co-occurrence matrix feature , entropy, contrast and homogeneity, and the contrast, angular second moment, and entropy in the gray difference statistical features are used as the recognition feature vector; for the shape feature, the circularity, rectangularity, eccentricity, sphericity ratio, compactness, Parameters such as breadth and inscribed circle radius are used as feature vectors for shape recognition.

As shown in Figure 1, the feature set is used for feature training, and the SVM statistical vector machine method is used to train the sample set data to obtain the disease image feature data model. During this training process, the radial basis kernel function is selected to train the sample vector. The basic kernel function maps the sample to the high-dimensional feature space H, and uses the function of the original space in this space to realize the inner product operation, and converts the nonlinear problem into a linear problem in another space to obtain the attribution of a sample.

The radial basis kernel function is as follows:

K(x,y)=exp{-|xy| ² /2σ ² }

The kernel function K(x,y) is a monotone function of the Euclidean distance between any point x in space and a certain center y, where y is the center of the kernel function, and σ is the width parameter of the function, which controls the radial effect of the function scope;

After generating the data model file, save the data model file as a .model file and store it in the raw folder of the client program as a pest diagnosis model;

Through the feature vector extraction and comparison of the preprocessed pictures of diseases and insect pests to be identified, the .model disease diagnosis model is called to perform SVM statistical vector machine classification, and the classification and diagnosis results of disease pictures are obtained, and the control methods are fed back to mobile phone users.

As shown in Figure 2, the system includes: a disease image acquisition module, which starts the Android mobile phone’s own camera to take pictures of diseases and insect pests, and stores them in the SD card of the Android mobile phone; an image preprocessing module, which performs grayscale transformation on the disease and insect pest images , median filtering, threshold segmentation, contour detection, and preprocessing of lesion extraction; the image feature extraction module extracts texture features, color features, and shape features from the preprocessed images of diseases and insect pests; the image pattern recognition module The pre-processed pictures of diseases and insect pests to be identified are extracted and compared through feature vectors, and the diagnosis model of diseases and insect pests is called to perform SVM statistical vector machine classification, and the classification and diagnosis results of disease pictures are obtained, and the control methods are fed back to mobile phone users.

The required environment and construction of the system: the present invention relates to the program development on the Android platform, especially the calling of the underlying JNI. Using the mixed programming of Java and C++ under the Android platform, the image is processed under the Android platform using the OpenCV image processing library and the subsequent related classification work, so the required development environment is different from the normal Android development environment. The common development environment for Andriod is: Eclipse+Android SDK+ADT. The development environment and building steps of this system are roughly as follows: you need to build a development environment, install Eclipse, Android SDK, and ADT, and this step can be completed by downloading the ADT Bundle integrated development kit provided by Google; secondly, you need to download and install NDK, Cygwin, CDT, The reason for this step is that the development of OpenCV needs to write local code (C/C++), and it is necessary to use NDK, Cygwin, etc. to establish a cross-compilation environment and call local code; finally, install OpenCV4Android, establish an image library processing development environment, and introduce OpenCV SDK to the Eclipse workspace.

In summary, the present invention has studied the characteristics of image processing on the Android system platform, by carrying out image preprocessing and feature extraction to disease images, and utilizing the statistical vector machine learning method SVM to classify disease images and establish disease diagnosis models to achieve disease The image recognition target only needs the mobile phone user to point and take a photo, without the need for networking, and without the need for a computer on the network server side to receive, identify, and process, and the identification efficiency is high. The interface of the present invention adopts humanized design, is easy to operate, adopts modular programming, and has good expandability.

Claims (10)

1. a method for identifying crop diseases and insect pests based on the Android mobile phone platform, the method comprises the steps in the following order: (1) Mobile phone users take images of pests and diseases through the camera that comes with the Android phone, and store them in the SD card of the Android phone; (2) Preprocessing the images of pests and diseases; (3) Extract features from the preprocessed images of pests and diseases; (4) Perform feature training on the feature set, use the SVM statistical vector machine method to train the sample set data, and obtain the pest diagnosis model; (5) Call the pest diagnosis model for SVM statistical vector machine classification, obtain the classification and diagnosis results of disease pictures, and feedback the control methods to mobile phone users. 2. the method for identifying crop diseases and insect pests based on the Android mobile phone platform according to claim 1, is characterized in that: carrying out pretreatment to disease and insect pest image comprises the steps of grayscale transformation, median filter, threshold segmentation, contour detection, lesion extraction deal with. 3. the method for identifying crop diseases and insect pests based on the Android mobile phone platform according to claim 1, is characterized in that: the feature extraction of three aspects is carried out to the image of diseases and insect pests through pretreatment, is respectively: texture feature, color feature and shape feature , by extracting the color features, texture features, and shape features of the pest image as the recognition feature vector; For color features, three color features of the first-order moment, second-order moment and third-order moment under the blue channel of the color image are extracted respectively; For texture features, construct seven texture feature parameters, namely, energy, entropy, contrast and homogeneity in gray-level co-occurrence matrix features, and contrast, angular second-order moment, and entropy in gray-level difference statistical features as identification feature vectors; For the shape features, the circularity, rectangularity, eccentricity, sphericity ratio, compactness, breadth, and inscribed circle radius parameters are constructed as shape recognition feature vectors. 4. the method for identifying crop diseases and insect pests based on Android mobile phone platform according to claim 1, is characterized in that: feature set is carried out feature training, uses SVM statistical vector machine method training sample set data, obtains disease image characteristic data model, here During the training process, the radial basis kernel function is selected to train the sample vector. The radial basis kernel function maps the sample to the high-dimensional feature space H, and uses the function of the original space in this space to realize the inner product operation. The nonlinear problem is transformed into a linear problem in another space to obtain the assignment of a sample, The radial basis kernel function is as follows: K(x,y)=exp{-|xy| ² /2σ ² } The kernel function K(x,y) is a monotone function of the Euclidean distance between any point x in space and a certain center y, where y is the center of the kernel function, and σ is the width parameter of the function, which controls the radial effect of the function scope; After generating the data model file, save the data model file as a .model file and store it in the raw folder of the client program as a pest diagnosis model; Through the feature vector extraction and comparison of the preprocessed pictures of diseases and insect pests to be identified, the .model disease diagnosis model is called to perform SVM statistical vector machine classification, and the classification and diagnosis results of disease pictures are obtained, and the control methods are fed back to mobile phone users. 5. the method for identifying crop diseases and insect pests based on the Android mobile phone platform according to claim 2, is characterized in that: said grayscale transformation refers to that the images of diseases and insect pests that are collected are all color images, and at first it is necessary to convert the images of disease and insect pests into corresponding To convert a color image into a grayscale image, it is necessary to decompose and extract the three image channels of red (R), green (G), and blue (B) in the image, and take the R, G, and B color components of the pixel , use the following formula to calculate the gray value: Gray (gray value)=R*0.3＋G*0.59＋B*0.11 By performing the above operations on each pixel of a pest image, a gray scale transformed image of the pest image can be obtained. 6. the method for identifying crop diseases and insect pests based on the Android mobile phone platform according to claim 2, is characterized in that: described smoothing process refers to, uses non-linear median filtering method to carry out enhancement process to image, and its basic principle is exactly to image Neighborhood operations are performed between each pixel in the image and its surrounding pixels; since the extraction of lesion shape features requires the determination of the edge position, the median filter method is used to process the image. 7. the method for identifying crop diseases and insect pests based on the Android mobile phone platform according to claim 2, is characterized in that: the threshold segmentation refers to that the segmentation image target is to separate disease spots and background leaves in the disease and insect pest image, to obtain Images containing only lesions are used to eliminate noise and obtain more accurate lesion features for subsequent feature extraction of lesions, and select thresholds on the gray histogram for segmentation. However, the performance of threshold segmentation depends on the selection of thresholds ; Using OTSU adaptive threshold segmentation algorithm: For an image f(x,y), assuming that the image size is M×N, the threshold used to segment the foreground (target) and background of the image is T, and the number of pixels whose gray value is less than the threshold T in the image is recorded as N ₁ , the number of pixels greater than the threshold T is recorded as N ₂ ; if the proportion of foreground pixels in the image is recorded as ω ₁ , the proportion of background in the image is ω ₂ , the average gray level of foreground pixels is μ ₁ , and the average gray level of background pixels is degree is μ ₂ , and the total average gray level of the image is μ, and the variance between classes is denoted as g, then: ${\mathrm{<span\; class="notranslate">\; \&omega;</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; N</span>}}$ ${\mathrm{<span\; class="notranslate">\; \&omega;</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; N</span>}}$ N ₁ +N ₂ =M×N μ=μ ₁ ×ω ₁ +μ ₂ ×ω ₂ 1) g=ω ₁ ×(μ-μ ₁ ) ² +ω ₂ ×(μ-μ ₂ ) ² 2) Substituting formula 1) into formula 2), we get: g=ω ₁ ×ω ₂ ×(μ ₁ -μ ₂ ) ² 3) In this way, the maximum inter-class variance is obtained, and the gray value corresponding to the maximum variance is the threshold value to be found. 8. the method for identifying crop diseases and insect pests based on the Android mobile phone platform according to claim 2, is characterized in that: the contour extraction refers to that the lesion of the diseased blade contains rich morphological information, and some shape features of the lesion contain In the lesion outline, the parameters of the shape feature are calculated accordingly, so it is necessary to further extract the lesion outline, and use the Canny algorithm to detect the lesion outline. The specific method is to smooth the lesion image with a Gaussian filter, and use the first-order The finite difference of partial derivatives calculates the gradient magnitude and direction of the lesion image, on this basis, the non-maximum suppression of the gradient magnitude is carried out, and finally the double threshold algorithm is used to detect and connect the edges. 9. the method for identifying crop diseases and insect pests based on the Android mobile phone platform according to claim 2, is characterized in that: the lesion extraction refers to, the contour image and the original image are superimposed and carried out AND operation, obtain the diseased spot that has removed the leaf background. Spot image, the lesion site is clearly separated. 10. A system for realizing the method according to any one of claims 1 to 9, characterized in that: comprising: The disease image acquisition module starts the camera that the Android mobile phone carries to shoot the disease and insect pest image, and stores it in the SD card of the Android mobile phone; The image preprocessing module performs grayscale transformation, median filtering, threshold segmentation, contour detection, and disease spot extraction on the images of diseases and insect pests; The image feature extraction module is used to extract texture features, color features and shape features from the preprocessed images of diseases and insect pests; The image pattern recognition module extracts and compares the feature vectors of the preprocessed pictures of diseases and insect pests to be identified, calls the disease and insect pest diagnosis model to perform SVM statistical vector machine classification, obtains the classification and diagnosis results of disease pictures, and feeds back the control methods to mobile phone users.

Images