CN109308697B - Leaf disease identification method based on machine learning algorithm - Google Patents
Leaf disease identification method based on machine learning algorithm Download PDFInfo
- Publication number
- CN109308697B CN109308697B CN201811087765.8A CN201811087765A CN109308697B CN 109308697 B CN109308697 B CN 109308697B CN 201811087765 A CN201811087765 A CN 201811087765A CN 109308697 B CN109308697 B CN 109308697B
- Authority
- CN
- China
- Prior art keywords
- image
- gray
- sample
- value
- pixel
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 201000010099 disease Diseases 0.000 title claims abstract description 51
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 title claims abstract description 51
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 28
- 238000010801 machine learning Methods 0.000 title claims abstract description 13
- 238000007781 pre-processing Methods 0.000 claims abstract description 9
- 239000011159 matrix material Substances 0.000 claims abstract description 8
- 238000012795 verification Methods 0.000 claims abstract description 5
- 238000013145 classification model Methods 0.000 claims description 14
- 230000006870 function Effects 0.000 claims description 11
- 230000011218 segmentation Effects 0.000 claims description 7
- 238000012360 testing method Methods 0.000 claims description 6
- 230000001186 cumulative effect Effects 0.000 claims description 5
- 238000001914 filtration Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 230000001131 transforming effect Effects 0.000 claims description 2
- 238000012706 support-vector machine Methods 0.000 abstract description 14
- 241000607479 Yersinia pestis Species 0.000 abstract description 11
- 241000238631 Hexapoda Species 0.000 abstract description 8
- 238000012545 processing Methods 0.000 abstract description 7
- 238000012549 training Methods 0.000 abstract description 6
- 235000013399 edible fruits Nutrition 0.000 description 14
- 230000000694 effects Effects 0.000 description 13
- 241000196324 Embryophyta Species 0.000 description 9
- 235000011430 Malus pumila Nutrition 0.000 description 8
- 235000015103 Malus silvestris Nutrition 0.000 description 8
- 244000141359 Malus pumila Species 0.000 description 7
- 238000000605 extraction Methods 0.000 description 6
- 230000003044 adaptive effect Effects 0.000 description 5
- 238000003709 image segmentation Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 239000000575 pesticide Substances 0.000 description 4
- 230000003595 spectral effect Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000003745 diagnosis Methods 0.000 description 2
- 238000006073 displacement reaction Methods 0.000 description 2
- 208000015181 infectious disease Diseases 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000002265 prevention Effects 0.000 description 2
- 239000007921 spray Substances 0.000 description 2
- 241000269627 Amphiuma means Species 0.000 description 1
- 240000008067 Cucumis sativus Species 0.000 description 1
- 235000010799 Cucumis sativus var sativus Nutrition 0.000 description 1
- 241000221785 Erysiphales Species 0.000 description 1
- 241000220225 Malus Species 0.000 description 1
- 241000233679 Peronosporaceae Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 235000013305 food Nutrition 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 239000002420 orchard Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30188—Vegetation; Agriculture
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a leaf disease identification method based on a machine learning algorithm, and belongs to the technical field of image processing. Firstly, carrying out preprocessing operations such as graying, image enhancement, denoising and the like on an acquired leaf sample image; then dividing the preprocessed image by a self-adaptive threshold algorithm to effectively represent texture information of the sample image; selecting RGB color space to extract color features of the sample image, and extracting texture features of the segmented image according to the gray level co-occurrence matrix; and finally, selecting a support vector machine model, classifying and identifying the sample image by using a cross verification algorithm, optimizing main parameters of the SVM model by using a grid optimizing method, and then selecting the parameters with optimal identification accuracy to establish an SVM classification and identification model. The invention can enable the computer to automatically identify the plant diseases and insect pests of the leaves through training, greatly reduces the space and time expenditure, improves the identification precision, and has the characteristics of rapidness, accuracy and strong robustness.
Description
Technical Field
The invention relates to the technical field of computer information image processing, in particular to a leaf disease identification method based on a machine learning algorithm.
Background
Along with the continuous high-speed development of the economy and the rapid population growth of China, the agriculture of China is correspondingly developed, fruits become necessary food in common people, and the planting area of the fruits is continuously enlarged due to the large demand of the fruits. Along with the increase of the production area of the fruit, the disease problem of the fruit gradually draws attention and importance of experts, and the disease type of the fruit tree can be accurately and efficiently identified, so that the fruit tree can be timely treated, unnecessary loss is reduced, and the yield of the fruit garden is increased.
Currently, image processing technology is rapidly developed, and applications in life and production are also gradually popularized. Therefore, the application of the image recognition technology in the agricultural field also greatly improves the treatment and prevention of crop diseases. The method is used for quickly distinguishing the disease images of the tree leaves, mainly relates to research and realization of aspects such as image acquisition, image preprocessing, image segmentation, image feature extraction, image classification recognition and the like, and obtains good results.
By way of retrieval, mohammede1-Helly et al (2004) developed a comprehensive image processing system that automatically detected leaf spots to determine disease type. The system uses an artificial neural network as a classifier to identify cucumber powdery mildew, downy mildew and leaf pests. However, the method has the disadvantages of large calculated amount, high time consumption and poor algorithm performance.
In 2015, chinese patent application No.: CN104422660a, publication date is 3 month 18 of 2015, discloses a hyperspectral plant leaf disease and pest diagnosis system, which comprises hyperspectral imager, lens, calibration white board, sample placement platform, linear light source support wall, electric control displacement platform, mobile platform control device, camera bellows, adjustable bracket foot and computer. The hyperspectral imager and the linear light source are fixedly connected to the electric control displacement table and controlled by the mobile platform control device to move together; the linear light source irradiates the calibration white board and the plant leaves to be tested in the moving process, and the hyperspectral imager collects data and uploads the data to the computer in real time; after being processed by image analysis and processing software, the three-layer diagnosis model of plant diseases and insect pests is input, and the infection degree of the plant diseases and insect pests is obtained quantitatively in real time, so that whether to spray the pesticide to the measured plant and the dosage and the spray head form of the pesticide are analyzed and decided. The application is mainly used for rapidly diagnosing the infection degree of plant diseases and insect pests, and plays an active role in improving the decision level of accurate pesticide application and realizing fine agriculture. However, the disadvantage of this application is that it is not possible to distinguish between different diseases and insect pests.
In 2015, chinese patent application No. CN201510321137.1, the application date is 11 days of 6 months in 2015, and the invention is named: the application obtains spectral image data based on a spectral and image information processing technology, and then performs leaf area and background separation, pest and disease spot extraction, spectral feature and image feature calculation. And finally, outputting the plant diseases and insect pests of the leaves by using a Fisher linear discriminant analysis model. The application can effectively identify different types of plant diseases and insect pests, and provides a basis for scientific decision making for pesticide application management and plant disease and insect pest green prevention and control. However, the spectral images used in the application require specialized personnel to obtain and are expensive in equipment, so that the popularization and the application are not facilitated.
Disclosure of Invention
1. Technical problem to be solved by the invention
The traditional method for treating the tree leaf diseases mainly depends on that fruit growers or agricultural specialists check the growth conditions of the fruit trees in the orchard in person, and the method judges the growth conditions and the diseases of the fruit trees through human eyes, so that certain errors exist in the identification result, and meanwhile, the time and the labor are consumed. In order to overcome the problems, the invention provides a leaf disease identification method based on a machine learning algorithm; the invention can automatically identify different types of fruit tree leaf diseases through the computer, greatly reduces space and time expenditure, and improves identification precision; aiming at uneven quality of the collected sample images and insufficient salient image characteristics, the invention carries out pretreatment on the sample images so as to lead the samples to be convenient for follow-up work; aiming at the situation that the texture difference represented by the sample image is not obvious and the extraction of texture features is inconvenient, the invention adopts an adaptive threshold algorithm to carry out image segmentation on the preprocessed image, thereby well highlighting the texture features of the sample image; aiming at the fact that a Support Vector Machine (SVM) model cannot meet the identification requirement under default parameters, the method and the device for preventing the fruit diseases utilize a grid optimizing method to obtain optimal parameters of the SVM model by combining a cross verification algorithm, so that an optimal effect is achieved, and support is provided for preventing the fruit diseases.
2. Technical proposal
In order to achieve the above purpose, the technical scheme provided by the invention is as follows:
the invention discloses a leaf disease identification method based on a machine learning algorithm, which comprises the following steps:
1) Collecting leaf sample images, collecting one hundred healthy images and one disease image respectively, and preprocessing the images;
2) Dividing the preprocessed sample image by adopting a self-adaptive threshold segmentation method, and highlighting texture features of the sample image;
3) Extracting color features of the sample image and texture features of the segmented image based on the RGB color space;
4) Extracting the characteristics of all sample images, setting the characteristics as a sample characteristic set m, and constructing an SVM classification model by adopting a cross verification method;
5) Searching optimal parameters by using a grid optimizing method, wherein the parameters are penalty factors C and kernel parameters g;
6) And 5) obtaining an optimal SVM classification model by utilizing the optimal parameters searched in the step 5), and inputting the test sample image into the trained SVM classification model to obtain a recognition result.
Further, the image preprocessing process in step 1) is as follows: firstly, carrying out size adjustment and graying treatment on the size of a sample image, wherein the size is 100 x 100, then carrying out image enhancement operation on the gray image by using a histogram equalization algorithm, and finally carrying out denoising treatment on the image by using a median filtering algorithm.
Further, the image enhancement operation process is as follows: firstly, normalizing pixel gray values of an original image and counting distribution probability of each gray level, then utilizing a cumulative transformation function to transform the normalized gray values and converting the normalized gray values into standard gray values according to the same gray level, thereby obtaining a new image after equalization processing.
Further, the process of segmenting the sample image in step 2) is as follows:
setting the gray value of an image at a pixel point (a, B) as k (a, B), taking the pixel point (a, B) as a B X B window with the center, wherein B represents the side length of the window, C represents the difference value, and calculating the threshold value v (a, B) of each pixel (a, B) in the image, wherein the calculation formula is as follows:
binarizing each pixel point (a, b) in the image point by using v (a, b) values:
further, the color features described in step 3) include six features of mean, variance, skewness, kurtosis, entropy and gradient of the three RGB channels, and the texture features include angular second moment ASM, entropy ENT, contrast CON and inverse differential IDM.
Further, the color characteristics described in step 3) are expressed as follows:
A. average value:
B. variance:
C. degree of deviation:
D. kurtosis:
E. entropy:
F. gradient:
in the above formula, p (k) represents the distribution probability of the pixel gray value k in the image, n k For its frequency, n is the total number of pixels, L is the number of gray levels,is the mean value, delta 2 Is variance, k k Is of a skewness, k F Kurtosis, k E For entropy, mxn represents the size of the image,represents the gradient of a pixel with gray value k in the horizontal direction,/or->Represents the gradient of a pixel having a gray value k in the vertical direction, k g Is its average gradient.
Further, the texture features described in step 3) are expressed as follows:
A. angular Second Moment (ASM):
B. entropy (ENT):
C. contrast (CON):
D. inverse Differential Matrix (IDM):
wherein G (i, j) is an element in the gray level co-occurrence matrix, and L is a gray level number.
Still further, the kernel function of the SVM classification model in step 4) is a radial basis kernel function:
wherein the kernel parameter g represents the width of the kernel function.
Still further, the default values of penalty factor C and kernel parameter g in step 5) are 500 and 0.5, respectively.
Still further, the optimum values of penalty factor C and kernel parameter g found in step 5) are 7000 and 0.006.
3. Advantageous effects
Compared with the prior art, the technical scheme provided by the invention has the following remarkable effects:
(1) According to the leaf disease identification method based on the machine learning algorithm, in the image preprocessing process, firstly, the image with the adjusted size is subjected to grey-scale treatment and image enhancement treatment, and then, the image is subjected to denoising treatment by using a median filtering algorithm. After the processing, the image quality can be improved, the image information can be enriched, the image impression can be enhanced, and the image recognition effect can be improved.
(2) According to the leaf disease identification method based on the machine learning algorithm, the self-adaptive threshold segmentation algorithm is used in image segmentation, so that the healthy image and the disease image in the sample image can be effectively distinguished in texture expression, and favorable conditions are provided for subsequent texture feature extraction of the sample image through image segmentation.
(3) According to the leaf disease identification method based on the machine learning algorithm, color features of sample images are extracted based on RGB color space in feature extraction, and a gray level co-occurrence matrix is generated by using the segmented sample images, so that texture features of the sample images are extracted, the sample images are converted into feature values representing the images through feature extraction, and a necessary condition is provided for training of a later sample image classification model.
(4) According to the leaf disease identification method based on the machine learning algorithm, training and testing are carried out on the SVM model through the cross verification algorithm in image identification, the grid optimizing method is used for obtaining the optimal parameters of the SVM model, and the optimal identification effect can be achieved after training.
Drawings
FIG. 1 is a flow chart of a leaf disease identification method based on a machine learning algorithm of the present invention;
FIG. 2 is a gray scale image (disease image) of apple tree leaves according to example 1 of the present invention;
FIG. 3 is an enhancement effect diagram of the image of FIG. 2;
FIG. 4 is a denoised image of the median filter of FIG. 3;
FIG. 5 is an effect graph of the adaptive thresholding of FIG. 4;
FIG. 6 is a graph showing the effect of the preprocessing and adaptive thresholding of FIG. 2;
FIG. 7 is a graph showing the effect of the classification model in judging illness;
fig. 8 is an effect diagram of the classification model judging healthy.
Detailed Description
For a further understanding of the present invention, the present invention will be described in detail with reference to the drawings and examples.
Example 1
Referring to fig. 1, in the leaf disease identification method based on the machine learning algorithm of the present embodiment, taking disease identification of apple leaves as an example, the steps of identifying disease images are as follows:
1) Image preprocessing: the method comprises the steps of collecting apple tree leaf sample images, wherein the sample images comprise healthy images and disease images, disease areas and certain healthy areas are contained in the disease images, and the disease images can be considered as the disease images only by the sample images containing the disease areas, and the healthy images and the disease images are collected for one hundred in total and preprocessed.
The image preprocessing process comprises the following steps: firstly, carrying out size adjustment and graying treatment on the size of a sample image to obtain a gray scale image shown in fig. 2, wherein the adjustment size is 100 x 100, and then carrying out image enhancement operation on the gray scale image by using a histogram equalization algorithm, wherein the image enhancement operation process comprises the following steps:
firstly, normalizing the pixel gray value k of an original image to obtain r k And calculates the distribution probability p (r) of each gray level k ) The normalized gray value r is then scaled by the cumulative transform function T (·) k Transforming to obtain s k And the image is converted into standard gray values according to the same gray level, so that a new image after equalization treatment is obtained, the effect obtained by image enhancement is shown in figure 3, and the contrast of the image can be improved and the details of the image are enriched after the treatment. The cumulative transform function is as follows:
wherein n is the total number of pixels, n j The number of pixels for different gray levels, L, is the number of gray levels, j=0, 1.
And finally, denoising by adopting a median filtering algorithm to obtain a denoised image shown in fig. 4, and filtering noise and simultaneously keeping details and contours of the image to the greatest extent after denoising.
2) Image segmentation:
setting the gray value of an image at a pixel point (a, B) as k (a, B) in a segmentation algorithm, and taking the pixel point (a, B) as a B X B window with the center, wherein B represents the side length of the window, C represents the difference value, and calculating the threshold value v (a, B) of each pixel (a, B) in the image, wherein the calculation formula is as follows:
binarizing each pixel point (a, b) in the image point by using v (a, b) values:
performing adaptive threshold segmentation on the denoised image to obtain an effect diagram shown in fig. 5, and performing step 1) and step 2) on the healthy image to obtain the effect diagram shown in fig. 6. As can be seen from a comparison of fig. 5 and fig. 6, the texture distribution of the healthy segmented image is uniform, the texture primitives are small, and the texture distribution of the diseased segmented image is disordered and the texture primitives are large, which indicates that the texture features of the sample image after step 1) and step 2) are effectively characterized.
3) Extracting image features: and (3) segmenting the apple tree leaf image preprocessed in the step (1) by using the field adaptive threshold segmentation algorithm in the step (2), and highlighting the texture characteristics of the healthy image and the disease image.
Firstly, based on RGB color space, selecting color moment in color characteristics of apple tree leaf disease images, respectively extracting six characteristic values of mean value, variance, skewness, kurtosis, entropy and gradient of three channels of R, G and B according to gray values after normalization of the apple tree leaf disease images, wherein each characteristic value represents a characteristic of a certain aspect of a picture in digital image processing. And then, according to the distribution rule of the image brightness and the spatial position characteristics of pixels with similar brightness, counting the frequency of the pixel combination occurrence of which the gray values are i-1 and j-1 respectively according to a certain direction and a certain distance in the sample image after the segmentation in the step 3), wherein i=1, 2.
The six color features are described as follows:
A. mean (Mean): reflecting the brightness of the image, the larger the average value is, the larger the brightness of the image is, and the smaller the average value is, the smaller the brightness of the image is.
B. Variance (Variance): is a scale for measuring the uniformity of sample distribution.
C. Skewness (Shewness): described is the symmetry of the overall sample value distribution, i.e. the torsion degree of the image.
D. Kurtosis (Kurtosis): described is the degree to which the peak of the sample is steep or gentle.
E. Entropy (Entropy): which characterizes the amount of information contained in the aggregated features of the gray distribution in the image.
F. Gradient (Gradient): the case of how fast the image gradation value changes is described.
Wherein p (k) represents the probability of distribution of pixel gray values k, n in the image k For its frequency, n is the total number of pixels,is the average value, L is the gray level number, delta 2 Is variance, k k Is of a skewness, k F Kurtosis, k E For entropy, MXN represents the size of the image, +.>Represents the gradient of a pixel with gray value k in the horizontal direction,/or->Represents the gradient of a pixel having a gray value k in the vertical direction, k g Is its average gradient.
The four texture features are described as follows:
A. angular Second Moment (ASM): the method is used for measuring whether the gray distribution is uniform or not and the width of the texture, and when the image texture is narrow and the gray distribution is uniform, the value is larger, and conversely, the value is smaller.
B. Entropy (ENT): describing the information quantity and the information complexity of the image, wherein when the complexity is high, the entropy value is larger, and conversely, the entropy value is smaller.
C. Contrast (CON): a measure between the maximum and minimum gray values of an image is described, i.e. the contrast of gray values. The larger the contrast, the smaller the contrast.
D. Inverse Differential Matrix (IDM): the degree of image texture resolvable and whether or not to periodically render are described. Easy to distinguish, and has strong regularity, the value is large, otherwise, the value is smaller.
Wherein G (i, j) is an element in the gray level co-occurrence matrix, and L is a gray level number.
4) Repeating the steps to extract the characteristics of all the sample images, and setting the characteristics as a sample characteristic set m;
5) Training an image classification model: the cross-validation method is adopted to divide the sample feature set m into n subsets, randomly select n-1 subsets to train the SVM classification model, and the rest 1 subsets are used for testing. The parameters to be optimized in the model training are mainly two, namely a penalty factor (C) and a kernel parameter (g), and the optimal parameters are found by using a grid optimizing method. The kernel function of the SVM in this embodiment is a Radial Basis Function (RBF):
wherein the kernel parameter g represents the width of the kernel function.
Default values for penalty factor (C) and kernel parameter (g) are 500 and 0.5. The optimum values of the found penalty factor (C) and kernel parameter (g) are 7000 and 0.006.
6) Classifying and identifying apple tree leaf diseases: and 5) obtaining an optimal SVM classification model by utilizing the optimal parameters searched in the step 5), and inputting the test sample image into the trained SVM classification model to obtain the identification result shown in the figures 7 and 8.
The prediction time and the recognition accuracy rate of the embodiment for recognizing the apple tree leaf diseases are shown in table 1, and the data in table 1 can show that the accuracy rate of the embodiment reaches 96.67%, and the prediction time can meet the real-time requirement.
Table 1 test image recognition results table
Parameter value | Prediction time | Accuracy of identification |
Default parameters | 3ms | 83.33% |
Optimum parameters | 2.3ms | 96.67% |
The invention and its embodiments have been described above by way of illustration and not limitation, and the invention is illustrated in the accompanying drawings and described in the drawings in which the actual structure is not limited thereto. Therefore, if one of ordinary skill in the art is informed by this disclosure, the structural mode and the embodiments similar to the technical scheme are not creatively designed without departing from the gist of the present invention.
Claims (1)
1. The leaf disease identification method based on the machine learning algorithm is characterized by comprising the following steps:
1) Acquisition treeLeaf sample images, collecting one hundred of healthy images and disease images respectively, and preprocessing the images; the specific process is as follows: firstly, carrying out size adjustment and graying treatment on the size of a sample image, adjusting the size to be 100 x 100, and then normalizing the pixel gray value k of the original image to obtain r k And calculates the distribution probability p (r) of each gray level k ) The normalized gray value r is then scaled by the cumulative transform function T (·) k Transforming to obtain s k Converting the gray level into a standard gray level according to the same gray level, thereby obtaining a new image after equalization treatment; the cumulative transform function is as follows:
wherein n is the total number of pixels, n j The number of pixels for different gray levels, L is the number of gray levels, j=0, 1,..k;
finally, denoising the image by adopting a median filtering algorithm;
2) Setting the gray value of an image at a pixel point (a, B) as k (a, B), taking the pixel point (a, B) as a B X B window with the center, wherein B represents the side length of the window, C represents the difference value, and calculating the threshold value v (a, B) of each pixel (a, B) in the image, wherein the calculation formula is as follows:
binarizing each pixel point (a, b) in the image point by using v (a, b) values:
dividing the preprocessed sample image by adopting a self-adaptive threshold segmentation method, and highlighting texture features of the sample image;
3) Extracting color features of the sample image and texture features of the segmented image based on the RGB color space; the color features comprise six features of mean value, variance, skewness, kurtosis, entropy and gradient of three RGB channels, and the texture features comprise angular second moment ASM, entropy ENT, contrast CON and inverse differential matrix IDM; the color characteristics are expressed as follows:
A. average value:
B. variance:
C. degree of deviation:
D. kurtosis:
E. entropy:
F. gradient:
in the above formula, p (k) represents the distribution of pixel gray values k in an imageProbability n k For its frequency, n is the total number of pixels, L is the number of gray levels,is the mean value, delta 2 Is variance, k k Is of a skewness, k F Kurtosis, k E For entropy, MXN represents the size of the image, +.>Represents the gradient of a pixel with gray value k in the horizontal direction,/or->Represents the gradient of a pixel having a gray value k in the vertical direction, k g Is its average gradient;
the texture features are expressed as follows:
A. angular second moment:
B. entropy:
C. contrast ratio:
D. inverse differential array:
wherein G (i, j) is an element in the gray level co-occurrence matrix, and L is a gray level number;
4) Extracting the characteristics of all sample images, setting the characteristics as a sample characteristic set m, and constructing an SVM classification model by adopting a cross verification method;
5) Searching optimal parameters by using a grid optimizing method, wherein the found optimal values of the penalty factor C and the core parameter g are 7000 and 0.006;
6) And 5) obtaining an optimal SVM classification model by utilizing the optimal parameters searched in the step 5), and inputting the test sample image into the trained SVM classification model to obtain a recognition result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811087765.8A CN109308697B (en) | 2018-09-18 | 2018-09-18 | Leaf disease identification method based on machine learning algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811087765.8A CN109308697B (en) | 2018-09-18 | 2018-09-18 | Leaf disease identification method based on machine learning algorithm |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109308697A CN109308697A (en) | 2019-02-05 |
CN109308697B true CN109308697B (en) | 2024-03-22 |
Family
ID=65225162
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811087765.8A Active CN109308697B (en) | 2018-09-18 | 2018-09-18 | Leaf disease identification method based on machine learning algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109308697B (en) |
Families Citing this family (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109858501A (en) * | 2019-02-20 | 2019-06-07 | 云南农业大学 | A kind of two phase flow pattern feature extracting method |
CN111652834A (en) * | 2019-03-02 | 2020-09-11 | 周口师范学院 | Steel coil strapping tape identification method based on support vector machine |
CN110349118A (en) * | 2019-03-28 | 2019-10-18 | 云南农业大学 | Quality of Gastrodia Elata Bl detection method based on computer vision technique |
CN112014398A (en) * | 2019-05-29 | 2020-12-01 | 天津中元百宜科技有限责任公司 | Method for classifying, establishing and identifying defective pipelines based on machine vision |
CN110346312B (en) * | 2019-07-19 | 2021-08-10 | 安徽大学 | Winter wheat head gibberellic disease identification method based on Fisher linear discrimination and support vector machine technology |
CN110378433A (en) * | 2019-07-24 | 2019-10-25 | 重庆大学 | The classifying identification method of bridge cable surface defect based on PSO-SVM |
CN110598732B (en) * | 2019-08-02 | 2023-04-28 | 佛山职业技术学院 | Plant health detection method and device based on image recognition |
CN110728286B (en) * | 2019-09-24 | 2023-02-10 | 西安理工大学 | Abrasive belt grinding material removal rate identification method based on spark image |
CN110874835B (en) * | 2019-10-25 | 2022-09-16 | 北京农业信息技术研究中心 | Crop leaf disease resistance identification method and system, electronic equipment and storage medium |
CN110771591B (en) * | 2019-11-13 | 2021-06-29 | 吉林省农业机械研究院 | Medicine device is spouted to automatic variable based on image recognition |
CN111507940A (en) * | 2020-03-18 | 2020-08-07 | 仲恺农业工程学院 | Image recognition-based citrus orchard pest and disease identification and alarm system and method |
CN111861103B (en) * | 2020-06-05 | 2024-01-12 | 中南民族大学 | Fresh tea classification method based on multiple features and multiple classifiers |
CN111967357A (en) * | 2020-08-05 | 2020-11-20 | 茅台学院 | Intelligent sorghum disease identification system and identification method based on machine vision |
CN112580659A (en) * | 2020-11-10 | 2021-03-30 | 湘潭大学 | Ore identification method based on machine vision |
CN113011467B (en) * | 2021-02-25 | 2024-06-14 | 南京中医药大学 | Method for identifying origin of angelica medicinal material based on image structure texture information |
CN113112451A (en) * | 2021-03-08 | 2021-07-13 | 潍坊科技学院 | Green leaf disease characteristic optimization and disease identification method based on image processing |
CN112959939A (en) * | 2021-03-17 | 2021-06-15 | 新阳荣乐(重庆)汽车电子有限公司 | Reinforced intelligent vehicle lamp control method |
CN113679082A (en) * | 2021-06-24 | 2021-11-23 | 中国烟草总公司郑州烟草研究院 | Intelligent cigar cutting method and device |
CN113379781B (en) * | 2021-07-29 | 2023-04-07 | 浙江大华技术股份有限公司 | Image-based fire monitoring method and device, storage medium and electronic equipment |
CN113627531B (en) * | 2021-08-11 | 2023-12-08 | 南京农业大学 | Method for determining pear ring rot resistance based on support vector machine classification algorithm |
CN113627528A (en) * | 2021-08-11 | 2021-11-09 | 江南大学 | Automatic Chinese painting owner identification method based on human visual deep learning |
CN113688959B (en) * | 2021-10-26 | 2022-02-18 | 寿光得峰生态农业有限公司 | Plant disease and insect pest diagnosis method and system based on artificial intelligence |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102214306A (en) * | 2011-06-16 | 2011-10-12 | 中国农业大学 | Leaf disease spot identification method and device |
CN103514459A (en) * | 2013-10-11 | 2014-01-15 | 中国科学院合肥物质科学研究院 | Method and system for identifying crop diseases and pests based on Android mobile phone platform |
CN106446942A (en) * | 2016-09-18 | 2017-02-22 | 兰州交通大学 | Crop disease identification method based on incremental learning |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI435234B (en) * | 2011-11-24 | 2014-04-21 | Inst Information Industry | Plant disease identification method, system and record media |
-
2018
- 2018-09-18 CN CN201811087765.8A patent/CN109308697B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102214306A (en) * | 2011-06-16 | 2011-10-12 | 中国农业大学 | Leaf disease spot identification method and device |
CN103514459A (en) * | 2013-10-11 | 2014-01-15 | 中国科学院合肥物质科学研究院 | Method and system for identifying crop diseases and pests based on Android mobile phone platform |
CN106446942A (en) * | 2016-09-18 | 2017-02-22 | 兰州交通大学 | Crop disease identification method based on incremental learning |
Non-Patent Citations (2)
Title |
---|
Python下opencv使用笔记(四) (图像的阈值处理);on2way;《CSDN https://blog.csdn.net/on2way/article/details/46812121》;20150709;第1-2页 * |
基于核函数支持向量机的植物叶部病害多分类检测方法;魏丽冉;岳峻;李振波;寇光杰;曲海平;;农业机械学报;20171225(第S1期);171-176 * |
Also Published As
Publication number | Publication date |
---|---|
CN109308697A (en) | 2019-02-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109308697B (en) | Leaf disease identification method based on machine learning algorithm | |
Gavhale et al. | An overview of the research on plant leaves disease detection using image processing techniques | |
Sandika et al. | Random forest based classification of diseases in grapes from images captured in uncontrolled environments | |
Wu et al. | Detection and counting of banana bunches by integrating deep learning and classic image-processing algorithms | |
CN107103306B (en) | Winter wheat powdery mildew remote sensing monitoring method based on wavelet analysis and support vector machine | |
CN106845497B (en) | Corn early-stage image drought identification method based on multi-feature fusion | |
Pereira et al. | Recent advances in image processing techniques for automated harvesting purposes: A review | |
CN111860330A (en) | Apple leaf disease identification method based on multi-feature fusion and convolutional neural network | |
Ji et al. | In-field automatic detection of maize tassels using computer vision | |
CN110827273A (en) | Tea disease detection method based on regional convolution neural network | |
Pinto et al. | Crop disease classification using texture analysis | |
CN112257702A (en) | Crop disease identification method based on incremental learning | |
CN109101891A (en) | A kind of rice pest detection system and its detection method merging artificial intelligence | |
Saha et al. | Orange fruit disease classification using deep learning approach | |
Kumar et al. | Automatic leaf disease detection and classification using hybrid features and supervised classifier | |
CN116524224A (en) | Machine vision-based method and system for detecting type of cured tobacco leaves | |
Patil | Pomegranate fruit diseases detection using image processing techniques: a review | |
Kumar et al. | Identification of plant diseases using image processing and image recognition | |
CN112085032A (en) | Grapefruit disease and insect pest dynamic monitoring model based on combination of Internet of things image processing and SVM (support vector machine) | |
Prabhu et al. | A deep learning approach to identify defects in coffee leaves using convoluional neural network | |
CN116524279A (en) | Artificial intelligent image recognition crop growth condition analysis method for digital agriculture | |
Poojari et al. | Identification and solutions for grape leaf disease using convolutional neural network (CNN) | |
Yang et al. | Intelligent collection of rice disease images based on convolutional neural network and feature matching | |
CN112070095A (en) | Grapefruit disease and insect pest monitoring method based on Internet of things technology | |
CN115862003A (en) | Lightweight YOLOv 5-based in-vivo apple target detection and classification method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |