CN109308697A - A kind of leaf disease recognition method based on machine learning algorithm - Google Patents
A kind of leaf disease recognition method based on machine learning algorithm Download PDFInfo
- Publication number
- CN109308697A CN109308697A CN201811087765.8A CN201811087765A CN109308697A CN 109308697 A CN109308697 A CN 109308697A CN 201811087765 A CN201811087765 A CN 201811087765A CN 109308697 A CN109308697 A CN 109308697A
- Authority
- CN
- China
- Prior art keywords
- image
- value
- machine learning
- method based
- learning algorithm
- 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.)
- Granted
Links
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 title claims abstract description 59
- 201000010099 disease Diseases 0.000 title claims abstract description 58
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000010801 machine learning Methods 0.000 title claims abstract description 22
- 238000012545 processing Methods 0.000 claims abstract description 15
- 230000003044 adaptive effect Effects 0.000 claims abstract description 7
- 238000002790 cross-validation Methods 0.000 claims abstract description 5
- 230000006870 function Effects 0.000 claims description 17
- 230000008569 process Effects 0.000 claims description 12
- 238000009826 distribution Methods 0.000 claims description 9
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 6
- 238000009825 accumulation Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- 230000001131 transforming effect Effects 0.000 claims description 4
- 241000269627 Amphiuma means Species 0.000 claims description 3
- 239000004744 fabric Substances 0.000 claims 1
- 241000607479 Yersinia pestis Species 0.000 abstract description 11
- 238000012549 training Methods 0.000 abstract description 5
- 230000008093 supporting effect Effects 0.000 abstract description 2
- 235000013399 edible fruits Nutrition 0.000 description 11
- 230000000694 effects Effects 0.000 description 9
- 241000219998 Philenoptera violacea Species 0.000 description 7
- 238000012706 support-vector machine Methods 0.000 description 7
- 238000000605 extraction Methods 0.000 description 6
- 239000000284 extract Substances 0.000 description 5
- 238000001914 filtration Methods 0.000 description 5
- 238000003709 image segmentation Methods 0.000 description 5
- 241000196324 Embryophyta Species 0.000 description 4
- 230000011218 segmentation Effects 0.000 description 4
- 238000001228 spectrum Methods 0.000 description 4
- 230000036541 health Effects 0.000 description 3
- 239000002420 orchard Substances 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000012512 characterization method Methods 0.000 description 2
- 238000003745 diagnosis Methods 0.000 description 2
- 238000006073 displacement reaction Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 208000015181 infectious disease Diseases 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 239000000575 pesticide Substances 0.000 description 2
- 230000002265 prevention Effects 0.000 description 2
- 230000003595 spectral effect Effects 0.000 description 2
- 208000031968 Cadaver Diseases 0.000 description 1
- 235000010799 Cucumis sativus var sativus Nutrition 0.000 description 1
- 244000299906 Cucumis sativus var. sativus Species 0.000 description 1
- 241000221785 Erysiphales Species 0.000 description 1
- 241000238631 Hexapoda Species 0.000 description 1
- 241000233679 Peronosporaceae Species 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 235000013305 food Nutrition 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 230000001976 improved effect Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 230000001151 other effect Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 230000003014 reinforcing effect Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 239000007921 spray Substances 0.000 description 1
- 238000012795 verification Methods 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)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The leaf disease recognition method based on machine learning algorithm that the invention discloses a kind of, belongs to technical field of image processing.The present invention carries out the pretreatment operations such as gray processing, image enhancement and denoising to the leaf sample image of acquisition first;Then pretreated image is split by adaptive thresholding algorithm, effectively characterizes the texture information of sample image;It selects RGB color to extract the color characteristic of sample image, while extracting the textural characteristics of segmented image according to gray level co-occurrence matrixes;Finally selection supporting vector machine model carries out Classification and Identification to sample image using cross validation algorithm, is first optimized using grid optimizing method to the major parameter of SVM model, then choose the optimal parameter of recognition accuracy and establish svm classifier identification model.The present invention can enable a computer to the pest and disease damage of automatic identification leaf by training, greatly reduce room and time expense, also improve the precision of identification, have the characteristics that quick, accurate and strong robustness.
Description
Technical field
The present invention relates to computerized information technical field of image processing, are based on machine learning more specifically to one kind
The leaf disease recognition method of algorithm.
Background technique
Quickly increase along with the continuous high speed development and population of China's economy, the agricultural in China also develops accordingly
Get up, fruit has become indispensable food in common people family, and since fruit demand is very big, its cultivated area is also constantly expanding
Greatly.With the increase of Production of fruit area, its disease problem also gradually causes the concern and attention of expert, correctly efficiently knows
The Damage Types of other fruit tree can be such that fruit tree is timely treated, to reduce unnecessary loss, increase the yield in orchard.
Instantly, image processing techniques is developed rapidly, while the application in life and production is also gradually popularized.Then
Image recognition technology is in agriculture field using but also the improvement and prevention and treatment of corps diseases achieve great improvement.To fruit
The disease geo-radar image for setting leaf carries out rapid identification, is mainly concerned with special to Image Acquisition, image preprocessing, image segmentation, image
Sign extraction, image classification identification etc. are studied and are realized, and achieve good achievement.
Through retrieving, MohammedE1-Helly etc. (2004) develops a comprehensive image processing system, can be automatic
Tikka is detected to determine disease type.The system uses artificial neural network as classifier to identify powdery mildew of cucumber, downy mildew
Disease and leaf pest.But this method calculation amount is larger, and time loss is more, and algorithm performance is poor.
2015, Chinese Patent Application No.: CN104422660A, publication date were on March 18th, 2015, are disclosed a kind of high
The plant leaf blade pest diagnosis system of spectrum, including hyperspectral imager, camera lens, calibration blank, sample placement platform, linear light
Source, linear light source supporting wall, automatically controlled displacement platform, mobile platform control device, camera bellows, adjustable support foot and computer.EO-1 hyperion at
As instrument and linear light source are connected in together on automatically controlled displacement platform, are controlled and moved together by mobile platform control device;In moving process
Middle linear light source irradiation calibration blank and measured leaf, hyperspectral imager acquisition data simultaneously upload to computer in real time;Through
After crossing image analysis and processing software processing, the triple diagnostic models of the pest and disease damage having built up are inputted, real-time quantitative obtains plant
Pest and disease damage gradient of infection, whether analysis decision to measurement plant spray and types of medicines dosage and nozzle type accordingly.It should
Application case is mainly used for the pest and disease damage gradient of infection of rapidly diagnosis of plant, to level of decision-making, the realization essence for improving accuracy pesticide applying
Thin agricultural plays positive effect.But the shortcoming of this application is, can not distinguish to different pest and disease damages.
2015, Chinese Patent Application No. CN201510321137.1, the applying date was on June 11st, 2015, innovation and creation
Title are as follows: fusion spectrum and the crop disease and insect of image information identify and distinguish between method, and this application is based on spectrum and image is believed
Cease processing technique and obtain spectral image data, carry out leaf area and background separation later, disease pest blemish extracts, spectral signature and
Box counting algorithm.The pest and disease damage type of final Shi Yongfeishi linear discriminant analysis model output blade.This application can have
Effect ground identifies different type pest and disease damage, for applications of pesticide management and pest and disease damage green prevention and control provide science decision according to
According to.But spectrum picture used in this application needs professional's acquisition and equipment is expensive, is unfavorable for promoting the use.
Summary of the invention
1. technical problems to be solved by the inivention
The method of traditional fruit tree leaf disease management relies primarily on orchard worker or agricultural experts in person to checking in orchard
The upgrowth situation of fruit tree, this method judges the upgrowth situation and disease of fruit tree by human eye, thus the result identified can exist
Certain error, while also taking time and effort.In order to overcome the above problem, the invention proposes a kind of based on machine learning algorithm
Leaf disease recognition method;The present invention can significantly be reduced by Computer Automatic Recognition different type fruit tree leaf disease
Room and time expense, while also improving accuracy of identification;Irregular for the sample image quality of acquisition, characteristics of image is not
Enough prominent, the present invention pre-processes sample image, to make sample convenient for the progress of follow-up work;For sample image
The texture difference of performance is unobvious, is not easy to texture feature extraction, and the present invention uses adaptive threshold to pretreated image
Algorithm carries out image segmentation, to highlight the textural characteristics of sample image well;For support vector machines (SVM) model silent
Recognize under parameter and be unable to satisfy identification requirement, present invention combination cross validation algorithm obtains SVM model most using grid optimizing method
Excellent parameter prevents from providing support to reach optimum efficiency for fruit disease evil.
2. technical solution
In order to achieve the above objectives, technical solution provided by the invention are as follows:
A kind of leaf disease recognition method based on machine learning algorithm of the invention, the steps include:
1) collect leaves sample image, collects healthy image and each 100 width of disease geo-radar image altogether, and located in advance to image
Reason;
2) pretreated sample image is split using adaptive threshold fuzziness method, the texture of prominent sample image
Feature;
3) color characteristic of sample image and the textural characteristics of segmented image are extracted based on RGB color;
4) feature for extracting all sample images is set as sample characteristics collection m, constructs SVM using the method for cross validation
Disaggregated model;
5) optimized parameter is found using grid optimizing method, the parameter is penalty factor and nuclear parameter g;
6) optimized parameter found using step 5), can be obtained optimal svm classifier model, and test sample image is defeated
Enter to trained svm classifier model, obtains recognition result.
Further, image preprocessing process described in step 1) are as follows: size tune is carried out to sample image size first
The processing of whole and gray processing is adjusted having a size of 100*100, then carries out image increasing to gray level image using algorithm of histogram equalization
Strong operation finally carries out denoising to image using median filtering algorithm.
Further, the image enhancement operation process are as follows: first simultaneously by the grey scale pixel value normalization of original image
Each gray-scale distribution probability is counted, then normalized gray value convert and according to identical using accumulation transforming function transformation function
Gray level be converted into the gray value of standard, to obtain equalisedization treated new images.
Further, the process that step 2) is split sample image are as follows:
Setting gray value of the image at pixel (a, b) is k (a, b), B × B window centered on pixel (a, b)
Mouthful, wherein B indicates the side length of window, and C indicates difference, calculates the threshold value v (a, b) of each pixel (a, b) in image, calculation formula
It is as follows:
Binaryzation is carried out point by point with v (a, b) value to pixel (a, b) each in image:
Further, color characteristic described in step 3) include the mean value in tri- channels RGB, variance, the degree of bias, kurtosis,
Six features of entropy and gradient, textural characteristics include angular second moment ASM, entropy ENT, contrast C ON and unfavourable balance battle array IDM.
Further, color characteristic described in step 3) is expressed as follows:
A. mean value:
B. variance:
C. the degree of bias:
D. kurtosis:
E. entropy:
F. gradient:
In above formula, p (k) indicates the distribution probability that grey scale pixel value is k in the picture, nkFor its frequency, n is that pixel is total
Number, L is number of greyscale levels,For mean value, δ2For variance, kkFor the degree of bias, kFFor kurtosis, kEFor entropy, M × N indicates the size of image,Indicate that gray value is the pixel gradient in the horizontal direction of k,Gradient of the pixel in vertical direction that expression gray value is k,
kgFor its average gradient.
Further, textural characteristics described in step 3) are expressed as follows:
A. angular second moment (ASM):
B. entropy (ENT):
C. contrast (CON):
D. unfavourable balance battle array (IDM):
Wherein, G (i, j) is the element in gray level co-occurrence matrixes, and L is number of greyscale levels.
Further, the kernel function of svm classifier model described in step 4) is Radial basis kernel function:
Wherein, nuclear parameter g represents the width of kernel function.
Further, the default value of penalty factor described in step 5) and nuclear parameter g are respectively 500 and 0.5.
Further, the optimal value of the penalty factor and nuclear parameter g that search out in step 5) is 7000 and 0.006.
3. beneficial effect
Using technical solution provided by the invention, compared with existing well-known technique, there is following remarkable result:
(1) a kind of leaf disease recognition method based on machine learning algorithm of the invention, during image preprocessing
Gray processing and image enhancement processing are carried out to the image after size adjusting first, then image is carried out using median filtering algorithm
Denoising.After the processing, it can reinforce image perception with improving image quality, rich image information and improve image knowledge
Other effect.
(2) a kind of leaf disease recognition method based on machine learning algorithm of the invention, using certainly in image segmentation
Adapt to Threshold Segmentation Algorithm can well by sample image healthy image and disease geo-radar image effectively distinguish texture expression is upper,
By image segmentation, the texture feature extraction for subsequent sample image provides advantage.
(3) a kind of leaf disease recognition method based on machine learning algorithm of the invention is based on RGB in feature extraction
Color space extracts the color characteristic of sample image, and generates gray level co-occurrence matrixes with the sample image after segmentation, to extract
Its textural characteristics converts sample image to the characteristic value of characterization image by feature extraction, classifies for subsequent sample image
The training of model provides necessary condition.
(4) a kind of leaf disease recognition method based on machine learning algorithm of the invention, passes through friendship in image recognition
Fork verification algorithm is trained and tests to SVM model, and the optimized parameter of SVM model is obtained using grid optimizing method, passes through
After the training, optimal recognition effect can achieve.
Detailed description of the invention
Fig. 1 is a kind of flow chart of leaf disease recognition method based on machine learning algorithm of the invention;
Fig. 2 is apple leaf gray level image (disease geo-radar image) described in the embodiment of the present invention 1;
Fig. 3 is the reinforcing effect figure of Fig. 2 image;
Fig. 4 is the denoising image after Fig. 3 median filtering;
Fig. 5 is the effect picture after Fig. 4 adaptive threshold fuzziness;
Fig. 6 is the preprocessed effect picture with after adaptive threshold fuzziness of Fig. 2;
Fig. 7 is that this disaggregated model is judged as sick effect picture;
Fig. 8 is the effect picture that this disaggregated model is judged as health.
Specific embodiment
To further appreciate that the contents of the present invention, the present invention is described in detail in conjunction with the accompanying drawings and embodiments.
Embodiment 1
In conjunction with Fig. 1, a kind of leaf disease recognition method based on machine learning algorithm of the present embodiment, to apple leaf
For carrying out disease recognition, the step of identifying to disease geo-radar image, is as follows:
1) image preprocessing: acquisition apple leaf sample image, the sample image include health and disease geo-radar image, disease
Containing disease region and certain healthy area in image, as long as the sample image containing disease region can regard as disease figure
Picture, the present embodiment has collected healthy image and each 100 width of disease geo-radar image altogether, and pre-processes to these images.
Described image preprocessing process are as follows: size adjusting is carried out to sample image size first and gray processing is handled, is obtained
Grayscale image as shown in Figure 2 adjusts having a size of 100*100, then carries out figure to gray level image using algorithm of histogram equalization
Image intensifying operation, the image enhancement operation process are as follows:
It normalizes the grey scale pixel value k of original image to obtain r firstkAnd calculate each gray-scale distribution probability p (rk), so
Afterwards using accumulation transforming function transformation function T () to normalized gray value rkIt is converted to obtain skAnd according to identical gray level by its
The gray value of standard is converted to, to obtain equalisedization treated new images, carries out effect such as Fig. 3 of image enhancement acquisition
It is shown, the contrast of image, the details of rich image can be improved after above-mentioned processing.The accumulation transforming function transformation function is as follows:
N is sum of all pixels, n in above formulajFor the number of pixels of different grey-scale, L is number of greyscale levels, j=0,1 ..., k.
Denoising is finally carried out using median filtering algorithm, denoising image as shown in Figure 4 is obtained, by denoising
Afterwards, while filtering out noise, the details and profile of image are remained to the greatest extent.
2) image segmentation:
It is k (a, b) that gray value of the image at pixel (a, b) is set in partitioning algorithm, with pixel (a, b) is
B × B window of the heart, wherein B indicate window side length, C indicate difference, calculate image in each pixel (a, b) threshold value v (a,
B), calculation formula is as follows:
Binaryzation is carried out point by point with v (a, b) value to pixel (a, b) each in image:
Adaptive threshold fuzziness is carried out to the image after denoising and obtains effect picture as shown in Figure 5 and to healthy image warp
Step 1) and step 2) obtain effect picture as shown in FIG. 6.From Fig. 5 and Fig. 6 comparison as can be seen that the segmented image texture of health
It is evenly distributed, texture primitive is smaller, and the segmented image grain distribution entanglement of disease, and texture primitive is larger, this shows sample graph
As after step 1) and step 2) textural characteristics by Efficient Characterization.
3) image characteristics extraction: after being pre-processed using domain-adaptive Threshold Segmentation Algorithm described in step 2) to step 1)
Apple leaf image be split, highlight the textural characteristics of healthy image and disease geo-radar image.
It is primarily based on RGB color, the color moment in the color characteristic of apple leaf disease geo-radar image is selected, according to apple
Gray value after the normalization of leaf disease geo-radar image extracts R, the mean value in tri- channels G, B, variance, the degree of bias, kurtosis, entropy, ladder respectively
Spend six characteristic values, each characteristic value, the characteristics of picture is all represented in Digital Image Processing in a certain respect.Then root
According to image brightness distribution rule and the pixel space position feature of similar luminance, counted in the sample image after step 3) segmentation
According to certain orientation and distance and gray value is respectively that the frequency that occurs of combination of pixels of i-1 and j-1 is denoted as G (i, j), wherein i
=1,2 ..., L, j=1,2 ..., L, and then generate gray level co-occurrence matrixes G, thus extract its second moment (ASM), entropy (ENT),
The textural characteristics such as contrast (CON) and unfavourable balance battle array (IDM).
Six color features are as follows:
A. mean value (Mean): reflecting the brightness of image, and mean value is bigger to illustrate that brightness of image is bigger, otherwise smaller.
B. variance (Variance): being a kind of scale calibration for measuring sample distribution uniformity.
C. the symmetry of sample totality value distribution, the i.e. torsion resistance of image the degree of bias (Skewness): are described.
D. kurtosis (Kurtosis): describe sample peak value is precipitous or gradual degree.
E. entropy (Entropy): it characterizes the information content that the aggregation characteristic of intensity profile in image is included.
F. gradient (Gradient): the case where describing the variation speed of gray value of image.
Wherein, p (k) indicates the distribution probability that grey scale pixel value is k in the picture, nkFor its frequency, n is sum of all pixels,
For mean value, L is number of greyscale levels, δ2For variance, kkFor the degree of bias, kFFor kurtosis, kEFor entropy, M × N indicates the size of image,It indicates
Gray value is the pixel gradient in the horizontal direction of k,Expression gray value is gradient of the pixel of k in vertical direction, kgFor it
Average gradient.
Four textural characteristics are described as follows:
A. angular second moment (ASM): the width for measuring the whether uniform texture of intensity profile, when image texture is relatively narrow,
When intensity profile is uniform, value is larger, conversely, smaller.
B. entropy (ENT): the information content and information complexity that description image has, when complexity is high, entropy is larger,
It is on the contrary then smaller.
C. contrast (CON): the measurement between description piece image gray value is minimum and maximum refers to the contrast of gray value
Situation.Contrast is bigger, and representative comparison is bigger, and the smaller representative comparison of contrast is smaller.
D. unfavourable balance battle array (IDM): the distinguishable degree of description image texture and whether period presentation.It is easy to differentiate, regularity is strong
, then value is big, otherwise value is smaller.
Wherein, G (i, j) is the element in gray level co-occurrence matrixes, and L is number of greyscale levels.
4) it repeats the above steps and extracts the feature of all sample images, be set as sample characteristics collection m;
5) image classification model training: the method for using cross validation divides sample characteristics collection m for n subset, at random
N-1 subset therein is chosen to train svm classifier model, remaining 1 subset is used to test.It is needed in model training excellent
There are two the parameter of change is main, it is penalty factor (C) and nuclear parameter (g) respectively, finds optimized parameter using grid optimizing method.This
The kernel function of SVM is Radial basis kernel function (RBF) in embodiment:
Wherein, nuclear parameter g represents the width of kernel function.
The default value of the penalty factor (C) and nuclear parameter (g) is 500 and 0.5.The penalty factor (C) that searches out and
The optimal value of nuclear parameter (g) is 7000 and 0.006.
6) apple leaf disease Classification and Identification: optimal svm classifier can be obtained in the optimized parameter found using step 5)
Test sample image is input to trained svm classifier model, obtains recognition result as shown in Figure 7 and Figure 8 by model.
See Table 1 for details for predicted time and recognition accuracy of the present embodiment to apple leaf disease recognition, can by data in table 1
To find out that the accuracy rate of the present embodiment has reached 96.67%, while predicted time also can satisfy requirement of real-time.
1 test image recognition result table of table
Parameter value | Predicted time | Recognition accuracy |
Default parameters | 3ms | 83.33% |
Optimized parameter | 2.3ms | 96.67% |
Schematically the present invention and embodiments thereof are described above, description is not limiting, institute in attached drawing
What is shown is also one of embodiments of the present invention, and actual structure is not limited to this.So if the common skill of this field
Art personnel are enlightened by it, without departing from the spirit of the invention, are not inventively designed and the technical solution
Similar frame mode and embodiment, are within the scope of protection of the invention.
Claims (10)
1. a kind of leaf disease recognition method based on machine learning algorithm, which is characterized in that the steps include:
1) collect leaves sample image, collects healthy image and each 100 width of disease geo-radar image altogether, and pre-process to image;
2) pretreated sample image is split using adaptive threshold fuzziness method, the texture of prominent sample image is special
Sign;
3) color characteristic of sample image and the textural characteristics of segmented image are extracted based on RGB color;
4) feature for extracting all sample images is set as sample characteristics collection m, constructs svm classifier using the method for cross validation
Model;
5) optimized parameter is found using grid optimizing method, the parameter is penalty factor and nuclear parameter g;
6) optimized parameter found using step 5), can be obtained optimal svm classifier model, test sample image is input to
Trained svm classifier model, obtains recognition result.
2. a kind of leaf disease recognition method based on machine learning algorithm according to claim 1, it is characterised in that: step
It is rapid 1) described in image preprocessing process are as follows: size adjusting and gray processing are carried out to sample image size first and handled, ruler is adjusted
Very little is 100*100, then carries out image enhancement operation to gray level image using algorithm of histogram equalization, is finally filtered using intermediate value
Wave algorithm carries out denoising to image.
3. a kind of leaf disease recognition method based on machine learning algorithm according to claim 2, it is characterised in that: institute
The image enhancement operation process stated are as follows: by the grey scale pixel value Normalized Grey Level value of original image and count each gray-scale point first
Then cloth probability convert and be converted according to identical gray level to normalized gray value using accumulation transforming function transformation function
For the gray value of standard, to obtain equalisedization treated new images.
4. a kind of leaf disease recognition method based on machine learning algorithm according to claim 1-3, special
Sign is: the process that step 2) is split sample image are as follows:
Setting gray value of the image at pixel (a, b) is k (a, b), B × B window centered on pixel (a, b),
Middle B indicates the side length of window, and C indicates difference, calculates the threshold value v (a, b) of each pixel (a, b) in image, and calculation formula is as follows:
Binaryzation is carried out point by point with v (a, b) value to pixel (a, b) each in image:
5. a kind of leaf disease recognition method based on machine learning algorithm according to claim 4, it is characterised in that: step
It is rapid 3) described in color characteristic include tri- channels RGB six mean value, variance, the degree of bias, kurtosis, entropy and gradient features, texture
Feature includes angular second moment ASM, entropy ENT, contrast C ON and unfavourable balance battle array IDM.
6. a kind of leaf disease recognition method based on machine learning algorithm according to claim 5, it is characterised in that: step
It is rapid 3) described in color characteristic be expressed as follows:
A. mean value:
B. variance:
C. the degree of bias:
D. kurtosis:
E. entropy:
F. gradient:
In above formula, p (k) indicates the distribution probability that grey scale pixel value is k in the picture, nkFor its frequency, n is sum of all pixels, and L is
Number of greyscale levels,For mean value, δ2For variance, kkFor the degree of bias, kFFor kurtosis, kEFor entropy, M × N indicates the size of image,Indicate ash
Angle value is the pixel gradient in the horizontal direction of k,Expression gray value is gradient of the pixel of k in vertical direction, kgIt is flat for it
Equal gradient.
7. a kind of leaf disease recognition method based on machine learning algorithm according to claim 6, it is characterised in that: step
It is rapid 3) described in textural characteristics be expressed as follows:
A. angular second moment:
B. entropy:
C. contrast:
D. unfavourable balance battle array:
Wherein, G (i, j) is the element in gray level co-occurrence matrixes, and L is number of greyscale levels.
8. a kind of leaf disease recognition method based on machine learning algorithm according to claim 7, it is characterised in that: step
It is rapid 4) described in svm classifier model kernel function be Radial basis kernel function:
Wherein, nuclear parameter g represents the width of kernel function.
9. a kind of leaf disease recognition method based on machine learning algorithm according to claim 8, it is characterised in that: step
It is rapid 5) described in penalty factor and the default value of nuclear parameter g be respectively 500 and 0.5.
10. a kind of leaf disease recognition method based on machine learning algorithm according to claim 8, it is characterised in that:
The optimal value of the penalty factor and nuclear parameter g that search out in step 5) is 7000 and 0.006.
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 true CN109308697A (en) | 2019-02-05 |
CN109308697B 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) |
Cited By (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 |
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 |
CN110346312A (en) * | 2019-07-19 | 2019-10-18 | 安徽大学 | Winter wheat fringe head blight recognition methods based on Fei Shi linear discriminant and support vector machines 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 |
CN110598732A (en) * | 2019-08-02 | 2019-12-20 | 佛山职业技术学院 | Plant health detection method and device based on image recognition |
CN110728286A (en) * | 2019-09-24 | 2020-01-24 | 西安理工大学 | Abrasive belt grinding material removal rate identification method based on spark image |
CN110771591A (en) * | 2019-11-13 | 2020-02-11 | 吉林省农业机械研究院 | Medicine device is spouted to automatic variable based on image recognition |
CN110874835A (en) * | 2019-10-25 | 2020-03-10 | 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 |
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 |
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 |
CN111861103A (en) * | 2020-06-05 | 2020-10-30 | 中南民族大学 | Fresh tea leaf 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 |
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 |
CN112580659A (en) * | 2020-11-10 | 2021-03-30 | 湘潭大学 | Ore identification method based on machine vision |
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 |
CN113011467A (en) * | 2021-02-25 | 2021-06-22 | 南京中医药大学 | Angelica sinensis medicinal material producing area identification method 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 |
CN113379781A (en) * | 2021-07-29 | 2021-09-10 | 浙江大华技术股份有限公司 | Image processing method and apparatus, storage medium, and electronic device |
CN113627531A (en) * | 2021-08-11 | 2021-11-09 | 南京农业大学 | Method for determining ring rot resistance of pear trees 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 |
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 |
CN113688959A (en) * | 2021-10-26 | 2021-11-23 | 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 (4)
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 |
US20130136312A1 (en) * | 2011-11-24 | 2013-05-30 | Shih-Mu TSENG | Method and system for recognizing plant diseases and recording medium |
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 |
-
2018
- 2018-09-18 CN CN201811087765.8A patent/CN109308697B/en active Active
Patent Citations (4)
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 |
US20130136312A1 (en) * | 2011-11-24 | 2013-05-30 | Shih-Mu TSENG | Method and system for recognizing plant diseases and recording medium |
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 (3)
Title |
---|
ON2WAY: "Python下opencv使用笔记(四) (图像的阈值处理)", 《CSDN HTTPS://BLOG.CSDN.NET/ON2WAY/ARTICLE/DETAILS/46812121》 * |
ON2WAY: "Python下opencv使用笔记(四) (图像的阈值处理)", 《CSDN HTTPS://BLOG.CSDN.NET/ON2WAY/ARTICLE/DETAILS/46812121》, 9 July 2015 (2015-07-09), pages 1 - 2 * |
魏丽冉;岳峻;李振波;寇光杰;曲海平;: "基于核函数支持向量机的植物叶部病害多分类检测方法", 农业机械学报, no. 1, 25 December 2017 (2017-12-25), pages 171 - 176 * |
Cited By (30)
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 |
CN110346312A (en) * | 2019-07-19 | 2019-10-18 | 安徽大学 | Winter wheat fringe head blight recognition methods based on Fei Shi linear discriminant and support vector machines technology |
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 |
CN110598732A (en) * | 2019-08-02 | 2019-12-20 | 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 |
CN110728286A (en) * | 2019-09-24 | 2020-01-24 | 西安理工大学 | Abrasive belt grinding material removal rate identification method based on spark image |
CN110874835A (en) * | 2019-10-25 | 2020-03-10 | 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 |
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 |
CN110771591A (en) * | 2019-11-13 | 2020-02-11 | 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 |
CN111861103A (en) * | 2020-06-05 | 2020-10-30 | 中南民族大学 | Fresh tea leaf classification method based on multiple features and multiple classifiers |
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 |
CN113011467A (en) * | 2021-02-25 | 2021-06-22 | 南京中医药大学 | Angelica sinensis medicinal material producing area identification method based on image structure texture information |
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 |
CN113379781A (en) * | 2021-07-29 | 2021-09-10 | 浙江大华技术股份有限公司 | Image processing method and apparatus, storage medium, and electronic device |
CN113627528A (en) * | 2021-08-11 | 2021-11-09 | 江南大学 | Automatic Chinese painting owner identification method based on human visual deep learning |
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 |
CN113627531A (en) * | 2021-08-11 | 2021-11-09 | 南京农业大学 | Method for determining ring rot resistance of pear trees based on support vector machine classification algorithm |
CN113627528B (en) * | 2021-08-11 | 2024-09-06 | 江南大学 | Automatic identification method for painters belonging to traditional Chinese painting based on human eye vision deep learning |
CN113688959A (en) * | 2021-10-26 | 2021-11-23 | 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 |
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 |
Also Published As
Publication number | Publication date |
---|---|
CN109308697B (en) | 2024-03-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109308697A (en) | A kind of leaf disease recognition method based on machine learning algorithm | |
Bhargava et al. | Fruits and vegetables quality evaluation using computer vision: A review | |
Kurtulmuş et al. | Classification of pepper seeds using machine vision based on neural network | |
CN106525732B (en) | Rapid nondestructive detection method for internal and external quality of apple based on hyperspectral imaging technology | |
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 | |
CN109034269A (en) | A kind of bollworm female male imago method of discrimination based on computer vision technique | |
CN105320970B (en) | A kind of potato disease diagnostic device, diagnostic system and diagnostic method | |
CN110827273A (en) | Tea disease detection method based on regional convolution neural network | |
Tumang | Pests and diseases identification in mango using MATLAB | |
CN109101891A (en) | A kind of rice pest detection system and its detection method merging artificial intelligence | |
CN112257702A (en) | Crop disease identification method based on incremental learning | |
Wang et al. | Improved rotation kernel transformation directional feature for recognition of wheat stripe rust and powdery mildew | |
CN116524224A (en) | Machine vision-based method and system for detecting type of cured tobacco leaves | |
Srivastav et al. | Tea leaf disease detection using deep learning-based convolutional neural networks | |
CN201041547Y (en) | Device for analyzing and recognizing different types of green teas based on multi-spectrum image texture | |
Ji et al. | Apple color automatic grading method based on machine vision | |
Prabhu et al. | A deep learning approach to identify defects in coffee leaves using convoluional neural network | |
Jerome et al. | An effective approach for plant disease detection using assessment-based convolutional neural networks (A-CNN) | |
CN115862003A (en) | Lightweight YOLOv 5-based in-vivo apple target detection and classification method | |
Ullagaddi et al. | Automatic robust segmentation scheme for pathological problems in mango crop | |
Singh et al. | A novel algorithm for segmentation of diseased apple leaf images | |
Radhakrishnan | Identification of citrus fruit diseases through intelligent computational approaches: a review | |
Sandhiya et al. | Disease Detection of Various Plant Leaf Using Image Processing Technique | |
Theerthagiri | Plant Leaf disease detection using supervised machine learning algorithm |
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 |