CN107330892A - A kind of sunflower disease recognition method based on random forest method - Google Patents
A kind of sunflower disease recognition method based on random forest method Download PDFInfo
- Publication number
- CN107330892A CN107330892A CN201710606339.XA CN201710606339A CN107330892A CN 107330892 A CN107330892 A CN 107330892A CN 201710606339 A CN201710606339 A CN 201710606339A CN 107330892 A CN107330892 A CN 107330892A
- Authority
- CN
- China
- Prior art keywords
- disease
- sunflower
- image
- color
- geo
- 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.)
- Pending
Links
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 title claims abstract description 205
- 201000010099 disease Diseases 0.000 title claims abstract description 201
- 238000000034 method Methods 0.000 title claims abstract description 116
- 241000208818 Helianthus Species 0.000 title claims abstract description 99
- 235000003222 Helianthus annuus Nutrition 0.000 title claims abstract description 99
- 238000007637 random forest analysis Methods 0.000 title claims abstract description 41
- 230000011218 segmentation Effects 0.000 claims abstract description 24
- 238000000605 extraction Methods 0.000 claims abstract description 22
- 238000003745 diagnosis Methods 0.000 claims abstract description 15
- 238000004458 analytical method Methods 0.000 claims abstract description 12
- 241000233679 Peronosporaceae Species 0.000 claims abstract description 7
- 230000001580 bacterial effect Effects 0.000 claims abstract description 7
- 206010027146 Melanoderma Diseases 0.000 claims abstract description 6
- 238000004422 calculation algorithm Methods 0.000 claims description 49
- 206010039509 Scab Diseases 0.000 claims description 31
- 238000012545 processing Methods 0.000 claims description 27
- 238000011160 research Methods 0.000 claims description 20
- 241000196324 Embryophyta Species 0.000 claims description 16
- 230000000694 effects Effects 0.000 claims description 14
- 238000003064 k means clustering Methods 0.000 claims description 12
- 238000005516 engineering process Methods 0.000 claims description 11
- 230000000877 morphologic effect Effects 0.000 claims description 11
- 230000008859 change Effects 0.000 claims description 9
- 239000000284 extract Substances 0.000 claims description 9
- 230000008901 benefit Effects 0.000 claims description 6
- 238000003672 processing method Methods 0.000 claims description 6
- 238000012216 screening Methods 0.000 claims description 6
- 238000002474 experimental method Methods 0.000 claims description 5
- 238000001914 filtration Methods 0.000 claims description 5
- 238000010835 comparative analysis Methods 0.000 claims description 4
- 238000003909 pattern recognition Methods 0.000 claims description 4
- 239000000463 material Substances 0.000 claims description 3
- 230000007170 pathology Effects 0.000 claims description 3
- 238000010921 in-depth analysis Methods 0.000 claims description 2
- 238000004519 manufacturing process Methods 0.000 claims 1
- 239000011159 matrix material Substances 0.000 claims 1
- 239000000843 powder Substances 0.000 claims 1
- 241000221785 Erysiphales Species 0.000 abstract description 5
- 238000003066 decision tree Methods 0.000 description 14
- 238000012549 training Methods 0.000 description 10
- 238000012360 testing method Methods 0.000 description 8
- 230000008569 process Effects 0.000 description 7
- 238000012706 support-vector machine Methods 0.000 description 6
- 238000007405 data analysis Methods 0.000 description 4
- 238000003709 image segmentation Methods 0.000 description 4
- 238000013138 pruning Methods 0.000 description 4
- 230000009466 transformation Effects 0.000 description 4
- 230000002708 enhancing effect Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 241000894007 species Species 0.000 description 3
- 240000008067 Cucumis sativus Species 0.000 description 2
- 235000010799 Cucumis sativus var sativus Nutrition 0.000 description 2
- 241000607479 Yersinia pestis Species 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 2
- 230000004069 differentiation Effects 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000003786 synthesis reaction Methods 0.000 description 2
- 208000031968 Cadaver Diseases 0.000 description 1
- 235000002566 Capsicum Nutrition 0.000 description 1
- 240000008574 Capsicum frutescens Species 0.000 description 1
- 244000141359 Malus pumila Species 0.000 description 1
- 235000011430 Malus pumila Nutrition 0.000 description 1
- 235000015103 Malus silvestris Nutrition 0.000 description 1
- 244000131316 Panax pseudoginseng Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 208000025865 Ulcer Diseases 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000032683 aging Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000001390 capsicum minimum Substances 0.000 description 1
- 239000003610 charcoal Substances 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000004040 coloring Methods 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000007596 consolidation process Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000010339 dilation Effects 0.000 description 1
- 208000035475 disorder Diseases 0.000 description 1
- 244000144992 flock Species 0.000 description 1
- 238000013467 fragmentation Methods 0.000 description 1
- 238000006062 fragmentation reaction Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000029052 metamorphosis Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 244000052769 pathogen Species 0.000 description 1
- 230000001717 pathogenic effect Effects 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000001932 seasonal effect Effects 0.000 description 1
- 238000007873 sieving Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000007920 subcutaneous administration Methods 0.000 description 1
- 238000001308 synthesis method Methods 0.000 description 1
- 231100000397 ulcer Toxicity 0.000 description 1
- 208000029257 vision disease Diseases 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 238000013316 zoning 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
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration by the use of local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration by the use of histogram techniques
-
- G06T5/70—
-
- G06T5/73—
-
- 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
- G06T7/12—Edge-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/40—Analysis of texture
- G06T7/41—Analysis of texture based on statistical description of texture
- G06T7/45—Analysis of texture based on statistical description of texture using co-occurrence matrix computation
-
- 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
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20004—Adaptive image processing
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20112—Image segmentation details
- G06T2207/20152—Watershed segmentation
-
- 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
Abstract
The invention discloses a kind of sunflower disease recognition method based on random forest method, including the common four kinds of diseases of sunflower leaf portion:Powdery mildew, bacterial leaf spot, black spot, downy mildew, A:Disease geo-radar image is gathered, the color that the leaf image color collected will try one's best close to leaf in itself, B:Disease geo-radar image is pre-processed, using the preprocess method for being suitable for the identification of sunflower disease geo-radar image, C:Disease geo-radar image is split, by analysis the various methods of contrast images segmentation, D:Disease geo-radar image feature extraction, color characteristic, the textural characteristics parameter for extracting disease geo-radar image is studied, E:The identifying and diagnosing of disease, carries out last diagnosis to sunflower disease using random forest method and recognizes.Present invention mainly solves the problem of the new expression distinguished in the subjectivity that there is naked eyes judgement during disease recognition, limitation, ambiguity and hardly possible, the accuracy of identification disease is improved, help well is provided to identification, the preventing and treating of sunflower disease for agriculturist.
Description
Technical field
The present invention relates to and agricultural pest identification field, more particularly to a kind of sunflower disease based on random forest method
Recognition methods.
Background technology
Recognized it is well known that traditional disease screening method is mainly plant protection personnel by naked eyes, and combine plant disease
The form of pathogen judged that this method diagnosis efficiency is low, it is difficult in time, accurately judge Damage Types." accurate agriculture
Industry ", new thinking is provided for agriculture cultivator, by fast and effectively recognizing plant disease with information technology, relative to biography
The recognition methods of system, recognition speed is fast, accuracy rate is high, also with ageing.By taking the crop diseases such as apple, cucumber and capsicum as an example
The image of plant leaf diseases is analyzed, level set is used for the colouring information of image and improves C-V scale-model investigations, warp
Verify that this method substantially increases the discrimination of disease.
China Patent No. under Various Seasonal, is acquired other just to be mentioned in 201510508653.5 open source literature
Condition identical plant disease blade, influence of the research environment information to disease, and use image processing techniques and statistical analysis
Theory, with attribute reduction method extraction environment information characteristics amount, and it is common to combine its shape facility, color characteristic, textural characteristics
40 characteristic parameters, based on maximum membership degree Function Criterion, downy mildew, brown spot and charcoal on diagnosis cucumber disease blade
Subcutaneous ulcer three kinds of scab species of disease, after diagnosing research draws the recognition accuracy of three kinds of diseases more than 90%.
The application text of sunflower leaf diseases determination methods of the Patent No. 201610064181.3 based on SVMs
It is scab ratio in R, G, B color component that sunflower disease geo-radar image is extracted according to RGB color model, G component images in offering
More apparent, pest and disease damage scab part gray value is larger compared with normal segments gray value deviation, and diagnosis is used as using the G components of coloured image
Have disease-free main study subject, determine that G components discrimination threshold comes, using based on b values and SVMs it is normal with it is non-
Normal foliar diagnosis, based on gray level co-occurrence matrixes and SVMs whether the improper foliar diagnosis containing disease both
Method inputs the affiliated disease classification of sunflower leaf portion image to identify.But not using random forest method come specifically do not study to
The detailed process of day certain herbaceous plants with big flowers leaf diseases identification.
From the point of view of research conditions in summary, be conducive to overcoming human eye vision disease instead of human eye using image recognition technology
Subjectivity, the low feature of empirical, efficiency existed during identification, but image recognition so far is also not directed to sunflower disease
Research in terms of diagnosis.Therefore, at home and abroad on the basis of achievement in research, further sunflower of the research based on image recognition is sick
Evil diagnosis has important practical significance, and traditional data classification has clustering algorithm, Bayesian Classification Arithmetic, SVMs
Deng, but the nicety of grading of these methods is not often high, and it is also easy to over-fitting problem occur.And random forest (RF) is to be based on
The theory of statistical learning, has more preferable effect to sunflower disease screening.
The content of the invention
The technical problem to be solved in the present invention be overcome the defect of prior art there is provided it is a kind of based on random forest method to
Day certain herbaceous plants with big flowers disease recognition method, solves to exist during disease recognition subjectivity, limitation, ambiguity and difficulty that naked eyes judge and distinguishes
New expression the problem of.
In order to solve the above-mentioned technical problem, the invention provides following technical scheme:
Technical scheme:For the sunflower being grown under the environment of crop field, IMAQ will be under natural lighting
Carry out, the leaf image collected first has to carry out image denoising operation by pretreatment, with reference to actual environment factor, using neighbour
The domain method of average, the denoising of morphology adaptive method simultaneously realize that image defogging is handled with histogram equalization method;Secondly, by comparing five
After kind of image partition method, the method segmentation figure being combined using K- means clustering algorithms with watershed algorithm is as disease, extraction
The scab of disease;Again, for random forest method the characteristics of, Disease Characters parameter is analysed in depth, chooses its hsv color
Nine color characteristic parameters of first moment, second moment, third moment of the color component of H, S, V tri- in space and based on gray level co-occurrence matrixes
The energy of textural characteristics, entropy, the moment of inertia, correlation, unfavourable balance away from ten characteristic parameters of average and standard deviation totally ten nine spies
Parameter is levied as the foundation for differentiating different diseases;Finally, random forest grader is built according to the characteristic parameter extracted, used
Vote decision-making method come realize to disease identification classification, and to disease recognition holistic approach analyze on the basis of, with reference to
The situation that different user is used, has built two kinds of sunflower leaf portion disease recognition analogue systems based on random forest method.
A kind of sunflower disease recognition method based on random forest method of the present invention, including the common four kinds of diseases of sunflower leaf portion
Evil:Powdery mildew, bacterial leaf spot, black spot, downy mildew, and it is used as research object:
A:Disease geo-radar image is gathered.The color that the leaf image color collected will try one's best close to leaf in itself.
B:Disease geo-radar image is pre-processed.In the case where combining actual conditions, using the pre- place for being suitable for the identification of sunflower disease geo-radar image
Reason method, strengthens image effect, and adaptively scheme based on morphologic weight using the histogram equalization defogging algorithm in spatial domain
As Denoising Algorithm carries out denoising to sunflower leaf portion disease geo-radar image.
C:Disease geo-radar image is split.The various methods of contrast images segmentation, choose suitable optimal coloured image by analysis
Dividing method, splits sunflower leaf diseases using K- means clustering algorithms after experiment with the method that watershed algorithm is combined
Coloured image.
D:Disease geo-radar image feature extraction.There is obvious change by sunflower color and texture, thus extract disease figure
Color characteristic, the textural characteristics parameter of picture are studied, and by the in-depth analysis to characteristic parameter, are preferably gone out suitable image and known
Other best features parameter.
E:The identifying and diagnosing of disease.With reference to the color characteristic and textural characteristics parameter extracted, random forest method pair is used
Sunflower disease carries out last diagnosis identification.
F:Hardware environment:Free background board, conventional background color is set to black, white two kinds, and subsequent treatment uses color portion
Point, using white background, and using blank sheet of paper as material, camera angle is adjusted during shooting, under the conditions of natural lighting, make to photograph as far as possible
Leaf color close to the color of itself.
As a preferred technical solution of the present invention, disease geo-radar image collection, targetedly collecting part is to day
Certain herbaceous plants with big flowers leaf diseases associated picture, incoming computer carries out subsequent treatment, it is ensured that the automatic recognition system pair of sunflower leaf diseases
Disease figure recognizes accurate and effective, professional plant protection personage will be invited to carry out the sunflower leaf diseases sample collected accurate
Distinguish.
As a preferred technical solution of the present invention, the disease geo-radar image pretreatment, using Computer Image Processing skill
Art, through analyzing sunflower leaf diseases own characteristic and all kinds of color image filtering methods of comparison, using the histogram in spatial domain
Equalize defogging algorithm enhancing image effect.
As a preferred technical solution of the present invention, the disease geo-radar image segmentation, comparative analysis color image color is empty
Between each model, by contrast, and actual conditions of the image collected according to crop field, final choice is sensitive to human eye vision
The color space split as sunflower disease of hsv color space.By comparing the processing method that five kinds of images are split, and it is comprehensive
Close and consider its respective advantage and disadvantage, final choice by K- means clustering algorithms and watershed algorithm combine it is common realize to
The segmentation of the colored disease geo-radar image of day certain herbaceous plants with big flowers blade, obtains scab segmentation figure picture.
As a preferred technical solution of the present invention, the disease geo-radar image feature extraction, choosing has notable difference special
Property color characteristic and textural characteristics analyzed and researched, it is special by the color characteristic and texture of analyzing four kinds of disease scab images
Levy, therefrom extract the first moment, second moment, three ranks of tri- color components of H, S, V in the hsv color space of four kinds of disease scabs
Energy, entropy, the moment of inertia, correlation and the unfavourable balance of d=1 gray level co-occurrence matrixes in nine color characteristic parameters of square and textural characteristics
Away from ten textural characteristics parameters of average and standard deviation, the differentiation that 19 characteristic parameters are diagnosed as disease recognition altogether according to
According to, be next step disease screening identification get ready.
As a preferred technical solution of the present invention, the identifying and diagnosing of the disease is analysed in depth to Disease Characters
On the basis of, with reference to this research practical study condition, choose herein and diagnosis knowledge is carried out to sunflower disease using random forest method
Not.
As a preferred technical solution of the present invention, the disease geo-radar image feature extraction, for sunflower leaf portion scab
When color, textural characteristics are started with, using computer image processing technology, integrated use image procossing, pattern-recognition, computer are regarded
Knowledge in terms of feel, Plant Pathology carries out automatic identification to sunflower leaf diseases, so as to deepen to sunflower disease
Understanding.
Compared with prior art, beneficial effects of the present invention are as follows:
1. present invention mainly solves during disease recognition exist naked eyes judge subjectivity, limitation, ambiguity and
The problem of new expression that difficulty is distinguished, the generation of these problems is avoided that, improves the accuracy of identification disease, present system operation letter
Single, easy to use, the precision of disease recognition is greatly improved, and identification, the preventing and treating of sunflower disease are provided for agriculturist
Good help.
2. the present invention is had outstanding performance using the recognition methods of random forest method in classification and regression problem, it is good at processing
High dimensional data, with very high predictablity rate, has good disposal ability, and be not easy to occur for abnormal conditions
Fitting phenomenon, reaches higher discrimination, and recognition effect is preferable.
3. multiple Models Sets can be combined together by the present invention, constitute a new sorter model to improve point of data
The method of class precision of prediction, improves classification performance by way of combination, and the classification and recurrence for sample have appearance well
Degree of bearing.
Brief description of the drawings
Accompanying drawing is used for providing a further understanding of the present invention, and constitutes a part for specification, the reality with the present invention
Applying example is used to explain the present invention together, is not construed as limiting the invention.In the accompanying drawings:
Fig. 1 is the Technology Roadmap of the present invention;
Embodiment
The preferred embodiments of the present invention are illustrated below in conjunction with accompanying drawing, it will be appreciated that preferred reality described herein
Apply example to be merely to illustrate and explain the present invention, be not intended to limit the present invention.
Embodiment 1
As shown in figure 1, the present invention provides a kind of sunflower disease recognition method based on random forest method, including sunflower
The common four kinds of diseases of leaf portion:Powdery mildew, bacterial leaf spot, black spot, downy mildew, and it is used as research object:
A:Disease geo-radar image is gathered.The color that the leaf image color collected will try one's best close to leaf in itself.
B:Disease geo-radar image is pre-processed.In the case where combining actual conditions, using the pre- place for being suitable for the identification of sunflower disease geo-radar image
Reason method, strengthens image effect, and adaptively scheme based on morphologic weight using the histogram equalization defogging algorithm in spatial domain
As Denoising Algorithm carries out denoising to sunflower leaf portion disease geo-radar image.
C:Disease geo-radar image is split.The various methods of contrast images segmentation, choose suitable optimal coloured image by analysis
Dividing method, splits sunflower leaf diseases using K- means clustering algorithms after experiment with the method that watershed algorithm is combined
Coloured image.
D:Disease geo-radar image feature extraction.Sunflower disease shape is disorderly and unsystematic, can be sought without track, and color and texture have compared with
Significantly change, thus extract color characteristic, the textural characteristics parameter of disease geo-radar image and studied, and by characteristic parameter
Analyse in depth, preferably go out the best features parameter of suitable image recognition.
E:The identifying and diagnosing of disease.With reference to the color characteristic and textural characteristics parameter extracted, random forest method pair is used
Sunflower disease carries out last diagnosis identification.
F:Hardware environment:Housing environment not science is used for the complicated field crop of environment, is examined for convenience of subsequent treatment
Disconnected, free background board, conventional background color is set to black, white two kinds, and subsequent treatment uses color part, black background overall brightness
It is partially dark, thus using white background, and using blank sheet of paper as material, camera angle is adjusted during shooting, under the conditions of natural lighting, as far as possible
Make the leaf color that photographed close to the color of itself.
Disease geo-radar image is gathered, targetedly collecting part sunflower leaf diseases associated picture, and incoming computer is carried out
Subsequent treatment, it is ensured that the automatic recognition system of sunflower leaf diseases recognizes accurate and effective to disease figure, will invite specialty to plant
Guarantor scholar carries out accurate discrimination to the sunflower leaf diseases sample collected.
Disease geo-radar image is pre-processed, using computer image processing technology, through analyze sunflower leaf diseases own characteristic with
And relatively more all kinds of color image filtering methods, image effect is strengthened using the histogram equalization defogging algorithm in spatial domain.
Disease geo-radar image is split, each model in comparative analysis color image color space, by contrast, and is adopted according to crop field
Split as sunflower disease in the actual conditions of the image collected, the final choice hsv color space sensitive to human eye vision
Color space.By comparing the processing method that five kinds of images are split, and consider its respective advantage and disadvantage, final choice is by K-
Means clustering algorithm and watershed algorithm combine segmentation of the common realization to the colored disease geo-radar image of sunflower blade, obtain disease
Spot segmentation figure picture.
Disease geo-radar image feature extraction, chooses the color characteristic with notable difference characteristic and textural characteristics progress analysis is ground
Study carefully, by analyzing the color characteristic and textural characteristics of four kinds of disease scab images, therefrom extract the HSV face of four kinds of disease scabs
D in nine color characteristic parameters of first moment, second moment, third moment and textural characteristics of tri- color components of H, S, V of the colour space
Energy, entropy, the moment of inertia, correlation and the unfavourable balance of=1 gray level co-occurrence matrixes away from average and standard deviation ten textural characteristics ginseng
Amount, the distinguishing rule that 19 characteristic parameters are diagnosed as disease recognition altogether is that standard has been carried out in next step disease screening identification
It is standby.
The identifying and diagnosing of disease, on the basis of being analysed in depth to Disease Characters, with reference to this research practical study condition, this
Place chooses carries out diagnosis identification using random forest method to sunflower disease.
Disease geo-radar image feature extraction, when starting with for sunflower leaf portion scab color, textural characteristics, using computer picture
Knowledge in terms for the treatment of technology, integrated use image procossing, pattern-recognition, computer vision, Plant Pathology is to sunflower
Leaf diseases carry out automatic identification, so as to deepen the understanding to sunflower disease.
Specifically, the present invention it is main using sunflower powdery mildew, bacterial leaf spot, black spot and downy mildew scab image as
Research object, extracts the characteristic parameter of disease.Because the various shapes of disease geo-radar image, and it is disorderly and unsystematic, no track can be sought, so
Do not consider shape facility in the feature extraction of scab, only choose color characteristic and textural characteristics with notable difference characteristic and enter
Row analysis and research.By analyzing the color characteristic and textural characteristics of four kinds of disease scab images, four kinds of disease diseases are therefrom extracted
Nine color characteristic parameters of first moment, second moment, third moment and line of tri- color components of H, S, V in the hsv color space of spot
Manage feature in d=1 gray level co-occurrence matrixes energy, entropy, the moment of inertia, correlation and unfavourable balance away from ten lines of average and standard deviation
The characteristic parameter distinguishing rule that 19 characteristic parameters are diagnosed as disease recognition altogether is managed, is that next step disease screening identification is done
Get well and prepared.
Disease geo-radar image feature extraction:The purpose of feature extraction is to extract important information in image, for follow-up
Image recognition processing, suitable characteristic parameter is particularly important, and it is the important evidence of disease recognition.Common Disease Characters have face
Color characteristic, textural characteristics and shape facility etc., color characteristic are the surface naturies of image, and textural characteristics are knot of tissue in kind
Structure, shape facility is its outside geometrical model, when sunflower leaf portion is ill, and significant change occurs for its leaf color, according to disease
The degree that infects of evil turns yellow or blackening and with fleck, and same blade disease sites color different from other parts, face
Color the change of divergence is larger, and its texture shows different tendencies with the increase of certain scab, with the original tendency phase of blade texture
The back of the body or intersect is carried out, and larger difference occur in scab texture and normal blade texture, and its spot pattern is nearly all rounded or ellipse
Shape, difference is not obvious, therefore, and the color characteristic and textural characteristics that the present invention chooses disease are used as sunflower leaf portion disease recognition
Foundation.
Conventional color model has RGB, HSV, XYZ, YUV etc..RGB and hsv color model are usually used in corps diseases figure
In the research of picture, RGB color model is mainly related to brightness, but the leaf image collected under the environment of crop field is easily by illumination
Etc. factor influence, so RGB color can not be used to describe the coloured image got under the environment of crop field.HSV is main
The bright-coloured degree of image is described, it is unrelated with brightness, and also its characteristic is more suitable for the observation of human eye, and image is described using the model
Color characteristic, its superior performance can be more embodied compared to RGB, and image color information is not lost, so in the present invention,
Using the color characteristic of hsv color spatial extraction scab, but image is RGB models in itself, and hsv color space can not directly make
With need to change and just may be used.
Therefore, for color image color feature extraction, top priority is that disease geo-radar image color space is carried out into one to turn
Change processing, space be converted into behind hsv color space, extract hsv color three-component study in detail, and using its characteristic parameter as
Color characteristic studies parameter.
Texture is some shape very littles, a series of patterns of regular arrangement, and textural characteristics are present in image, are figures
It is its internal organizational structure as the attribute of itself, is the description to internal pixel, numerous pixels flocks together to form table
Face characteristic, is the change of gradation of image grade, is one of the principal character of image procossing and pattern-recognition.
Structure-based method, the method based on statistics, the method based on synthesis are three kinds of texture feature extraction
Conventional method.Structure-based method it is most basic be Fourier transformation and wavelet transformation, by data conversion with certain
Method is classified to obtain certain measurement, and Statistics-Based Method is mainly based upon gray level co-occurrence matrixes method, and it is handled
Functional, the feature extracted is perfect, and different its gray level co-occurrence matrixes difference of image is very big.Close grain scab, its
Structure is closeer, arranges consolidation, and gray scale is smoother;Open grain scab, its structure is sparse, and gray value four dissipates distribution.Based on synthesis
Method be substantially Structure Method and statistic law combination, by being contrasted with the value in texture searching, to show its superior function,
It is obtained to measure.With reference to sunflower grown in field environment, the present invention chooses the method based on gray level co-occurrence matrixes to disease geo-radar image
Scab texture carry out extraction research, analyze the feature of its energy, entropy, the moment of inertia, correlation and unfavourable balance away from 5 characteristic parameters
Value, by carefully comparing the textural characteristics parameter for choosing optimal characteristics parameter as disease recognition.
The identifying and diagnosing of disease:In data mining processing, sorting technique is using most extensive, most efficient method, tradition
Data classification have clustering algorithm, Bayesian Classification Arithmetic, SVMs etc., but these methods nicety of grading often not
Height, and it is also easy to over-fitting problem occur.Thus, whether many experts and scholars expect can be by multiple model sets to one
Rise, one new sorter model of composition improves the classification precision of prediction of data, just have today we have seen that combination or
The processing method of classifiers combination.It is used for training data firstly the need of a base categories model is set up, to what is trained
Data are voted, and determine which kind of it belongs to, by that analogy, obtain multiple base disaggregated models, eventually through multiple point
Class, which predicts the outcome, determines its classification.
Random forest (RF) is exactly a kind of such disaggregated model, and classification performance is improved by way of combination.By many
Decision tree combines generation grader, determines that it predicts the outcome by the way of multiple backing up, reaches the purpose of classification, as
A kind of Statistical Learning Theory, random forest has good tolerance for the classification and recurrence of sample.
The essence of decision tree (decision tree) is conducted interviews differentiation according to condition for a certain problem, through sieving layer by layer
After choosing, an answer is given, is a tree (y-bend or multiway tree).Each nonleaf node on tree represents one
Test on individual sample characteristics attribute, each branch represents output of this characteristic attribute in some codomain, and each
Leaf node can all deposit the classification of a data.The decision process of decision tree is started to walk from root node, and layering successively is carried out, by surveying
The data characteristics of sample determines which kind of it belongs in examination test set, is exported according to generic, until test point is reached
Leaf node, test terminates, and final classification result is based on the classification deposited in the leaf node where last.
Decision tree method is easier to understand compared to methods such as SVMs, because when using decision tree, be not required at all
Priori is wanted, but easily there is over-fitting in decision tree method, so when carrying out data analysis using decision tree, it is necessary to elder generation
Cut operator is carried out to decision tree.
Decision tree beta pruning has two methods:Beta pruning and post pruning in advance, however, being which kind of pruning method, in processing
The complexity of algorithm will be increased.
Post-class processing (CART) includes two parts:Classification tree and regression tree.This two parts is all applied to target variable
Data analysis, but the scope of application is different, and classification tree is applied to the data analysis that target variable is class variable, regression tree application
In target variable be the data analysis of continuous variable.In post-class processing, its fragmentation criterion is gini index, gini index energy
It is enough to pick out the attribute for reducing the data degree of disorder.CART trees are when building, by being marked off under different criterions
Result good and bad performance come choose its division attribute.
Random forest (RF) is a new classifiers combination model, with very high classification performance.Current random forest
Apply to the various aspects such as medical image, remote sensing images, by the substantial amounts of analysis and research of scholar and facts have proved and draw:It is random gloomy
Woods has a very high predictablity rate, and noise for burst and the exceptional value that occurs can be received well, and not
Easily there is the phenomenon of overfitting, random forest method is easily understood, and simple and clear draws a conclusion as desired, as instantly
One of information processing most popular research field.
Random forest classifies (RFC) as classifiers combination model, and its basic thought for classifying processing is:In original training
Multiple sample generation subsample collection, the sample size of subsample collection and original training set sample size one are randomly selected in sample set
Sample, sets up corresponding decision tree by subsample collection and is used as test sample, obtain a variety of classification results, multiple according to classification results
Its final classification result is voted, classification is completed.
Sunflower sample image cannot be directly used to classifier training, so needing to extract and can describe from image
The parameter of sunflower feature is used to train, and has been introduced before feature extraction, and the quantity of primitive character is a lot, how with less
And the parameter of essence accurately describes sunflower classification again, it is necessary to feature is carried out preferably, so be able to can both to carry with exact classification
High-class operation ratio.
Embodiment 2
In scab segmentation:Comparative analysis each model in color image color space, by contrast, and according to crop field
The actual conditions of the image collected, final choice is split to the sensitive hsv color space of human eye vision as sunflower disease
Color space.By comparing the processing method that five kinds of images are split, and consider its respective advantage and disadvantage, final choice will
K- means clustering algorithms and watershed algorithm combine segmentation of the common realization to the colored disease geo-radar image of sunflower blade, obtain
Preferable scab segmentation figure picture, accurate good base image is provided for sunflower leaf portion disease geo-radar image signature analysis.
On identifying and diagnosing:On the basis of being analysed in depth to Disease Characters, choose using random forest method to sunflower
Disease carries out diagnosis identification, and through simulation results show, this method accurate can be diagnosed to be four kinds of disease species of sunflower leaf portion.
In disease geo-radar image pretreatment:The disease leaf image got by camera, inevitably because of various factors
The influence of (noise) and occur pixel qualities it is poor, it is smudgy situations such as, this be unfavorable for behind us operation processing.Therefore,
In order to be smoothed out operation below, and identification error is smaller, and what is primarily done is exactly to carry out enhancing processing to image, is made
Image information is more complete.
Purpose based on the processing of morphologic weight adaptive denoising morphological image is reasonable by means of collection
By analyzing image and being recognized, gone to measure information important in image with element, and these elements will have certain shape
State structure, the algorithm has following characteristics:
The holding of image information:In morphological image process, it can be selected by the geometric properties information of existing target
Select based on morphologic morphological filter, so both can effectively be filtered when being handled, image Central Plains can be kept again
Some Information invariabilities.
The extraction of image border:Theoretical based on mathematical morphology is handled, and can be prevented effectively to a certain extent
The interference of noise, has higher stability for the technology of differential operator.Extracted and schemed using Mathematical Morphology technology
As edge, its structure is relatively smooth, and can more embody the detailed information of image.
The extraction of image framework:, can be preferably with set using morphological method when being extracted to image framework
The characteristics of computing, it is to avoid substantial amounts of breakpoint occur, skeleton is also more continuous.
The efficiency of image procossing:Mathematical Morphology Method processing image easily can enter using the technology of parallel processing
Row set computing.In calculating process, efficiency is significantly improved, and is also realized very well for the processing of hardware.Mathematical Morphology becomes
Change and be divided into two kinds of forms of binary transform and greyscale transformation by application scenarios.Binary transform is for the collection of collection disease geo-radar image and pre-
Conjunction problem is handled, greyscale transformation is directed to function problem.Basic metamorphosis includes burn into expansion, opening operation, closed operation.Open
Closed operation is the computing that corrosion is combined with expansion phase, with more preferable treatment effect.
Test result indicates that, denoising is filtered to image if only by connecting, then often there is certain limitation
Property, and also maintain more obvious noise in result images after processing, but image filtering is carried out by parallel filter and go
Obtained result PSNR (Y-PSNR) value of making an uproar is higher, and result images in visual effect than only carrying out series filtering
The result that device denoising is obtained is even more ideal.Therefore, the present invention chooses parallel filter and completed to sunflower leaf portion disease geo-radar image
Denoising.
Embodiment 3
Disease geo-radar image is split:The purpose of image segmentation is very clear and definite, is exactly in order to by extracting section interested in image
Out it is used to research handle, its basis is image recognition, and premise is then to carry out disease recognition to the disease geo-radar image of input, specifically
Apply in example.Image splits the committed step as sunflower disease geo-radar image subsequent treatment, it whether accurate directly right
The identification of disease has a very big impact effect.The key of the algorithm of image segmentation is the region part in figure, according to each area
Difference between domain is split.Each image segmentation algorithm is not comprehensive, and its dividing method of target difference can
It can be not quite similar.Thus, the multiple images protruded for features only cannot be split using same process.
So, it will meet following condition for each partitioning algorithm:One pixel must only belong to a sub-regions (i.e.
One pixel can not belong to two regions simultaneously, and region is not overlapping), segmentation also must be definitely sum thoroughly;For same
It has a same or like attribute for pixel in individual region, and pixel its attribute in different zones differs greatly.
Pass through contrast:Scab effect that watershed algorithm is partitioned into preferably, but be vulnerable in processing procedure noise and
, there is segmentation fault phenomenon in the influence of the factors such as quantization error;Clustering algorithm will first give a cluster centre before cluster, no
Same cluster centre can cause different segmentation effects, and clusters number must give in advance, not so easily segmentation fault.Synthesis is examined
Consider the advantage and disadvantage of algorithm above, used when the present invention is split for coloured image and mutually tie watershed algorithm with clustering algorithm
The method of conjunction.
1. K- means clustering algorithms:
As shown in Figure 1, cluster is a kind of method that system is classified for data, according between of a sort data
With maximized similitude, there is the principle of maximized otherness between inhomogeneity data.As in statistical method without prison
A kind of division of progress is superintended and directed, sample need not be trained in processing procedure.And K- means clustering algorithms are all
It is that degree of error is minimum in clustering algorithm, a kind of result method the most accurate.
Drawn after experiment, choose K- means clustering algorithms and clustering behaviour is carried out to sunflower leaf portion disease geo-radar image target
Make, when using 2 clusters, the scab of blade significantly can make a distinction with background.
2. watershed algorithm:
, it is necessary to note some following requirement when being operated using watershed algorithm:
(1) image is carried out smooth
There is an extreme value in the region of image, effectively extract the value for convenience, it is necessary to be carried out to original image
Enhancing is handled, and major way is to use image gray processing, extracts edge, and disappeared with morphological dilations, burn into open and close operator method
Except various noises present in image, it is to avoid because the influence of noise causes fault segmentation phenomenon occur in cutting procedure.
(2) zoning distance
The pixel value of image is higher at gradient edge, and relatively low elsewhere, in the two poles of the earth distribution.In perfect condition
Under, watershed transform will generate several watershed separator bars along the border of the target in image, in separator bar both sides gray value
Differ greatly.Calculate the amplitude of image gradient, using linear filter method or Morphological Gradient method calculate the region of image away from
From.
(3) watershed is marked
Because there are many minimum points in image, directly image is split using watershed method, can be produced
Split phenomenon, that is, being partitioned into many does not have the discreet region of any effect for graphical analysis.Labeling method has two kinds:Inside mark
Note, external label.Central concept for the not positive image of gradient, is modified processing by introducing mark, thus made
The minimum value of a certain position is only present in the region by mark, it is to avoid small range segmentation phenomenon occur.And by setting threshold
Value is filtered to pixel value, removes the region less than threshold value.
Algorithm combination segmentation result and analysis, to K- means clustering algorithms are combined into dividing processing with watershed algorithm
There is over-segmentation phenomenon when avoiding the independent dividing processing of each algorithm in disease geo-radar image, the mode so combined, preferably retains
The information of scab.It can see from experiment process result, the sunflower bacterial leaf spot scab after this algorithm process
Separated well with background, the target being partitioned into is independent and clear.
Embodiment 4
Sunflower leaf portion disease recognition grader, the foundation point three below step of grader are built using random forest method
Suddenly:
(1) currently without sunflower Sample Storehouse, so being acquired firstly the need of to sunflower sample, sample is manually set up
Storehouse.
(2) every decision tree in forest is constructed by obtained training sample of sampling every time, the classification of leaf node
Criterion is determined by putting to comparative approach.The present invention chooses sunflower color characteristic and textural characteristics, and totally 19 characteristic parameters are set up
Decision tree, during achievement, is required for from these characteristic parameters randomly selecting different types of characteristic parameter each time,
It is used as the alternative features of this decision-making.
(3) the decision tree number for needing to set up is determined according to system requirements, the tree established is entered using random forest method
Row classification prediction, classification results are regarded as by the class label for predicting the outcome most, i.e., by being ground to the analysis of sunflower characteristic parameter
Study carefully, automatic identification is carried out suitable for equipment so as to extract, and be difficult the characteristic parameter that is influenceed by picture quality, by entering to parameter
Row training and the grader for designing needs, finally output predict the outcome.
Sunflower disease Classification and Identification result:Take sunflower powdery mildew, bacterial leaf spot, black spot, downy mildew disease
Each 13 width of image, 10 width disease geo-radar images are used as training sample, and 3 width disease geo-radar images are used as test sample, according to color characteristic, texture
The parameter training grader that feature preferably goes out.Increase number of samples, make grader nicety of grading higher, disease geo-radar image sample respectively takes
100, wherein 60 are used as training sample, 40 are used as test sample.Sunflower disease blade is entered using random forest method
Row identification, discrimination is up to more than 92.5%, and recognition effect is preferable.Illustrate have using random forest method identification sunflower disease
Certain researching value, but in actual applications, it is numerous and diverse due to sunflower disease species, it need to further increase training sample
Storehouse, increases its contrast range, improves the accuracy rate of sunflower disease recognition.
This research considers its respective advantage and disadvantage by comparing the processing method that five kinds of different images are split,
K- means clustering algorithms and watershed algorithm are combined common realize to the colored disease geo-radar image of sunflower blade by final choice
Segmentation, obtain preferable scab segmentation figure picture.
Finally it should be noted that:The preferred embodiments of the present invention are the foregoing is only, are not intended to limit the invention,
Although the present invention is described in detail with reference to the foregoing embodiments, for those skilled in the art, it still may be used
To be modified to the technical scheme described in foregoing embodiments, or equivalent substitution is carried out to which part technical characteristic.
Within the spirit and principles of the invention, any modification, equivalent substitution and improvements made etc., should be included in the present invention's
Within protection domain.
Claims (7)
1. a kind of sunflower disease recognition method based on random forest method, including the common four kinds of diseases of sunflower leaf portion:White powder
Disease, bacterial leaf spot, black spot, downy mildew, and it is used as research object, it is characterised in that:
A:Disease geo-radar image is gathered.The color that the leaf image color collected will try one's best close to leaf in itself.
B:Disease geo-radar image is pre-processed.In the case where combining actual conditions, using the pretreatment side for being suitable for the identification of sunflower disease geo-radar image
Method, strengthens image effect, and go based on morphologic weight adapting to image using the histogram equalization defogging algorithm in spatial domain
Method of making an uproar carries out denoising to sunflower leaf portion disease geo-radar image.
C:Disease geo-radar image is split.The various methods of contrast images segmentation, choose suitable optimal color images by analysis
Method, it is colored with the method segmentation sunflower leaf diseases that watershed algorithm is combined using K- means clustering algorithms after experiment
Image.
D:Disease geo-radar image feature extraction.There is obvious change by sunflower color and texture, thus extract disease geo-radar image
Color characteristic, textural characteristics parameter are studied, and by the in-depth analysis to characteristic parameter, preferably go out suitable image recognition
Best features parameter.
E:The identifying and diagnosing of disease.With reference to the color characteristic and textural characteristics parameter extracted, using random forest method to day
Certain herbaceous plants with big flowers disease carries out last diagnosis identification.
F:Hardware environment:Free background board, conventional background color is set to black, white two kinds, and subsequent treatment uses color part, adopts
With white background, and using blank sheet of paper as material, camera angle is adjusted during shooting, under the conditions of natural lighting, the leaf photographed is made as far as possible
Color of the piece color close to itself.
2. a kind of sunflower disease recognition method based on random forest method according to claim 1, it is characterised in that institute
Disease geo-radar image collection is stated, targetedly collecting part sunflower leaf diseases associated picture, incoming computer is subsequently located
Reason, it is ensured that the automatic recognition system of sunflower leaf diseases recognizes accurate and effective to disease figure, will invite professional plant protection personage
Accurate discrimination is carried out to the sunflower leaf diseases sample collected.
3. a kind of sunflower disease recognition method based on random forest method according to claim 1, it is characterised in that institute
Disease geo-radar image pretreatment is stated, using computer image processing technology, through analyzing sunflower leaf diseases own characteristic and comparing
All kinds of color image filtering methods, image effect is strengthened using the histogram equalization defogging algorithm in spatial domain.
4. a kind of sunflower disease recognition method based on random forest method according to claim 1, it is characterised in that institute
State disease geo-radar image to split, each model in comparative analysis color image color space, by contrast, and collected according to crop field
The actual conditions of image, the colour that final choice is split to the sensitive hsv color space of human eye vision as sunflower disease is empty
Between.By comparing the processing method that five kinds of images are split, and consider its respective advantage and disadvantage, final choice gathers K- averages
Class algorithm and watershed algorithm combine segmentation of the common realization to the colored disease geo-radar image of sunflower blade, obtain scab segmentation
Image.
5. a kind of sunflower disease recognition method based on random forest method according to claim 1, it is characterised in that institute
Disease geo-radar image feature extraction is stated, color characteristic and textural characteristics with notable difference characteristic is chosen and is analyzed and researched, passed through
The color characteristic and textural characteristics of four kinds of disease scab images are analyzed, the hsv color space of four kinds of disease scabs is therefrom extracted
Tri- color components of H, S, V nine color characteristic parameters of first moment, second moment, third moment and the ash of d=1 in textural characteristics
Spend co-occurrence matrix energy, entropy, the moment of inertia, correlation and unfavourable balance away from ten textural characteristics parameters of average and standard deviation, altogether
The distinguishing rule that 19 characteristic parameters are diagnosed as disease recognition, is that next step disease screening identification is got ready.
6. a kind of sunflower disease recognition method based on random forest method according to claim 1, it is characterised in that institute
The identifying and diagnosing of disease is stated, on the basis of being analysed in depth to Disease Characters, with reference to this research practical study condition, is chosen herein
Diagnosis identification is carried out to sunflower disease using random forest method.
7. a kind of sunflower disease recognition method based on random forest method according to claim 5, it is characterised in that institute
Disease geo-radar image feature extraction is stated, when starting with for sunflower leaf portion scab color, textural characteristics, using Computer Image Processing skill
Knowledge in terms of art, integrated use image procossing, pattern-recognition, computer vision, Plant Pathology is to sunflower leaf portion disease
Evil carries out automatic identification, so as to deepen the understanding to sunflower disease.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710606339.XA CN107330892A (en) | 2017-07-24 | 2017-07-24 | A kind of sunflower disease recognition method based on random forest method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710606339.XA CN107330892A (en) | 2017-07-24 | 2017-07-24 | A kind of sunflower disease recognition method based on random forest method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107330892A true CN107330892A (en) | 2017-11-07 |
Family
ID=60199840
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710606339.XA Pending CN107330892A (en) | 2017-07-24 | 2017-07-24 | A kind of sunflower disease recognition method based on random forest method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107330892A (en) |
Cited By (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108428067A (en) * | 2018-04-09 | 2018-08-21 | 东华大学 | A kind of printing quality analysis of Influential Factors method based on historical data |
CN108564092A (en) * | 2018-04-12 | 2018-09-21 | 内蒙古工业大学 | Sunflower disease recognition method based on SIFT feature extraction algorithm |
CN108573502A (en) * | 2018-03-06 | 2018-09-25 | 安徽大学 | A kind of method at the angles automatic measurement Cobb |
CN108596254A (en) * | 2018-04-25 | 2018-09-28 | 福州大学 | The rigid bamboo poison moth hazard detection method of coupled vane multilist sign |
CN108665431A (en) * | 2018-05-16 | 2018-10-16 | 南京信息工程大学 | Fractional order image texture Enhancement Method based on K- mean clusters |
CN108921835A (en) * | 2018-06-28 | 2018-11-30 | 深圳市诚品鲜智能科技股份有限公司 | Crop control method and relevant apparatus and storage medium based on machine vision |
CN109635846A (en) * | 2018-11-16 | 2019-04-16 | 哈尔滨工业大学(深圳) | A kind of multiclass medical image judgment method and system |
CN110188806A (en) * | 2019-05-21 | 2019-08-30 | 华侨大学 | A kind of large circle machine fabric defects detection and classification method based on machine vision |
CN110310291A (en) * | 2019-06-25 | 2019-10-08 | 四川省农业科学院农业信息与农村经济研究所 | A kind of rice blast hierarchy system and its method |
CN110363119A (en) * | 2019-06-28 | 2019-10-22 | 广东中烟工业有限责任公司 | Based on wavelet transformation-random forests algorithm tobacco mildew method for quickly identifying |
CN110597172A (en) * | 2019-09-26 | 2019-12-20 | 无锡利信能源科技有限公司 | Resin regeneration real-time online measurement and control system based on machine vision and working method thereof |
CN110926430A (en) * | 2019-11-22 | 2020-03-27 | 海南省林业科学研究所 | Air-ground integrated mangrove forest monitoring system and control method |
CN111429455A (en) * | 2020-04-24 | 2020-07-17 | 杭州皓京云信息技术有限公司 | Cotton disease identification method and system based on rough set and BP neural network |
CN111523516A (en) * | 2020-05-14 | 2020-08-11 | 宁波工程学院 | Forest harmful wood identification method |
CN112749766A (en) * | 2021-02-02 | 2021-05-04 | 上海第二工业大学 | Method for identifying and classifying agaricus bisporus brown spot |
CN113011355A (en) * | 2021-03-25 | 2021-06-22 | 东北林业大学 | Pine wood nematode disease image recognition detection method and device |
CN113177911A (en) * | 2021-04-13 | 2021-07-27 | 沈阳大学 | Method for nondestructive evaluation of ozone sensitivity of plants by leaves |
CN113261428A (en) * | 2021-06-18 | 2021-08-17 | 湖北省农业科学院经济作物研究所 | Whole-process quality control method for machine-harvested cotton |
CN113284098A (en) * | 2021-05-11 | 2021-08-20 | 中国农业大学 | Method and device for classifying and judging nitrogen content deficiency of crops and electronic equipment |
CN115170542A (en) * | 2022-07-26 | 2022-10-11 | 中国农业科学院农业信息研究所 | Potato early-late blight classification model construction method based on GLCM feature extraction |
CN115297243A (en) * | 2022-08-04 | 2022-11-04 | 江西省农业科学院农业工程研究所 | Rice image recognition and nitrogen fertilizer quantitative diagnosis system and method |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102208099A (en) * | 2011-05-30 | 2011-10-05 | 华中科技大学 | Illumination-change-resistant crop color image segmentation method |
US20130136312A1 (en) * | 2011-11-24 | 2013-05-30 | Shih-Mu TSENG | Method and system for recognizing plant diseases and recording medium |
CN104598908A (en) * | 2014-09-26 | 2015-05-06 | 浙江理工大学 | Method for recognizing diseases of crop leaves |
CN106777845A (en) * | 2017-03-22 | 2017-05-31 | 南京农业大学 | The method that sensitive parameter builds wheat leaf blade powdery mildew early monitoring model is extracted based on subwindow rearrangement method |
-
2017
- 2017-07-24 CN CN201710606339.XA patent/CN107330892A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102208099A (en) * | 2011-05-30 | 2011-10-05 | 华中科技大学 | Illumination-change-resistant crop color image segmentation method |
US20130136312A1 (en) * | 2011-11-24 | 2013-05-30 | Shih-Mu TSENG | Method and system for recognizing plant diseases and recording medium |
CN104598908A (en) * | 2014-09-26 | 2015-05-06 | 浙江理工大学 | Method for recognizing diseases of crop leaves |
CN106777845A (en) * | 2017-03-22 | 2017-05-31 | 南京农业大学 | The method that sensitive parameter builds wheat leaf blade powdery mildew early monitoring model is extracted based on subwindow rearrangement method |
Non-Patent Citations (3)
Title |
---|
张延丽,李春贵: ""视频车辆检测中改进的阈值分割算法研究"", 《广西科技大学学报》 * |
李小林,景林,黄世国,黄习培: ""一种两阶段彩色图像分割方法的原木识别"", 《福建农林大学学报( 自然科学版)》 * |
田杰: ""基于图像分析的小麦叶部病害识别方法研究"", 《中国博士学位论文全文数据库 信息科技辑》 * |
Cited By (29)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108573502A (en) * | 2018-03-06 | 2018-09-25 | 安徽大学 | A kind of method at the angles automatic measurement Cobb |
CN108573502B (en) * | 2018-03-06 | 2021-07-06 | 安徽大学 | Method for automatically measuring Cobb angle |
CN108428067A (en) * | 2018-04-09 | 2018-08-21 | 东华大学 | A kind of printing quality analysis of Influential Factors method based on historical data |
CN108564092A (en) * | 2018-04-12 | 2018-09-21 | 内蒙古工业大学 | Sunflower disease recognition method based on SIFT feature extraction algorithm |
CN108596254A (en) * | 2018-04-25 | 2018-09-28 | 福州大学 | The rigid bamboo poison moth hazard detection method of coupled vane multilist sign |
CN108665431A (en) * | 2018-05-16 | 2018-10-16 | 南京信息工程大学 | Fractional order image texture Enhancement Method based on K- mean clusters |
CN108921835A (en) * | 2018-06-28 | 2018-11-30 | 深圳市诚品鲜智能科技股份有限公司 | Crop control method and relevant apparatus and storage medium based on machine vision |
CN109635846A (en) * | 2018-11-16 | 2019-04-16 | 哈尔滨工业大学(深圳) | A kind of multiclass medical image judgment method and system |
CN110188806A (en) * | 2019-05-21 | 2019-08-30 | 华侨大学 | A kind of large circle machine fabric defects detection and classification method based on machine vision |
CN110310291A (en) * | 2019-06-25 | 2019-10-08 | 四川省农业科学院农业信息与农村经济研究所 | A kind of rice blast hierarchy system and its method |
CN110363119A (en) * | 2019-06-28 | 2019-10-22 | 广东中烟工业有限责任公司 | Based on wavelet transformation-random forests algorithm tobacco mildew method for quickly identifying |
CN110597172B (en) * | 2019-09-26 | 2020-12-01 | 无锡利信能源科技有限公司 | Resin regeneration real-time online measurement and control system based on machine vision and working method thereof |
CN110597172A (en) * | 2019-09-26 | 2019-12-20 | 无锡利信能源科技有限公司 | Resin regeneration real-time online measurement and control system based on machine vision and working method thereof |
CN110926430A (en) * | 2019-11-22 | 2020-03-27 | 海南省林业科学研究所 | Air-ground integrated mangrove forest monitoring system and control method |
CN111429455A (en) * | 2020-04-24 | 2020-07-17 | 杭州皓京云信息技术有限公司 | Cotton disease identification method and system based on rough set and BP neural network |
CN111523516A (en) * | 2020-05-14 | 2020-08-11 | 宁波工程学院 | Forest harmful wood identification method |
CN111523516B (en) * | 2020-05-14 | 2024-02-02 | 宁波工程学院 | Forest harmful wood identification method |
CN112749766A (en) * | 2021-02-02 | 2021-05-04 | 上海第二工业大学 | Method for identifying and classifying agaricus bisporus brown spot |
CN113011355B (en) * | 2021-03-25 | 2022-10-11 | 东北林业大学 | Pine wood nematode disease image recognition detection method and device |
CN113011355A (en) * | 2021-03-25 | 2021-06-22 | 东北林业大学 | Pine wood nematode disease image recognition detection method and device |
CN113177911A (en) * | 2021-04-13 | 2021-07-27 | 沈阳大学 | Method for nondestructive evaluation of ozone sensitivity of plants by leaves |
CN113284098A (en) * | 2021-05-11 | 2021-08-20 | 中国农业大学 | Method and device for classifying and judging nitrogen content deficiency of crops and electronic equipment |
CN113284098B (en) * | 2021-05-11 | 2024-01-19 | 中国农业大学 | Method and device for classifying and judging nitrogen content deficiency of crops and electronic equipment |
CN113261428B (en) * | 2021-06-18 | 2022-04-12 | 湖北省农业科学院经济作物研究所 | Whole-process quality control method for machine-harvested cotton |
CN113261428A (en) * | 2021-06-18 | 2021-08-17 | 湖北省农业科学院经济作物研究所 | Whole-process quality control method for machine-harvested cotton |
CN115170542A (en) * | 2022-07-26 | 2022-10-11 | 中国农业科学院农业信息研究所 | Potato early-late blight classification model construction method based on GLCM feature extraction |
CN115170542B (en) * | 2022-07-26 | 2023-02-28 | 中国农业科学院农业信息研究所 | Potato early-late blight classification model construction method based on GLCM feature extraction |
CN115297243A (en) * | 2022-08-04 | 2022-11-04 | 江西省农业科学院农业工程研究所 | Rice image recognition and nitrogen fertilizer quantitative diagnosis system and method |
CN115297243B (en) * | 2022-08-04 | 2023-05-23 | 江西省农业科学院农业工程研究所 | Rice image recognition and nitrogen fertilizer quantitative diagnosis system and method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107330892A (en) | A kind of sunflower disease recognition method based on random forest method | |
CN104598908B (en) | A kind of crops leaf diseases recognition methods | |
CN107977671A (en) | A kind of tongue picture sorting technique based on multitask convolutional neural networks | |
CN107016405B (en) | A kind of pest image classification method based on classification prediction convolutional neural networks | |
Gavhale et al. | An overview of the research on plant leaves disease detection using image processing techniques | |
CN105718945B (en) | Apple picking robot night image recognition method based on watershed and neural network | |
CN102842032B (en) | Method for recognizing pornography images on mobile Internet based on multi-mode combinational strategy | |
CN108734719A (en) | Background automatic division method before a kind of lepidopterous insects image based on full convolutional neural networks | |
CN107194937A (en) | Tongue image partition method under a kind of open environment | |
CN102214306B (en) | Leaf disease spot identification method and device | |
Patil et al. | Grape leaf disease detection using k-means clustering algorithm | |
CN109344699A (en) | Winter jujube disease recognition method based on depth of seam division convolutional neural networks | |
CN101383008A (en) | Image classification method based on visual attention model | |
CN108537239A (en) | A kind of method of saliency target detection | |
CN107341813A (en) | SAR image segmentation method based on structure learning and sketch characteristic inference network | |
CN108288279A (en) | A kind of article discriminating method based on the extraction of x-ray image foreground target | |
Ouyang et al. | The research of the strawberry disease identification based on image processing and pattern recognition | |
CN108710916A (en) | The method and device of picture classification | |
CN106874825A (en) | The training method of Face datection, detection method and device | |
CN109190456A (en) | Pedestrian detection method is overlooked based on the multiple features fusion of converging channels feature and gray level co-occurrence matrixes | |
CN107154044A (en) | A kind of dividing method of Chinese meal food image | |
CN108364011A (en) | PolSAR image multi-stage characteristics extract and unsupervised segmentation method | |
Ji et al. | Apple color automatic grading method based on machine vision | |
Thella et al. | A comparative analysis on machine learning models for accurate identification of medical plants | |
Dong et al. | Dragon fruit disease image segmentation based on FCM algorithm and two-dimensional OTSU 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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171107 |
|
RJ01 | Rejection of invention patent application after publication |