CN106709505A - Dictionary learning-based grain type classification method and system for corn grains - Google Patents
Dictionary learning-based grain type classification method and system for corn grains Download PDFInfo
- Publication number
- CN106709505A CN106709505A CN201611059052.1A CN201611059052A CN106709505A CN 106709505 A CN106709505 A CN 106709505A CN 201611059052 A CN201611059052 A CN 201611059052A CN 106709505 A CN106709505 A CN 106709505A
- Authority
- CN
- China
- Prior art keywords
- dictionary
- classification
- training sample
- type
- corn kernel
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2136—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on sparsity criteria, e.g. with an overcomplete basis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/28—Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a dictionary learning-based grain type classification method and system for corn grains. The method comprises the steps of taking grain images of corn grains with different grain types as training samples; extracting eigenvectors used for performing grain type classification in the training samples; combining the eigenvectors to obtain a training sample dictionary; performing multi-type dictionary learning on the training sample dictionary based on a dictionary training algorithm K-SVD; and taking the types with small reconstruction errors as classification types, and performing sparse classification on the grain types of the corn grains according to the classification types. The system comprises a training sample obtaining unit, an eigenvector extraction unit, a training sample dictionary obtaining unit, a dictionary learning unit and a sparse classification unit. According to the method and the system, the corn grains can be subjected to grain type classification quickly and accurately, and an accurate and reliable basis is provided for corn yield assessment.
Description
Technical field
The present invention relates to modern agriculture identification technology field, and in particular to a kind of grain of the corn kernel based on dictionary learning
Type sorting technique and system.
Background technology
The grain type of corn ear is to assess one of important phenotypic parameter of corn yield, it and corn variety, Relationship with Yield
Closely.Grain type is broadly divided into Hard grain type, dent type, three kinds of half dent type, and different grain type seeds are in spies such as color, texture, shapes
Levy and have differences.
The current detection to corn fringe portion seed grain type is mainly by artificial sense in seed coat color, profile, sinking degree
Overall merit is carried out as classification foundation, there is the shortcomings of efficiency is low, subjective error is big.Based on species test material can be retained, carry
Breeding process high.
Therefore, how to design a kind of method for type classification that accurately can be carried out to corn kernel is this area urgently to be resolved hurrily
Problem.
The content of the invention
For defect of the prior art, the present invention provides a kind of grain type classification side of the corn kernel based on dictionary learning
Method and system, can quickly, accurately carry out a type classification, for the assessment of corn yield is provided accurately and securely to corn kernel
Basis.
In order to solve the above technical problems, the present invention provides following technical scheme：
On the one hand, the invention provides a kind of grain type sorting technique of corn kernel based on dictionary learning, including：
Step 1. obtains the seed image of each corn kernel of different grain types, and using the seed image as training sample
This；
Step 2. is used to carry out a characteristic vector for type classification in extracting each training sample；
Step 3. combines the characteristic vector, obtains training sample dictionary；
Step 4. carries out multi-class dictionary learning based on dictionary training algorithm K-SVD to the training sample dictionary；
Step 5. enters the small classification of reconstructed error as class categories according to the class categories to the grain type of corn kernel
The sparse classification of row.
Further, also include before the step 1：
Step A. gathers the corn ear of different grain types respectively, and generates the fringe portion image of corn ear；
Corn ear fringe portion core in the step B. extractions seed image is used as region of interest ROI；
Step C. extracts the red channel information of the ROI, and carries out denoising to the ROI according to median filtering method；
Step D. according to Otsu algorithm OTSU to denoising after the ROI split, obtain the seed of the corn kernel
Grain image.
Further, the characteristic vector in the step 2, including：
The characteristic vector includes：Seed energy, contrast, entropy, auto-correlation, unfavourable balance square, area, elongation, circularity,
Rectangular degree, L * channel average, a passages average and b passage averages.
Further, the step 2 includes：Extract in each training sample for carrying out a characteristic vector for type classification
Step 2-1. extracts seed L * channel average, a passages average and b passage average color characteristics as carrying out grain
The classification foundation of type classification；
Step 2-2. extracts seed energy, contrast, entropy, auto-correlation and unfavourable balance square as carrying out what a type was classified
Classification foundation；
Step 2-3. extracts seed area, elongation, circularity and rectangular degree as carrying out the classification that a type is classified
Foundation；
Step 2-4. integrates all classification foundations, obtains the characteristic vector.
Further, the step 2-1 includes：
Step a. extracts single corn kernel cromogram and segmentation binary picture；
Corn kernel image RGB color is converted to Lab color spaces by step b.；
Step c. calculates corn kernel L * channel, a passages, b passage averages respectively using binary picture as template.
Further, the step 3 includes：
Each characteristic vector is divided into different training sample dictionaries according to the different grain type of the corn kernel.
Further, the grain type includes half dent type, dent type and Hard grain type.
Further, the step 4 includes：
Step 4-1. extracts the data sample of the corn kernel of different grain types in each training sample dictionary, and
Using the data sample as dictionary learning foundation；
Step 4-2. is normalized to the training sample dictionary and data sample；
Step 4-3. according to the data sample after normalized, to normalized after the training sample dictionary enter
The multi-class dictionary training algorithm K-SVD dictionary learnings of row.
Further, the step 5 includes：
Step 5-1. carries out sparse according to the training sample dictionary after the dictionary learning to the test sample of corn kernel
Solve, and according to sparse solution coefficient, the reconstructed error of all categories of corn kernel is calculated respectively；
Step 5-2. using the minimum classification of middle reconstructed error of all categories as class categories, and according to the class categories to jade
The grain type of rice seed carries out sparse classification.
On the other hand, present invention also offers a kind of grain type categorizing system of corn kernel based on dictionary learning, including：
Training sample acquiring unit, the seed image of each corn kernel for obtaining different grain types, and by the seed
Image is used as training sample；
Characteristic vector pickup unit, for extracting in each training sample for carrying out a characteristic vector for type classification；
Training sample dictionary acquiring unit, for combining the characteristic vector, obtains training sample dictionary；
Dictionary learning unit, it is multi-class for being carried out to the training sample dictionary based on dictionary training algorithm K-SVD
Dictionary learning；
Sparse taxon, for using the small classification of reconstructed error as class categories, according to the class categories to corn
The grain type of seed carries out sparse classification.
As shown from the above technical solution, the grain type classification side of a kind of corn kernel based on dictionary learning of the present invention
Method and system, the method is using the seed image of each corn kernel of different grain types as training sample；In extracting each training sample
For carrying out a characteristic vector for type classification；Assemblage characteristic vector, obtains training sample dictionary；Based on dictionary training algorithm K-
SVD carries out multi-class dictionary learning to training sample dictionary；Using the small classification of reconstructed error as class categories, according to this point
Class classification carries out sparse classification to the grain type of corn kernel.The system includes training sample acquiring unit, characteristic vector pickup list
Unit, training sample dictionary acquiring unit, dictionary learning unit and sparse taxon；The present invention can quickly, it is accurate to corn
Seed carries out a type classification, for the assessment of corn yield provides basis accurately and securely；Non- threshing fruit ear is based on simultaneously
The grain type detection of fringe portion seed can retain species test material, improve breeding process.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
The accompanying drawing to be used needed for having technology description is briefly described, it should be apparent that, drawings in the following description are the present invention
Some embodiments, for those of ordinary skill in the art, on the premise of not paying creative work, can also basis
These accompanying drawings obtain other accompanying drawings.
Fig. 1 is a kind of schematic flow sheet of the grain type sorting technique of corn kernel based on dictionary learning of the invention；
Fig. 2 be in sorting technique of the invention before step 100 the step of A00 to D00 schematic flow sheet；
Fig. 3 is a kind of schematic flow sheet of specific embodiment of step 200 in control method of the invention；
Fig. 4 is a kind of schematic flow sheet of specific embodiment of step 400 in control method of the invention；
Fig. 5 is a kind of schematic flow sheet of specific embodiment of step 500 in sorting technique of the invention；
Fig. 6 is a kind of schematic flow sheet of application example of sorting technique of the invention；
Fig. 7 is a kind of structural representation of the grain type categorizing system of corn kernel based on dictionary learning of the invention.
Specific embodiment
To make the purpose, technical scheme and advantage of the embodiment of the present invention clearer, below in conjunction with the embodiment of the present invention
In accompanying drawing, clear, complete description is carried out to the technical scheme in the embodiment of the present invention, it is clear that described embodiment is
A part of embodiment of the present invention, rather than whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art
The every other embodiment obtained under the premise of creative work is not made, belongs to the scope of protection of the invention.
The principle of rarefaction representation grader SRC (Sparse Representation Classifier) is based on sparse table
Show, it regards in same linear subspaces same class sample as, and test sample can be by similar other sample optimum linearity tables
Show.Non-test sample class tends to 0 to the linear contributions rate of test sample simultaneously, therefore this linear solution of total training sample is most
Sparse, classification problem can be converted into coefficient and represent Solve problems.Experimental results demonstrate SRC is in image classification and target
Detection field achieves good effect.
But when training sample is less, SRC is poor to the expression ability of test sample, and increases training sample method operation
Complicated and can increase sparse solving complexity, the present invention proposes to use multi-class K-SVD Algorithm for Training sample dictionary, utilizes afterwards
SRC classifies.
The embodiment of the present invention one provides a kind of grain type sorting technique of corn kernel based on dictionary learning.Referring to Fig. 1,
The sorting technique specifically includes following steps：
Step 100：Obtain the seed image of each corn kernel of different grain types, and using seed image as training sample.
In this step, the seed image of each corn kernel of different grain types is obtained first, and grain type therein includes half horse
Three kinds of flute profile, dent type and Hard grain type, wherein, half dent type is also osculant, is hybridized by Hard grain type and dent corn,
Seed top depression it is shallow compared with dent type, have without depression and it is only white mottled, the farinaceous albumen at top it is few compared with dent type but
More than Hard grain type, quality is good compared with dent type；Dent shape seed again is flat to be rectangle, because the top of silty is than both sides cutin
It is dry soon, so the centre at top is recessed, it is similar to dent；Seed creasy surface is coarse opaque, mostly yellow, white, a small number of
It is poor in purple or red, edible quality；Hard grain type seed is generally square circular, and top and surrounding endosperm are all cutin, and only center is near
Embryo part is silty, therefore appearance is translucent glossy, hard full, and grain color is generally yellow, or have by, the color, seed such as red, purple
Quality better；Then using seed image as training sample, wherein, training sample is used for designing classification in image procossing.
Step 200：Extract in each training sample for carrying out a characteristic vector for type classification.
In this step, characteristic vector includes：Seed energy, contrast, entropy, auto-correlation, unfavourable balance square, area, elongation,
Circularity, rectangular degree, L * channel average, a passages average and b passage averages.
Step 300：Assemblage characteristic vector, obtains training sample dictionary.
In this step, each characteristic vector is divided into different training sample words according to the different grain type of corn kernel
Allusion quotation, the characteristic vector in the corn kernel extraction step 200 that will be partitioned into step 100, according to half dent type, dent type,
Three kinds of grain type composition training sample dictionaries of Hard grain type, extract three kinds of data samples of grain type as dictionary learning foundation.
Step 400：Multi-class dictionary learning is carried out to training sample dictionary based on dictionary training algorithm K-SVD.
In this step, first to training sample and data samples normalization, then to the training sample word after normalization
Allusion quotation carries out multi-class K-SVD dictionary learnings, wherein, K-SVD is a kind of classical dictionary training algorithm, minimum former according to error
Then, SVD decomposition is carried out to error term, selection makes the minimum decomposition item of error as the dictionary atom and corresponding atom system for updating
Number, by continuous iteration so as to the solution for being optimized.
Step 500：Using the small classification of reconstructed error as class categories, according to the class categories to the grain type of corn kernel
Carry out sparse classification.
In this step, test sample is selected first, and sparse solution, and root are carried out using the dictionary obtained in step 400
According to sparse solution coefficient, three classification reconstructed errors are calculated respectively, and using the minimum classification of reconstructed error as class categories.
Knowable to foregoing description, embodiments of the invention one realize the seed in each corn kernel for obtaining different grain types
Image, and using seed image as after training sample, accurately and securely extracting in each training sample for carrying out a type classification
Characteristic vector, obtain training sample dictionary, multi-class dictionary learning is carried out to training sample dictionary and to corn kernel
Grain type carry out the process of sparse classification, it is complete and accurately realize sparse classification is carried out to the grain type of corn kernel.
One kind of the step of embodiment of the present invention two is there is provided before step 100 in above-mentioned sorting technique A00 to D00 is specific
Implementation method.Referring to Fig. 2, following steps were also specifically included before above-mentioned steps 100：
Step A00：The corn ear of different grain types is gathered respectively, and generates the fringe portion image of corn ear.
Step B00：The corn ear fringe portion core in seed image is extracted as region of interest ROI.
In this step, ROI (region of interest) marks area-of-interest, i.e. machine vision, image procossing
In, being sketched the contours of with modes such as square frame, circle, ellipse, irregular polygons from processed image needs region to be processed, referred to as feels
Interest region, ROI.Commonly used on the machine vision softwares such as Halcon, OpenCV, Matlab various operators (Operator) and
Function tries to achieve region of interest ROI, and carries out the next step treatment of image.
Step C00：The red channel information of ROI is extracted, and denoising is carried out to ROI according to median filtering method.
Step D00：The ROI after denoising is split according to Otsu algorithm OTSU, obtains the seed image of corn kernel.
Knowable to foregoing description, embodiments of the invention two realize the seed image of each corn kernel to different grain types
Accurate and quick acquisition process, it is ensured that the grain type to corn kernel carries out the quick realization of sparse classification.
The embodiment of the present invention three provides a kind of specific embodiment of step 200 in above-mentioned grain type sorting technique.Referring to
Fig. 3, the step 200 specifically includes following steps：
Step 201：Seed L * channel average, a passages average and b passage average color characteristics are extracted as carrying out grain
The classification foundation of type classification；
In this step, specifically include：
Step a. extracts single corn kernel cromogram and segmentation binary picture；
Corn kernel image RGB color is converted to Lab color spaces by step b.；
Step c. calculates corn kernel L * channel, a passages, b passage averages respectively using binary picture as template.
Step 202：Seed energy, contrast, entropy, auto-correlation and unfavourable balance square are extracted as carrying out what a type was classified
Classification foundation；
Step 203：Seed area, elongation, circularity and rectangular degree are extracted as carrying out the classification that a type is classified
Foundation；
Step 204：All classification foundation is integrated, characteristic vector is obtained.
Knowable to foregoing description, embodiments of the invention three are realized to being used to carry out a type classification in each training sample
The accurate and efficient acquisition of characteristic vector.
The embodiment of the present invention four provides a kind of specific embodiment of step 400 in above-mentioned grain type sorting technique.Referring to
Fig. 4, the step 400 specifically includes following steps：
Step 401：Extract the data sample of the corn kernel of different grain types in each training sample dictionary, and by the data
Sample is used as dictionary learning foundation.
Step 402：Training sample dictionary and data sample are normalized.
Step 403：According to the data sample after normalized, the training sample dictionary after normalized is carried out many
Classification dictionary training algorithm K-SVD dictionary learnings.
Knowable to foregoing description, embodiments of the invention four are realized based on dictionary training algorithm K-SVD to training sample
Dictionary carries out multi-class dictionary learning, and data basis are provided for the follow-up grain type to corn kernel carries out sparse classification.
The embodiment of the present invention five provides a kind of specific embodiment of step 500 in above-mentioned grain type sorting technique.Referring to
Fig. 5, the step 500 specifically includes following steps：
Step 501. carries out sparse solution according to the training sample dictionary after dictionary learning to the test sample of corn kernel,
And according to sparse solution coefficient, the reconstructed error of all categories of corn kernel is calculated respectively.
Step 502. using the minimum classification of middle reconstructed error of all categories as class categories, and according to the class categories to jade
The grain type of rice seed carries out sparse classification.
Knowable to foregoing description, embodiments of the invention five are realized the small classification of reconstructed error as class categories,
Sparse classification is carried out to the grain type of corn kernel according to the class categories, rapidly and precisely corn kernel can be classified, can
To meet corn seed investigating demand.
It is further instruction technical scheme, the present invention also provides a kind of corn kernel based on dictionary learning
Grain type sorting technique application example.Referring to Fig. 6, following content is specifically included：
S1：The corn ear image of the different grain types of three kinds of collection simultaneously splits seed.
S2：Extract the grain type characteristic vector of classification corn kernel.
S3：By combination of eigenvectors is into training sample dictionary and carries out multi-class K-SVD dictionary learnings.
S4：Dictionary after being trained using S3 carries out rarefaction representation classification to the grain type of seed, obtains grain type classification
Further, step S1 includes following sub-step：
S11：Corn ear core is extracted as ROI.
S12：Extract R channel informations, medium filtering denoising.
S13：Split using OTSU algorithms, obtained seed profile.
Further, step S2 includes following sub-step：
S21：Seed L * channel average, a passages average, b passage average color characteristics are extracted as classification foundation.
S22：Seed energy, contrast, entropy, auto-correlation, unfavourable balance square textural characteristics are extracted as classification foundation.
S23：Seed area, elongation, circularity, rectangular degree are extracted as classification foundation.
S24：The parameter that S21, S22, S23 are extracted is vectorial as characteristic of division.
Further, step S21 includes following sub-step：
S211：Extract single corn kernel cromogram and segmentation binary picture
S212：Corn kernel image RGB color is converted into Lab color spaces
S213：Using binary picture as template, corn kernel L * channel, a passages, b passage averages are calculated respectively
Further, step S212 color space conversions step is：
RGB color is converted into XYZ color space, conversion formula is：
XYZ color space is converted into Lab color spaces, conversion formula is：
Wherein,
Further, step S213 is specifically expressed as：
Corn kernel cromogram midpoint is traveled through using scanning method, using binary picture as template, if point is in seed profile
(point is white in Prototype drawing), cumulative statistics color feature value；If point is not counted not in seed profile.It is beautiful after the completion of traversal
Rice seed L * channel, a passages, b passage mean value computation formula are：
Wherein, count is white portion pixel total number in binary picture,Respectively seed is colored
L, a, b component value of the image in kth pixel.
Further, step S22 includes following sub-step：
S221：Extract single corn kernel gray-scale map and segmentation binary picture.
S222：Gray scale normalization process is carried out to seed gray-scale map.
S223：Using binary picture as template, seed gray level co-occurrence matrixes are calculated.
S224：The gray level co-occurrence matrixes being calculated according to S223 calculate seed energy, contrast, entropy, auto-correlation, unfavourable balance
Square textural characteristics.
Further, step S222 is specifically expressed as：
The gray-scale map of former seed RGB image conversion is defaulted as 256 gray levels, and excessive gray scale can cause amount of calculation larger, and
And because seed profile is smaller, it is therefore desirable to reduce gray level.It is by gray value after normalization：
Wherein, gmaxFor the highest gray value 255, L that seed image is likely to occurnewIt is the gray level for specifying herein, through examination
Value 32 is determined after testing.
Further, step S223 is specifically expressed as：
Corn kernel gray-scale map midpoint is traveled through using scanning method, using binary picture as template, and if only if gray value two
(white in profile), gray scale when two points of first formula are 255 all not less than minimum enclosed rectangle border and corresponding templates pixel value
The corresponding elements of statistical matrix G add one.Four gray level co-occurrence matrixes in orientation of single grain are calculated respectively, calculate the energy of four direction
Amount, contrast, unfavourable balance square, entropy, auto-correlation.
Further, step S224 is specifically expressed as：
Five textural characteristics statistics for taking four direction do average as the final seed texture feature vector for extracting.
Further, step S23 includes following sub-step；
S231：Extract single corn kernel profile.
S232：Corn kernel profile minimum enclosed rectangle is extracted, single grain area, elongation, circularity, rectangle is calculated
Degree.
Further, step S3 includes following sub-step：
S31：The corn kernel that will be partitioned into S1 steps extracts the characteristic vector in S2 steps, according to half dent type, horse
Flute profile, three kinds of grain type composition training sample dictionaries of Hard grain type, extract three kinds of data sample Y={ Y of grain type1, Y2, Y3As word
Allusion quotation learns foundation.
S32：To training sample and data samples normalization.
S33：Multi-class K-SVD dictionary learnings are carried out to the training sample dictionary after normalization
Further, step S33 includes following sub-step：
S331：Dictionary after normalization is divided into three sub- dictionaries according to classification：
A'=[A'1, A'2, A'3]
S332：To each type dictionary A'i, K-SVD dictionary learnings are carried out by correspondence class data sample.
Further, K-SVD dictionary learnings follow formula in step S332
Further, step S4 includes following sub-step：
S41：Selection test sample, the dictionary obtained using S3 carries out sparse solution.
S42：According to sparse solution coefficient, three classification reconstructed errors are calculated respectively.
S43：Using the minimum classification of reconstructed error as class categories.
Further, S41 it is sparse solution follow the principles for：
The problem of minimal L1 norm is converted to by more than, is expressed as
Further, step S42 is specifically described as：
Solved using minimal L1 norm after obtaining sparse vector, introduce reconstructed error to judge optimal classification classification.It is right
I-th class, definitionIts formation rule isThe coefficient entry unrelated with the i-th class is multiplied by 0 in vector, related to the i-th class
Coefficient entry be multiplied by 1.UseReconstruct represents y.
Further, step S43 is specifically described as：
The norm of reconstructed error 2 is calculated, minimum is grain type class categories.
Knowable to foregoing description, application example of the invention provides a kind of rarefaction representation corn kernel based on dictionary learning
Grain type sorting technique, can quickly, accurately carry out a type classification to corn kernel.
The present invention also provides a kind of grain of the corn kernel based on dictionary learning that can realize above method full content
Type categorizing system.Referring to Fig. 7, the system specifically includes following content：
Training sample acquiring unit 10, for using the seed image of each corn kernel of different grain types as training sample.
Characteristic vector pickup unit 20, for extracting in each training sample for carrying out a characteristic vector for type classification.
Training sample dictionary acquiring unit 30, for assemblage characteristic vector, obtains training sample dictionary.
Dictionary learning unit 40, for carrying out multi-class word to training sample dictionary based on dictionary training algorithm K-SVD
Allusion quotation learns.
Sparse taxon 50, for using the small classification of reconstructed error as class categories, according to the class categories to jade
The grain type of rice seed carries out sparse classification.
Knowable to foregoing description, system of the invention realizes the seed figure in each corn kernel for obtaining different grain types
Picture, and using seed image as after training sample, accurately and securely extracting in each training sample for carrying out a type classification
Characteristic vector, obtains training sample dictionary, multi-class dictionary learning is carried out to training sample dictionary and to corn kernel
Grain type carries out the process of sparse classification, and realize completely and accurately carries out sparse classification to the grain type of corn kernel.
Above example is merely to illustrate technical scheme, rather than its limitations；Although with reference to the foregoing embodiments
The present invention has been described in detail, it will be understood by those within the art that：It still can be to foregoing each implementation
Technical scheme described in example is modified, or carries out equivalent to which part technical characteristic；And these are changed or replace
Change, do not make the spirit and scope of the essence disengaging various embodiments of the present invention technical scheme of appropriate technical solution.
Claims (10)
1. the grain type sorting technique of a kind of corn kernel based on dictionary learning, it is characterised in that including：
Step 1. obtains the seed image of each corn kernel of different grain types, and using the seed image as training sample；
Step 2. is used to carry out a characteristic vector for type classification in extracting each training sample；
Step 3. combines the characteristic vector, obtains training sample dictionary；
Step 4. carries out multi-class dictionary learning based on dictionary training algorithm K-SVD to the training sample dictionary；
Step 5. carries out dilute the small classification of reconstructed error as class categories according to the class categories to the grain type of corn kernel
Dredge classification.
2. method according to claim 1, it is characterised in that also include before the step 1：
Step A. gathers the corn ear of different grain types respectively, and generates the fringe portion image of corn ear；
Corn ear fringe portion core in the step B. extractions seed image is used as region of interest ROI；
Step C. extracts the red channel information of the ROI, and carries out denoising to the ROI according to median filtering method；
Step D. according to Otsu algorithm OTSU to denoising after the ROI split, obtain the seed figure of the corn kernel
Picture.
3. method according to claim 1, it is characterised in that the characteristic vector in the step 2, including：
The characteristic vector includes：Seed energy, contrast, entropy, auto-correlation, unfavourable balance square, area, elongation, circularity, rectangle
Degree, L * channel average, a passages average and b passage averages.
4. method according to claim 3, it is characterised in that the step 2 includes：Extract be used in each training sample into
The characteristic vector of row grain type classification
Step 2-1. extracts seed L * channel average, a passages average and b passage average color characteristics as carrying out a type point
The classification foundation of class；
Step 2-2. extracts seed energy, contrast, entropy, auto-correlation and unfavourable balance square as carrying out the classification that a type is classified
Foundation；
Step 2-3. extracts seed area, elongation, circularity and rectangular degree as carrying out the classification foundation that a type is classified；
Step 2-4. integrates all classification foundations, obtains the characteristic vector.
5. method according to claim 4, it is characterised in that the step 2-1 includes：
Step a. extracts single corn kernel cromogram and segmentation binary picture；
Corn kernel image RGB color is converted to Lab color spaces by step b.；
Step c. calculates corn kernel L * channel, a passages, b passage averages respectively using binary picture as template.
6. method according to claim 1, it is characterised in that the step 3 includes：
Each characteristic vector is divided into different training sample dictionaries according to the different grain type of the corn kernel.
7. method according to claim 6, it is characterised in that the grain type includes half dent type, dent type and Hard grain type.
8. method according to claim 1, it is characterised in that the step 4 includes：
Step 4-1. extracts the data sample of the corn kernel of different grain types in each training sample dictionary, and should
Data sample is used as dictionary learning foundation；
Step 4-2. is normalized to the training sample dictionary and data sample；
Step 4-3. according to the data sample after normalized, to normalized after the training sample dictionary carry out it is many
Classification dictionary training algorithm K-SVD dictionary learnings.
9. method according to claim 1, it is characterised in that the step 5 includes：
Step 5-1. carries out sparse solution according to the training sample dictionary after the dictionary learning to the test sample of corn kernel,
And according to sparse solution coefficient, the reconstructed error of all categories of corn kernel is calculated respectively；
Step 5-2. using the minimum classification of middle reconstructed error of all categories as class categories, and according to the class categories to Corn Seeds
The grain type of grain carries out sparse classification.
10. the grain type categorizing system of a kind of corn kernel based on dictionary learning, it is characterised in that including：
Training sample acquiring unit, for the seed image of each corn kernel by different grain types are obtained, and by the seed figure
As training sample；
Characteristic vector pickup unit, for extracting in each training sample for carrying out a characteristic vector for type classification；
Training sample dictionary acquiring unit, for combining the characteristic vector, obtains training sample dictionary；
Dictionary learning unit, for carrying out multi-class dictionary to the training sample dictionary based on dictionary training algorithm K-SVD
Study；
Sparse taxon, for using the small classification of reconstructed error as class categories, according to the class categories to corn kernel
Grain type carry out sparse classification.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611059052.1A CN106709505A (en) | 2016-11-23 | 2016-11-23 | Dictionary learning-based grain type classification method and system for corn grains |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611059052.1A CN106709505A (en) | 2016-11-23 | 2016-11-23 | Dictionary learning-based grain type classification method and system for corn grains |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106709505A true CN106709505A (en) | 2017-05-24 |
Family
ID=58934666
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611059052.1A Pending CN106709505A (en) | 2016-11-23 | 2016-11-23 | Dictionary learning-based grain type classification method and system for corn grains |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106709505A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110246133A (en) * | 2019-06-24 | 2019-09-17 | 中国农业科学院农业信息研究所 | A kind of corn kernel classification method, device, medium and equipment |
| CN111507418A (en) * | 2020-04-21 | 2020-08-07 | 中国科学技术大学 | Encaustic tile quality detection method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103093456A (en) * | 2012-12-25 | 2013-05-08 | 北京农业信息技术研究中心 | Corn ear character index computing method based on images |
| CN104102922A (en) * | 2014-07-15 | 2014-10-15 | 中国科学院合肥物质科学研究院 | Pest image classification method based on context sensing dictionary learning |
-
2016
- 2016-11-23 CN CN201611059052.1A patent/CN106709505A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103093456A (en) * | 2012-12-25 | 2013-05-08 | 北京农业信息技术研究中心 | Corn ear character index computing method based on images |
| CN104102922A (en) * | 2014-07-15 | 2014-10-15 | 中国科学院合肥物质科学研究院 | Pest image classification method based on context sensing dictionary learning |
Non-Patent Citations (2)
| Title |
|---|
| 冯丽娟: "基于稀疏表示的小麦品种识别方法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 杨蜀秦 等: "基于稀疏表示的大米品种识别", 《农业工程学报》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110246133A (en) * | 2019-06-24 | 2019-09-17 | 中国农业科学院农业信息研究所 | A kind of corn kernel classification method, device, medium and equipment |
| CN111507418A (en) * | 2020-04-21 | 2020-08-07 | 中国科学技术大学 | Encaustic tile quality detection method |
| CN111507418B (en) * | 2020-04-21 | 2022-09-06 | 中国科学技术大学 | Encaustic tile quality detection method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109360206B (en) | Method for dividing rice ears in field based on deep learning | |
| CN104598908B (en) | A kind of crops leaf diseases recognition methods | |
| Cinar et al. | Identification of rice varieties using machine learning algorithms | |
| CN104091179B (en) | Intelligent blumeria graminis spore picture identification method | |
| CN108960257A (en) | A kind of diabetic retinopathy grade stage division based on deep learning | |
| CN105844285A (en) | Cucumber disease identification method and apparatus based on image information | |
| CN109919177A (en) | Feature selection approach based on stratification depth network | |
| CN109344699A (en) | Winter jujube disease recognition method based on depth of seam division convolutional neural networks | |
| CN104102928B (en) | A kind of Classifying Method in Remote Sensing Image based on texture primitive | |
| CN110110596A (en) | High spectrum image feature is extracted, disaggregated model constructs and classification method | |
| Goriewa-Duba et al. | An evaluation of the variation in the morphometric parameters of grain of six Triticum species with the use of digital image analysis | |
| CN109101891A (en) | A kind of rice pest detection system and its detection method merging artificial intelligence | |
| Prasher et al. | Analysis of DenseNet201 with SGD optimizer for diagnosis of multiple rice leaf diseases | |
| CN106709505A (en) | Dictionary learning-based grain type classification method and system for corn grains | |
| Lee et al. | Intelligent classification methods of grain kernels using computer vision analysis | |
| Suseendran et al. | Automatic seed classification by multi-layer neural network with spatial-feature extraction | |
| Bharadwaj | Classification and grading of arecanut using texture based block-wise local binary patterns | |
| CN110288041A (en) | Chinese herbal medicine classification model construction method and system based on deep learning | |
| CN115862003A (en) | Lightweight YOLOv 5-based in-vivo apple target detection and classification method | |
| CN111709427B (en) | Fruit segmentation method based on sparse convolution kernel | |
| Feng et al. | Research of image recognition of plant diseases and pests based on deep learning | |
| Li et al. | Rapeseed seeds classification by machine vision | |
| Xing et al. | Traits expansion and storage of soybean phenotypic data in computer vision-based test | |
| Kotwaliwale et al. | Machine vision for characterisation of some phenomic features of plant parts in distinguishing varieties-a review | |
| Li et al. | Recognition Algorithm of Rice Germ Integrity Base on Improved Inception V3 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170524 |
|
| RJ01 | Rejection of invention patent application after publication |