CN106846334A - Field corn plant recognition methods based on Support Vector data description - Google Patents
Field corn plant recognition methods based on Support Vector data description Download PDFInfo
- Publication number
- CN106846334A CN106846334A CN201710036915.1A CN201710036915A CN106846334A CN 106846334 A CN106846334 A CN 106846334A CN 201710036915 A CN201710036915 A CN 201710036915A CN 106846334 A CN106846334 A CN 106846334A
- Authority
- CN
- China
- Prior art keywords
- feature
- picture
- region
- milpa
- support vector
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/188—Vegetation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30188—Vegetation; Agriculture
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Image Processing (AREA)
Abstract
The present invention relates to a kind of field corn plant recognition methods based on Support Vector data description, including two stages：Training stage, by obtaining initial stage intraday field corn picture, the picture feature of Maize Regional in picture is extracted, set up Support Vector data description（SVDD）Model；The segmentation stage, including：（1）Image preprocessing to be split：Based on super green feature extraction picture Green region；（2）Calculating the picture feature of regional respectively and bringing SVDD models into carries out categorizing selection Maize Regional, removes weed regions.The present invention for the actual weeds species in field it is various the characteristics of, corn identification is carried out from SVDD graders, the automatic identification of field corn weeds can be realized, include vegetation index feature and textural characteristics using picture feature, calculate easy and effective, and picture to different year has the advantages that robustness.
Description
Technical field
The present invention relates to the milpa automatic obtaining method that a kind of field shoots picture, especially it is a kind of based on support to
The field corn plant recognition methods of amount data description, the partitioning algorithm based on Support Vector data description, for corn to be planted
Strain is accurately extracted from complex background.
Background technology
Corn is Gross World Product highest cereal crops and main livestock feed source, while being also industrial production
In important source material, with developing rapidly for computer hardware technique and digital image processing techniques, it agriculturally should
There is larger progress with research.The machine vision and image analysis technology for being currently based on image are obtained extensively in agriculture field
General application, it especially has significant advantage in terms of the extraction of feature in terms of color, shape and texture.Machine is regarded
Feel technology is obtained including image, image procossing and pattern-recognition etc., and can simulate that human eye carries out to crops with machine vision can
See the shooting at close range of spectrum, research object then is analyzed to the image information for obtaining with technologies such as Digital Image Processing
Required information.For a long time, mainly by way of manually carrying out, this mode not only wastes time and energy Field observation, and receives
Influence to people's subjective factor is larger.Therefore it is very necessary monitor in real time to be carried out to field corn by the means of image procossing
's.
The adaptability and reproductive capacity of weeds are all very strong, they can with crops and soil contention nutrient, moisture, sunlight with
And growing space, crop ventilation, ventilative, the temperature of influence upper soll layer are hindered, the Severe Reduction of crops is directly contributed, less
Beneficial to the sustainable use of soil.It can be seen that, in order to maintain the high yield of crops, it is very necessary to go to cut weeds.Intelligent weeding
While conservation of nature environment, it is unlikely to make crop production reduction again.If that is, operator knows which place has miscellaneous
Grass, which place is without weeds, then herbicide will be sprayed to the place for having weeds when medicament is sprayed, to not miscellaneous
Herbicide is not sprayed in the place of grass, so both saves cost, reduces input, and the dirt to agroecological environment is reduced again
Dye, be conducive to agricultural sustainable development, this be also nowadays precision agriculture aspect compare a clearly trend.It is miscellaneous in order to reduce
The careless extent of injury, is preserved the ecological environment not contaminated, improves the quality and yield of product, and the automation of Weeds distribution is realized as early as possible
A very real problem is turned into scientific.Therefore, the development of intelligent weeding and application, to promoting China's agricultural modernization
Development have highly important economical and environmentally friendly meaning.
The content of the invention
The purpose of the present invention is to overcome the deficiencies in the prior art, there is provided a kind of based on Support Vector data description
Field corn plant recognition methods, can realize that field shoots the automatic identification of milpa in picture, and algorithm calculates simple, and
With robustness higher.
According to the technical scheme that the present invention is provided, the field corn plant identification side based on Support Vector data description
Method, it is characterized in that, comprise the following steps：
(1) training stage：The corn field picture of different time sections is gathered as training sample, corn field picture is extracted
In milpa region, obtain the picture feature in milpa region；Calculate tribute of the original image feature to first principal component
Rate W is offered, given threshold selects the feature corresponding to larger contribution rate W as optimal characteristics, exports optimal characteristics set, sets up
SVDD models；
(2) stage is split：The background area of test pictures is removed first, extracts the figure of test pictures residue connected region
Piece feature；Bring the picture feature that each connected region of test pictures is extracted into SVDD models to be tested, determine milpa area
Domain and non-Maize Regional.
Further, the picture feature in the milpa region for being obtained in the step (1) includes picture connected region
Vegetation index featureAnd textural characteristicsWherein comprising L sample with
Num feature.
Further, the vegetation index feature include Rn, Gn, Bn, Gray, ExG, ExR, CIVE, ExGR, GB, ERI,
12 vegetation index features of EGI, EBI；The textural characteristics include contrast on four direction, correlation, energy and same
Matter textural characteristics.
Further, in the step (1) in one day corn growth initial stage of training sample selection different time sections picture.
Further, picture pretreatment, the removal picture back of the body are carried out using super green feature in the step (1) and step (2)
Scape.
Further, when setting up SVDD models in the step (1), feature selecting is carried out using principal component analysis method,
Calculate the contribution rate W of picture feature and each primitive character to first principal component in each milpa region, threshold value setting
For 0.2).
Further, the detailed process of the step (2) is：Test pictures are removed into background area, retains milpa
Connected region；The vegetation index characteristic set of each pixel each connected region Nei is calculated, it is special to pixel in same area
Levying carries out summation and takes averaging operation, obtains the vegetation index feature of connected region；By the vegetation index feature and line of connected region
The SVDD models that reason feature substitution step (1) is obtained are tested, and determine milpa region, and non-milpa region is carried out
Deletion action.
Field corn plant recognition methods based on Support Vector data description of the present invention, mainly make use of crop map
As characteristic rule carries out milpa identification, it is possible to prevente effectively from the problems such as the error of artificial observation, while being directed to actual environment
The characteristics of middle weeds species is various, using SVDD algorithms, finally can exactly recognize the milpa region of field picture.Should
Method for crop automatic observation, weeds in field identification etc. field have great importance.
Brief description of the drawings
Fig. 1 is the overall flow figure of field corn plant recognition methods of the present invention.
Fig. 2 is to carry out the schematic diagram that plant region is extracted in picture pretreatment removal background area according to super green special piece.
Fig. 3-1~Fig. 3-3 is the training sample picture under different time.
Fig. 4-1, Fig. 4-2 carry out the shot chart and load diagram after PCA for training data.
Fig. 5-1, Fig. 5-2 are that test pictures are carried out with the result figure after coarse segmentation and final decision.
Specific embodiment
With reference to specific accompanying drawing, the invention will be further described.
Field corn plant recognition methods based on Support Vector data description of the present invention, it is therefore intended that for field
Under complex environment, the automatic identification plant region in colour picture；As shown in figure 1, specifically including following steps：
(1) training stage：Picture collection system is placed on corn field to be identified, in one day corn growth initial stage not
Collected with the corn field picture of time period, as the historical data of corn picture, entered using super green feature (2G-R-B)
Row picture is pre-processed, and is extracted milpa region and is set up training pattern；
As shown in Fig. 2 removal background area, extracts milpa region and sets up comprising the following steps that for training pattern：
A, sample image generation, choose one day in some width of different time sections picture (such as Fig. 3-1, Fig. 3-2, Fig. 3-3 institutes
Show), the picture feature in milpa region is extracted, as training data in milpa region at selection 197；The picture is special
Levy including Rn, Gn, Bn (Rn, Gn, Bn are normalization RGB feature), Gray (gray feature), ExG (super green feature), ExR
(exceedingly popular feature), CIVE (vegetation color index), ExGR (super green subtract red), GB (green to subtract red), ERI (excessively red feature), EGI (mistakes
Green feature), 12 vegetation index features of EBI (excessively blue feature), (0 degree, 45 degree, 90 degree and 135 of contrast on four direction
Degree), correlation, energy, homogeney textural characteristics；The eigenmatrix of each pixel L × Num is obtained, this feature matrix includes L
Individual sample and Num feature, can be expressed as the vegetation index feature of picture connected regionAnd line
Reason featureWherein textural characteristics matrix comprising the contrast on four direction, correlation, energy,
Homogeney textural characteristics；Summation is carried out to the vegetation index feature of regional and takes averaging operation；
B, step a generation training data after, carry out feature selecting using principal component analysis (PCA) method, calculate each
The picture feature in milpa region and each primitive character are to the contribution rate W of first principal component, given threshold (general threshold
0.2) value is set as that the feature corresponding to the larger feature of selection contribution rate W values exports optimal characteristics collection as optimal characteristics
Close, set up Support Vector data description (SVDD) model；
As shown in figure 4, wherein finding that corn weeds can be classified by first principal component；Therefore choose wherein to first
The big feature of principal component weight carries out SVDD modelings as optimal characteristics, obtains the hyperspherical center μ and radius R that classifies；
(2) stage is split：Using the SVDD models through step (1) training generation, test pictures are carried out into dividing processing, carried
The picture feature for taking regional is classified, and finally leaves milpa region；Concrete operation step is as follows：
A, corn map are as pre-treatment step：First coloured image to be split is pre-processed based on super green feature, is removed
Background area, retains green connected region, and non-plant region is set to black；
B, milpa region recognition step：Calculate the vegetation index feature set of each pixel each connected region Nei
Close, summation is carried out to pixel point feature in same area and takes averaging operation, obtain the vegetation index feature of connected region；Will connection
The SVDD models that the vegetation index feature and textural characteristics substitution step (1) in region are obtained are tested, and determine milpa area
Domain, deletion action is carried out to non-milpa region；
When being tested by SVDD models, by the picture feature average of each connected region with SVDD hyperspheres center
Distance, and hypersphere radius surface is compared, and according to the smaller principle of identical type plant difference, selection is in hypersphere
Region is used as milpa.
Specifically, as shown in Fig. 5-1, Fig. 5-2, Fig. 5-1, Fig. 5-2 are respectively two test pictures and carry out coarse segmentation and most
Schematic diagram after whole decision-making, wherein, a regions are original test pictures, and b regions are to remove background area through image preprocessing step
Schematic diagram, c regions are the schematic diagram in milpa region of cutting weeds, determine.
Claims (7)
1. a kind of field corn plant recognition methods based on Support Vector data description, it is characterized in that, comprise the following steps：
(1) training stage：The corn field picture of different time sections is gathered as training sample, in the picture of extraction corn field
Milpa region, obtains the picture feature in milpa region；Calculate contribution rate of the original image feature to first principal component
W, given threshold selects the feature corresponding to larger contribution rate W as optimal characteristics, exports optimal characteristics set, sets up SVDD
Model；
(2) stage is split：The background area of test pictures is removed first, the picture for extracting test pictures residue connected region is special
Levy；The picture feature that each connected region of test pictures is extracted is brought into SVDD models and is tested, determine milpa region and
Non- Maize Regional.
2. the field corn plant recognition methods based on Support Vector data description as claimed in claim 1, it is characterized in that：Institute
Stating the picture feature in the milpa region obtained in step (1) includes the vegetation index feature of picture connected regionAnd textural characteristicsWherein include L sample and Num feature.
3. the field corn plant recognition methods based on Support Vector data description as claimed in claim 2, it is characterized in that：Institute
Vegetation index feature is stated to refer to including 12 vegetation of Rn, Gn, Bn, Gray, ExG, ExR, CIVE, ExGR, GB, ERI, EGI, EBI
Number feature；The textural characteristics include the contrast on four direction, correlation, energy and homogeney textural characteristics.
4. the field corn plant recognition methods based on Support Vector data description as claimed in claim 1, it is characterized in that：Institute
State the picture of different time sections in one day corn growth initial stage of training sample selection in step (1).
5. the field corn plant recognition methods based on Support Vector data description as claimed in claim 1, it is characterized in that：Institute
State and carry out picture pretreatment, removal picture background using super green feature in step (1) and step (2).
6. the field corn plant recognition methods based on Support Vector data description as claimed in claim 1, it is characterized in that：Institute
Stating when set up SVDD models in step (1), feature selecting is carried out using principal component analysis method, calculate each milpa region
Picture feature and each primitive character to the contribution rate W of first principal component, 0.2) threshold value is set as.
7. the field corn plant recognition methods based on Support Vector data description as claimed in claim 1, it is characterized in that：Institute
The detailed process for stating step (2) is：Test pictures are removed into background area, retains milpa connected region；Calculate each company
The vegetation index characteristic set of each pixel in logical region, summation is carried out to pixel point feature in same area and takes average behaviour
Make, obtain the vegetation index feature of connected region；The vegetation index feature and textural characteristics of connected region are substituted into step (1) to obtain
To SVDD models tested, determine milpa region, deletion action is carried out to non-milpa region.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710036915.1A CN106846334A (en) | 2017-01-19 | 2017-01-19 | Field corn plant recognition methods based on Support Vector data description |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710036915.1A CN106846334A (en) | 2017-01-19 | 2017-01-19 | Field corn plant recognition methods based on Support Vector data description |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106846334A true CN106846334A (en) | 2017-06-13 |
Family
ID=59123281
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710036915.1A Pending CN106846334A (en) | 2017-01-19 | 2017-01-19 | Field corn plant recognition methods based on Support Vector data description |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106846334A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108537267A (en) * | 2018-04-02 | 2018-09-14 | 江南大学 | A kind of corncob automatic identifying method based on significance test |
| CN108629289A (en) * | 2018-04-11 | 2018-10-09 | 千寻位置网络有限公司 | The recognition methods in farmland and system, applied to the unmanned plane of agricultural |
| CN108776803A (en) * | 2018-04-20 | 2018-11-09 | 中国农业大学 | The method and system of weeds in a kind of removal farmland |
| CN109859099A (en) * | 2019-01-22 | 2019-06-07 | 华中农业大学 | The quick minimizing technology of potting corn weeds based on SFM point cloud depth degree |
| WO2019179270A1 (en) * | 2018-03-23 | 2019-09-26 | 广州极飞科技有限公司 | Plant planting data measuring method, working route planning method, device and system |
| CN110889420A (en) * | 2018-09-07 | 2020-03-17 | 华南农业大学 | Plant category identification method and system |
| CN111767903A (en) * | 2020-08-03 | 2020-10-13 | 吉林大学 | Corn core identification system and method based on machine vision |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101458522A (en) * | 2009-01-08 | 2009-06-17 | 浙江大学 | Multi-behavior process monitoring method based on pivot analysis and vectorial data description support |
| CN102708354A (en) * | 2011-12-31 | 2012-10-03 | 中国科学院遥感应用研究所 | Method for identifying golf course |
| CN106203452A (en) * | 2016-07-18 | 2016-12-07 | 江南大学 | Corn seed high spectrum image multiple features conversion method based on polyteny discriminant analysis |
-
2017
- 2017-01-19 CN CN201710036915.1A patent/CN106846334A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101458522A (en) * | 2009-01-08 | 2009-06-17 | 浙江大学 | Multi-behavior process monitoring method based on pivot analysis and vectorial data description support |
| CN102708354A (en) * | 2011-12-31 | 2012-10-03 | 中国科学院遥感应用研究所 | Method for identifying golf course |
| CN106203452A (en) * | 2016-07-18 | 2016-12-07 | 江南大学 | Corn seed high spectrum image multiple features conversion method based on polyteny discriminant analysis |
Non-Patent Citations (1)
| Title |
|---|
| KAZMI W,ET AL: "Detecting creeping thistle in sugar beet fields using vegetation indices", 《COMPUTERS AND ELECTRONICS IN AGRICULTURE》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019179270A1 (en) * | 2018-03-23 | 2019-09-26 | 广州极飞科技有限公司 | Plant planting data measuring method, working route planning method, device and system |
| US11321942B2 (en) | 2018-03-23 | 2022-05-03 | Guangzhou Xaircraft Technology Co., Ltd. | Method for measuring plant planting data, device and system |
| AU2019238712B2 (en) * | 2018-03-23 | 2022-07-21 | Guangzhou Xaircraft Technology Co., Ltd | Plant planting data measuring method, working route planning method, device and system |
| CN108537267A (en) * | 2018-04-02 | 2018-09-14 | 江南大学 | A kind of corncob automatic identifying method based on significance test |
| CN108629289A (en) * | 2018-04-11 | 2018-10-09 | 千寻位置网络有限公司 | The recognition methods in farmland and system, applied to the unmanned plane of agricultural |
| CN108629289B (en) * | 2018-04-11 | 2021-07-30 | 千寻位置网络有限公司 | Farmland identification method and system and agricultural unmanned aerial vehicle |
| CN108776803A (en) * | 2018-04-20 | 2018-11-09 | 中国农业大学 | The method and system of weeds in a kind of removal farmland |
| CN110889420A (en) * | 2018-09-07 | 2020-03-17 | 华南农业大学 | Plant category identification method and system |
| CN109859099A (en) * | 2019-01-22 | 2019-06-07 | 华中农业大学 | The quick minimizing technology of potting corn weeds based on SFM point cloud depth degree |
| CN111767903A (en) * | 2020-08-03 | 2020-10-13 | 吉林大学 | Corn core identification system and method based on machine vision |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106846334A (en) | Field corn plant recognition methods based on Support Vector data description | |
| Sannakki et al. | Diagnosis and classification of grape leaf diseases using neural networks | |
| Singh et al. | Detection and classification of plant leaf diseases in image processing using MATLAB | |
| Burgos-Artizzu et al. | Improving weed pressure assessment using digital images from an experience-based reasoning approach | |
| CN109376728A (en) | A kind of weeds in paddy field recognition methods and its application based on multiple features fusion and BP neural network | |
| Pastrana et al. | Novel image processing approach for solving the overlapping problem in agriculture | |
| de Luna et al. | Size classification of tomato fruit using thresholding, machine learning, and deep learning techniques | |
| Selvi et al. | Weed detection in agricultural fields using deep learning process | |
| Pawar et al. | Pomogranite disease detection and classification | |
| Gulavnai et al. | Deep learning for image based mango leaf disease detection | |
| Ullagaddi et al. | Disease recognition in Mango crop using modified rotational kernel transform features | |
| Arefi et al. | Development of an expert system based on wavelet transform and artificial neural networks for the ripe tomato harvesting robot | |
| CN114067207A (en) | Vegetable seedling field weed detection method based on deep learning and image processing | |
| CN107680098A (en) | A kind of recognition methods of sugarcane sugarcane section feature | |
| CN116129260A (en) | Forage grass image recognition method based on deep learning | |
| CN108037123A (en) | A kind of hybrid paddy rice disc type sows performance parameter accurate detecting method | |
| CN115861686A (en) | Litchi key growth period identification and detection method and system based on edge deep learning | |
| CN104504736A (en) | Method and system for identifying green plants in field crop growth monitoring image | |
| Tran et al. | Automatic dragon fruit counting using adaptive thresholds for image segmentation and shape analysis | |
| Kumar et al. | Detection of Sugarcane Disease and Classification using Image Processing | |
| Zhang et al. | Application of computer vision technology in agricultural field | |
| Ibaraki et al. | Image analysis for plants: basic procedures and techniques | |
| Zahan et al. | A deep learning-based approach for mushroom diseases classification | |
| Cui et al. | Phenotypic characteristics of ramie (Boehmeria nivea L) germplasm resources based on UAV remote sensing | |
| CN110633720A (en) | Corn disease identification method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170613 |