CN107957264A - A kind of tractor rotary tillage vision navigation method based on new and old native boundary line - Google Patents
A kind of tractor rotary tillage vision navigation method based on new and old native boundary line Download PDFInfo
- Publication number
- CN107957264A CN107957264A CN201710389147.8A CN201710389147A CN107957264A CN 107957264 A CN107957264 A CN 107957264A CN 201710389147 A CN201710389147 A CN 201710389147A CN 107957264 A CN107957264 A CN 107957264A
- Authority
- CN
- China
- Prior art keywords
- image
- new
- rotary tillage
- shearlet
- method based
- 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
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/20—Instruments for performing navigational calculations
Abstract
The present invention proposes a kind of tractor rotary tillage vision navigation method based on new and old native boundary line,Crop row variation and the characteristics of uneven illumination in working environment during for tractor rotary tillage operation,It is proposed that a kind of method based on Steerable filter (Guided Image Filter) and shearing wave conversion (Shearlet Transform) is used to extract new and old native boundary line to complete tractor vision guided navigation,First,Image is quickly transferred to YCrCb color spaces,Steerable filter is carried out to the image of gray processing,Then the new and old native marginal information of Shearlet canny operator extractions is used,Most vision guided navigation line is provided through Hough transform afterwards,Tractor rotary tillage vision navigation method based on new and old native boundary line proposed by the invention can be used in the intelligent navigation under farm environment.
Description
Technical field
The present invention relates to a kind of intelligent tractor vision navigation method, the new and old soil of Field-working Tractor-implement is based especially on
Boundary line vision navigation method, belongs to agricultural engineering technology field.
Background technology
Under the background that mechanization of agriculture and intellectualized technology develop rapidly, precision agriculture has obtained significant progress, special
It is not intelligent tractor automatic navigation technology.Under the farmland operation environment such as crop row variation, uneven illumination, automatic vision is led
Boat technology provides good solution for the restriction of existing technology development level.Existing automatic navigation method is mainly two
Kind, a kind of is the Centimeter Level satellite precision navigation (GPS, the Big Dipper) that can be achieved farmland, and another kind is that algorithm is complicated but of low cost
Vision guided navigation.GPS and Big Dipper technology are both needed to strengthen differential technique using ground, with high costs, and because of geographical location and meteorology
, there are interruption and delay in the factors such as environment, whens farmland satellite navigation signals.Vision guided navigation technology is widely used, but current base
In the automatic steering control of farm mechanism technology of machine vision mostly leading line is extracted by studying the distributional pattern of crop row, and agriculture
Field crops have harvested during industry machinery rotary tillage process, it is difficult to are navigated based on crop row.Therefore there is an urgent need for study a kind of energy
The vision guided navigation algorithm of field rotary tillage process in no crops.
The content of the invention
The defects of to overcome the prior art, the present invention propose that a kind of tractor rotary tillage vision based on new and old native boundary line is led
Boat method, be suitable for crop row variation, uneven illumination working environment under tractor intelligent navigation rotary tillage process.
To achieve the above object, the present invention uses following technical scheme:
A kind of tractor rotary tillage vision navigation method based on new and old native boundary line of the present invention, it is according to following step
It is rapid to implement:
Step 1:Visual pattern p (x, y) in front of tractor is gathered by camera;
Step 2:Image is subjected to gray processing, and is changed by formula f (x, y)=(R (x, y)+G (x, y)+B (x, y))/3
To YCrCb color spaces, wherein Y=0.299*R+0.587*G+0.114*B, Cr=(R-Y) * 0.713+128, Cb=(B-
Y)*0.564+128;
Step 3:Steerable filter processing is carried out to image under YCrCb color spaces;
The wherein processing of Steerable filter, i.e. " Local Linear Model " solve the course of work:
Make the value that q is output pixel;I and k is pixel index;I is the value of input picture, i.e., image to be filtered or other
The navigational figure of image;A and b is the coefficient that window center is located at linear function when at k;P is image to be filtered;It is anti-that ε, which is,
Only a values it is excessive and introduce have adjust filter effect parameter, ε is bigger, and filter effect is more obvious;μk, σ2K is respectively that I exists
Average value and variance in window;It is averages of the image p to be filtered in window;| w | it is the quantity that pixel is included in window;i
It is pixel with j;wijIt is a filtering core, for the function being oriented between image I and independent variable p;
Step is 1.:Image under YCrCb color spaces represents that the input of this function passes through a two dimension with two-dimensional function
The output and function input that window obtains meets linear relationship, i.e.,:qi=akIi+bk,
Step is 2.:To step 1. in formula both sides do gradient algorithm, i.e.,
Step is 3.:Gap between the actual value p and real output value of digital simulation function is
Step is 4.:A is calculated based on least square methodkAnd bk,bk=pk-akμk;
Step is 5.:All linear function values comprising k points are done weighted average to obtainStep 4:After Steerable filter processing, ground using Shearlet-Canny operator extractions
Study carefully the marginal information of image;
Algorithm therein is as follows:
Order research image is f [n1, n2];AaFor anisotropic expansion matrix;BsTo shear matrix.A > 0 are scale parameter;
S ∈ R are shear parameters;T ∈ R are translation parameters;
Step a:Read in image f [n1, n2];
Step b:By laplacian pyramid by image f [n1, n2] it is decomposed into low pass subbandWith high pass subband
Step c:By high pass subbandPseudo- polar coordinate system is transformed into from cartesian coordinate system, produced matrix passes through a frequency
Domain sub-filter, pseudo- polar coordinate system convert back cartesian coordinate system;
Step d:Some subgraph outputs are tried to achieve in the multi-direction characteristic converted using Shearlet, conversion, wherein
Shearlet systems are represented byShearlet is transformed to SHψF (a, s, t)=
< f, ψA, s, t >, carry out canny edge detections to the subgraph of multiple directions, obtain respective edge image respectively;
Step e:Shearlet inverse transformations are carried out to the subgraph of all directions in step 2;
Step f:Principle that can be complementary according to different images edge, carries out image after inverse transformation using logical operator
Fusion;
Step 5:Hough transformation extracts target navigation line;
The algorithm steps of wherein Hough transformation are as follows:
The slope that m is straight line is made, c is intercept;
(i) in image X-Y, all conllinear points (x, y) are described as y=mx+c with linear equation;
(ii) straight line is regarded to the straight line equation in parameter space M-C as, the slope of its cathetus is x, and intercept is
y;
(iii), instead of former linear equation, ρ=x cos θ are expressed as with the straight line polar equation of Duda and Hart propositions
+ y sin θs, ρ are distance of the origin to straight line, and θ crosses the vertical line of origin and the angle of positive direction of the x-axis for straight line;
(iiii), it is necessary to carry out discretization to parameter space during calculating, the center point coordinate of each unit is:
Step 6:Go to step 1.
Compared with prior art, beneficial effects of the present invention are as follows:
In line drawing result of navigating, Steerable filter treatment effect becomes apparent from compared to other filtering algorithms, and algorithm takes
It is most short;The leading line extracted by Shearlet-Canny operator edge detections is the most accurate.Subjective evaluation result also indicates that,
When uneven illumination and crop row variation, leading line extraction method effect is preferable in text.
The present invention identifies new and old native boundary line under rotary tillage environment by Steerable filter and Shearlet-Canny algorithms
There is the advantages of time-consuming short and precision is high, disclosure satisfy that vision guided navigation needs during intelligent tractor field rotary tillage process, there is weight
The application value wanted.
Brief description of the drawings
Fig. 1 is a kind of tractor rotary tillage vision navigation method flow chart based on new and old native boundary line.
Fig. 2 is the Steerable filter fate map in a kind of tractor rotary tillage vision navigation method based on new and old native boundary line.
Fig. 3 is that the Shearlet-Canny in a kind of tractor rotary tillage vision navigation method based on new and old native boundary line is calculated
Method flow chart.
Embodiment
Below in conjunction with the accompanying drawings and specific embodiment the invention will be further described.
Embodiment 1
As shown in Figure 1, a kind of tractor rotary tillage vision navigation method based on new and old native boundary line, it is real according to following step
Apply:
Step 1:Visual pattern p (x, y) in front of tractor is gathered by camera;
Step 2:Gray processing is carried out, and is changed by formula f (x, y)=(R (x, y)+G (x, y)+B (x, y))/3 pairs of images
To YCrCb color spaces, wherein Y=0.299*R+0.587*G+0.114*B, Cr=(R-Y) * 0.713+128, Cb=(B-
Y)*0.564+128;
Step 3:Steerable filter processing is carried out to image under YCrCb color spaces;
As shown in Fig. 2, the wherein processing of Steerable filter, i.e. " Local Linear Model " the solution course of work are:
Make the value that q is output pixel;I and k is pixel index;I is the value of input picture, i.e., image to be filtered or other
The navigational figure of image;A and b is the coefficient that window center is located at linear function when at k;P is image to be filtered;It is anti-that ε, which is,
Only a values it is excessive and introduce have adjust filter effect parameter, ε is bigger, and filter effect is more obvious;μk, σ2K is respectively that I exists
Average value and variance in window;It is averages of the image p to be filtered in window;| w | it is the quantity that pixel is included in window;i
It is pixel with j;wijIt is a filtering core, for the function being oriented between image I and independent variable p;
Step is 1.:Image under YCrCb color spaces represents that the input of this function passes through a two dimension with two-dimensional function
The output and function input that window obtains meets linear relationship, i.e.,:qi=akIi+bk,
Step is 2.:To step 1. in formula both sides do gradient algorithm, i.e.,
Step is 3.:Gap between the actual value p and real output value of digital simulation function is
Step is 4.:A is calculated based on least square methodkAnd bk,bk=pk-akμk;
Step is 5.:All linear function values comprising k points are done weighted average to obtain
Step 4:After Steerable filter processing, using the marginal information of Shearlet-Canny operator extraction research images;
As shown in figure 3, algorithm therein is as follows:
Order research image is f [n1, n2];AaFor anisotropic expansion matrix;BsTo shear matrix;A > 0 are scale parameter;
S ∈ R are shear parameters;T ∈ R are translation parameters;
Step a:Read in image f [n1, n2];
Step b:By laplacian pyramid by image f [n1, n2] it is decomposed into low pass subbandWith high pass subband
Step c:By high pass subbandPseudo- polar coordinate system is transformed into from cartesian coordinate system, produced matrix passes through a frequency
Domain sub-filter, pseudo- polar coordinate system convert back cartesian coordinate system;
Step d:Some subgraph outputs are tried to achieve in the multi-direction characteristic converted using Shearlet, conversion, wherein
Shearlet systems are represented byShearlet is transformed to SHψF (a, s, t)=
< f, ψA, s, t>, carries out canny edge detections to the subgraph of multiple directions, obtains respective edge image respectively;
Step e:Shearlet inverse transformations are carried out to the subgraph of all directions in step 2;
Step f:Principle that can be complementary according to different images edge, carries out image after inverse transformation using logical operator
Fusion;
Step 5:Hough transformation extracts target navigation line;
The algorithm steps of wherein Hough transformation are as follows:
The slope that m is straight line is made, c is intercept;
(i) in image X-Y, all conllinear points (x, y) are described as y=mx+c with linear equation;
(ii) straight line is regarded to the straight line equation in parameter space M-C as, the slope of its cathetus is x, and intercept is
y;
(iii), instead of former linear equation, ρ=x cos θ are expressed as with the straight line polar equation of Duda and Hart propositions
+ y sin θs, ρ are distance of the origin to straight line, and θ crosses the vertical line of origin and the angle of positive direction of the x-axis for straight line;
(iiii), it is necessary to carry out discretization to parameter space during calculating, the center point coordinate of each unit is:
Step 6:Go to step 1.
Claims (4)
1. a kind of tractor rotary tillage vision navigation method based on new and old native boundary line, it is characterized in that:
Step 1:Visual pattern p (x, y) in front of tractor is gathered by camera;
Step 2:Gray processing is carried out, and is transformed into by formula f (x, y)=(R (x, y)+G (x, y)+B (x, y))/3 pairs of images
YCrCb color spaces, wherein Y=0.299*R+0.587*G+0.114*B, Cr=(R-Y) * 0.713+128, Cb=(B-Y) *
0.564+128;
Step 3:Steerable filter processing is carried out to image under YCrCb color spaces;
Step 4:To the image after Steerable filter processing, believed using the edge of Shearlet-Canny operator extraction research images
Breath;
Step 5:The marginal information in image is fitted by Hough transformation, target navigation line is extracted, with Duda and Hart
The straight line polar equation of proposition replaces former linear equation, be ρ=x cos θ+y sin θs, and ρ is origin to the distance of straight line, θ
The vertical line of origin and the angle of positive direction of the x-axis are crossed for straight line;During calculating, discretization is carried out to parameter space, each
The center point coordinate of unit is:
Step 6:Go to step 1.
2. in a kind of tractor rotary tillage vision navigation method based on new and old native boundary line described in claim 1 at Steerable filter
The calculating of reason, it is characterized in that, calculated according to following steps:
Make the value that q is output pixel;I and k is pixel index;I is the value of input picture, i.e., image to be filtered or other images
Navigational figure;A and b is the coefficient that window center is located at linear function when at k;P is image to be filtered;ε is to prevent a values
Excessive and introducing to have the parameter for adjusting filter effect, ε is bigger, and filter effect is more obvious;μk, σ2 kRespectively I is in the window
Average value and variance;It is averages of the image p to be filtered in window;| w | it is the quantity that pixel is included in window.I and j is picture
Element;wijIt is a filtering core, for the function being oriented between image I and independent variable p;
Step is 1.:Image under YCrCb color spaces represents that the input of this function passes through a two-dimentional window with two-dimensional function
Obtained output and function input meets linear relationship, i.e.,:qi=akIi+bk,
Step is 2.:To step 1. in formula both sides do gradient algorithm, i.e.,
Step is 3.:Gap between the actual value p and real output value of digital simulation function is
Step is 4.:A is calculated based on least square methodkAnd bk,bk=pk-akμk;
Step is 5.:All linear function values comprising k points are done weighted average to obtain。
3. in a kind of tractor rotary tillage vision navigation method based on new and old native boundary line described in claim 1
The calculating of Shearlet-Canny operators, it is characterized in that, calculated according to following steps:
Order research image is f [n1, n2];AaFor anisotropic expansion matrix;BsTo shear matrix.A > 0 are scale parameter;s∈R
For shear parameters;T ∈ R are translation parameters;
Step a:Read in image f [n1, n2];
Step b:By laplacian pyramid by image f [n1, n2] it is decomposed into low pass subband fa jWith high pass subband fa j;
Step c:By high pass subband fa jPseudo- polar coordinate system is transformed into from cartesian coordinate system, the matrix of generation passes through frequency domain
Band filter, pseudo- polar coordinate system convert back cartesian coordinate system;
Step d:Some subgraph outputs, wherein Shearlet systems are tried to achieve in the multi-direction characteristic converted using Shearlet, conversion
System is represented byShearlet is transformed to SHψF (a, s, t)=<F, ψA, s, t>, to more
The subgraph in a direction carries out canny edge detections respectively, obtains respective edge image;
Step e:Shearlet inverse transformations are carried out to the subgraph of all directions in step 2;
Step f:Principle that can be complementary according to different images edge, melts image after inverse transformation using logical operator
Close.
A kind of 4. Hough transformation in tractor rotary tillage vision navigation method based on new and old native boundary line described in claim 1
Algorithm, it is characterized in that:
The slope that m is straight line is made, c is intercept;
(i) in image X-Y, all conllinear points (x, y) are described as y=mx+c with linear equation;
(ii) straight line is regarded to the straight line equation in parameter space M-C as, the slope of its cathetus is x, intercept y;
(iii), instead of former linear equation, it is ρ=x cos θ+y sin θs with the straight line polar equation of Duda and Hart propositions;
(iiii), it is necessary to carry out discretization to parameter space during calculating, the center point coordinate of each unit is:
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2016109072729 | 2016-10-17 | ||
CN201610907272 | 2016-10-17 |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107957264A true CN107957264A (en) | 2018-04-24 |
Family
ID=61954615
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710389147.8A Pending CN107957264A (en) | 2016-10-17 | 2017-05-25 | A kind of tractor rotary tillage vision navigation method based on new and old native boundary line |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107957264A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110310239A (en) * | 2019-06-20 | 2019-10-08 | 四川阿泰因机器人智能装备有限公司 | It is a kind of to be fitted the image processing method for eliminating illumination effect based on characteristic value |
-
2017
- 2017-05-25 CN CN201710389147.8A patent/CN107957264A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110310239A (en) * | 2019-06-20 | 2019-10-08 | 四川阿泰因机器人智能装备有限公司 | It is a kind of to be fitted the image processing method for eliminating illumination effect based on characteristic value |
CN110310239B (en) * | 2019-06-20 | 2023-05-05 | 四川阿泰因机器人智能装备有限公司 | Image processing method for eliminating illumination influence based on characteristic value fitting |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Zhang et al. | New research methods for vegetation information extraction based on visible light remote sensing images from an unmanned aerial vehicle (UAV) | |
Reza et al. | Rice yield estimation based on K-means clustering with graph-cut segmentation using low-altitude UAV images | |
Samiappan et al. | Using unmanned aerial vehicles for high-resolution remote sensing to map invasive Phragmites australis in coastal wetlands | |
Lati et al. | Estimating plant growth parameters using an energy minimization-based stereovision model | |
KR102053582B1 (en) | Method of ground coverage classification by using image pattern learning based on deep learning | |
CN110020635A (en) | Growing area crops sophisticated category method and system based on unmanned plane image and satellite image | |
EP2923333B1 (en) | Method for the automatic creation of two- or three-dimensional building models | |
CN109376728A (en) | A kind of weeds in paddy field recognition methods and its application based on multiple features fusion and BP neural network | |
Khan et al. | UAV’s agricultural image segmentation predicated by clifford geometric algebra | |
JP7344987B2 (en) | Convolutional neural network construction method and system based on farmland images | |
Ospina et al. | Simultaneous mapping and crop row detection by fusing data from wide angle and telephoto images | |
Peng et al. | Binocular-vision-based structure from motion for 3-D reconstruction of plants | |
CN115687850A (en) | Method and device for calculating irrigation water demand of farmland | |
CN107957264A (en) | A kind of tractor rotary tillage vision navigation method based on new and old native boundary line | |
CN107578447B (en) | A kind of crop ridge location determining method and system based on unmanned plane image | |
CN105844264A (en) | Oil peony fruit image identification method based on stress | |
CN113569772A (en) | Remote sensing image farmland instance mask extraction method, system, equipment and storage medium | |
Hu et al. | Optimal scale extraction of farmland in coal mining areas with high groundwater levels based on visible light images from an unmanned aerial vehicle (UAV) | |
Mohammed Amean et al. | Automatic plant branch segmentation and classification using vesselness measure | |
CN100480628C (en) | Stereo image row tree 3-D information fetching method based on image division technology | |
Bupathy et al. | Optimizing low-cost UAV aerial image mosaicing for crop growth monitoring | |
CN104567872B (en) | A kind of extracting method and system of agricultural machinery and implement leading line | |
Karydas et al. | Fine scale mapping of agricultural landscape features to be used in environmental risk assessment in an olive cultivation area | |
CN113870278A (en) | Improved Mask R-CNN model-based satellite remote sensing image farmland block segmentation method | |
Avetisyan et al. | Modification in landscape horizontal structure, induced by changing environmental conditions: a case study of Haskovo region (Southeastern Bulgaria) |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180424 |