CN109848984A - A kind of visual servo method controlled based on SVM and ratio - Google Patents
A kind of visual servo method controlled based on SVM and ratio Download PDFInfo
- Publication number
- CN109848984A CN109848984A CN201811643848.0A CN201811643848A CN109848984A CN 109848984 A CN109848984 A CN 109848984A CN 201811643848 A CN201811643848 A CN 201811643848A CN 109848984 A CN109848984 A CN 109848984A
- Authority
- CN
- China
- Prior art keywords
- image
- svm
- target
- vector
- target image
- 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
Landscapes
- Image Analysis (AREA)
Abstract
The present invention is suitable for automatic control technology field, provides a kind of visual servo method controlled based on SVM and ratio, this method comprises: S1, being trained to sample data based on SVM training pattern, exports the Jacobian matrix based on SVM；S2, characteristics of image difference value vector and Jacobian matrix based on present image and target image obtain expectation joint angle vector when robot reaches target position；S3, robot arm joint is controlled based on desired joint angle vector towards target position movement.It replaces gloomy matrix to solve optimization problem with proportional controlling means, greatly reduces calculation amount；Contacting for picture position feature and target joint angle is directly established by Jacobian matrix, and using joint angle as output quantity, so that control mode is more direct.
Description
Technical field
The invention belongs to automatic control technology fields, provide a kind of visual servo method controlled based on SVM and ratio.
Background technique
Robotics development, flexibility of the machine man-based development to robot, versatility etc. require higher and higher, machine
Visual development has developed out SVR from SVM.Without calibration visual servo, when target object location is unknown and camera parameter does not pass through
In the case where calibration, it still is able to control end effector of robot and is accurately positioned to right above target object.Existing is to utilize ox
The method combination SVR that pauses is iterated the method approached, and needs to obtain matrix (Jacobian matrix) of the control amount relative to changing features,
And need to calculate the change rate of two variable quantities, that is, gloomy matrix, the calculating of gloomy matrix very very complicated, so that
Calculation amount greatly increases.
Summary of the invention
The embodiment of the present invention provides a kind of visual servo method controlled based on SVM and ratio, it is desirable to provide a kind of calculating
Measure relatively small visual servo method.
In order to solve above-mentioned purpose, the present invention provides a kind of visual servo method controlled based on SVM and ratio, the side
Method includes the following steps:
S1, sample data is trained based on SVM training pattern, exports the Jacobian matrix based on SVM；
S2, picture position feature difference vector and Jacobian matrix based on present image and target image obtain machine
People reaches expectation joint angle vector when target position；
S3, robot arm joint is controlled based on the difference that target joint is angularly measured towards target position movement.
Further, the calculation formula of the expectation joint angle vector is specific as follows:
Wherein, θgFor the expectation joint angle vector of p degree of freedom robot, θkFor the pass at p degree of freedom robot current k moment
The angular amount of section, J is Jacobian matrix,For the picture position feature difference vector of target image and present image,fgFor the picture position feature vector of target image, fkFor the picture position feature vector of present image, KpFor
Proportionality coefficient.
Further, after step s 3 further include:
Detection robot arm joint currently whether have been located in target position, if testing result be it is no, then follow the steps S2.
Further, the picture position feature vector f of the present imagekAcquisition methods include the following steps:
S21, gray proces are carried out to the original image of shooting, obtains image one；
S22, Threshold segmentation is carried out based on threshold value a pair of image one, obtains image two, closed side is extracted in image two
Edge profile；
S23, Threshold segmentation is carried out to image two based on threshold value two, obtains target image, threshold value two is the side of target image
The ratio of edge profile enclosing region area and edge contour perimeter；
S24, coordinate of all pixels point in image coordinate system in target image region is calculated；
S25, position feature of the target image in image coordinate system, institute's rheme are calculated based on normalization geometry Moment Methods
Feature is set to be made of the centre coordinate and target image drift angle of target image.
The visual servo method provided by the invention controlled based on SVM and ratio is had the following beneficial effects:
1. replacing gloomy matrix to solve optimization problem with proportional controlling means, greatly reduce calculation amount；
2. directly establishing contacting for picture position feature and target joint angle by Jacobian matrix, and joint angle is made
For output quantity, so that control mode is more direct.
Detailed description of the invention
Fig. 1 is the visual servo method flow diagram provided in an embodiment of the present invention controlled based on SVM and ratio；
Fig. 2 is the schematic diagram without calibration Visual servoing control provided in an embodiment of the present invention based on SVR.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
Image information and joint angle informational linkage are got up by Jacobian matrix, and use support vector regression
It is fitted the value of Jacobian matrix, finally proportional controlling means is combined to drive robot arm step by step by continuous interative computation
Target position is reached, the parameter in support vector regression therein is needed through processes such as sample collection, the training of SVR algorithm,
By designing final control algolithm experimental verification, the feasibility and robustness of this control method ensure that.
Fig. 1 is the visual servo method flow diagram provided in an embodiment of the present invention controlled based on SVM and ratio, this method packet
Include following steps:
S1, sample data is trained based on SVM training pattern, exports the Jacobian matrix based on SVM；
In embodiments of the present invention, camera is parallel to target object and shoots, and in the sample image of acquisition, target object is being schemed
As referred to as target image, target image be must be it is complete, therefore, sample data is the image that record has complete object object
Data carry out the deletion of redundant samples data, and the deletion of lopsided sample data for the sample data of acquisition.
In embodiments of the present invention, the acquisition methods of Jacobian matrix are specific as follows: acquisition data are carried out at matrixing
Reason obtains training data；Training data is trained based on SVM training pattern, until error is located at tolerance range
Either reach specified the number of iterations, exports the Jacobian matrix based on SVM.Training data selection SVM type be
CvSVM::EPS_SVR is obtained using gaussian kernel function, ε-insensitive loss function, and by the crosscheck of K folding, grid data service
To in preferably kernel function parameter gamma, penalty factor, should be less than distance P with hyperplane, it is straight to carry out automation training
To the termination condition for meeting error requirements or arrival given number of iterations.
S2, picture position feature difference vector and Jacobian matrix based on present image and target image obtain machine
People reaches expectation joint angle vector when target position；
In embodiments of the present invention, it is expected that the calculation formula of joint angle vector is specific as follows:
Wherein, θgFor the expectation joint angle vector of p degree of freedom robot, θkFor the pass at p degree of freedom robot current k moment
The angular amount of section, J is Jacobian matrix,For the picture position feature difference vector of target image and present image, it is expressed asfgIt is scalar quantity, f for the picture position feature vector of target imagekFor the picture position feature of present image
Vector, KpIt is empirical parameter for proportionality coefficient.
S3, robot arm joint is controlled based on desired joint angle vector towards target position movement, control process tool
Body is as shown in Figure 2.
After step s 3 further include:
Whether detection robot arm joint currently has moved to target position, if testing result is no, thens follow the steps
S2, if testing result be it is yes, control terminate.
The visual servo method provided by the invention controlled based on SVM and ratio is had the following beneficial effects:
1. replacing gloomy matrix to solve optimization problem with proportional controlling means, greatly reduce calculation amount；
2. directly establishing contacting for picture position feature and target joint angle by Jacobian matrix, and joint angle is made
For output quantity, so that control mode is more direct.
In embodiments of the present invention, the picture position feature vector f of present imagekAcquisition methods include the following steps:
S21, gray proces are carried out to the original image of shooting, obtains image one；
S22, Threshold segmentation is carried out based on threshold value a pair of image one, obtains image two, closed side is extracted in image two
Edge profile；
In embodiments of the present invention, threshold value carries out threshold value point based on threshold value a pair of image one first is that histogram method is obtained
It cutting, eliminates most of noise in background, the noise as similar in brightness and target object not can be removed, extract figure
As in closed edge contour, partially it is made of noise.
S23, Threshold segmentation is carried out to image two based on threshold value two, obtains target image, threshold value two is the side of target image
The ratio of edge profile enclosing region area and edge contour perimeter；
In embodiments of the present invention, Threshold segmentation is carried out to image again based on threshold value two, brightness and target can be removed
Noise similar in object, obtains target image.
S24, coordinate of all pixels point in image coordinate system in target image region is calculated；
S25, position feature of the target image in image coordinate system, institute's rheme are calculated based on normalization geometry Moment Methods
Feature is set to be made of the centre coordinate and target image drift angle of target image.
Picture position feature of the invention is divided into two parts during the extraction process: utilizing image border and object region
Information extracts accurately target image；Target image region is only traversed, to calculate picture position feature, relative to whole
The ratio of a image-region, region shared by target image is small, greatly subtracts the extraction time for shortening image object feature, target figure
As the precision extracted ensure that the precision of picture position feature.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within mind and principle.
Claims (4)
1. a kind of visual servo method controlled based on SVM and ratio, which is characterized in that described method includes following steps:
S1, sample data is trained based on SVM training pattern, exports the Jacobian matrix based on SVM；
S2, picture position feature difference vector and Jacobian matrix based on present image and target image obtain robot and reach
Expectation joint angle vector when to target position；
S3, it is moved based on desired joint angle vector majorization robot arm joint towards target position.
2. the visual servo method controlled as described in claim 1 based on SVM and ratio, which is characterized in that after step s 3
Further include:
Detection robot arm joint currently whether have been located in target position, if testing result be it is no, then follow the steps S2.
3. the visual servo method controlled as described in claim 1 based on SVM and ratio, which is characterized in that the expectation joint
The calculation formula angularly measured is specific as follows:
θg=θk+Kp·JT(JJT)-1Δfgk
Wherein, θgFor the expectation joint angle vector of p degree of freedom robot, θkFor the joint angle at p degree of freedom robot current k moment
Vector, J are Jacobian matrix, Δ fgkFor the picture position feature difference vector of target image and present image, Δ fgk=fg-
fk, fgFor the picture position feature vector of target image, fkFor the picture position feature vector of present image, KpFor proportionality coefficient.
4. the visual servo method controlled as described in claim 1 based on SVM and ratio, which is characterized in that the present image
Picture position feature vector fkAcquisition methods include the following steps:
S21, gray proces are carried out to the original image of shooting, obtains image one；
S22, Threshold segmentation is carried out based on threshold value a pair of image one, obtains image two, closed edge wheel is extracted in image two
It is wide；
S23, Threshold segmentation is carried out to image two based on threshold value two, obtains target image, threshold value two is the edge wheel of target image
The ratio of wide enclosing region area and edge contour perimeter；
S24, coordinate of all pixels point in image coordinate system in target image region is calculated；
S25, position feature of the target image in image coordinate system is calculated based on normalization geometry Moment Methods, the position is special
Sign is made of the centre coordinate and target image drift angle of target image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811643848.0A CN109848984A (en) | 2018-12-29 | 2018-12-29 | A kind of visual servo method controlled based on SVM and ratio |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811643848.0A CN109848984A (en) | 2018-12-29 | 2018-12-29 | A kind of visual servo method controlled based on SVM and ratio |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109848984A true CN109848984A (en) | 2019-06-07 |
Family
ID=66893513
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811643848.0A Pending CN109848984A (en) | 2018-12-29 | 2018-12-29 | A kind of visual servo method controlled based on SVM and ratio |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109848984A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111546344A (en) * | 2020-05-18 | 2020-08-18 | 北京邮电大学 | Mechanical arm control method for alignment |
| CN114946403A (en) * | 2022-07-06 | 2022-08-30 | 青岛科技大学 | Tea picking robot based on calibration-free visual servo and tea picking control method thereof |
| CN116079697A (en) * | 2022-12-23 | 2023-05-09 | 北京纳通医用机器人科技有限公司 | Monocular vision servo method, device, equipment and medium based on image |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104808490A (en) * | 2015-03-10 | 2015-07-29 | 浙江工业大学 | Uncalibrated visual servoing control method for estimating image Jacobian matrix based on echo state network facing mold protection |
| CN104898421A (en) * | 2015-05-05 | 2015-09-09 | 上海大学 | Uncalibrated hand-eye coordination fussy control method based on support vector regression (SVR) learning |
| CN106981075A (en) * | 2017-05-31 | 2017-07-25 | 江西制造职业技术学院 | The skeleton point parameter acquisition devices of apery motion mimicry and its recognition methods |
| CN107901041A (en) * | 2017-12-15 | 2018-04-13 | 中南大学 | A kind of robot vision servo control method based on image blend square |
-
2018
- 2018-12-29 CN CN201811643848.0A patent/CN109848984A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104808490A (en) * | 2015-03-10 | 2015-07-29 | 浙江工业大学 | Uncalibrated visual servoing control method for estimating image Jacobian matrix based on echo state network facing mold protection |
| CN104898421A (en) * | 2015-05-05 | 2015-09-09 | 上海大学 | Uncalibrated hand-eye coordination fussy control method based on support vector regression (SVR) learning |
| CN106981075A (en) * | 2017-05-31 | 2017-07-25 | 江西制造职业技术学院 | The skeleton point parameter acquisition devices of apery motion mimicry and its recognition methods |
| CN107901041A (en) * | 2017-12-15 | 2018-04-13 | 中南大学 | A kind of robot vision servo control method based on image blend square |
Non-Patent Citations (1)
| Title |
|---|
| 陆雪强: ""基于无标定视觉伺服的机器人视觉引导装配作业研究"", 《中国优秀硕士学位论文全文数据库（信息科技辑）》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111546344A (en) * | 2020-05-18 | 2020-08-18 | 北京邮电大学 | Mechanical arm control method for alignment |
| CN114946403A (en) * | 2022-07-06 | 2022-08-30 | 青岛科技大学 | Tea picking robot based on calibration-free visual servo and tea picking control method thereof |
| CN116079697A (en) * | 2022-12-23 | 2023-05-09 | 北京纳通医用机器人科技有限公司 | Monocular vision servo method, device, equipment and medium based on image |
| CN116079697B (en) * | 2022-12-23 | 2024-05-28 | 北京纳通医用机器人科技有限公司 | Monocular vision servo method, device, equipment and medium based on image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109202912B (en) | Method for registering target contour point cloud based on monocular depth sensor and mechanical arm | |
| CN109035204B (en) | Real-time detection method for weld joint target | |
| CN108460779B (en) | Mobile robot image visual positioning method in dynamic environment | |
| CN109848984A (en) | A kind of visual servo method controlled based on SVM and ratio | |
| CN104463851B (en) | A kind of sole edge line automatic tracking method based on robot | |
| CN107186708B (en) | Hand-eye servo robot grabbing system and method based on deep learning image segmentation technology | |
| CN109102525B (en) | Mobile robot following control method based on self-adaptive posture estimation | |
| CN111707272B (en) | Underground garage automatic driving laser positioning system | |
| CN111784778B (en) | Binocular camera external parameter calibration method and system based on linear solving and nonlinear optimization | |
| CN103198477B (en) | Apple fruitlet bagging robot visual positioning method | |
| CN109102522A (en) | A kind of method for tracking target and device | |
| CN112001955A (en) | Point cloud registration method and system based on two-dimensional projection plane matching constraint | |
| CN106650701B (en) | Binocular vision-based obstacle detection method and device in indoor shadow environment | |
| CN112964186B (en) | Device and method for measuring clearance in automatic shaft hole assembly process | |
| Grigorescu et al. | Robust camera pose and scene structure analysis for service robotics | |
| Geng et al. | A novel welding path planning method based on point cloud for robotic welding of impeller blades | |
| CN110516548B (en) | Iris center positioning method based on three-dimensional eyeball model and Snakucle | |
| CN106595601B (en) | Accurate repositioning method for camera pose with six degrees of freedom without hand-eye calibration | |
| EP4322020A1 (en) | Terminal device positioning method and related device therefor | |
| CN111624875A (en) | Visual servo control method and device and unmanned equipment | |
| CN106097277B (en) | A kind of rope substance point-tracking method that view-based access control model measures | |
| Delabarre et al. | Visual servoing using the sum of conditional variance | |
| CN112767481B (en) | High-precision positioning and mapping method based on visual edge features | |
| CN112268548B (en) | Airplane local appearance measuring method based on binocular vision | |
| CN114419259B (en) | Visual positioning method and system based on physical model imaging simulation |
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: 20190607 |