CN107818577A - A kind of Parts Recognition and localization method based on mixed model - Google Patents
A kind of Parts Recognition and localization method based on mixed model Download PDFInfo
- Publication number
- CN107818577A CN107818577A CN201711021221.7A CN201711021221A CN107818577A CN 107818577 A CN107818577 A CN 107818577A CN 201711021221 A CN201711021221 A CN 201711021221A CN 107818577 A CN107818577 A CN 107818577A
- Authority
- CN
- China
- Prior art keywords
- image
- pose
- point set
- template
- likelihood function
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/001—Industrial image inspection using an image reference approach
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30164—Workpiece; Machine component
Abstract
The invention discloses a kind of Parts Recognition and localization method based on mixed model, belongs to technical field of image processing, and method includes S1, collection part image, and the part in part image is identified using template matching algorithm, obtains pose corresponding to each part;S2, the edge point set for extracting each part identified respectively, and the edge point set and its direction vector of extraction template image;S3, based on gauss hybrid models, build the likelihood function of the geometric error between each part edge point set and template image edge point set respectively;S4, using EM algorithms constructed likelihood function is optimized, obtain the optimal value of the geometric error;S5, using the optimal value of the geometric error pose of corresponding part is modified, obtains the amendment pose of each part so that each part in the part image to be identified.The present invention has accurate identification and locating effect, and has preferable robustness.
Description
Technical field
The present invention relates to vision Industrial Robot Technology field, more particularly to a kind of Parts Recognition based on mixed model with
Localization method.
Background technology
Industrial robot is the comprehensive integration of the key technologies such as mechanical mechanism, motion planning and control system, has work
Efficiency high, it is reliable and stable, repeatable accuracy is good the advantages that, be widely used in welding, stacking, assembling, processing, detection, logistics carry
With spraying etc. field, played a significant role in labor-intensive production transition and upgrade.China's industrial robot sales volume in 2014
5.7 ten thousand, it has also become global first big Industrial Robot Market, what State Council in 2015 printed and distributed《Made in China 2025》It is proposed to add
Fast development intelligence equipment, encourages developing industry robot technology energetically.At present, Switzerland ABB, Japanese Fa Nake companies, day intrinsic safety
The giant such as river motor and German Ku Ka robots occupies the Chinese market share of Robot industry 60%, monopolizes high-end industrial robot
Core technology, therefore it is significant to develop proprietary technology.
Vision accurately identifies that location technology is industrial robot autonomous classification and locating element in unstructured moving grids, enters
The core technology of row flexible job, overcome conventional industrial robot by teaching program operation, lack adaptability to changes the problem of,
The development of intelligent industrial robot is significantly facilitated, has adapted to high accuracy, high speed and intelligent operation demand.Therefore research and develop
The vision of industrial robot accurately identifies that location technology is significant, has a extensive future.
At present, Parts Recognition generally uses template matching algorithm, and it is that one is moved on search image as template
The similarity of the window of size, calculation window image and template image, and find the process of the optimal pose of similarity.But the calculation
On the one hand method can only obtain discrete translation and running accuracy;On the other hand it is according to change in camera angles change, nonlinear optical
Change, block or the operating mode such as mixed and disorderly background under, part is commonly present geometry deformation with template image.Above-mentioned two aspects reason causes template
Matching algorithm identification positioning precision is low.
The content of the invention
It is an object of the invention to provide a kind of Parts Recognition and localization method based on mixed model, to improve vision work
Industry robot identifies and positioning precision.
To realize object above, the present invention uses a kind of Parts Recognition and localization method based on mixed model, including such as
Lower step:
S1, part image is gathered, and the part in part image is identified using template matching algorithm, obtain each part
Corresponding pose;
S2, the edge point set for extracting each part identified respectively, and extraction template image edge point set and its
Direction vector;
S3, based on gauss hybrid models, build respectively several between each part edge point set and template image edge point set
The likelihood function of what error;
S4, using EM algorithms constructed likelihood function is optimized, obtain the optimal value of the geometric error;
S5, using the optimal value of the geometric error pose of corresponding part is modified, obtains the amendment of each part
Pose is so that each part in the part image to be identified.
Wherein, after the step S5, in addition to:
S6, the amendment pose of each part changed into world coordinate system, obtain each part in world coordinate system
Pose so that each part in the part image to be identified.
Wherein, the step S1, is specifically included:
In the part during part image is identified using the template matching algorithm, plan is searched for based on Pyramid technology
Slightly, the pose of each part in the part image is obtained.
Wherein, it is described in the part during part image is identified using the template matching algorithm, based on pyramid point
Layer search strategy, obtains the pose of each part in the part image, specifically includes:
The rough pose of each part in the part image is identified using the template matching algorithm;
Calculate the image pyramid that the number of plies is adapted with the part image and template image;
Rough pose based on each part, in the pyramidal top progress once complete template matches of described image,
Each part in the top search part image;
Each part obtained in top search is tracked into the pyramidal bottom of described image, obtains the part drawing
The fine pose of each part as in.
Wherein, described step S3, is specifically included:
According to the scale factor of the edge point set of each part, the direction vector of template image edge point set and setting, meter
Calculate the probability that the template image marginal point concentrates arbitrfary point;
The probability of arbitrfary point is concentrated according to the template image marginal point, builds the part edge point set and template image
The negative log-likelihood function of the geometric error of edge point set part.
Wherein, described step S4, is specifically included:
S41, the parameter to the negative log-likelihood function carry out initialization process, obtain parameter initialization value, wherein negative
The parameter of log-likelihood function includes scale factor and geometric error;
S42, according to the parameter initialization value or the parameter of last negative log-likelihood function iteration, calculate recessive become
The posterior probability of amount;
S43, according to the posterior probability likelihood function of the geometric error and scale factor is optimized, obtained new
Parameter value;
S44, repeat step S42~S43, until the convergence of the likelihood function of the geometric error and scale factor, is obtained several
What error amount and scale factor.
Wherein, described step S6, is specifically included:
World coordinate system is fixed on plane reference plate;
Position auto―control and Intrinsic Matrix between plane reference plate and video camera is obtained by demarcation;
According to the position auto―control and Intrinsic Matrix, the part in the part image is calculated in world coordinate system
Pose.
Compared with prior art, there is following technique effect in the present invention:The present invention is first by based on normalized crosscorrelation
The part image that the template matching algorithm of coefficient collects slightly is identified, is obtained the rough pose of part, is then based on Gauss
Mixed model builds the likelihood function on geometric error between part and template image edge point set, and geometry is obtained using EM algorithms
The optimal value of error, and the pose of part is corrected using the optimal value of geometric error, realize zero in accurate identification part image
The purpose of part.Because edge has consistency to illumination variation, and the pose adjustment algorithm based on gauss hybrid models is to noise Shandong
Rod is high, so present invention can be suitably applied to non-linear illumination variation, accurate identification positioning that is mixed and disorderly or blocking part under operating mode, tool
There is good robustness.
Brief description of the drawings
Below in conjunction with the accompanying drawings, the embodiment of the present invention is described in detail:
Fig. 1 is a kind of schematic flow sheet of Parts Recognition and localization method based on mixed model in the present invention;
Fig. 2 is the part image collected in the present invention;
Fig. 3 is the template image in the present invention;
Fig. 4 is that the result identified in the present invention using the template matching algorithm of normalized-cross-correlation function to part image is shown
It is intended to;
Fig. 5 is the distribution schematic diagram of each part edge point set and template image edge point set in the present invention;
Fig. 6 is to obtain the result schematic diagram of part geometry error exact value based on gauss hybrid models in the present invention;
Fig. 7 is to all part poses are modified in part image result in the present invention using geometric error exact value
Schematic diagram;
Fig. 8 is the schematic flow sheet of Parts Recognition and localization method of the another kind based on mixed model in the present invention;
Fig. 9 is camera coordinate system and the schematic diagram of world coordinate system conversion in the present invention;
Figure 10 is the schematic diagram tested in the present invention using template matching algorithm and the inventive method.
Embodiment
In order to illustrate further the feature of the present invention, please refer to the following detailed descriptions related to the present invention and accompanying drawing.Institute
Accompanying drawing is only for reference and purposes of discussion, is not used for being any limitation as protection scope of the present invention.
As shown in figure 1, present embodiment discloses a kind of Parts Recognition and localization method based on mixed model, including it is as follows
Step S1~S5:
S1, part image is gathered, and the part in part image is identified using template matching algorithm, obtain each part
Corresponding pose;
It should be noted that the model MV120SC of Visual Co., Ltd's production, imaging resolution are regarded in the present embodiment using dimension
For the pixel of 1280 pixels × 960, the industrial camera that focal length is 8mm gathers image, and the image of collection is as shown in Figure 2.Then create
Template image is built as shown in figure 3, searching for plan using the template matching algorithm based on normalized-cross-correlation function and Pyramid technology
Slightly, part pose (x is identifiedi,θi), i=1,2,3 ..., k, k represent the number of parts identified, θiRepresent i-th part
Angle, xiThe positional value of i-th of part is represented, is a two-dimensional coordinate value, obtained recognition result is as shown in Figure 4.
It should be noted that the template matching algorithm concrete principle based on normalized-cross-correlation function is as follows:In part drawing
As upper one window as template image size of movement, the normalized crosscorrelation system of calculation window image and template image
Number, i.e. similarity;Find similarity and be more than pose of the position of some threshold value as part.Wherein, normalized crosscorrelation system
Number is:
In formula, Q is the quantity of template image pixel, and s (x, y) is any pixel coordinate point (x, y) on part image I
Similarity, (r, c) are the pixel coordinate on template image t, and t (r, c) is gray value of the template image at pixel (r, c) place.
Because template matching algorithm computing cost is big, therefore, accelerated using Pyramid technology search strategy, it is specific former
Reason is as follows:First, calculate has the image pyramid for being applicable the number of plies with part image and template image;Secondly, top
Pyramid carries out once complete template matches;Then, the part obtained in the top search of image pyramid is all tracked
The bottom of image pyramid, obtain the fine pose of part.Reach reduction computing cost, improve Parts Recognition speed.
S2, the edge point set for extracting each part identified respectively, and extraction template image edge point set and its
Direction vector;
It should be noted that as shown in figure 5, extract each part edge point set, X is designated asi={ xi,1,xi,2,…,xi,M,
M is part edge point total quantity;Template image edge point set and its direction vector are extracted, is designated as Y={ y respectively1,y2,…,yN}
With v={ v1,v2,…,vN, N is template image marginal point total quantity.
S3, based on gauss hybrid models, build respectively several between each part edge point set and template image edge point set
The likelihood function of what error;
It should be noted that the geometric error in the present embodiment includes rotation error matrix Ri, translation error matrix tiAnd contracting
Put error si.The building process of the likelihood function of the geometric error is:
(1) the collection Y that sets up an office derives from a gauss hybrid models, and the probability of arbitrfary point is in point set Y:
In formula:σ is scale factor and is more than zero, ynIt is a two-dimensional coordinate value for n-th of marginal point of template image, yn∈
Y, xi,m∈Xi, vnIt is ynDirection vector, vn∈ v, p () represent probability function, and T is transposition symbol.
(2) point set Y negative log-likelihood function is built:
In formula:xi,mFor i-th of part, m-th of marginal point.
S4, using EM algorithms constructed likelihood function is optimized, obtain the optimal value of the geometric error;
Wherein, rotation error matrix Ri, translation error matrix tiWith scaled error siOptimal value it is as shown in Figure 6.Fig. 6 and
Fig. 5 is compared, it is known that above-mentioned negative log-likelihood function can be effectively minimized in EM algorithms, calculate optimal geometric error value.
S5, using the optimal value of the geometric error pose of corresponding part is modified, obtains the amendment of each part
Pose is so that each part in the part image to be identified.
Wherein, schematic diagram after each part pose is corrected in part image is based on as shown in fig. 7, Fig. 7 compares with Fig. 4
The part pose correction algorithm of gauss hybrid models can effectively correct part pose, realize accurate identifying purpose.Need what is illustrated
It is that the present embodiment is slightly known by using the template matching algorithm of Normalized Cross Correlation Function to the part in part image
Not, the rough pose of part is obtained, is then based on the geometric error of gauss hybrid models structure part and template point set part seemingly
Right function, expectation-maximization algorithm (Expectation Maximization Algorithm, EM), obtains the excellent of geometric error
Change value, recycle the optimal value of geometric error to be modified the pose of corresponding part, can accurately identify zero in part image
Part.
Further, step S4, specifically comprise the following steps:
S41, the parameter to the negative log-likelihood function carry out initialization process, obtain parameter initialization value, wherein negative
The parameter of log-likelihood function includes scale factor σ and Ri、tiAnd si;
S42, according to the parameter initialization value or the parameter of last negative log-likelihood function iteration, calculate recessive become
The posterior probability of amount, wherein, posterior probability Q (xi,m|yn) calculation formula be:
S43, according to the posterior probability likelihood function of the geometric error and scale factor is optimized, obtained new
Parameter value;
It should be noted that the likelihood formula of geometric error and scale factor is:
S44, repeat step S42~S43, until the convergence of the likelihood function of the geometric error and scale factor, is obtained several
What error amount and scale factor.
It should be noted that utilize the geometric error value R after linear least square calculation optimizationi',ti',si', then
It is that zero can solve scale factor σ to seek scale factor σ derivative and make it.
Further, in step S5, the pose of corresponding part is modified using the optimal value of the geometric error, had
Body is:
The R obtained using optimizationi',ti',si', the pose of corresponding part is corrected, correction formula is specially:
In formula:R'i(1,1) the spin matrix R' of i-th of part is representediElement value on 1st row the 1st row, x'iRefer to amendment
Part position value afterwards, θ 'iRefer to revised part angle value, t'iRepresent the translation matrix of i-th of part, s'iRepresent i-th
Scaled error after part optimization, is a constant.
Further, as shown in figure 8, a kind of Parts Recognition based on mixed model disclosed in the present embodiment and positioning side
Method, on the basis of above-described embodiment disclosure, also include S6 after step s 5:
S6, the amendment pose of each part changed into world coordinate system, obtain each part in world coordinate system
Pose so that each part in the part image to be identified, its detailed process is:
(1) world coordinate system is fixed on plane reference plate, the plane reference Board position by user according to the actual requirements
Set;
(2) position auto―control (R, T) and the Intrinsic Matrix K between plane reference plate and video camera are obtained by demarcation;
(3) according to the position auto―control (R, T) and Intrinsic Matrix K, the part in the part image is calculated in the world
Pose in coordinate system, order:
In formula:K refers to camera intrinsic parameter matrix, and R refers to the spin matrix between world coordinate system and camera coordinate system, and T refers to
Translation matrix between world coordinate system and camera coordinate system, H are homography matrixes, and [R T] representing matrix R and matrix T are carried out
Multiplication operation, hpfRefer to homography matrix pth row f column element values, 1≤p≤3,1≤f≤4.
(the X' being calculated according to equation belowi,Y'i) it is position of the part in world coordinate system:
In formula:x'i(1)、x'i(2) it is revised x'iTwo-dimensional coordinate value.
It should be noted that the present embodiment is according to industrial camera and the position auto―control and industrial camera of world coordinate system
Interior parameter matrix, part pose is changed into world coordinate system, transition diagram is as shown in figure 9, be easy to industrial robot
Part is picked up, can realize and part in part image is accurately positioned.
Tested as follows using template matching algorithm of the prior art and the inventive method:
(1) with the straight line in Hough transform detection template image long side and the ruler being next to part subject to registration,
L is designated as respectively1And L2, L1And L2Actual angle of the angle theta as part subject to registration, as shown in Figure 10;
(2) service precision 0.02mm vernier caliper measurements Central of the parts to ruler distance be d=17.60mm, such as Figure 10 institutes
Show;
(3) part pose value is calculated respectively using form fit algorithm and technical scheme, be designated as (x1,θ1) and
(x2,θ2), then calculate x1And x2To L2 distance, d is designated as respectively1And d2.What template matching algorithm and the inventive method obtained determines
Position result is as shown in table 1:
Table 1
From the experimental data in table 1:Using the inventive method measurement part distance and the error point of part angle
Not Wei -0.01mm ± 0.57mm and -0.04 ° ± 0.36 °, be superior to form fit algorithm, show the inventive method have it is higher,
Relatively stable positioning precision.And Fig. 4, Fig. 7 are combined, from visual angle as can be seen that using gauss hybrid models and EM algorithms
After being modified to template matching algorithm, even in non-linear light application ratio that word, mixed and disorderly or under the conditions of blocking under operating mode, to part
Identification positioning it is also more accurate.
The foregoing is only presently preferred embodiments of the present invention, be not intended to limit the invention, it is all the present invention spirit and
Within principle, any modification, equivalent substitution and improvements made etc., it should be included in the scope of the protection.
Claims (7)
1. a kind of Parts Recognition and localization method based on mixed model, it is characterised in that including:
S1, part image is gathered, and the part in part image is identified using template matching algorithm, it is corresponding to obtain each part
Pose;
S2, the edge point set for extracting each part identified respectively, and the edge point set of extraction template image and its direction
Vector;
S3, based on gauss hybrid models, the geometry built respectively between each part edge point set and template image edge point set misses
The likelihood function of difference;
S4, using EM algorithms constructed likelihood function is optimized, obtain the optimal value of the geometric error;
S5, using the optimal value of the geometric error pose of corresponding part is modified, obtains the amendment pose of each part
So that each part in the part image to be identified.
2. the method as described in claim 1, it is characterised in that after the step S5, in addition to:
S6, the amendment pose of each part changed into world coordinate system, obtain position of each part in world coordinate system
Appearance is so that each part in the part image to be identified.
3. the method as described in claim 1, it is characterised in that the step S1, specifically include:
In the part during part image is identified using the template matching algorithm, based on Pyramid technology search strategy, obtain
The pose of each part into the part image.
4. method as claimed in claim 3, it is characterised in that described to identify part drawing using the template matching algorithm
As in part when, based on Pyramid technology search strategy, obtain the pose of each part in the part image, specifically wrap
Include:
The rough pose of each part in the part image is identified using the template matching algorithm;
Calculate the image pyramid that the number of plies is adapted with the part image and template image;
Rough pose based on each part, in the pyramidal top progress once complete template matches of described image, most
Each part that high level is searched in the part image;
Each part obtained in top search is tracked into the pyramidal bottom of described image, obtained in the part image
The fine pose of each part.
5. method as claimed in claim 3, it is characterised in that described step S3, specifically include:
According to the scale factor of the edge point set of each part, the direction vector of template image edge point set and setting, institute is calculated
State the probability that template image marginal point concentrates arbitrfary point;
The probability of arbitrfary point is concentrated according to the template image marginal point, builds the part edge point set and template image edge
The negative log-likelihood function of the geometric error of point set part.
6. method as claimed in claim 5, it is characterised in that described step S4, specifically include:
S41, the parameter to the negative log-likelihood function carry out initialization process, obtain parameter initialization value, wherein negative logarithm
The parameter of likelihood function includes scale factor and geometric error;
S42, according to the parameter initialization value or the parameter of last negative log-likelihood function iteration, calculate recessive variable
Posterior probability;
S43, according to the posterior probability likelihood function of the geometric error and scale factor is optimized, obtain new ginseng
Numerical value;
S44, repeat step S42~S43, until the convergence of the likelihood function of the geometric error and scale factor, obtains geometry mistake
Difference and scale factor.
7. method as claimed in claim 2, it is characterised in that described step S6, specifically include:
World coordinate system is fixed on plane reference plate;
Position auto―control and Intrinsic Matrix between plane reference plate and video camera is obtained by demarcation;
According to the position auto―control and Intrinsic Matrix, position of the part in the part image in world coordinate system is calculated
Appearance.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711021221.7A CN107818577A (en) | 2017-10-26 | 2017-10-26 | A kind of Parts Recognition and localization method based on mixed model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711021221.7A CN107818577A (en) | 2017-10-26 | 2017-10-26 | A kind of Parts Recognition and localization method based on mixed model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107818577A true CN107818577A (en) | 2018-03-20 |
Family
ID=61604245
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711021221.7A Pending CN107818577A (en) | 2017-10-26 | 2017-10-26 | A kind of Parts Recognition and localization method based on mixed model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107818577A (en) |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108996268A (en) * | 2018-08-01 | 2018-12-14 | 上海主线科技有限公司 | A kind of container tractor based on camera and suspension bridge are mutually located method |
CN109726481A (en) * | 2018-12-29 | 2019-05-07 | 深圳市优必选科技有限公司 | A kind of householder method that robot is built, device and terminal device |
CN110532897A (en) * | 2019-08-07 | 2019-12-03 | 北京科技大学 | The method and apparatus of components image recognition |
CN110706285A (en) * | 2019-10-08 | 2020-01-17 | 中国人民解放军陆军工程大学 | Object pose prediction method based on CAD model |
CN111013883A (en) * | 2019-11-25 | 2020-04-17 | 浙江明泉工业涂装有限公司 | Robot control method for intelligent spraying of multiple vehicle types |
CN111553952A (en) * | 2020-05-08 | 2020-08-18 | 中国科学院自动化研究所 | Industrial robot visual image identification method and system based on survival countermeasure |
CN111604909A (en) * | 2020-06-24 | 2020-09-01 | 辽宁工业大学 | Visual system of four-axis industrial stacking robot |
CN112308121A (en) * | 2020-10-16 | 2021-02-02 | 易思维(杭州)科技有限公司 | Template image edge point optimization method |
CN113537237A (en) * | 2021-06-25 | 2021-10-22 | 西安交通大学 | Intelligent sensing method, system and device for multi-feature part quality information |
CN113610815A (en) * | 2021-08-10 | 2021-11-05 | 宿迁旺春机械制造有限公司 | Structural member production safety early warning method and system based on image processing |
CN113681560A (en) * | 2021-08-26 | 2021-11-23 | 浙江大学 | Method for mechanical arm to operate hinged object based on vision fusion |
CN113758439A (en) * | 2021-08-23 | 2021-12-07 | 武汉理工大学 | Method and device for measuring geometric parameters on line in hot ring rolling forming process |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101369346A (en) * | 2007-08-13 | 2009-02-18 | 北京航空航天大学 | Tracing method for video movement objective self-adapting window |
CN102637298A (en) * | 2011-12-31 | 2012-08-15 | 辽宁师范大学 | Color image segmentation method based on Gaussian mixture model and support vector machine |
CN105021124A (en) * | 2015-04-16 | 2015-11-04 | 华南农业大学 | Planar component three-dimensional position and normal vector calculation method based on depth map |
CN105957342A (en) * | 2016-05-30 | 2016-09-21 | 武汉大学 | Lane-level road mapping method and system based on crowdsourcing space-time big data |
CN106056124A (en) * | 2016-05-31 | 2016-10-26 | 佛山市天目工业测量科技有限公司 | Relief extraction method based on three-dimensional curved surface loop guidance |
CN106611413A (en) * | 2016-11-30 | 2017-05-03 | 上海联影医疗科技有限公司 | Image segmentation method and system |
-
2017
- 2017-10-26 CN CN201711021221.7A patent/CN107818577A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101369346A (en) * | 2007-08-13 | 2009-02-18 | 北京航空航天大学 | Tracing method for video movement objective self-adapting window |
CN102637298A (en) * | 2011-12-31 | 2012-08-15 | 辽宁师范大学 | Color image segmentation method based on Gaussian mixture model and support vector machine |
CN105021124A (en) * | 2015-04-16 | 2015-11-04 | 华南农业大学 | Planar component three-dimensional position and normal vector calculation method based on depth map |
CN105957342A (en) * | 2016-05-30 | 2016-09-21 | 武汉大学 | Lane-level road mapping method and system based on crowdsourcing space-time big data |
CN106056124A (en) * | 2016-05-31 | 2016-10-26 | 佛山市天目工业测量科技有限公司 | Relief extraction method based on three-dimensional curved surface loop guidance |
CN106611413A (en) * | 2016-11-30 | 2017-05-03 | 上海联影医疗科技有限公司 | Image segmentation method and system |
Non-Patent Citations (3)
Title |
---|
彭磊 等: "基于邻域结构和高斯混合模型的非刚性点集配准算法", 《电子与信息学报》 * |
李春: "基于机器视觉的焊接工件识别与焊接轨迹校正方法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
邹湘军 等: "改进迭代最近点法的亚像素级零件图像配准", 《计算机辅助设计与图形学学报》 * |
Cited By (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108996268A (en) * | 2018-08-01 | 2018-12-14 | 上海主线科技有限公司 | A kind of container tractor based on camera and suspension bridge are mutually located method |
CN109726481A (en) * | 2018-12-29 | 2019-05-07 | 深圳市优必选科技有限公司 | A kind of householder method that robot is built, device and terminal device |
CN109726481B (en) * | 2018-12-29 | 2023-08-04 | 深圳市优必选科技有限公司 | Auxiliary method and device for robot construction and terminal equipment |
US11636779B2 (en) | 2018-12-29 | 2023-04-25 | Ubtech Robotics Corp Ltd | Method, apparatus and terminal device for constructing parts together |
CN110532897B (en) * | 2019-08-07 | 2022-01-04 | 北京科技大学 | Method and device for recognizing image of part |
CN110532897A (en) * | 2019-08-07 | 2019-12-03 | 北京科技大学 | The method and apparatus of components image recognition |
CN110706285A (en) * | 2019-10-08 | 2020-01-17 | 中国人民解放军陆军工程大学 | Object pose prediction method based on CAD model |
CN111013883A (en) * | 2019-11-25 | 2020-04-17 | 浙江明泉工业涂装有限公司 | Robot control method for intelligent spraying of multiple vehicle types |
CN111553952A (en) * | 2020-05-08 | 2020-08-18 | 中国科学院自动化研究所 | Industrial robot visual image identification method and system based on survival countermeasure |
CN111604909A (en) * | 2020-06-24 | 2020-09-01 | 辽宁工业大学 | Visual system of four-axis industrial stacking robot |
CN112308121A (en) * | 2020-10-16 | 2021-02-02 | 易思维(杭州)科技有限公司 | Template image edge point optimization method |
CN112308121B (en) * | 2020-10-16 | 2022-06-14 | 易思维(杭州)科技有限公司 | Template image edge point optimization method |
CN113537237A (en) * | 2021-06-25 | 2021-10-22 | 西安交通大学 | Intelligent sensing method, system and device for multi-feature part quality information |
CN113537237B (en) * | 2021-06-25 | 2024-01-16 | 西安交通大学 | Multi-feature part quality information intelligent sensing method, system and device |
CN113610815A (en) * | 2021-08-10 | 2021-11-05 | 宿迁旺春机械制造有限公司 | Structural member production safety early warning method and system based on image processing |
CN113610815B (en) * | 2021-08-10 | 2022-07-05 | 宿迁旺春机械制造有限公司 | Structural member production safety early warning method and system based on image processing |
CN113758439A (en) * | 2021-08-23 | 2021-12-07 | 武汉理工大学 | Method and device for measuring geometric parameters on line in hot ring rolling forming process |
CN113681560A (en) * | 2021-08-26 | 2021-11-23 | 浙江大学 | Method for mechanical arm to operate hinged object based on vision fusion |
CN113681560B (en) * | 2021-08-26 | 2023-07-04 | 浙江大学 | Method for operating articulated object by mechanical arm based on vision fusion |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107818577A (en) | A kind of Parts Recognition and localization method based on mixed model | |
CN103886619B (en) | A kind of method for tracking target merging multiple dimensioned super-pixel | |
CN110966991A (en) | Single unmanned aerial vehicle image positioning method without control point | |
CN107741234A (en) | The offline map structuring and localization method of a kind of view-based access control model | |
US11847803B2 (en) | Hand trajectory recognition method for following robot based on hand velocity and trajectory distribution | |
CN105354531A (en) | Marking method for facial key points | |
CN103440667B (en) | The automaton that under a kind of occlusion state, moving target is stably followed the trail of | |
CN109472816B (en) | Point cloud registration method | |
CN105354841B (en) | A kind of rapid remote sensing image matching method and system | |
CN105335973A (en) | Visual processing method for strip steel processing production line | |
CN104091350B (en) | A kind of object tracking methods of utilization motion blur information | |
CN112484746B (en) | Monocular vision auxiliary laser radar odometer method based on ground plane | |
CN112132874A (en) | Calibration-board-free different-source image registration method and device, electronic equipment and storage medium | |
CN106327528A (en) | Moving object tracking method and operation method of unmanned aerial vehicle | |
Ma et al. | Crlf: Automatic calibration and refinement based on line feature for lidar and camera in road scenes | |
CN110553650B (en) | Mobile robot repositioning method based on small sample learning | |
CN109784297A (en) | A kind of Three-dimensional target recognition based on deep learning and Optimal Grasp method | |
Wu et al. | Location recognition algorithm for vision-based industrial sorting robot via deep learning | |
Yang et al. | A method for wafer defect detection using spatial feature points guided affine iterative closest point algorithm | |
CN113295171B (en) | Monocular vision-based attitude estimation method for rotating rigid body spacecraft | |
CN102324043B (en) | Image matching method based on DCT (Discrete Cosine Transformation) through feature description operator and optimization space quantization | |
CN113420590B (en) | Robot positioning method, device, equipment and medium in weak texture environment | |
CN113420648B (en) | Target detection method and system with rotation adaptability | |
CN107610120B (en) | A kind of multiple dimensioned building surface Methodology for Entities Matching and system | |
CN111914832B (en) | SLAM method of RGB-D camera under dynamic scene |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180320 |
|
RJ01 | Rejection of invention patent application after publication |