CN109903332A - A kind of object's pose estimation method based on deep learning - Google Patents
A kind of object's pose estimation method based on deep learning Download PDFInfo
- Publication number
- CN109903332A CN109903332A CN201910016293.5A CN201910016293A CN109903332A CN 109903332 A CN109903332 A CN 109903332A CN 201910016293 A CN201910016293 A CN 201910016293A CN 109903332 A CN109903332 A CN 109903332A
- Authority
- CN
- China
- Prior art keywords
- target object
- body surface
- plane
- obtains
- acquired
- 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 object's pose estimation method based on deep learning that the present invention relates to a kind of.It needs to grab target object using mechanical arm in actual industrial environment, needs to obtain the spatial positional information and posture information of target object first.Camera chain is cheap, so most widely used come the method for carrying out the Attitude estimation of target object using visual information.It carries out Attitude estimation using traditional vision algorithm to be difficult to extract effective feature, ratio of precision is relatively limited.The advantage of neural network is utilized in the method for the present invention, and the method that models after being extracted target object important area using neural network algorithm to body surface estimates gestures of object.This method is adaptable, for different types of object, need to only collect data set and carry out re -training to neural network, without redesigning feature extractor.And this method determines that posture is accurate, using the powerful ability in feature extraction of neural network, can estimate to analyze the object in most of scene.
Description
Technical field
The invention belongs to computer vision fields, and in particular to a kind of object's pose estimation method based on deep learning.
Background technique
The posture information that target object is obtained using visual information is the vital task of computer vision.In actual work
In industry environment, needs to grab target object using mechanical arm, need to obtain the spatial three-dimensional position of target object
Information and posture information.So the accuracy of Attitude estimation is extremely important for mechanical arm crawl object.Currently, posture is estimated
The method of meter mainly has: 1) based on the Attitude estimation of study;2) based on the Attitude estimation of model.
Estimated based on the Attitude estimation of model mainly in conjunction with the characteristic point information of object or the geometrical relationship of object
Count the posture of object.It is the structure and shape that object is indicated by the geometrical characteristic or geometrical model of object, is then passed through
Image and the direct matching relationship of model are established to determine the posture information of object in three dimensions.This method is usually to compare
The relationship between feature that the model relatively set up and input picture extract, goes to determine model by Feature Correspondence Algorithm
Posture.
Attitude estimation method based on study refers to using machine learning or deep learning, from training dataset middle school
The disaggregated model succeeded in school either regression model is used test data by the relationship for practising three-dimensional pose posture and two dimension observation figure
Detection on, for obtaining the posture of target object.What such method was generally extracted is the information of whole picture, does not only use object
A part of feature of body, so comparatively, the Attitude estimation based on study has better anti-interference ability, and robustness is more
By force.
Summary of the invention
The present invention models fitting ability powerful using neural network, proposes a kind of Attitude estimation based on deep learning
Method.This method extracts the effective coverage of target object surface using neural network, and establishes model to its surface and go to estimate
Obtain gestures of object information.It comprises the concrete steps that:
Step (1) obtains color image and depth image using RGB-D camera
The scene comprising target object is shot using RGB-D camera, obtain a color image and with colour
The one-to-one depth image of image pixel.
Step (2) extracts interested pixel region in image using semantic segmentation network
Using semantic segmentation neural network, input picture is extracted into feature by convolution pondization and then is up-sampled
It is merged with shallow-layer profile information, finally obtains one and the consistent segmentation result characteristic pattern of input image resolution, it is right
The classification of a pixel scale is done in input picture, is determined in the interested pixel region of target object surface.It is defeated extracting
Enter in image after interested target object region, so that it may which the body surface for being directed to the part determines the appearance of target object
State.
After semantic segmentation network, the image information combination depth information that extraction obtains is acquired to the appearance of target object
State.
Step (3) models body surface and determines gestures of object
The depth on body surface plane domain and corresponding fusion depth image detected using semantic segmentation network
Information establishes areal model to body surface plane:
Ax+By+Cz+D=0
The parameter of plane equation is acquired using Principal Component Analysis Algorithm, calculation is as follows:
1. utilizing formula ∑=E (aaT)-E(a)E(aT), the covariance matrix of all the points on object plane is acquired, is it
Middle a is sample data, and E is with regard to average operation.
2. seeking the characteristic value and feature vector of covariance matrix, the corresponding feature vector of minimal eigenvalue is chosen, with this
As the normal vector of body surface plane, plane equation parameter A, B, C are obtained.
3. seeking the average coordinates of all surface coordinate points, bring into plane equation, solution obtains equation parameter D.
By Principal Component Analysis Algorithm, the plane equation of body surface is obtained, estimation obtains the posture of target object.But
Be, during seeking plane equation using depth integration figure because depth camera collect there are much noise, serious shadows
The equation for finally acquiring plane is rung, needs first to filter out the noise spot in initial data.
Plane is acquired after filtering out the noise spot in data using random sample unification algorism combination Principal Component Analysis Algorithm
Model.The step of process, is as follows:
1. randomly selecting four sample datas that can determine that equation parameter, the parameter of plane equation is acquired using data point,
Obtain areal model.
2. calculating the error of all data points using the areal model acquired.When error is less than given threshold value, it is believed that be
It is interior, it is otherwise exterior point.
3. the number put in statistics, if interior number is greater than setting quantity, using Principal Component Analysis Algorithm in all interior points
On acquire "current" model.
4. the mean error of all interior points is calculated using the model that interior point set acquires, if error current is less than the optimal of storage
When the error of model, optimal models are updated, and update the error of optimal models.
5. the step more than constantly repeating obtains final areal model until meeting maximum the number of iterations.
After acquiring the normal vector of body surface, the mean value of all spatial points of body surface is calculated as object table
The central point in face.Determine that planar central point and plane normal vector have determined that the posture information of target object later.
Beneficial effects of the present invention: the advantage of neural network is utilized in the method for the present invention, using neural network algorithm by mesh
Method that mark object important area models body surface after extracting estimates gestures of object.This method adaptability
By force, for different types of object, it need to only collect data set and re -training is carried out to neural network, be mentioned without redesigning feature
Take device.And this method determines that posture is accurate, using the powerful ability in feature extraction of neural network, can estimate that analysis is most of
Object in scene.
Detailed description of the invention
Fig. 1 is semantic segmentation network;
Specific implementation step
Step (1) obtains color image and depth image using RGB-D camera
The scene comprising target object is shot using RGB-D camera, obtain a color image and with colour
The one-to-one depth image of image pixel.
Step (2) extracts interested pixel region in image using semantic segmentation network
As shown in Figure 1, using semantic segmentation neural network, by input picture by convolution pondization extraction feature and then
It carries out up-sampling to be merged with shallow-layer profile information, finally obtains one and the consistent segmentation result of input image resolution
Characteristic pattern makees input picture the classification of one pixel scale, determines in the interested pixel region of target object surface.?
It extracts in input picture after interested target object region, so that it may which the body surface for being directed to the part determines target
The posture of object.
After semantic segmentation network, the image information combination depth information that extraction obtains is acquired to the appearance of target object
State.
Step (3) models body surface and determines gestures of object
The depth on body surface plane domain and corresponding fusion depth image detected using semantic segmentation network
Information establishes areal model to body surface plane:
Ax+By+Cz+D=0
The parameter of plane equation is acquired using Principal Component Analysis Algorithm, calculation is as follows:
1. utilizing formula ∑=E (aaT)-E(a)E(aT), the covariance matrix of all the points on object plane is acquired, is it
Middle a is sample data, and E is with regard to average operation.
2. seeking the characteristic value and feature vector of covariance matrix, the corresponding feature vector of minimal eigenvalue is chosen, with this
As the normal vector of body surface plane, plane equation parameter A, B, C are obtained.
3. seeking the average coordinates of all surface coordinate points, bring into plane equation, solution obtains equation parameter D.
By Principal Component Analysis Algorithm, the plane equation of body surface is obtained, estimation obtains the posture of target object.But
Be, during seeking plane equation using depth integration figure because depth camera collect there are much noise, serious shadows
The equation for finally acquiring plane is rung, needs first to filter out the noise spot in initial data.
Plane is acquired after filtering out the noise spot in data using random sample unification algorism combination Principal Component Analysis Algorithm
Model.The step of process, is as follows:
1. randomly selecting four sample datas that can determine that equation parameter, the parameter of plane equation is acquired using data point,
Obtain areal model.
2. calculating the error of all data points using the areal model acquired.When error is less than given threshold value, it is believed that be
It is interior, it is otherwise exterior point.
3. the number put in statistics, if interior number is greater than setting quantity, using Principal Component Analysis Algorithm in all interior points
On acquire "current" model.
4. the mean error of all interior points is calculated using the model that interior point set acquires, if error current is less than the optimal of storage
When the error of model, optimal models are updated, and update the error of optimal models.
5. the step more than constantly repeating obtains final areal model until meeting maximum the number of iterations.
After acquiring the normal vector of body surface, the mean value of all spatial points of body surface is calculated as object table
The central point in face.Determine that planar central point and plane normal vector have determined that the posture information of target object later.
Claims (1)
1. a kind of object's pose estimation method based on deep learning, it is characterised in that this method comprises the concrete steps that:
Step (1) obtains color image and depth image using RGB-D camera
The scene comprising target object is shot using RGB-D camera, obtains a color image and and color image
The one-to-one depth image of pixel;
Step (2) extracts interested pixel region in image using semantic segmentation network
Using semantic segmentation neural network, input picture by convolution pondization is extracted into feature and then carry out up-sampling with shallowly
Layer profile information is merged, and one and the consistent segmentation result characteristic pattern of input image resolution is finally obtained, for defeated
Enter the classification that picture does a pixel scale, determines in the interested pixel region of target object surface;Extracting input figure
As in after interested target object region, so that it may which the body surface for being directed to the part determines the posture of target object;
After semantic segmentation network, the image information combination depth information that extraction obtains is acquired to the posture of target object;
Step (3) models body surface and determines gestures of object
The depth information on body surface plane domain and corresponding fusion depth image detected using semantic segmentation network,
Areal model is established to body surface plane:
Ax+By+Cz+D=0
The parameter of plane equation is acquired using Principal Component Analysis Algorithm, calculation is as follows:
1. utilizing formula ∑=E (aaT)-E(a)E(aT), the covariance matrix of all the points on object plane is acquired, for wherein a
For sample data, E is with regard to average operation;
2. seeking the characteristic value and feature vector of covariance matrix, the corresponding feature vector of minimal eigenvalue is chosen, in this, as
The normal vector of body surface plane obtains plane equation parameter A, B, C;
3. seeking the average coordinates of all surface coordinate points, bring into plane equation, solution obtains equation parameter D;
By Principal Component Analysis Algorithm, the plane equation of body surface is obtained, estimation obtains the posture of target object;
Areal model is acquired after filtering out the noise spot in data using random sample unification algorism combination Principal Component Analysis Algorithm;
The step of process, is as follows:
1. randomly selecting four sample datas that can determine that equation parameter, the parameter of plane equation is acquired using data point, is obtained
Areal model;
2. calculating the error of all data points using the areal model acquired;When error is less than given threshold value, it is believed that be interior
Otherwise point is exterior point;
3. the number put in statistics is asked on all interior points if interior number is greater than setting quantity using Principal Component Analysis Algorithm
Obtain "current" model;
4. the mean error of all interior points is calculated using the model that interior point set acquires, if error current is less than the optimal models of storage
Error when, update optimal models, and update the error of optimal models;
5. the step more than constantly repeating obtains final areal model until meeting maximum the number of iterations;
After acquiring the normal vector of body surface, the mean value of all spatial points of body surface is calculated as body surface
Central point;The posture information of target object has been determined that using planar central point and plane normal vector.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910016293.5A CN109903332A (en) | 2019-01-08 | 2019-01-08 | A kind of object's pose estimation method based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910016293.5A CN109903332A (en) | 2019-01-08 | 2019-01-08 | A kind of object's pose estimation method based on deep learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109903332A true CN109903332A (en) | 2019-06-18 |
Family
ID=66943698
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910016293.5A Pending CN109903332A (en) | 2019-01-08 | 2019-01-08 | A kind of object's pose estimation method based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109903332A (en) |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110276805A (en) * | 2019-06-28 | 2019-09-24 | 联想(北京)有限公司 | A kind of data processing method and electronic equipment |
CN110322510A (en) * | 2019-06-27 | 2019-10-11 | 电子科技大学 | A kind of 6D position and orientation estimation method using profile information |
CN110910447A (en) * | 2019-10-31 | 2020-03-24 | 北京工业大学 | Visual odometer method based on dynamic and static scene separation |
CN110992372A (en) * | 2019-11-21 | 2020-04-10 | 浙江大华技术股份有限公司 | Article grabbing method and device, storage medium and electronic device |
CN110992422A (en) * | 2019-11-04 | 2020-04-10 | 浙江工业大学 | Medicine box posture estimation method based on 3D vision |
CN111524184A (en) * | 2020-04-21 | 2020-08-11 | 湖南视普瑞智能科技有限公司 | Intelligent unstacking method and system based on 3D vision |
CN111597976A (en) * | 2020-05-14 | 2020-08-28 | 杭州相芯科技有限公司 | Multi-person three-dimensional attitude estimation method based on RGBD camera |
CN113313810A (en) * | 2021-06-18 | 2021-08-27 | 广东工业大学 | 6D attitude parameter calculation method for transparent object |
CN113370217A (en) * | 2021-06-29 | 2021-09-10 | 华南理工大学 | Method for recognizing and grabbing object posture based on deep learning for intelligent robot |
CN113393421A (en) * | 2021-05-08 | 2021-09-14 | 深圳市识农智能科技有限公司 | Fruit evaluation method and device and inspection equipment |
CN113643433A (en) * | 2020-04-27 | 2021-11-12 | 成都术通科技有限公司 | Form and attitude estimation method, device, equipment and storage medium |
CN114612939A (en) * | 2022-03-25 | 2022-06-10 | 珠海视熙科技有限公司 | Sitting posture identification method and device based on TOF camera and intelligent desk lamp |
CN115578461A (en) * | 2022-11-14 | 2023-01-06 | 之江实验室 | Object attitude estimation method and device based on bidirectional RGB-D feature fusion |
CN116968022A (en) * | 2023-07-14 | 2023-10-31 | 武汉纺织大学 | Method and system for grabbing target object by mechanical arm based on visual guidance |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170277981A1 (en) * | 2016-01-08 | 2017-09-28 | Siemens Healthcare Gmbh | Deep Image-to-Image Network Learning for Medical Image Analysis |
CN107330439A (en) * | 2017-07-14 | 2017-11-07 | 腾讯科技(深圳)有限公司 | A kind of determination method, client and the server of objects in images posture |
CN107403430A (en) * | 2017-06-15 | 2017-11-28 | 中山大学 | A kind of RGBD image, semantics dividing method |
CN108280856A (en) * | 2018-02-09 | 2018-07-13 | 哈尔滨工业大学 | The unknown object that network model is inputted based on mixed information captures position and orientation estimation method |
KR20180096355A (en) * | 2017-02-21 | 2018-08-29 | 서강대학교산학협력단 | Depth Map Filtering and Two-level Predictor-Corrector Method for Precision Enhancement of RGB-D Camera Pose Estimation |
CN108491880A (en) * | 2018-03-23 | 2018-09-04 | 西安电子科技大学 | Object classification based on neural network and position and orientation estimation method |
CN109035327A (en) * | 2018-06-25 | 2018-12-18 | 北京大学 | Panorama camera Attitude estimation method based on deep learning |
-
2019
- 2019-01-08 CN CN201910016293.5A patent/CN109903332A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170277981A1 (en) * | 2016-01-08 | 2017-09-28 | Siemens Healthcare Gmbh | Deep Image-to-Image Network Learning for Medical Image Analysis |
KR20180096355A (en) * | 2017-02-21 | 2018-08-29 | 서강대학교산학협력단 | Depth Map Filtering and Two-level Predictor-Corrector Method for Precision Enhancement of RGB-D Camera Pose Estimation |
CN107403430A (en) * | 2017-06-15 | 2017-11-28 | 中山大学 | A kind of RGBD image, semantics dividing method |
CN107330439A (en) * | 2017-07-14 | 2017-11-07 | 腾讯科技(深圳)有限公司 | A kind of determination method, client and the server of objects in images posture |
CN108280856A (en) * | 2018-02-09 | 2018-07-13 | 哈尔滨工业大学 | The unknown object that network model is inputted based on mixed information captures position and orientation estimation method |
CN108491880A (en) * | 2018-03-23 | 2018-09-04 | 西安电子科技大学 | Object classification based on neural network and position and orientation estimation method |
CN109035327A (en) * | 2018-06-25 | 2018-12-18 | 北京大学 | Panorama camera Attitude estimation method based on deep learning |
Non-Patent Citations (6)
Title |
---|
ERIC BRACHMANN等: "Learning 6D Object Pose Estimation Using 3D Object Coordinates", 《EUROPEAN CONFERENCE ON COMPUTER VISION 2014》 * |
JAY M. WONG等: "SegICP: Integrated Deep Semantic Segmentation and Pose Estimation", 《2017 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS (IROS)》 * |
JONATHAN LONG等: "Fully convolutional networks for semantic segmentation", 《2015 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR)》 * |
杨扬: "基于机器视觉的服务机器人智能抓取研究", 《中国优秀博士学位论文全文数据库 信息科技辑(月刊)》 * |
石广升: "基于Kinect的物体三维模型构建和姿态估计方法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑(月刊)》 * |
翁健: "基于全卷积神经网络的全向场景分割研究与算法实现", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110322510A (en) * | 2019-06-27 | 2019-10-11 | 电子科技大学 | A kind of 6D position and orientation estimation method using profile information |
CN110322510B (en) * | 2019-06-27 | 2021-08-27 | 电子科技大学 | 6D pose estimation method using contour information |
CN110276805A (en) * | 2019-06-28 | 2019-09-24 | 联想(北京)有限公司 | A kind of data processing method and electronic equipment |
CN110910447A (en) * | 2019-10-31 | 2020-03-24 | 北京工业大学 | Visual odometer method based on dynamic and static scene separation |
CN110992422A (en) * | 2019-11-04 | 2020-04-10 | 浙江工业大学 | Medicine box posture estimation method based on 3D vision |
CN110992422B (en) * | 2019-11-04 | 2023-11-07 | 浙江工业大学 | Medicine box posture estimation method based on 3D vision |
CN110992372B (en) * | 2019-11-21 | 2023-08-29 | 浙江大华技术股份有限公司 | Article grabbing method and device, storage medium and electronic device |
CN110992372A (en) * | 2019-11-21 | 2020-04-10 | 浙江大华技术股份有限公司 | Article grabbing method and device, storage medium and electronic device |
CN111524184A (en) * | 2020-04-21 | 2020-08-11 | 湖南视普瑞智能科技有限公司 | Intelligent unstacking method and system based on 3D vision |
CN111524184B (en) * | 2020-04-21 | 2024-01-16 | 湖南视普瑞智能科技有限公司 | Intelligent unstacking method and unstacking system based on 3D vision |
CN113643433A (en) * | 2020-04-27 | 2021-11-12 | 成都术通科技有限公司 | Form and attitude estimation method, device, equipment and storage medium |
CN111597976A (en) * | 2020-05-14 | 2020-08-28 | 杭州相芯科技有限公司 | Multi-person three-dimensional attitude estimation method based on RGBD camera |
CN113393421A (en) * | 2021-05-08 | 2021-09-14 | 深圳市识农智能科技有限公司 | Fruit evaluation method and device and inspection equipment |
CN113313810A (en) * | 2021-06-18 | 2021-08-27 | 广东工业大学 | 6D attitude parameter calculation method for transparent object |
CN113370217B (en) * | 2021-06-29 | 2023-06-16 | 华南理工大学 | Object gesture recognition and grabbing intelligent robot method based on deep learning |
CN113370217A (en) * | 2021-06-29 | 2021-09-10 | 华南理工大学 | Method for recognizing and grabbing object posture based on deep learning for intelligent robot |
CN114612939B (en) * | 2022-03-25 | 2023-01-10 | 珠海视熙科技有限公司 | Sitting posture identification method and device based on TOF camera and intelligent desk lamp |
CN114612939A (en) * | 2022-03-25 | 2022-06-10 | 珠海视熙科技有限公司 | Sitting posture identification method and device based on TOF camera and intelligent desk lamp |
CN115578461A (en) * | 2022-11-14 | 2023-01-06 | 之江实验室 | Object attitude estimation method and device based on bidirectional RGB-D feature fusion |
CN115578461B (en) * | 2022-11-14 | 2023-03-10 | 之江实验室 | Object attitude estimation method and device based on bidirectional RGB-D feature fusion |
CN116968022A (en) * | 2023-07-14 | 2023-10-31 | 武汉纺织大学 | Method and system for grabbing target object by mechanical arm based on visual guidance |
CN116968022B (en) * | 2023-07-14 | 2024-05-28 | 武汉纺织大学 | Method and system for grabbing target object by mechanical arm based on visual guidance |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109903332A (en) | A kind of object's pose estimation method based on deep learning | |
CN109255813B (en) | Man-machine cooperation oriented hand-held object pose real-time detection method | |
CN102657532B (en) | Height measuring method and device based on body posture identification | |
CN105023010B (en) | A kind of human face in-vivo detection method and system | |
US9019267B2 (en) | Depth mapping with enhanced resolution | |
CN110866969B (en) | Engine blade reconstruction method based on neural network and point cloud registration | |
CN104794737B (en) | A kind of depth information Auxiliary Particle Filter tracking | |
CN106971406B (en) | Object pose detection method and device | |
JP6016716B2 (en) | Bin picking performance evaluation apparatus and method | |
US20230085384A1 (en) | Characterizing and improving of image processing | |
CN109410248A (en) | A kind of flotation froth motion feature extracting method based on r-K algorithm | |
CN110555908A (en) | three-dimensional reconstruction method based on indoor moving target background restoration | |
CN107194985A (en) | A kind of three-dimensional visualization method and device towards large scene | |
CN105869166A (en) | Human body action identification method and system based on binocular vision | |
CN105279769A (en) | Hierarchical particle filtering tracking method combined with multiple features | |
CN114494594B (en) | Deep learning-based astronaut operation equipment state identification method | |
Migniot et al. | 3D human tracking from depth cue in a buying behavior analysis context | |
CN108388854A (en) | A kind of localization method based on improvement FAST-SURF algorithms | |
Migniot et al. | 3d human tracking in a top view using depth information recorded by the xtion pro-live camera | |
CN106933976B (en) | Method for establishing human body 3D net model and application thereof in 3D fitting | |
Singh et al. | Fusing semantics and motion state detection for robust visual SLAM | |
Sokhib et al. | A combined method of skin-and depth-based hand gesture recognition. | |
CN113470073A (en) | Animal center tracking method based on deep learning | |
CN111531546B (en) | Robot pose estimation method, device, equipment and storage medium | |
Haggag et al. | LGT/VOT tracking performance evaluation of depth images |
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: 20190618 |
|
RJ01 | Rejection of invention patent application after publication |