CN111008627A - Method for detecting mark code frame under boundary shielding condition - Google Patents
Method for detecting mark code frame under boundary shielding condition Download PDFInfo
- Publication number
- CN111008627A CN111008627A CN201911233398.2A CN201911233398A CN111008627A CN 111008627 A CN111008627 A CN 111008627A CN 201911233398 A CN201911233398 A CN 201911233398A CN 111008627 A CN111008627 A CN 111008627A
- Authority
- CN
- China
- Prior art keywords
- code
- visual
- data set
- training data
- mark code
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/24—Aligning, centring, orientation detection or correction of the image
- G06V10/245—Aligning, centring, orientation detection or correction of the image by locating a pattern; Special marks for positioning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06K—GRAPHICAL DATA READING; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
- G06K7/00—Methods or arrangements for sensing record carriers, e.g. for reading patterns
- G06K7/10—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation
- G06K7/14—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation using light without selection of wavelength, e.g. sensing reflected white light
- G06K7/1404—Methods for optical code recognition
- G06K7/146—Methods for optical code recognition the method including quality enhancement steps
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Multimedia (AREA)
- Electromagnetism (AREA)
- General Health & Medical Sciences (AREA)
- Toxicology (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a method for detecting a mark code frame under the condition of boundary shielding, which comprises the following steps of S1, obtaining a training data set; s2, estimating the rotation angle of the boundary occlusion visual marker code; and S3, detecting the border of the border occlusion visual mark code. The invention has the beneficial effects that: different from the traditional marker code contour extraction method using image processing, the marker code in the image is directly extracted by using a method based on target detection, so that the boundary shielding condition of the marker code can be dealt with, and the whole identification method has better robustness.
Description
Technical Field
The present invention relates to a mark code, and more particularly, to a method for detecting a mark code frame under a boundary occlusion condition.
Background
Visual marker codes (marker codes) are artificial visual features designed to simplify automatic detection of machines, and are widely used in the fields of computer vision, augmented reality and robotics. Common visual mark codes on the market comprise AprilTag mark codes, ArUco mark codes and the like, and the classical mark codes can utilize black frames of the classical mark codes to finish outline extraction and camera attitude estimation of the mark codes, but the classical mark codes also become a key factor for restricting the recognition robustness of the classical mark codes, once the boundary of the mark codes is shielded, the frame detection work of the mark codes cannot be finished, and further the recognition and the positioning of the mark codes cannot be realized.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a method for detecting a mark code frame under the condition of boundary shielding.
The invention provides a method for detecting a mark code frame under the condition of boundary shielding, which comprises the following steps:
s1, acquiring a training data set;
s2, estimating the rotation angle of the boundary occlusion visual marker code;
and S3, detecting the border of the border occlusion visual mark code.
As a further improvement of the present invention, step S1 includes: firstly, determining the rotation angle of the visual mark code by using a classification method; then, determining the specific position of the visual mark code in the image; and finally, carrying out boundary occlusion processing on the visual mark codes in the images of the training data set.
As a further refinement of the present invention, in step S1, the rotation angle information of the visual marker is given by the label of the training data set.
As a further improvement of the present invention, in step S1, coordinate information of four corner points of the visual mark code is obtained first, and then a specific position of the visual mark code in the image is obtained.
As a further improvement of the present invention, in step S1, the generation method of the training data set is expressed as follows: modifying an identification program of the ArUco code, so that the identification program stores the picture acquired by the camera as a training data set only containing the picture according to a set naming rule on the basis of identifying the visual tag code, and simultaneously stores coordinate information of four corner points of the visual tag code in the acquired picture as a csv file by taking the picture name as an index, wherein the coordinate information is used as a label of the training data set.
As a further improvement of the invention, the obtained corner coordinates are preprocessed to obtain the rotation angle information, random boundary shielding is added to the mark codes in the training data set, and the complete training data set acquisition work is completed.
As a further improvement of the present invention, in step S2, a classification method is used to estimate the rotation angle of the visual marker.
As a further improvement of the invention, the classification targets are set to 91 categories which respectively represent that the visual mark codes are rotated by 0-90 degrees, during training, when a VGG16 network is used for transfer learning, the convolutional layer and the pooling layer in front of the network are reserved, the last full-connection layer is removed, and the full-connection network is added to complete the complete training process.
As a further improvement of the present invention, in step S3, the visual tag is corrected by using the predicted rotation angle, and the coordinates of the tag are transformed correspondingly to obtain a new training data set, and then the new training picture and the tag are used as the input of the RetinaNet network and output as the frame of the occlusion tag, thereby completing the frame detection work of the whole occlusion tag.
The invention has the beneficial effects that: different from the traditional marker code contour extraction method using image processing, the marker code in the image is directly extracted by using a method based on target detection, so that the boundary shielding condition of the marker code can be dealt with, and the whole identification method has better robustness.
Drawings
FIG. 1 is a flow chart illustrating the identification of a mark code by a method for detecting a mark code border under a boundary occlusion condition according to the present invention.
Detailed Description
The invention is further described with reference to the following description and embodiments in conjunction with the accompanying drawings.
A method for detecting a mark code frame under the condition of boundary shielding obtains a training data set through a normal recognition condition result, simplifies the data set collection process and improves the accuracy of a data set label; then, based on the data sets, adding boundary shielding treatment of the mark codes to ensure that the mark codes in the whole data set have boundary shielding conditions; then, the rotation angle prediction of the mark code is completed, and preparation is made for the final mark code frame detection; and finally, finishing the detection work of the shielding mark code frame by using a RetinaNet target detection method. The invention improves on the basis of the identification of Aruco codes.
A method for detecting a mark code frame under a boundary shielding condition specifically comprises the following steps:
(1) acquisition of a training data set: firstly, in order to determine the rotation angle of the visual marker code by using a classification method, the rotation angle information of the visual marker code must be given in the label of the training data set; then, in order to realize the frame detection of the visual mark code, the specific position of the visual mark code in the image needs to be known, and the most intuitive idea is to obtain the coordinate information of four corner points of the visual mark code first and further obtain the specific position of the visual mark code in the image; finally, because the frame detection method based on target detection is mainly used for dealing with the boundary occlusion condition existing in the visual mark code, the training data set needs to perform boundary occlusion processing on the mark code in the image.
According to the ideas given above, the way in which the training data set is generated can be expressed as follows. And modifying the recognition program of the Aruco code, so that the new program can store the picture acquired by the camera as a training data set only containing the picture according to a certain naming rule on the basis of recognizing the visual mark code. And simultaneously, storing coordinate information of four corner points of the visual marker codes in the acquired picture as a csv file by taking the picture name as an index, wherein the coordinate information is used as a label of a training data set. Note that the obtained tag does not directly include the rotation angle information of the visual mark code, and the obtained corner coordinates can be simply preprocessed to obtain the rotation angle information. And finally, adding random boundary shielding for the mark codes in the training data set to complete the acquisition work of the complete training data set.
(2) And (3) estimating the rotation angle of the boundary occlusion marker code: the rotation angle information for acquiring the visual mark code is meaningful for assisting the camera positioning and subsequently acquiring the position of the visual mark code in the image. Only when the rotation angle information of the visual mark code can be obtained, the training data set is rotated by a corresponding angle, and the correct visual mark code can be obtained, so that the target is detected to be a vertical rectangular frame, and the frame detection work of the visual mark code can be finished by directly using a mature target detection method.
For the rotation angle estimation of the visual marker code, a method using classification may be considered. Regarding the rotation angle, it can be considered that the rotation angle is just rotated, and what angle needs to be rotated is, so that the classification target can be set to 91 categories, which respectively represent that the visual mark code is rotated by 0-90 degrees. During training, when the VGG16 network is used for transfer learning, the convolutional layer and the pooling layer in front of the network are reserved, the last full-connection layer is removed, and the full-connection network of the user is added to complete the complete training process.
(3) And (3) detecting a frame of the boundary shielding mark code: firstly, the mark code is corrected by the predicted rotation angle, and the coordinates of the label are correspondingly transformed to obtain a new training data set. And then, taking the new training picture and the new label as the input of the RetinaNet network, outputting the new training picture and the new label as a frame of the shielding mark code, and finishing the frame detection work of the whole shielding mark code.
The method for detecting the mark code frame under the condition of boundary shielding has the following advantages that:
1. as shown in fig. 1, unlike the conventional method for extracting a contour of a mark code by using image processing, the border detection method based on target detection is used to directly extract the mark code in an image, so that the border occlusion condition of the mark code can be dealt with, and the whole recognition method has better robustness.
2. The frame of the mark code is directly identified, so that the process of eliminating irrelevant candidate mark codes one by one in the traditional method is omitted, and the whole identification process is more accurate and more targeted.
The foregoing is a more detailed description of the invention in connection with specific preferred embodiments and it is not intended that the invention be limited to these specific details. For those skilled in the art to which the invention pertains, several simple deductions or substitutions can be made without departing from the spirit of the invention, and all shall be considered as belonging to the protection scope of the invention.
Claims (9)
1. A method for detecting a mark code frame under the condition of boundary shielding is characterized by comprising the following steps:
s1, acquiring a training data set;
s2, estimating the rotation angle of the boundary occlusion visual marker code;
and S3, detecting the border of the border occlusion visual mark code.
2. The method for detecting the mark code border under the boundary occlusion condition as claimed in claim 1, wherein the step S1 comprises: firstly, determining the rotation angle of the visual mark code by using a classification method; then, determining the specific position of the visual mark code in the image; and finally, carrying out boundary occlusion processing on the visual mark codes in the images of the training data set.
3. The method of claim 2, wherein the method comprises: in step S1, the rotation angle information of the visual marker is given by the label of the training data set.
4. The method of claim 2, wherein the method comprises: in step S1, coordinate information of four corner points of the visual mark code is obtained first, and then a specific position of the visual mark code in the image is obtained.
5. The method of claim 1, wherein the method comprises: in step S1, the training data set is generated in the following manner: modifying an identification program of the ArUco code, so that the identification program stores the picture acquired by the camera as a training data set only containing the picture according to a set naming rule on the basis of identifying the visual tag code, and simultaneously stores coordinate information of four corner points of the visual tag code in the acquired picture as a csv file by taking the picture name as an index, wherein the coordinate information is used as a label of the training data set.
6. The method of claim 5, wherein the method comprises: and preprocessing the obtained corner point coordinates to obtain rotation angle information, and adding random boundary shielding to the mark codes in the training data set to complete the acquisition work of the complete training data set.
7. The method of claim 1, wherein the method comprises: in step S2, the rotation angle of the visual marker is estimated by a classification method.
8. The method of claim 7, wherein the method comprises: the classification target is set to 91 categories which respectively represent that the visual mark codes are rotated by 0-90 degrees, during training, when the VGG16 network is used for transfer learning, the convolution layer and the pooling layer in front of the network are reserved, the last full-connection layer is removed, and the full-connection network is added to complete the complete training process.
9. The method of claim 1, wherein the method comprises: in step S3, the predicted rotation angle is used to correct the visual tag, and the coordinates of the tag are transformed accordingly to obtain a new training data set, and then the new training picture and the tag are used as the input of the RetinaNet network and output as the frame of the shielding tag, thereby completing the frame detection of the whole shielding tag.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911233398.2A CN111008627B (en) | 2019-12-05 | 2019-12-05 | Method for detecting marking code frame under boundary shielding condition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911233398.2A CN111008627B (en) | 2019-12-05 | 2019-12-05 | Method for detecting marking code frame under boundary shielding condition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111008627A true CN111008627A (en) | 2020-04-14 |
| CN111008627B CN111008627B (en) | 2023-09-05 |
Family
ID=70113887
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911233398.2A Active CN111008627B (en) | 2019-12-05 | 2019-12-05 | Method for detecting marking code frame under boundary shielding condition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111008627B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108764186A (en) * | 2018-06-01 | 2018-11-06 | 合肥工业大学 | Personage based on rotation deep learning blocks profile testing method |
| CN109165538A (en) * | 2018-07-18 | 2019-01-08 | 北京飞搜科技有限公司 | Bar code detection method and device based on deep neural network |
| CN109859158A (en) * | 2018-11-27 | 2019-06-07 | 邦鼓思电子科技（上海）有限公司 | A kind of detection system, method and the machinery equipment on the working region boundary of view-based access control model |
| CN109977718A (en) * | 2019-03-21 | 2019-07-05 | 连尚（新昌）网络科技有限公司 | A kind of method and apparatus of two dimensional code for identification |
| CN110390302A (en) * | 2019-07-24 | 2019-10-29 | 厦门大学 | A kind of objective detection method |
-
2019
- 2019-12-05 CN CN201911233398.2A patent/CN111008627B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108764186A (en) * | 2018-06-01 | 2018-11-06 | 合肥工业大学 | Personage based on rotation deep learning blocks profile testing method |
| CN109165538A (en) * | 2018-07-18 | 2019-01-08 | 北京飞搜科技有限公司 | Bar code detection method and device based on deep neural network |
| CN109859158A (en) * | 2018-11-27 | 2019-06-07 | 邦鼓思电子科技（上海）有限公司 | A kind of detection system, method and the machinery equipment on the working region boundary of view-based access control model |
| CN109977718A (en) * | 2019-03-21 | 2019-07-05 | 连尚（新昌）网络科技有限公司 | A kind of method and apparatus of two dimensional code for identification |
| CN110390302A (en) * | 2019-07-24 | 2019-10-29 | 厦门大学 | A kind of objective detection method |
Non-Patent Citations (1)
| Title |
|---|
| RUIJIE CHANG ETC: "Robust Identification of Visual Markers Under Boundary Occlusion Condition" * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111008627B (en) | 2023-09-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109815850B (en) | Iris image segmentation and positioning method, system and device based on deep learning | |
| CN111667520B (en) | Registration method and device for infrared image and visible light image and readable storage medium | |
| CN111178250B (en) | Object identification positioning method and device and terminal equipment | |
| CN105069799B (en) | Angular point positioning method and apparatus | |
| CN108230383B (en) | Hand three-dimensional data determination method and device and electronic equipment | |
| CN109376631B (en) | Loop detection method and device based on neural network | |
| CN109426814B (en) | Method, system and equipment for positioning and identifying specific plate of invoice picture | |
| CN108369741A (en) | Method and system for registration data | |
| CN103136525B (en) | High-precision positioning method for special-shaped extended target by utilizing generalized Hough transformation | |
| CN110930374A (en) | Acupoint positioning method based on double-depth camera | |
| CN108335331A (en) | A kind of coil of strip binocular visual positioning method and apparatus | |
| US10025977B2 (en) | Method for identifying a sign on a deformed document | |
| CN104517101A (en) | Game poker card recognition method based on pixel square difference matching | |
| CN107203742B (en) | Gesture recognition method and device based on significant feature point extraction | |
| CN112418216A (en) | Method for detecting characters in complex natural scene image | |
| CN111191582B (en) | Three-dimensional target detection method, detection device, terminal device and computer readable storage medium | |
| CN111190595A (en) | Method, device, medium and electronic equipment for automatically generating interface code based on interface design drawing | |
| WO2022170896A9 (en) | Key point detection method and system, intelligent terminal, and storage medium | |
| CN113903024A (en) | Handwritten bill numerical value information identification method, system, medium and device | |
| CN111199198B (en) | Image target positioning method, image target positioning device and mobile robot | |
| CN113283466A (en) | Instrument reading identification method and device and readable storage medium | |
| CN115471748A (en) | Monocular vision SLAM method oriented to dynamic environment | |
| CN114119695A (en) | Image annotation method and device and electronic equipment | |
| CN107729863B (en) | Human finger vein recognition method | |
| CN118015190A (en) | Autonomous construction method and device of digital twin model |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |