CN103942554A - Image identifying method and device - Google Patents
Image identifying method and device Download PDFInfo
- Publication number
- CN103942554A CN103942554A CN201410190472.8A CN201410190472A CN103942554A CN 103942554 A CN103942554 A CN 103942554A CN 201410190472 A CN201410190472 A CN 201410190472A CN 103942554 A CN103942554 A CN 103942554A
- Authority
- CN
- China
- Prior art keywords
- image
- characteristic
- cut apart
- multiple image
- image blocks
- 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
Abstract
The invention discloses an image identifying method and device and belongs to the technical field of image processing. The image identifying method includes: using a cloud model to divide an image to be identified to obtain a plurality of image blocks; using a parallel reinforcement learning method to learn the image blocks to obtain a multi-characteristic combination; fusing the multi-characteristic combination based on the D-S evidence theory to confirm objects in the image to be identified. The image identifying method is based on the multi-characteristic combination optimization method of the parallel reinforcement learning, makes the most of the advantage characteristic of the single characteristics of the image, and comprehensively considers multiple characteristics to describe the image so as to identify the image more precisely.
Description
Technical field
The present invention relates to technical field of image processing, particularly a kind of image-recognizing method and device.
Background technology
Flourish along with internet shopping, the demand that people divide into groups to product picture is more and more huger.At present for product pictures taken and divide into groups generally by manually completing, its flow process is roughly: the product picture of taking in camera is copied to after PC, by artificial cognition, several similar pictures are classified as to one group, and move in a new folder, then new folder is named with corresponding product bar code.
Said method relates to artificial cognition product picture, picture sorts out and figure organizes the links such as name, and in the time that the quantity of product picture increases, the probability making a mistake in the process of picture recognition and picture classification also increases thereupon.In addition, the method efficiency of artificial component is also lower.
Summary of the invention
In order to solve the problem of prior art, the embodiment of the present invention provides a kind of image-recognizing method and device.Described technical scheme is as follows:
On the one hand, provide a kind of image-recognizing method, described method comprises:
Adopt cloud model to treat recognition image and cut apart, obtain multiple image blocks;
Adopt parallel intensified learning method to learn described multiple image blocks, obtain multiclass feature combination;
Based on D-S evidence theory, described multiclass feature combination is merged, determine article that described image to be identified comprises.
On the other hand, provide a kind of pattern recognition device, described device comprises:
Image is cut apart module, cuts apart for adopting cloud model to treat recognition image, obtains multiple image blocks;
Characteristic extracting module, for adopting parallel intensified learning device to learn described multiple image blocks, obtains multiclass feature combination;
Fusion Features module, for based on D-S evidence theory, described multiclass feature combination being merged, determines article that described image to be identified comprises.
The beneficial effect that the technical scheme that the embodiment of the present invention provides is brought is:
Based on the multiclass feature combined optimization method of parallel intensified learning, make full use of the advantageous characteristic between the each single features of image, consider multiclass feature image is described, make image recognition more accurate.
Brief description of the drawings
In order to be illustrated more clearly in the technical scheme in the embodiment of the present invention, below the accompanying drawing of required use during embodiment is described is briefly described, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skill in the art, do not paying under the prerequisite of creative work, can also obtain according to these accompanying drawings other accompanying drawing.
Fig. 1 is the image-recognizing method process flow diagram that the embodiment of the present invention provides;
Fig. 2 is the image-recognizing method process flow diagram that the embodiment of the present invention provides;
Fig. 3 is the pattern recognition device structural representation that the embodiment of the present invention provides.
Embodiment
For making the object, technical solutions and advantages of the present invention clearer, below in conjunction with accompanying drawing, embodiment of the present invention is described further in detail.
Fig. 1 is the image-recognizing method process flow diagram that the embodiment of the present invention provides.Referring to Fig. 1, the method comprises:
101, adopt cloud model to treat recognition image and cut apart, obtain multiple image blocks;
102, adopt parallel intensified learning method to learn the plurality of image block, obtain multiclass feature combination;
103, based on D-S evidence theory, this multiclass feature combination is merged, determine article that this image to be identified comprises.
Alternatively, this employing cloud model is treated recognition image and is cut apart, and before obtaining multiple image blocks, the method also comprises:
Obtain this image to be identified.
Alternatively, this employing cloud model is treated recognition image and is cut apart, and obtains multiple image blocks and comprises:
Pixel in image space is represented by characteristic of correspondence spatial point;
Feature space is cut apart in the gathering of feature space according to this characteristic of correspondence spatial point;
This characteristic of correspondence spatial point is shone upon go back to original image space, obtain multiple image blocks.
Alternatively, the parallel intensified learning method of this employing is learnt the plurality of image block, obtains multiclass feature combination and comprises:
For each intensified learning agency of multiple intensified learnings agency, by the angular setting of the plurality of image block to preset angles;
Extract respectively the plurality of image block characteristics information, obtain the Feature Combination of this intensified learning agency study, this characteristic information comprises color characteristic information, textural characteristics information and shape facility information.
The method that the embodiment of the present invention provides, based on the multiclass feature combined optimization method of parallel intensified learning, makes full use of the advantageous characteristic between the each single features of image, considers multiclass feature image is described, and makes image recognition more accurate.
Fig. 2 is the image-recognizing method process flow diagram that the embodiment of the present invention provides.Referring to Fig. 2, the method comprises:
201, obtain this image to be identified;
In embodiments of the present invention, this image to be identified can be to obtain from cloud storage, can be also to obtain from terminal, can be also that terminal taking arrives after this image to be identified, pushes to server side.
202, adopt cloud model to treat recognition image and cut apart, obtain multiple image blocks;
User uses the object edge part of the image that mobile phone photograph obtains owing to being subject to the interference of other classification pixels, has higher uncertainty.In addition, the subjectivity of user to image cognition and the demand of retrieval also often have uncertainty.Cloud model is uncertain transformation model between the qualitative, quantitative proposing for the subordinate function of fuzzy set theory, adopts cloud model to carry out the uncertain problem in specification and analysis image, more can reflect intuitively that pixel cluster is the ambiguity existing behind region.Embodiment of the present invention employing cloud model carries out image cuts apart to solve the uncertain problem that image is cut apart.
Alternatively, adopt cloud model to treat recognition image and cut apart, the embodiment that obtains multiple image blocks comprises: the pixel in image space is represented by characteristic of correspondence spatial point; Feature space is cut apart in the gathering of feature space according to this characteristic of correspondence spatial point; This characteristic of correspondence spatial point is shone upon go back to original image space, obtain multiple image blocks.
203, adopt parallel intensified learning method to learn the plurality of image block, obtain multiclass feature combination;
Due to mass property and the exponential Feature Combination of view data feature, need parallelization to improve operation efficiency.Therefore, this project adopts parallel intensified learning to learn the Feature Combination of Images Classification.
Alternatively, adopt parallel intensified learning method to learn the plurality of image block, the embodiment that obtains multiclass feature combination comprises: for each intensified learning agency of multiple intensified learnings agencies, by the angular setting of the plurality of image block to preset angles; Extract respectively the plurality of image block characteristics information, obtain the Feature Combination of this intensified learning agency study, this characteristic information comprises color characteristic information, textural characteristics information and shape facility information.
The multiclass feature combined optimization method of the embodiment of the present invention based on parallel intensified learning, makes full use of the mutual supplement with each other's advantages characteristic between the each single features of image, considers multiclass feature image is described, and has higher image clustering precision.The method makes multiple intensified learning agencies learning characteristic combination independently in environment separately, avoid in traditional Q-study, when each iteration, only upgrade the right Q value of a state-move, the multiple states-right Q value renewal of moving can concurrently be carried out, improved the efficiency of Feature Combination optimization.Due in learning process, multiple intensified learning agencies independent study in environment separately, makes the Feature Combination optimization of large nuber of images have very strong concurrency and extensibility.
204, based on D-S evidence theory, this multiclass feature combination is merged, determine article that this image to be identified comprises.
The embodiment of the present invention adopts D-S evidence theory to merge each intensified learning agency's learning outcome, eliminates redundancy and the contradiction of the Q value that may exist between many intensified learnings agencies, and in addition complementation, reduces its uncertainty, obtains consistent Q value.In supposing the system, there is n intensified learning agency, after all intensified learnings have been acted on behalf of the study of one-period, their Q table is merged.(s, a) is worth Q in normalization Q table, utilizes D-S evidence to merge, and each fusion is all two fusions between evidence.
The method that the embodiment of the present invention provides, based on the multiclass feature combined optimization method of parallel intensified learning, makes full use of the advantageous characteristic between the each single features of image, considers multiclass feature image is described, and makes image recognition more accurate.
Fig. 3 is the structural representation of the pattern recognition device of embodiment of the present invention body.Referring to Fig. 3, this device comprises: image is cut apart module 301, characteristic extracting module 302 and Fusion Features module 303.Wherein:
Image is cut apart module 301 and is cut apart for adopting cloud model to treat recognition image, obtains multiple image blocks; Image is cut apart module 301 and is connected with characteristic extracting module 302, and characteristic extracting module 302, for adopting parallel intensified learning device to learn the plurality of image block, obtains multiclass feature combination; Characteristic extracting module 302 is connected with Fusion Features module 303, and Fusion Features module 303, for based on D-S evidence theory, this multiclass feature combination being merged, is determined article that this image to be identified comprises.
Alternatively, this device also comprises: image collection module, and for obtaining this image to be identified.
Alternatively, this image is cut apart module 301 also for the pixel of image space characteristic of correspondence spatial point is represented; Feature space is cut apart in the gathering of feature space according to this characteristic of correspondence spatial point; This characteristic of correspondence spatial point is shone upon go back to original image space, obtain multiple image blocks.
Alternatively, this characteristic extracting module 302 is also for each intensified learning agency for multiple intensified learnings agency, by the angular setting of the plurality of image block to preset angles; Extract respectively the plurality of image block characteristics information, obtain the Feature Combination of this intensified learning agency study, this characteristic information comprises color characteristic information, textural characteristics information and shape facility information.
The device that the embodiment of the present invention provides, based on the multiclass feature combined optimization method of parallel intensified learning, makes full use of the advantageous characteristic between the each single features of image, considers multiclass feature image is described, and makes image recognition more accurate.
It should be noted that: the pattern recognition device that above-described embodiment provides is in the time of image recognition, only be illustrated with the division of above-mentioned each functional module, in practical application, can above-mentioned functions be distributed and completed by different functional modules as required, be divided into different functional modules by the inner structure of device, to complete all or part of function described above.In addition, the pattern recognition device that above-described embodiment provides and image-recognizing method embodiment belong to same design, and its specific implementation process refers to embodiment of the method, repeats no more here.
One of ordinary skill in the art will appreciate that all or part of step that realizes above-described embodiment can complete by hardware, also can carry out the hardware that instruction is relevant by program completes, described program can be stored in a kind of computer-readable recording medium, the above-mentioned storage medium of mentioning can be ROM (read-only memory), disk or CD etc.
The foregoing is only preferred embodiment of the present invention, in order to limit the present invention, within the spirit and principles in the present invention not all, any amendment of doing, be equal to replacement, improvement etc., within all should being included in protection scope of the present invention.
Claims (8)
1. an image-recognizing method, is characterized in that, described method comprises:
Adopt cloud model to treat recognition image and cut apart, obtain multiple image blocks;
Adopt parallel intensified learning method to learn described multiple image blocks, obtain multiclass feature combination;
Based on D-S evidence theory, described multiclass feature combination is merged, determine article that described image to be identified comprises.
2. method according to claim 1, is characterized in that, described employing cloud model is treated recognition image and cut apart, and before obtaining multiple image blocks, described method also comprises:
Obtain described image to be identified.
3. method according to claim 1, is characterized in that, described employing cloud model is treated recognition image and cut apart, and obtains multiple image blocks and comprises:
Pixel in image space is represented by characteristic of correspondence spatial point;
Feature space is cut apart in the gathering of feature space according to described characteristic of correspondence spatial point;
Described characteristic of correspondence spatial point is shone upon go back to original image space, obtain multiple image blocks.
4. method according to claim 1, is characterized in that, the parallel intensified learning method of described employing is learnt described multiple image blocks, obtains multiclass feature combination and comprises:
For each intensified learning agency of multiple intensified learnings agency, by the angular setting of described multiple image blocks to preset angles;
Extract respectively described multiple image block characteristics information, obtain the Feature Combination of described intensified learning agency study, described characteristic information comprises color characteristic information, textural characteristics information and shape facility information.
5. a pattern recognition device, is characterized in that, described device comprises:
Image is cut apart module, cuts apart for adopting cloud model to treat recognition image, obtains multiple image blocks;
Characteristic extracting module, for adopting parallel intensified learning device to learn described multiple image blocks, obtains multiclass feature combination;
Fusion Features module, for based on D-S evidence theory, described multiclass feature combination being merged, determines article that described image to be identified comprises.
6. device according to claim 5, is characterized in that, described device also comprises:
Image collection module, for obtaining described image to be identified.
7. device according to claim 5, is characterized in that, image is cut apart module also for the pixel of image space characteristic of correspondence spatial point is represented; Feature space is cut apart in the gathering of feature space according to described characteristic of correspondence spatial point; Described characteristic of correspondence spatial point is shone upon go back to original image space, obtain multiple image blocks.
8. device according to claim 5, is characterized in that, characteristic extracting module is also for each intensified learning agency for multiple intensified learnings agency, by the angular setting of described multiple image blocks to preset angles; Extract respectively described multiple image block characteristics information, obtain the Feature Combination of described intensified learning agency study, described characteristic information comprises color characteristic information, textural characteristics information and shape facility information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410190472.8A CN103942554A (en) | 2014-05-07 | 2014-05-07 | Image identifying method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410190472.8A CN103942554A (en) | 2014-05-07 | 2014-05-07 | Image identifying method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103942554A true CN103942554A (en) | 2014-07-23 |
Family
ID=51190217
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410190472.8A Pending CN103942554A (en) | 2014-05-07 | 2014-05-07 | Image identifying method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103942554A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104766080A (en) * | 2015-05-06 | 2015-07-08 | 苏州搜客信息技术有限公司 | Image multi-class feature recognizing and pushing method based on electronic commerce |
| CN105046191A (en) * | 2015-05-13 | 2015-11-11 | 信阳师范学院 | Texture image identifying method |
| CN105321141A (en) * | 2015-09-17 | 2016-02-10 | 清华大学深圳研究生院 | Image load balancing processing method and apparatus |
| CN107667382A (en) * | 2015-06-26 | 2018-02-06 | 莱克斯真株式会社 | Vehicle number code recognition device and its method |
| CN113283516A (en) * | 2021-06-01 | 2021-08-20 | 西北工业大学 | Multi-sensor data fusion method based on reinforcement learning and D-S evidence theory |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101727657A (en) * | 2008-10-31 | 2010-06-09 | 李德毅 | Image segmentation method based on cloud model |
| US20140029856A1 (en) * | 2012-07-30 | 2014-01-30 | Microsoft Corporation | Three-dimensional visual phrases for object recognition |
-
2014
- 2014-05-07 CN CN201410190472.8A patent/CN103942554A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101727657A (en) * | 2008-10-31 | 2010-06-09 | 李德毅 | Image segmentation method based on cloud model |
| US20140029856A1 (en) * | 2012-07-30 | 2014-01-30 | Microsoft Corporation | Three-dimensional visual phrases for object recognition |
Non-Patent Citations (4)
| Title |
|---|
| 俞志富等: ""一种基于云模型和证据理论的融合识别方法"", 《信息与控制》 * |
| 孟伟 等: ""并行强化学习算法及其应用研究"", 《计算机工程与应用》 * |
| 李先锋等: ""基于SVM和D-S证据理论的多特征融合杂草识别方法"", 《农业机械学报》 * |
| 風尘荏苒: ""图像分割"", 《百度百科：BAIKE.BAIDU.COM/HISTORY/图像分割/20761127#5》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104766080A (en) * | 2015-05-06 | 2015-07-08 | 苏州搜客信息技术有限公司 | Image multi-class feature recognizing and pushing method based on electronic commerce |
| CN105046191A (en) * | 2015-05-13 | 2015-11-11 | 信阳师范学院 | Texture image identifying method |
| CN107667382A (en) * | 2015-06-26 | 2018-02-06 | 莱克斯真株式会社 | Vehicle number code recognition device and its method |
| CN105321141A (en) * | 2015-09-17 | 2016-02-10 | 清华大学深圳研究生院 | Image load balancing processing method and apparatus |
| CN105321141B (en) * | 2015-09-17 | 2018-07-20 | 清华大学深圳研究生院 | A kind of picture load equalization processing method and device |
| CN113283516A (en) * | 2021-06-01 | 2021-08-20 | 西北工业大学 | Multi-sensor data fusion method based on reinforcement learning and D-S evidence theory |
| CN113283516B (en) * | 2021-06-01 | 2023-02-28 | 西北工业大学 | Multi-sensor data fusion method based on reinforcement learning and D-S evidence theory |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110321910B (en) | Point cloud-oriented feature extraction method, device and equipment | |
| US20200012888A1 (en) | Image annotating method and electronic device | |
| EP2806356A1 (en) | Internet of things (iot) application development | |
| CN108416003A (en) | A kind of picture classification method and device, terminal, storage medium | |
| CN103942554A (en) | Image identifying method and device | |
| CN108986801A (en) | A kind of man-machine interaction method, device and human-computer interaction terminal | |
| CN104915351A (en) | Picture sorting method and terminal | |
| CN106203490A (en) | Based on attribute study and the image ONLINE RECOGNITION of interaction feedback, search method under a kind of Android platform | |
| DE102019102866A1 (en) | Identification of a manufacturing part using computer vision and machine learning | |
| CN110610169B (en) | Picture marking method and device, storage medium and electronic device | |
| CN103942220A (en) | Method used for intelligently linking work orders with knowledge of knowledge base and suitable for IT operation and maintenance system | |
| CN109726248A (en) | Vehicle driving track display method, device, system and computer equipment | |
| CN105095479A (en) | Mobile terminal and method for achieving photo classification management | |
| CN109710788A (en) | Image pattern mark and management method and equipment | |
| CN111598899A (en) | Image processing method, image processing apparatus, and computer-readable storage medium | |
| CN108256520B (en) | Method for identifying coin year, terminal equipment and computer readable storage medium | |
| US20200111013A1 (en) | Methods and systems for collecting and analyzing data using a cognitive fabric to provide cognitive agility for the analytic tradecraft | |
| CN104766080A (en) | Image multi-class feature recognizing and pushing method based on electronic commerce | |
| CN111047387B (en) | Recovery management method and device | |
| CN109857878B (en) | Article labeling method and device, electronic equipment and storage medium | |
| CN107077500A (en) | Extract similar group's element | |
| CN103929460A (en) | Method for obtaining state information of contact and mobile device | |
| CN106202440A (en) | Data processing method, device and equipment | |
| CN113743434A (en) | Training method of target detection network, image augmentation method and device | |
| CN105718471A (en) | User preference modeling method, system, and user preference evaluation method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into 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: 20140723 |