CN103632143B - Cloud computing combined object identifying system on basis of images - Google Patents

Cloud computing combined object identifying system on basis of images Download PDF

Info

Publication number
CN103632143B
CN103632143B CN201310651359.0A CN201310651359A CN103632143B CN 103632143 B CN103632143 B CN 103632143B CN 201310651359 A CN201310651359 A CN 201310651359A CN 103632143 B CN103632143 B CN 103632143B
Authority
CN
China
Prior art keywords
image data
module
image
collection thing
cloud computing
Prior art date
Application number
CN201310651359.0A
Other languages
Chinese (zh)
Other versions
CN103632143A (en
Inventor
陈志沅
黄暐皓
Original Assignee
冠捷显示科技(厦门)有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 冠捷显示科技(厦门)有限公司 filed Critical 冠捷显示科技(厦门)有限公司
Priority to CN201310651359.0A priority Critical patent/CN103632143B/en
Publication of CN103632143A publication Critical patent/CN103632143A/en
Application granted granted Critical
Publication of CN103632143B publication Critical patent/CN103632143B/en

Links

Abstract

The invention discloses a cloud computing combined object identifying system on the basis of images. The cloud computing combined object identifying system is characterized in that image data captured by an image capturing module are characterized and processed by a preceding-stage processing module, acquired object image data divided by a dividing module are subjected to similarity comparison computation in a local knowledge base module, and compared and screened acquired object image data are uploaded to a cloud knowledge base module and are identified and analyzed after characteristic values of the acquired object image data are extracted by a representation and description module. The cloud computing combined object identifying system has the advantages that the cloud computing combined object identifying system is high in identification ability and wide in identification scope, while powerful hardware for identifying gestures, a large storage space for storing gesture templates and shortening of operation time of processors are required in the prior art, and accordingly the cloud computing combined object identifying system with simplified hardware, large storage space and short operation time of processors can be widely applied to electronic equipment with the requirement of image identifying and matching.

Description

In conjunction with the article identification based on image for the cloud computing

Technical field

The invention belongs to image identification field, it relates in particular to recognition of face, Activity recognition, gesture knowledge can be used in Not, the article identification based on image for a kind of combination cloud computing on the electronic equipment such as Object identifying.

Background technology

Image recognition technology has tended to ripe at present, and the speed of detection is fast, and accuracy is high, most of existing systems In, express the different meanings by different gestures, the shape of gesture is highly useful for man-machine interaction, but current System in, images of gestures is identified mate, it is desirable to have the hardware device of very high configuration and memory space, processor transport Evaluation time is long, and the identification for images of gestures can only be limited to several patterns of minority, majority can only unit using it is impossible to Complicated image is identified with coupling, the development with image recognition technology and the demand of user, realizes long-range, intellectuality Online Sharing Service be trend of the times.

Content of the invention

In view of this, the present invention provides a kind of article identification based on image for combination cloud computing.This system identification energy Power is high, identification category is wide and hardware requirement that is not needing high standard.

For realizing object above, the present invention employs the following technical solutions:A kind of object based on image for combination cloud computing is known Other system, applies to need to carry out on the electronic equipment of image recognition coupling, this system includes,

Image acquiring module, for capturing video data;

Prime processing module, for receiving the described image acquiring module image data capturing the spy extracting image data Levy;

Segmentation module, the image data that described segmentation module is extracted prime processing module after feature divides into collection thing shadow As data and non-acquired thing image data, and collection thing image data cutting is stored in local side base module.

Local side base module, gathers thing image data for access and is added with several collection thing templates, will deposit The collection thing image data of storage does preliminary similarity contrast conting with the collection thing template added, and will filter out more afterwards Collection thing image data is uploaded to high in the clouds base module;

Represent and describing module, extract the collection thing image data in local side base module further eigenvalue Analysis;

High in the clouds base module, for access collection thing template, and updates interpolation collection thing to local side base module Template;

Identification and parsing module, the collection thing image number after further Eigenvalues analysis is done in described expression and describing module Do similarity contrast conting according to the collection thing template in the base module of described high in the clouds, identify collection thing image data.

Further, after described image acquiring module first leaves the image data of capture in local side knowledge base in, then Image data is sent to prime processing module.

Further, the image data feature that described prime processing module is extracted includes:Edge, COLOR COMPOSITION THROUGH DISTRIBUTION, material or One of angle point or more than one combination.

Further, described prime processing module extract image data feature before, first by described image capturing mould The image data of block capture carries out size reduction or other accelerate identification to process, such as taking-up low-frequency data or statistic histogram, Vector projection etc. is processed.

Further, the collection thing template in described local side base module be updated by high in the clouds base module or Add.

Further, the collection thing image data of described segmentation module cutting is passed through image by described Indigenous knowledge library module Original ROI image in acquisition module is uploaded to high in the clouds base module.

Further, described identification and parsing module do eigenvalue sum operation to the content in collection thing image data, Do different descriptions using different eigenvalues by gathering thing image data.

Further, the data used of the collection thing template contrast in described high in the clouds base module, by according to different Image data feature does corresponding renewal or interpolation.

Further, the collection thing image data identifying in described expression and describing module is stored in the knowledge base of high in the clouds Use as collection thing template contrast.

Further, the collection thing image data that described identification and parsing module will identify that explains, and returns solution The result released.

The present invention adopts above technical scheme, and the image data that image acquiring module is captured is carried out through prime processing module Characterization, the collection thing image data that segmentation module is cut out carries out similarity comparison in local side base module Calculate, and the collection filtering out more afterwards thing image data is uploaded to high in the clouds base module, carried by expression and describing module Take identification and parsing collection thing image data after eigenvalue;The system identification ability that the present invention provides is high, identification category is wide, and works as Front technology needs to do gesture identification it is also desirable to a very big storage area to store gesture template by a powerful hardware And need to consume the operation time of processor, the operation time of the hardware simplified, storage area and processor of the present invention. Can widely apply to need to carry out on the electronic equipment of image recognition coupling.

Brief description

In conjunction with accompanying drawing, the present invention is further described:

Fig. 1 is that the present invention combines the article identification structural representation based on image for the cloud computing;

Fig. 2 is article identification local side system processor schematic diagram of the present invention;

Fig. 3 is article identification cloud system treatment mechanism schematic diagram of the present invention.

Specific embodiment

Refer to shown in Fig. 1, a kind of article identification based on image for the combination cloud computing that the present invention provides, apply to Need to carry out on the electronic equipment of image recognition coupling, this system includes,

Image acquiring module 101, for capturing video data;Prime processing module 102, for receiving described image capturing The image data of module 101 capture simultaneously extracts the feature of image data;Segmentation module, described segmentation module is prime processing module Image data after 102 extraction features divides into collection thing image data and non-acquired thing image data, and will gather thing image Data cutting is stored in local side base module 104.

Local side base module 104, gathers thing image data for access and is added with several collection thing templates, will Storage collection thing image data with add collection thing template do preliminary similarity contrast conting, and by after filter out Collection thing image data be uploaded to high in the clouds base module 106;

Represent and describing module 105, the collection thing image data in extraction local side base module is simultaneously further special Value indicative is analyzed;

High in the clouds base module 106, for access collection thing template, and updates interpolation to local side base module 104 Collection thing template;

Identification and parsing module 107, the collection thing after further Eigenvalues analysis is done in described expression and describing module 105 Image data does similarity contrast conting with the collection thing template in described high in the clouds base module 106, identifies collection thing shadow As data.

Refer to Fig. 1 and Fig. 2 the process operation of local end system is described in detail.

In step 201, image acquiring module 101 first leaves the image data capturing in local side base module In 104, the image data constituting for collection thing and non-acquired thing when depositing stores as an image data, then transfers to Prime processing module 102 is done preliminary image data and is processed;

In step 202., prime processing module 102 is first the image data size reduction receiving or other acceleration identifications Process, for example, take out low-frequency data, statistic histogram or vector projection etc. and process.Fall during preliminary computing is done in itself with sharp hardware Low efficiency needed for it;

In step 203, image data is done characterization by prime processing module 102, such as Sobel, Canny Edge ... waits edge extraction or other eigenvalue technology such as color Histogram etc. to take its strong feature as comparison foundation, The feature of required extraction comprises one of features such as edge, COLOR COMPOSITION THROUGH DISTRIBUTION, material, Corner or more than one combination;

In step 204, segmentation module, for the scope of collection thing, does dividing processing, specially:Mould is processed by prime Block 102 processes the data of coming, and here is compared by simple comparison mechanism such as eigenvalue, the cutting of front background, dynamically cutting such as MHI Deng or using grader such as SVM, Adaboost etc. by image data divide into may collection thing and non-acquired thing image, and May gather in the cutting of thing image data locally stored end base module 106, here ignores the image of non-acquired thing.

In step 205, local side base module 104 will split the possible collection thing image data that module cutting obtains With storage collection thing template do preliminary similarity contrast conting, by with collection thing template matching, by collection thing do tentatively Classification, local side base module 104 is stored with and gathers thing template, and described collection thing template is by high in the clouds base module 106 To update and to add;

In step 206, the relative position of segmentation collection thing place block is obtained by image acquiring module 101 Original ROI image data;

In step 207, the collection thing image data that the original ROI image data obtaining filters out more afterwards is uploaded High in the clouds base module 106 does fine comparison.

By this kind of mode, image can be not necessary to whole Zhang Shangchuan high in the clouds base module 106 and make comparison, only need to will may adopt Collection thing image data block uploads.

Refer to Fig. 3 cloud system process operation is described in detail.

In a step 208, content analysis are carried out to the collection thing image data receiving, do more complicated eigenvalue and be added Computing.As SIFT, HOG, AAM, LBP etc., draw the eigenvalue after conversion.Thing will be gathered using the different eigenvalue after conversion Image data does different descriptions, using different description contents as the foundation recognizing.

In step 209, by the graders such as such as neural network, SVM by the collection thing image data of required differentiation by Image acquiring module 101, prime processing module 102, segmentation module, expression and describing module 105 and identification and parsing module 107 The strong eigenvalue resolution information training gained that the collection thing image data getting is converted to judges that information is stored in local side and knows Know library module 104, the template matching as collection thing uses.The data of thing template matching is gathered in high in the clouds base module 106, Can be updated according to required identification objects feature and add.

In step 210, the collection thing image data after Eigenvalues analysis will be carried out by expression and describing module 105, with Collection thing template stored by high in the clouds base module 106 carries out similarity and finely compares, and collection thing image data is distinguished Come.

In step 211, the collection thing image data result after passback identification is explained.

Although specifically show and describe this patent in conjunction with preferred embodiment, those skilled in the art should be bright In vain, in the spirit and scope of the present invention being limited without departing from appended claims, in the form and details can be right The present invention makes a variety of changes, and is protection scope of the present invention.

Claims (8)

1. the article identification based on image for a kind of combination cloud computing, applies to carry out the electronic equipment of image recognition coupling Upper it is characterised in that:This system includes,
Image acquiring module, for capturing video data;
Prime processing module, for receiving the described image acquiring module image data capturing the feature extracting image data, The image data feature that described prime processing module is extracted includes:One of edge, COLOR COMPOSITION THROUGH DISTRIBUTION, material or angle point or one Plant above combination;
Segmentation module, the image data that described segmentation module is extracted prime processing module after feature divides into collection thing image number According to non-acquired thing image data, and by collection thing image data cutting be stored in local side base module;
Local side base module, gathers thing image data for access and is added with several collection thing templates, by storage Collection thing image data does preliminary similarity contrast conting with the collection thing template added, and by the collection filtering out more afterwards Thing image data is uploaded to high in the clouds base module;
Represent and describing module, extract the collection thing image data in local side base module and further eigenvalue divides Analysis, further feature value analyze be to collection thing image data in content do eigenvalue sum operation, using different spies Value indicative does different descriptions by gathering thing image data;
High in the clouds base module, for access collection thing template, and updates interpolation collection thing template to local side base module;
Identification and parsing module, by described represent and describing module do the collection thing image data after further Eigenvalues analysis with Collection thing template in the base module of described high in the clouds does similarity contrast conting, identifies collection thing image data.
2. the article identification based on image for the combination cloud computing according to claim 1 it is characterised in that:Described shadow After first leaving the image data of capture in local side knowledge base in as acquisition module, then image data is sent to prime process Module.
3. the article identification based on image for the combination cloud computing according to claim 1 and 2 it is characterised in that:Described Prime processing module before extracting image data feature, first the image data of described image acquiring module capture is carried out Size reduction is processed.
4. the article identification based on image for the combination cloud computing according to claim 1 it is characterised in that:Described basis Collection thing template in ground terminal base module is updated or is added by high in the clouds base module.
5. the article identification based on image for the combination cloud computing according to claim 1 it is characterised in that:Described segmentation The collection thing image data of module cutting obtains original ROI image by image acquiring module, described Indigenous knowledge library module This original ROI image is uploaded to high in the clouds base module.
6. the article identification based on image for the combination cloud computing according to claim 1 it is characterised in that:Described cloud Collection thing template contrast in the base module of end data used, will do corresponding renewal according to different image data features Or add.
7. the article identification based on image for the combination cloud computing according to claim 1 it is characterised in that:Described table Show and describing module in the collection thing image data that identifies be stored in the knowledge base of high in the clouds and use as collection thing template contrast.
8. the article identification based on image for the combination cloud computing according to claim 1 it is characterised in that:Described knowledge The collection thing image data that other and parsing module will identify that explains, and returns the result of explanation.
CN201310651359.0A 2013-12-05 2013-12-05 Cloud computing combined object identifying system on basis of images CN103632143B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201310651359.0A CN103632143B (en) 2013-12-05 2013-12-05 Cloud computing combined object identifying system on basis of images

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201310651359.0A CN103632143B (en) 2013-12-05 2013-12-05 Cloud computing combined object identifying system on basis of images

Publications (2)

Publication Number Publication Date
CN103632143A CN103632143A (en) 2014-03-12
CN103632143B true CN103632143B (en) 2017-02-08

Family

ID=50213172

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201310651359.0A CN103632143B (en) 2013-12-05 2013-12-05 Cloud computing combined object identifying system on basis of images

Country Status (1)

Country Link
CN (1) CN103632143B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105224066A (en) * 2014-06-03 2016-01-06 北京创思博德科技有限公司 A kind of gesture identification method based on high in the clouds process
CN105205436B (en) * 2014-06-03 2019-09-06 北京创思博德科技有限公司 A kind of gesture recognition system based on forearm bioelectricity multisensor
CN107292223A (en) * 2016-04-13 2017-10-24 芋头科技(杭州)有限公司 A kind of online verification method and system of real-time gesture detection
CN106844569A (en) * 2017-01-03 2017-06-13 陕西甄宝网络有限公司 A kind of frame selects identifying system and method

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101216884A (en) * 2007-12-29 2008-07-09 北京中星微电子有限公司 A method and system for face authentication
CN103310466A (en) * 2013-06-28 2013-09-18 安科智慧城市技术(中国)有限公司 Single target tracking method and achievement device thereof
CN103324955A (en) * 2013-06-14 2013-09-25 浙江智尔信息技术有限公司 Pedestrian detection method based on video processing

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8942483B2 (en) * 2009-09-14 2015-01-27 Trimble Navigation Limited Image-based georeferencing

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101216884A (en) * 2007-12-29 2008-07-09 北京中星微电子有限公司 A method and system for face authentication
CN103324955A (en) * 2013-06-14 2013-09-25 浙江智尔信息技术有限公司 Pedestrian detection method based on video processing
CN103310466A (en) * 2013-06-28 2013-09-18 安科智慧城市技术(中国)有限公司 Single target tracking method and achievement device thereof

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
医学图像处理与检索技术的研究;曹阳;《中国优秀硕士学位论文全文数据库 信息科技辑》;20130315(第03期);论文正文第5-51页 *
图像特征提取与匹配算法研究;赵艳丽;《中国优秀硕士学位论文全文数据库 信息科技辑》;20110715(第07期);论文正文第1-59页 *
基于内容图像检索与敏感图像过滤的若干算法研究;孙艳;《中国博士学位论文全文数据库 信息科技辑》;20120515(第05期);论文正文第1-80页 *

Also Published As

Publication number Publication date
CN103632143A (en) 2014-03-12

Similar Documents

Publication Publication Date Title
Kadir et al. Leaf classification using shape, color, and texture features
Keskin et al. Hand pose estimation and hand shape classification using multi-layered randomized decision forests
Pflug et al. Ear biometrics: a survey of detection, feature extraction and recognition methods
Lempitsky et al. Learning to count objects in images
US8792722B2 (en) Hand gesture detection
Kim et al. Spatiotemporal saliency detection and its applications in static and dynamic scenes
Khan Image segmentation techniques: A survey
Yun et al. An automatic hand gesture recognition system based on Viola-Jones method and SVMs
Wu et al. One shot learning gesture recognition from rgbd images
Ikizler et al. Recognizing actions from still images
TWI395145B (en) Hand gesture recognition system and method
Alsmadi et al. Fish recognition based on robust features extraction from size and shape measurements using neural network
KR100847143B1 (en) System and Method for analyzing of human motion based silhouettes of real-time video stream
Baraldi et al. Gesture recognition in ego-centric videos using dense trajectories and hand segmentation
Levine et al. Saliency detection based on frequency and spatial domain analysis
Gao et al. Multi-view discriminative and structured dictionary learning with group sparsity for human action recognition
Cope et al. Plant texture classification using gabor co-occurrences
CN102214298B (en) Method for detecting and identifying airport target by using remote sensing image based on selective visual attention mechanism
CN103679147A (en) Method and device for identifying model of mobile phone
CN102332092A (en) Flame detection method based on video analysis
Kim et al. Spatiotemporal saliency detection using textural contrast and its applications
Schwartz et al. A novel feature descriptor based on the shearlet transform
Salahat et al. Recent advances in features extraction and description algorithms: A comprehensive survey
Chen et al. Pedestrian detection with deep convolutional neural network
CN104036287A (en) Human movement significant trajectory-based video classification method

Legal Events

Date Code Title Description
PB01 Publication
SE01 Entry into force of request for substantive examination
C10 Entry into substantive examination
GR01 Patent grant
C14 Grant of patent or utility model