CN107045642A - A kind of logo image-recognizing method and device - Google Patents

A kind of logo image-recognizing method and device Download PDF

Info

Publication number
CN107045642A
CN107045642A CN201710311843.7A CN201710311843A CN107045642A CN 107045642 A CN107045642 A CN 107045642A CN 201710311843 A CN201710311843 A CN 201710311843A CN 107045642 A CN107045642 A CN 107045642A
Authority
CN
China
Prior art keywords
logo
image
identified
grader
cluster
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
Application number
CN201710311843.7A
Other languages
Chinese (zh)
Inventor
程良伦
陈仿雄
佘爽
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangdong University of Technology
Original Assignee
Guangdong University of Technology
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 Guangdong University of Technology filed Critical Guangdong University of Technology
Priority to CN201710311843.7A priority Critical patent/CN107045642A/en
Publication of CN107045642A publication Critical patent/CN107045642A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • 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
    • G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
    • G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G06V20/54—Surveillance or monitoring of activities, e.g. for recognising suspicious objects of traffic, e.g. cars on the road, trains or boats
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/08—Detecting or categorising vehicles

Abstract

The invention discloses a kind of logo image-recognizing method and device, logo grader is pre-established, logo grader is the grader set up after being trained to multiple logo sample images；By obtaining logo image to be identified；Extract the SURF features of logo image to be identified；The SURF feature sets of logo image to be identified are clustered using hard clustering method, cluster result is generated；Cluster result is inputted into grader, corresponding logo classification of type result is exported.Logo image-recognizing method and device provided by the present invention, have preferable recognition result, accuracy of identification is higher for low-quality logo image, have certain robustness for obtaining logo image under complex environment.

Description

A kind of logo image-recognizing method and device
Technical field
The present invention relates to image identification technical field, more particularly to a kind of logo image-recognizing method and device.
Background technology
Vehicle-logo recognition can provide effective letter as the important part of intelligent transportation system to vehicle identification Breath, particularly relates to increasing for car crime, and configuration high-tech detective work is even more extremely urgent.Some offenders rob vehicle in robber After, escape and chase often through modes such as replacing car plate, change vehicle colors, considerably increase difficulty of solving a case.Logo conduct The important symbol of vehicle, provide not only vehicle model information, additionally provides the information of manufacturer and is difficult change.These are heavy Want information that investigation scope of the police to suspect vehicle during solving a case is reduced rapidly, accelerate the speed solved a case.And With developing rapidly for image processing techniques and computer technology, the related ITS products based on vehicle-logo recognition are bound to have more To be more widely applied prospect.
Although the exploration and research to vehicle-logo recognition technology deepen continuously in recent years, and achieve certain achievement, It is ripe not enough to the paractical research of vehicle-logo recognition technology, however it remains many stubborn problems are difficult to solve.In complicated ring Under border, such as sleet, haze, uneven illumination, in addition logo there is a situation where defaced, partial occlusion in itself, these constantly become The factor of change have impact on the collection of logo image, logo image is produced larger change, to the accurate of logo Positioning and identification bring very big difficulty.
Conventional automobile logo identification method is template matches or characteristic matching at present.Phyllos et al. proposes a kind of base In the automobile logo identification method of the enhancing matching of Scale invariant features transform (SIFT).Although having certain Shandong under complex environment Rod, but because intrinsic dimensionality is excessive, amount of calculation is excessive, and real-time is not strong.The direction gradient Nogata more commonly used at present in addition Figure HOG and support vector machines grader are identified, but because HOG description are quite sensitive to noise, easily by environment The influence of condition, causes discrimination relatively low.
The content of the invention
It is an object of the invention to provide a kind of logo image-recognizing method and device, to solve existing vehicle-logo recognition in complexity Influenceed under environment by environment factor, cause the problem of discrimination is relatively low.
In order to solve the above technical problems, the present invention provides a kind of logo image-recognizing method, including：
Logo grader is pre-established, the logo grader is set up after being trained to multiple logo sample images Grader；
Obtain logo image to be identified；
Extract the SURF features of the logo image to be identified；
The SURF feature sets of logo image to be identified are clustered using hard clustering method, cluster result is generated；
The cluster result is inputted into the grader, corresponding logo classification of type result is exported.
Alternatively, the acquisition logo image to be identified includes：
Original vehicle image to be identified is pre-processed；
The car mark region in the original vehicle image to be identified after pretreatment is determined, the logo to be identified is used as Image.
Alternatively, it is described that original vehicle image progress pretreatment to be identified is included：
The original vehicle image to be identified is converted into gray level image；
Smothing filtering is carried out using the method for gaussian filtering；
Rim detection is carried out using Sobel operators；
Binary conversion treatment is carried out to the image after detection；
The image after binary conversion treatment is filtered using morphologic filtering.
Alternatively, the car mark region determined in the original vehicle image to be identified after pretreatment includes：
Determine the car plate position in the original vehicle image to be identified after pretreatment；
According to the corresponding relation of the car plate position that pre-establishes and logo position, after pretreatment described original treat is determined Recognize the car mark region in vehicle image.
Alternatively, also include after the SURF features for extracting the logo image to be identified：
Calculate and extract each characteristic point entropy after SURF features；
Characteristic point information contained amount is detected according to the characteristic point entropy, the feature that information content is less than predetermined threshold value is screened out Point.
Alternatively, described that the SURF feature sets of logo image to be identified are clustered using hard clustering method, generation is poly- Class result includes：
From the SURF feature sets, selection is N number of to be characterized as initial cluster center μ；
Calculate each feature d in addition to cluster centre μiTo center μjEurope it is several inner apart from Dij, and return distance is most short Class is corresponding cluster CjIn；Dij=| | di-μj||2
Update the barycenter of cluster：
The process with updating is computed repeatedly, untill reaching default precision, N number of cluster centre is obtained.
Alternatively, the grader is logistic regression grader, described to input the cluster result to the grader In, exporting corresponding logo classification of type result includes：
According to the probable value size of each class logo type matching in the grader, the maximum classification of output probability value is made For logo classification of type result.
Present invention also offers a kind of logo pattern recognition device, including：
Built in advance module, for pre-establishing logo grader, the logo grader is that multiple logo sample images are entered The grader set up after row training；
Image collection module, for obtaining logo image to be identified；
Characteristic extracting module, the SURF features for extracting the logo image to be identified；
Cluster module, for being clustered using hard clustering method to the SURF feature sets of logo image to be identified, is generated Cluster result；
Classification and Identification module, for the cluster result to be inputted into the grader, exports corresponding logo type Classification results.
Alternatively, described image acquisition module includes：
Pretreatment unit, for being pre-processed to original vehicle image to be identified；
Car mark region determining unit, for determining the logo area in the original vehicle image to be identified after pretreatment Domain, is used as the logo image to be identified.
Alternatively, the pretreatment unit includes：
Conversion subunit, for the original vehicle image to be identified to be converted into gray level image；
Filtering subunit, for carrying out smothing filtering using the method for gaussian filtering；
Rim detection subelement, for carrying out rim detection using Sobel operators；
Binary conversion treatment subelement, for carrying out binary conversion treatment to the image after detection；
Morphologic filtering subelement, for being filtered using morphologic filtering to the image after binary conversion treatment.
Logo image-recognizing method and device provided by the present invention, pre-establish logo grader, and logo grader is The grader set up after being trained to multiple logo sample images；By obtaining logo image to be identified；Extract car to be identified The SURF features of logo image；The SURF feature sets of logo image to be identified are clustered using hard clustering method, generation cluster As a result；Cluster result is inputted into grader, corresponding logo classification of type result is exported.Logo figure provided by the present invention As recognition methods and device, there is preferable recognition result for low-quality logo image, accuracy of identification is higher, for complex environment Lower acquisition logo image has certain robustness.
Brief description of the drawings
, below will be to embodiment or existing for the clearer explanation embodiment of the present invention or the technical scheme of prior art The accompanying drawing used required in technology description is briefly described, it should be apparent that, drawings in the following description are only this hair Some bright embodiments, for those of ordinary skill in the art, on the premise of not paying creative work, can be with root Other accompanying drawings are obtained according to these accompanying drawings.
Fig. 1 is a kind of flow chart of embodiment of logo image-recognizing method provided by the present invention；
Fig. 2 is to obtain the process schematic of logo image to be identified in logo image-recognizing method provided by the present invention；
Fig. 3 is the flow chart of another embodiment of logo image-recognizing method provided by the present invention；
Fig. 4 is the structured flowchart of logo pattern recognition device provided in an embodiment of the present invention.
Embodiment
In order that those skilled in the art more fully understand the present invention program, with reference to the accompanying drawings and detailed description The present invention is described in further detail.Obviously, described embodiment is only a part of embodiment of the invention, rather than Whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art are not making creative work premise Lower obtained every other embodiment, belongs to the scope of protection of the invention.
A kind of flow chart of embodiment of logo image-recognizing method provided by the present invention was as shown in figure 1, should Method includes：
Step S101：Logo grader is pre-established, the logo grader is that multiple logo sample images are instructed The grader set up after white silk；
Step S102：Obtain logo image to be identified；
As a kind of embodiment, reference picture 2 obtains in logo image-recognizing method provided by the present invention and waits to know Other logo image can be specifically included：
Step S1021：Original vehicle image to be identified is pre-processed；
Specifically, the process pre-processed can include：The original vehicle image to be identified is converted into gray-scale map Picture；Smothing filtering is carried out using the method for gaussian filtering；Rim detection is carried out using Sobel operators；Image after detection is entered Row binary conversion treatment；The image after binary conversion treatment is filtered using morphologic filtering.
Step S1022：The car mark region in the original vehicle image to be identified after pretreatment is determined, as described Logo image to be identified.
Determine the car plate position in the original vehicle image to be identified after pretreatment；According to the car plate pre-established The corresponding relation of position and logo position, determines the car mark region in the original vehicle image to be identified after pretreatment.
Step S103：Extract the SURF features of the logo image to be identified；
Step S104：The SURF feature sets of logo image to be identified are clustered using hard clustering method, generation cluster As a result；
Step S105：The cluster result is inputted into the grader, corresponding logo classification of type result is exported.
Logo image-recognizing method provided by the present invention, pre-establishes logo grader, and logo grader is to multiple The grader that logo sample image is set up after being trained；By obtaining logo image to be identified；Extract logo image to be identified SURF features；The SURF feature sets of logo image to be identified are clustered using hard clustering method, cluster result is generated；Will Cluster result is inputted into grader, exports corresponding logo classification of type result.Logo image recognition provided by the present invention Method, has preferable recognition result, accuracy of identification is higher for low-quality logo image, for obtaining logo figure under complex environment As having certain robustness.
As shown in the flow chart of another embodiment of Fig. 3 logo image-recognizing methods provided by the present invention, Grader is logistic regression grader in this method, and it is specifically included：
Step S201：Pre-establish logo grader；
Pre-establishing the process of logo grader includes：
Step S2011：The extraction of SURF features is carried out for training logo sample image, SURF feature sets are obtained；
The embodiment of the present invention can use characteristic point screening strategy after feature is extracted using SURF algorithm：Calculate just Walk the characteristic point entropy of extraction to detect characteristic point information contained amount, remove the too low characteristic point of information content.
Step S2012：Clustered using K-means methods for SURF features, set up corresponding visual dictionary, obtain N number of keyword and corresponding weight；
Step S2013：Using the N number of keyword and its weight of generation as the input of logistic regression grader, train many The logic of class returns grader.
The specific steps that visual dictionary is set up include：
(1) from SURF feature sets, arbitrarily take and N number of be characterized as initial cluster center μ.
(2) each feature d in addition to cluster centre μ is calculatedi, to center μjEurope it is several inner apart from Dij, and will be apart from most Short is classified as corresponding cluster CjIn.
Dij=| | di-μj||2
(3) barycenter of cluster is updated.
(4) repetitive process (2) and (3), the precision specified until reaching obtain N number of cluster centre, as N number of visual word Allusion quotation.
(5) define one and calculate feature with the degree of correlation of visual dictionary as corresponding weight apart from kernel function.
In training grader step, specific steps include：
The logistic regression of two classification is expanded into multicategory classification, it is assumed that training data X is given from N classes.Classifying more In each class is defined as：
Define a target function g (a), enabling be applied to two-sorted logic regressive object function E and cause minimum Assess Wk and bk.
Using improved logistic regression grader, according to the probable value size of each class, output probability greatly to be final Classification results.
Step S202：Obtain logo image to be identified；
Step S203：The image of acquisition is pre-processed, according to priori, license plate area is finally filtered out；
In order to strengthen image information, image is pre-processed first, image will be obtained and be converted into gray level image；Using height The method of this filtering carries out smooth noise reduction；There is certain rule vertical direction edge is intensive using car plate, calculated using Sobel Son carries out rim detection.Then a threshold value is set, binary conversion treatment is carried out to image；Finally use morphologic filtering so that License plate area one connection shape substantially of formation.
According to priori, because the size, length, width of license plate area have certain limit, it may reduce and differentiate model Enclose.
L and H is the length and width of connected region, and A is the area of connected region, and Hmin and Hmax are expressed as width Bound, similarly, thus length and area are also.
Further screened according to the priori between characters on license plate, there is certain rule, seven between characters on license plate At least there is 6 saltus steps between individual character, according to this characteristic, by counting the Gray Level Jump of license plate candidate area, less than 6 times Screened, select final license plate area.
Step S204：According to the geometrical relationship between car plate and logo, rim detection, Morphological scale-space are carried out, level is hung down Shadow is delivered directly, accurate car mark region is obtained；
According to the geometrical relationship between car plate and logo.Determine logo region interested.
Wherein, (ROIl.x, ROIl.y) is the top left co-ordinate of logo area-of-interest, and (ROIr.x, ROIr.y) is the right side Lower angular coordinate, coordinate value is all greater than 0 here, and rect.x is car plate top left co-ordinate, is the length and height of car plate, α and β For normalization coefficient, rational logo region interested is obtained.
After being pre-processed to region interested, its processing procedure is with carrying out floor projection and vertical throwing to image Shadow.The up-and-down boundary of horizontal column projection can substantially determine the up-and-down boundary of car mark region, and similarly vertical pillar projection is determined Logo right boundary.Consider in pretreatment, the edge feature information of a part, therefore the threshold value choosing of projection histogram can be lost Select, carry out appropriate expansion, obtain accurately logo position.
Step S205：Extract the SURF feature sets of car mark region；
Step S206：Using K-means clustering methods, N keywords and corresponding weight are obtained, it is many as what is trained The logic of class returns the input of grader, exports corresponding classification results.
Feature is proposed using SURF algorithm to the car mark region of candidate, then using K-means, obtain n keyword and Corresponding weighted value, the input of grader is returned as many-sorted logic, for the data point X* of test, for the general of K classes mark Rate is P (y*=k), so that the larger as final classification results of output probability.
The present invention can specifically use database for 10 class logos, and each class logo image is 1000, including Buick, Honda, popular, Toyota etc..The pixel of each image is 70*70, and resolution ratio is low, and each image is in different weather and light Generated according under the conditions of.The image of the various acquisitions under actual conditions is met, as shown in Figure 4.
The method that the embodiment of the present invention is provided carries out SURF feature extractions for training logo sample image, using k- Means algorithms produce visual dictionary, are trained by using improved logistic regression grader；Then to collecting vehicle figure Picture, first orients the position of car plate, and the positioning for logo is realized according to the priori between logo and car plate, then to fixed The car mark region that position goes out produces feature set using SURF.Using K-means clustering methods, N number of keyword and corresponding power are obtained Weight, as the input of the grader trained, exports corresponding result.The embodiment of the present invention uses improved logistic regression Grader, effectively can be identified under complex environment for the logo database of acquisition, improve accuracy of identification and robustness.
Logo pattern recognition device provided in an embodiment of the present invention is introduced below, logo image described below is known Other device can be mutually to should refer to above-described logo image-recognizing method.
Fig. 4 is the structured flowchart of logo pattern recognition device provided in an embodiment of the present invention, the logo image recognition of reference picture 4 Device can include：
Built in advance module 100, for pre-establishing logo grader, the logo grader is to multiple logo sample images The grader set up after being trained；
Image collection module 200, for obtaining logo image to be identified；
Characteristic extracting module 300, the SURF features for extracting the logo image to be identified；
Cluster module 400, it is raw for being clustered using hard clustering method to the SURF feature sets of logo image to be identified Into cluster result；
Classification and Identification module 500, for the cluster result to be inputted into the grader, exports corresponding logo class Type classification results.
As a kind of embodiment, in logo pattern recognition device provided by the present invention, described image obtains mould Block can be specifically included：
Pretreatment unit, for being pre-processed to original vehicle image to be identified；
Car mark region determining unit, for determining the logo area in the original vehicle image to be identified after pretreatment Domain, is used as the logo image to be identified.
As a kind of embodiment, in logo pattern recognition device provided by the present invention, the pretreatment unit Including：
Conversion subunit, for the original vehicle image to be identified to be converted into gray level image；
Filtering subunit, for carrying out smothing filtering using the method for gaussian filtering；
Rim detection subelement, for carrying out rim detection using Sobel operators；
Binary conversion treatment subelement, for carrying out binary conversion treatment to the image after detection；
Morphologic filtering subelement, for being filtered using morphologic filtering to the image after binary conversion treatment.
The logo pattern recognition device of the present embodiment is used to realize foregoing logo image-recognizing method, therefore logo image The embodiment part of the visible logo image-recognizing method hereinbefore of embodiment in identifying device, for example, pre- modeling Block 100, image collection module 200, characteristic extracting module 300, cluster module 400, Classification and Identification module 500 is respectively used to reality Step S101, S102, S103, S104 and S105 in existing above-mentioned logo image-recognizing method, so, its embodiment can be with With reference to the description of corresponding various pieces embodiment, it will not be repeated here.
Logo pattern recognition device provided by the present invention, pre-establishes logo grader, and logo grader is to multiple The grader that logo sample image is set up after being trained；By obtaining logo image to be identified；Extract logo image to be identified SURF features；The SURF feature sets of logo image to be identified are clustered using hard clustering method, cluster result is generated；Will Cluster result is inputted into grader, exports corresponding logo classification of type result.Logo image recognition provided by the present invention Device, has preferable recognition result, accuracy of identification is higher for low-quality logo image, for obtaining logo figure under complex environment As having certain robustness.
The embodiment of each in this specification is described by the way of progressive, what each embodiment was stressed be with it is other Between the difference of embodiment, each embodiment same or similar part mutually referring to.For being filled disclosed in embodiment For putting, because it is corresponded to the method disclosed in Example, so description is fairly simple, related part is referring to method part Explanation.
Professional further appreciates that, with reference to the unit of each example of the embodiments described herein description And algorithm steps, can be realized with electronic hardware, computer software or the combination of the two, in order to clearly demonstrate hardware and The interchangeability of software, generally describes the composition and step of each example according to function in the above description.These Function is performed with hardware or software mode actually, depending on the application-specific and design constraint of technical scheme.Specialty Technical staff can realize described function to each specific application using distinct methods, but this realization should not Think beyond the scope of this invention.
Directly it can be held with reference to the step of the method or algorithm that the embodiments described herein is described with hardware, processor Capable software module, or the two combination are implemented.Software module can be placed in random access memory (RAM), internal memory, read-only deposit Reservoir (ROM), electrically programmable ROM, electrically erasable ROM, register, hard disk, moveable magnetic disc, CD-ROM or technology In any other form of storage medium well known in field.
Logo image-recognizing method provided by the present invention and device are described in detail above.It is used herein Specific case is set forth to the principle and embodiment of the present invention, and the explanation of above example is only intended to help and understands The method and its core concept of the present invention.It should be pointed out that for those skilled in the art, not departing from this On the premise of inventive principle, some improvement and modification can also be carried out to the present invention, these are improved and modification also falls into the present invention In scope of the claims.

Claims (10)

1. a kind of logo image-recognizing method, it is characterised in that including：
Logo grader is pre-established, the logo grader is the classification set up after being trained to multiple logo sample images Device；
Obtain logo image to be identified；
Extract the SURF features of the logo image to be identified；
The SURF feature sets of logo image to be identified are clustered using hard clustering method, cluster result is generated；
The cluster result is inputted into the grader, corresponding logo classification of type result is exported.
2. logo image-recognizing method as claimed in claim 1, it is characterised in that the acquisition logo image bag to be identified Include：
Original vehicle image to be identified is pre-processed；
The car mark region in the original vehicle image to be identified after pretreatment is determined, the logo figure to be identified is used as Picture.
3. logo image-recognizing method as claimed in claim 2, it is characterised in that described to enter to original vehicle image to be identified Row pretreatment includes：
The original vehicle image to be identified is converted into gray level image；
Smothing filtering is carried out using the method for gaussian filtering；
Rim detection is carried out using Sobel operators；
Binary conversion treatment is carried out to the image after detection；
The image after binary conversion treatment is filtered using morphologic filtering.
4. logo image-recognizing method as claimed in claim 2, it is characterised in that the determination original after pretreatment The car mark region begun in vehicle image to be identified includes：
Determine the car plate position in the original vehicle image to be identified after pretreatment；
According to the corresponding relation of the car plate position pre-established and logo position, determine after pretreatment described original to be identified Car mark region in vehicle image.
5. the logo image-recognizing method as described in any one of Claims 1-4, it is characterised in that treated described in the extraction Also include after the SURF features for recognizing logo image：
Calculate and extract each characteristic point entropy after SURF features；
Characteristic point information contained amount is detected according to the characteristic point entropy, the characteristic point that information content is less than predetermined threshold value is screened out.
6. the logo image-recognizing method as described in any one of Claims 1-4, it is characterised in that described using hard cluster side Method is clustered to the SURF feature sets of logo image to be identified, and generation cluster result includes：
From the SURF feature sets, selection is N number of to be characterized as initial cluster center μ；
Calculate each feature d in addition to cluster centre μiTo center μjEurope it is several inner apart from Dij, and it is classified as phase by distance is most short The cluster C answeredjIn；Dij=| | di-μj||2
Update the barycenter of cluster：
The process with updating is computed repeatedly, untill reaching default precision, N number of cluster centre is obtained.
7. the logo image-recognizing method as described in any one of Claims 1-4, it is characterised in that the grader is logic Grader is returned, it is described to input the cluster result into the grader, export corresponding logo classification of type result bag Include：
According to the probable value size of each class logo type matching in the grader, the maximum classification of output probability value is used as car Mark classification of type result.
8. a kind of logo pattern recognition device, it is characterised in that including：
Built in advance module, for pre-establishing logo grader, the logo grader is that multiple logo sample images are instructed The grader set up after white silk；
Image collection module, for obtaining logo image to be identified；
Characteristic extracting module, the SURF features for extracting the logo image to be identified；
Cluster module, for being clustered using hard clustering method to the SURF feature sets of logo image to be identified, generation cluster As a result；
Classification and Identification module, for the cluster result to be inputted into the grader, exports corresponding logo classification of type As a result.
9. logo pattern recognition device as claimed in claim 8, it is characterised in that described image acquisition module includes：
Pretreatment unit, for being pre-processed to original vehicle image to be identified；
Car mark region determining unit, for determining the car mark region in the original vehicle image to be identified after pretreatment, It is used as the logo image to be identified.
10. logo pattern recognition device as claimed in claim 9, it is characterised in that the pretreatment unit includes：
Conversion subunit, for the original vehicle image to be identified to be converted into gray level image；
Filtering subunit, for carrying out smothing filtering using the method for gaussian filtering；
Rim detection subelement, for carrying out rim detection using Sobel operators；
Binary conversion treatment subelement, for carrying out binary conversion treatment to the image after detection；
Morphologic filtering subelement, for being filtered using morphologic filtering to the image after binary conversion treatment.
CN201710311843.7A 2017-05-05 2017-05-05 A kind of logo image-recognizing method and device Pending CN107045642A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710311843.7A CN107045642A (en) 2017-05-05 2017-05-05 A kind of logo image-recognizing method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710311843.7A CN107045642A (en) 2017-05-05 2017-05-05 A kind of logo image-recognizing method and device

Publications (1)

Publication Number Publication Date
CN107045642A true CN107045642A (en) 2017-08-15

Family

ID=59546072

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710311843.7A Pending CN107045642A (en) 2017-05-05 2017-05-05 A kind of logo image-recognizing method and device

Country Status (1)

Country Link
CN (1) CN107045642A (en)

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103077407A (en) * 2013-01-21 2013-05-01 信帧电子技术（北京）有限公司 Car logo positioning and recognition method and car logo positioning and recognition system
CN103279738A (en) * 2013-05-09 2013-09-04 上海交通大学 Automatic identification method and system for vehicle logo
ES2492315A1 (en) * 2013-03-05 2014-09-08 Universidad De Alcalá Procedure for the recognition of vehicle brands by classification of the logo and device for its realization (Machine-translation by Google Translate, not legally binding)
CN104832418A (en) * 2015-05-07 2015-08-12 北京航空航天大学 Hydraulic pump fault diagnosis method based on local mean conversion and Softmax
CN105160292A (en) * 2015-07-07 2015-12-16 深圳市捷顺科技实业股份有限公司 Vehicle identification recognition method and system
CN105678304A (en) * 2015-12-30 2016-06-15 浙江宇视科技有限公司 Vehicle-logo identification method and apparatus
CN105740886A (en) * 2016-01-25 2016-07-06 宁波熵联信息技术有限公司 Machine learning based vehicle logo identification method
CN106250812A (en) * 2016-07-15 2016-12-21 汤平 A kind of model recognizing method based on quick R CNN deep neural network
CN106405640A (en) * 2016-08-26 2017-02-15 中国矿业大学(北京) Automatic microseismic signal arrival time picking method based on depth belief neural network

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103077407A (en) * 2013-01-21 2013-05-01 信帧电子技术（北京）有限公司 Car logo positioning and recognition method and car logo positioning and recognition system
ES2492315A1 (en) * 2013-03-05 2014-09-08 Universidad De Alcalá Procedure for the recognition of vehicle brands by classification of the logo and device for its realization (Machine-translation by Google Translate, not legally binding)
CN103279738A (en) * 2013-05-09 2013-09-04 上海交通大学 Automatic identification method and system for vehicle logo
CN104832418A (en) * 2015-05-07 2015-08-12 北京航空航天大学 Hydraulic pump fault diagnosis method based on local mean conversion and Softmax
CN105160292A (en) * 2015-07-07 2015-12-16 深圳市捷顺科技实业股份有限公司 Vehicle identification recognition method and system
CN105678304A (en) * 2015-12-30 2016-06-15 浙江宇视科技有限公司 Vehicle-logo identification method and apparatus
CN105740886A (en) * 2016-01-25 2016-07-06 宁波熵联信息技术有限公司 Machine learning based vehicle logo identification method
CN106250812A (en) * 2016-07-15 2016-12-21 汤平 A kind of model recognizing method based on quick R CNN deep neural network
CN106405640A (en) * 2016-08-26 2017-02-15 中国矿业大学(北京) Automatic microseismic signal arrival time picking method based on depth belief neural network

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
杨飚 等: "基于HOG和ASIFT特征的车标二次识别", 《计算机仿真》 *
毛颖: "大空间建筑火源的精确定位方法", 《计算机应用与软件》 *

Similar Documents

Publication Publication Date Title
CN109753914B (en) License plate character recognition method based on deep learning
Pan et al. A robust system to detect and localize texts in natural scene images
CN107273896A (en) A kind of car plate detection recognition methods based on image recognition
CN106650553A (en) License plate recognition method and system
US20180114337A1 (en) Method and system of detecting and recognizing a vehicle logo based on selective search
CN102663348B (en) Marine ship detection method in optical remote sensing image
CN105160330B (en) A kind of automobile logo identification method and vehicle-logo recognition system
CN107103317A (en) Fuzzy license plate image recognition algorithm based on image co-registration and blind deconvolution
CN107122776A (en) A kind of road traffic sign detection and recognition methods based on convolutional neural networks
CN107480620B (en) Remote sensing image automatic target identification method based on heterogeneous feature fusion
CN110298376B (en) Bank bill image classification method based on improved B-CNN
CN109101924A (en) A kind of pavement marking recognition methods based on machine learning
CN104680127A (en) Gesture identification method and gesture identification system
CN105574063A (en) Image retrieval method based on visual saliency
CN106529532A (en) License plate identification system based on integral feature channels and gray projection
CN107122777A (en) A kind of vehicle analysis system and analysis method based on video file
CN105740886B (en) A kind of automobile logo identification method based on machine learning
Ma et al. Text detection in natural images based on multi-scale edge detetion and classification
CN104182763A (en) Plant type identification system based on flower characteristics
CN110659374A (en) Method for searching images by images based on neural network extraction of vehicle characteristic values and attributes
CN105787475A (en) Traffic sign detection and identification method under complex environment
CN112200186A (en) Car logo identification method based on improved YOLO _ V3 model
CN110633727A (en) Deep neural network ship target fine-grained identification method based on selective search
CN111507344A (en) Method and device for recognizing characters from image
Li et al. The research on traffic sign recognition based on deep learning

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: 20170815

RJ01 Rejection of invention patent application after publication