CN106446850A - Station logo recognition method and device - Google Patents
Station logo recognition method and device Download PDFInfo
- Publication number
- CN106446850A CN106446850A CN201610870776.8A CN201610870776A CN106446850A CN 106446850 A CN106446850 A CN 106446850A CN 201610870776 A CN201610870776 A CN 201610870776A CN 106446850 A CN106446850 A CN 106446850A
- Authority
- CN
- China
- Prior art keywords
- station symbol
- template
- image
- unit
- station
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/435—Processing of additional data, e.g. decrypting of additional data, reconstructing software from modules extracted from the transport stream
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Signal Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a station logo recognition method and device. Station logo recognition is carried out by establishing a template image and carrying out rough matching, refined matching and rough matching and refined matching fusion. The method and device can reduce identification time, reduce identification complexity and greatly improves identification accuracy and speed.
Description
Technical field
The present invention relates to image procossing, human-computer interaction technique field, specifically TV station's station symbol or program LOGO are known
Method for distinguishing and device.
Background technology
In the last few years, the programme variety of TV station gets more and more, program category more and more extensively and TV station quantity
Also increase is very rapid.Therefore, by substantial amounts of TV programme carry out according to TV station according to classification of type, cataloguing be one urgently
Problem to be solved.The method exactly combines the image processing techniques quickly growing now and Video Analysis Technology is carried out to station symbol
Identification, good solution problem above.Existing TV station symbol recognition method is a lot, the TV station symbol recognition such as based on wavelet character, base
In the Hu not TV station symbol recognition method of bending moment and the TV station symbol recognition method based on neutral net, but the accuracy of identification and speed
All have much room for improvement.
HOG is the validity feature of description contour of object.It is considered that different station symbols has different profiles, therefore use
HOG feature and histogram intersection core are carefully mated.The algorithm of HOG is first image to be carried out color normalization, then will be whole
Width image is divided into much individual units, calculates the gradient of each pixel in each unit.Afterwards gradient direction is divided into several areas
Between, voted in each direction as weights with gradient magnitude.Eventually form characteristic vector it is simply that the gradient of all units
The vector that histogram is constituted.
Content of the invention
It is an object of the invention to overcoming above-mentioned deficiency, one kind is provided quick and precisely to identify platform calibration method.
The object of the present invention is achieved like this：A kind of station identification method for distinguishing, the method comprising the steps of：
（1）Obtain image,
（2）Set up template image；
（3）The cross-correlation coefficient of calculation template image and test image is slightly mated；
（4）The HOG characteristic value of image and the test image of seeking template calculates histogram intersection distance and is carefully mated；
（5）Described thick coupling and described thin coupling are carried out linear fusion.
Wherein, described step（2）Specifically include following steps：
（101）It is poor that the random frame that every kind of station symbol is chosen 20 to do frame；
（102）Use half threshold method, by step（101）The result of middle frame difference constructs black background template and white background template；
Further, described step（3）Specifically include following steps：
（201）Intercept the station symbol in the video upper left corner, and the station symbol of intercepting is converted into gray-scale map.
（202）Calculate（201）Middle station symbol and be converted into average between the black background template of gray-scale map and white background template
Distance, formula is as follows：
（203）Average distance is compared with threshold value, removes the part more than threshold value, image after being processed, computing formula is as follows：
（204）Solution procedure（203）In the image that obtains and black background template cross-correlation coefficient, solution formula is as follows：
When cross-correlation coefficient is larger, then test image and template are alike, complete slightly to mate, use threshold valueTo determine candidate's mould
The space of plate, threshold valueComputing formula be：
Wherein N is station symbol template sum, and M is candidate marker number.
Further, described step（4）Specifically include following steps：
（301）Using Sobel operator extraction test image marginal information；
（302）Test image station symbol part is divided into n region, described region is referred to as unit, the HOG calculating described unit is special
Levy, calculate gradient and the formula in direction is：
（303）The histogram of unit calculating in test image intersects distance, meter with the histogrammic of the unit in candidate template
Calculating formula is：
（304）By step（303）The histogram intersection distance obtaining calculates the histogram average similarity of n unit, compares survey
Attempt the similarity of picture and template, computing formula is：
Further, described step（5）It is specially：
The similarity of thick coupling and thin coupling is merged, is expressed as with a linear function：
The template station symbol making linear function shown in above formula maximum is recognition result.
And a kind of TV station symbol recognition device, described device includes：
Image acquisition unit, for obtaining the station symbol region of station symbol to be identified；
Station symbol template sets up unit, by half threshold method, builds the template of every kind of station symbol；
The thick matching unit of station symbol, is slightly mated by the cross-correlation coefficient of calculation template image and test image；
The thin matching unit of station symbol, averagely similar to histogram by HOG characteristic value and histogram intersection core histogram intersection distance
Property is carefully mated；
Thickness mates linear integrated unit, selects best matching result for both similarities are carried out fusion.
It is an advantage of the current invention that：Invention introduces normalized-cross-correlation function calculates and HOG characteristic value coupling meter
Calculate, mate the complexity of the time decreasing identification by thick and thin two-layer, the degree of accuracy and the speed of identification are greatly improved.
Brief description
Fig. 1 is TV station symbol recognition flow chart of the present invention；
Fig. 2 is that the station symbol template of the present invention sets up unit flow chart；
Fig. 3 be the present invention TV station symbol recognition during thick matching unit flow chart；
Fig. 4 be the present invention TV station symbol recognition during thin matching unit flow chart；
The black background template of Tu5Shi Guangdong satellite TV and white background template；
The image interception figure of Tu6Shi Guangdong satellite TV station symbol.
Specific embodiment
The present invention is Guangdong satellite TV station identification method for distinguishing, and the present invention is broadly divided into four parts, referring to the drawings 1, distinguishes
It is to set up template image, thick coupling, thin coupling and thickness coupling to merge.During setting up template image, referring to the drawings 2,
Method using inter-frame difference；During thick coupling, referring to the drawings 3, the cross-correlation of calculation template image and test image
Coefficient；During thin coupling, referring to the drawings 4, the HOG characteristic value of seek template image and test image, and calculate them
Histogram intersection distance and average similarity.Finally thickness coupling is carried out linear fusion, specifically：
（1）Set up template image
By the station symbol of Guangdong satellite TV, select the random frame of 20 about poor to do frame.Use half threshold method, the image that frame difference is obtained
It is configured to black background template and white background template, referring to the drawings 5.
（2）Thick coupling
In frame of video, the upper left corner is usually the position of station symbol, by the image interception of this position, referring to the drawings 6, and it is converted into gray scale
Figure.
Calculate station symbol and the average distance being converted between the black background template of gray-scale map and white background template first, public
Formula is as follows：
Then average distance is compared with threshold value, remove the part more than threshold value, image after being processed, computing formula is as follows：
Finally, using equation below, seek the cross-correlation coefficient of the image that above formula obtains and black background template, solution formula is as follows：
When cross-correlation coefficient is larger, then test image and template are alike, complete slightly to mate, use threshold valueTo determine candidate's mould
The space of plate, threshold valueComputing formula be：
Wherein N is station symbol template sum, and M is candidate marker number.
（3）Thin coupling
First, using Sobel operator extraction test image marginal information.Secondly, the pixel of test image station symbol part is divided into n
Individual region, referred to as unit.For each unit, calculate HOG characteristic value.The size of gradient is used as carrying out on histogram
The weight of ballot.
Following formula is the formula calculating gradient and direction.
Then, calculate the histogram in region and the unit in candidate template in test image histogrammic intersect away from
From such as following formula.
Finally, the histogram intersection distance being obtained using above formula calculates the histogram average similarity of n unit（As follows
Formula）, compare the similarity of test image and template.
（4）Thickness merges
After thick coupling and thin coupling, both similarities are carried out linear fusion, is represented with a linear function（As follows
Formula）
Choose and the maximum Schaltisch of linear function shown in above formula is denoted as recognition result.
Further, it is also possible to apply the invention to following two aspect：
（1）Program classification based on station symbol：If TV station will carry out the automated cataloging of TV programme, by different TV programme
Classified according to different TV stations with material, then TV station symbol recognition just can be applied in categorizing system.By to each
A certain frame in program is extracted, then inputs in TV station symbol recognition device it is possible to export TV station information, by TV programme
Automatically classify so that TV station staff is more convenient for by computer retrieval TV program information with material.
（2）TV interaction based on station symbol：In field of virtual reality, if comprising virtual existing in the program of a certain TV station
Real scene, spectators need wearing spectacles to be watched, can embed the program of TV station symbol recognition in glasses.When glasses are to platform
After mark is identified, this virtual reality scenario information of Network Capture can be passed through, to increase the interactive experience of spectators；Wide
During accusing promotion, if TV station and advertising company need to spectators' automatic promotion advertisement, then to pass through station identification
Do not go out the audient crowd of TV programme.As this channel of CCTV-5 is sports cast it is possible to add motion in advertisement
The advertisement of brand is so that the advertisement that pushes has and more accurately located.
Finally it should be noted that：Obviously, examples detailed above is only intended to clearly illustrate the application example, and not right
The restriction of embodiment.For those of ordinary skill in the field, can also be made it on the basis of the above description
The change of its multi-form or variation.There is no need to be exhaustive to all of embodiment.And thus amplified out
Obvious change or change among protection domain still in the application type.
Claims (9)
1. a kind of TV station symbol recognition method is it is characterised in that the method comprising the steps of：
（1）Obtain image,
（2）Set up template image；
（3）The cross-correlation coefficient of calculation template image and test image is slightly mated；
（4）The HOG characteristic value of calculation template image and test image is carefully mated；
（5）Described thick coupling and described thin coupling are carried out linear fusion.
2. a kind of TV station symbol recognition method according to claim 1 is it is characterised in that described step（2）Specifically include following
Step：
（101）Every kind of station symbol 20 random frame of selection to be done frame poor；
（102）Use half threshold method, by step（101）The result of middle frame difference constructs black background template and white background template.
3. a kind of TV station symbol recognition method according to claim 1 is it is characterised in that described step（3）Specifically include following
Step：
（201）Intercept the image of test image upper left corner area, and truncated picture is converted into gray-scale map；
（202）Calculation procedure（201）Average distance between middle gray-scale map and black background template and white background template, formula is such as
Under：
（203）Average distance is compared with threshold value, removes the part more than threshold value, image after being processed, computing formula is as follows：
（204）Solution procedure（203）In the image that obtains and black background template cross-correlation coefficient, solution formula is as follows：
Then illustrate that when cross-correlation coefficient is larger test image and this station symbol template are alike, complete slightly to mate work；Use threshold valueTo determine the space of candidate template, threshold valueComputing formula be：
Wherein N is station symbol template sum, the conventional number that M goes out for candidate.
4. a kind of TV station symbol recognition method according to claim 1 is it is characterised in that described step（4）Specifically include following
Step：
（301）Using Sobel operator extraction test image marginal information；
（302）Test image station symbol part is divided into n region, described region is referred to as unit, the HOG calculating described unit is special
Levy, calculate gradient and the formula in direction is：
（303）The histogram of unit calculating in test image intersects distance, meter with the histogrammic of the unit in candidate template
Calculating formula is：
（304）By step（303）The histogram intersection distance obtaining calculates the histogram average similarity of n unit, compares survey
Attempt the similarity of picture and this candidate template, computing formula is：
.
5. a kind of TV station symbol recognition method according to claim 1 is it is characterised in that described step（5）It is specially：
The similarity of thick coupling and thin coupling is carried out linear fusion, is expressed as with a linear function：
The template station symbol making linear function shown in above formula maximum is recognition result.
6. a kind of TV station symbol recognition device is it is characterised in that described device includes：
Image acquisition unit, for obtaining the station symbol region of station symbol to be identified；
Station symbol template sets up unit, by half threshold method, builds the template of every kind of station symbol；
The thick matching unit of station symbol, is slightly mated by the cross-correlation coefficient of calculation template image and test image；
The thin matching unit of station symbol, averagely similar to histogram by HOG characteristic value and histogram intersection core histogram intersection distance
Property is carefully mated；
Thickness mates linear integrated unit, selects best matching result for both similarities are carried out fusion.
7. a kind of TV station symbol recognition device according to claim 6 is it is characterised in that described station symbol template is set up unit and included
Construction black background formwork module, construction white background formwork module and station symbol template area acquisition module.
8. a kind of TV station symbol recognition device according to claim 6 is it is characterised in that the thick matching unit of described station symbol includes platform
Mark region acquisition module, cross-correlation coefficient computing module and candidate template choose module.
9. a kind of TV station symbol recognition device according to claim 6 is it is characterised in that the thin matching unit of described station symbol includes
Sobel edge edge detection module, HOG characteristic value calculating module, histogram intersection distance calculation module and histogram average similarity
Computing module.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610870776.8A CN106446850B (en) | 2016-09-30 | 2016-09-30 | Station caption identification method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610870776.8A CN106446850B (en) | 2016-09-30 | 2016-09-30 | Station caption identification method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106446850A true CN106446850A (en) | 2017-02-22 |
| CN106446850B CN106446850B (en) | 2020-05-19 |
Family
ID=58171965
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610870776.8A Expired - Fee Related CN106446850B (en) | 2016-09-30 | 2016-09-30 | Station caption identification method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106446850B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109977859A (en) * | 2019-03-25 | 2019-07-05 | 腾讯科技（深圳）有限公司 | A kind of map logo method for distinguishing and relevant apparatus |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102289663A (en) * | 2011-07-29 | 2011-12-21 | 四川长虹电器股份有限公司 | Method for identifying station caption based on color and shape |
| CN103955713A (en) * | 2014-04-24 | 2014-07-30 | 海信集团有限公司 | Icon recognition method and device |
| CN104185069A (en) * | 2013-05-23 | 2014-12-03 | 北京下周科技有限公司 | Station icon identification method and identification system |
| MX2014009067A (en) * | 2013-11-06 | 2015-08-27 | Xiaomi Inc | Station caption recognition method, device, television and system. |
| CN105760842A (en) * | 2016-02-26 | 2016-07-13 | 北京大学 | Station caption identification method based on combination of edge and texture features |
-
2016
- 2016-09-30 CN CN201610870776.8A patent/CN106446850B/en not_active Expired - Fee Related
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102289663A (en) * | 2011-07-29 | 2011-12-21 | 四川长虹电器股份有限公司 | Method for identifying station caption based on color and shape |
| CN104185069A (en) * | 2013-05-23 | 2014-12-03 | 北京下周科技有限公司 | Station icon identification method and identification system |
| MX2014009067A (en) * | 2013-11-06 | 2015-08-27 | Xiaomi Inc | Station caption recognition method, device, television and system. |
| CN103955713A (en) * | 2014-04-24 | 2014-07-30 | 海信集团有限公司 | Icon recognition method and device |
| CN105760842A (en) * | 2016-02-26 | 2016-07-13 | 北京大学 | Station caption identification method based on combination of edge and texture features |
Non-Patent Citations (1)
| Title |
|---|
| 贾国利: "《基于语义的跨媒体搜索》", 《中国硕士学位论文全文数据库信息科技辑》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109977859A (en) * | 2019-03-25 | 2019-07-05 | 腾讯科技（深圳）有限公司 | A kind of map logo method for distinguishing and relevant apparatus |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106446850B (en) | 2020-05-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9495754B2 (en) | Person clothing feature extraction device, person search device, and processing method thereof | |
| CN105005766B (en) | A kind of body color recognition methods | |
| CN105160317B (en) | One kind being based on area dividing pedestrian gender identification method | |
| WO2018072233A1 (en) | Method and system for vehicle tag detection and recognition based on selective search algorithm | |
| CN104850850B (en) | A kind of binocular stereo vision image characteristic extracting method of combination shape and color | |
| US8340420B2 (en) | Method for recognizing objects in images | |
| WO2016066038A1 (en) | Image body extracting method and system | |
| CN108197587A (en) | A kind of method that multi-modal recognition of face is carried out by face depth prediction | |
| CN109740572B (en) | Human face living body detection method based on local color texture features | |
| CN105913082B (en) | Method and system for classifying targets in image | |
| CN108629319B (en) | Image detection method and system | |
| CN106897681A (en) | A kind of remote sensing images comparative analysis method and system | |
| CN109145708A (en) | A kind of people flow rate statistical method based on the fusion of RGB and D information | |
| CN108600865A (en) | A kind of video abstraction generating method based on super-pixel segmentation | |
| CN110232331B (en) | Online face clustering method and system | |
| WO2013075295A1 (en) | Clothing identification method and system for low-resolution video | |
| CN109583452A (en) | A kind of identification of individual method and system based on barefoot footprint | |
| CN104143077A (en) | Pedestrian target retrieving method and system based on images | |
| CN114463724A (en) | Lane extraction and recognition method based on machine vision | |
| CN107832667A (en) | A kind of face identification method based on deep learning | |
| CN106446850A (en) | Station logo recognition method and device | |
| CN109635679B (en) | Real-time target paper positioning and loop line identification method | |
| CN106066887A (en) | A kind of sequence of advertisements image quick-searching and the method for analysis | |
| CN112581495A (en) | Image processing method, device, equipment and storage medium | |
| CN105550706A (en) | Method of recognizing 2D image and 3D image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20200519 Termination date: 20200930 |