CN106650755A - Feature extraction method based on color feature - Google Patents
Feature extraction method based on color feature Download PDFInfo
- Publication number
- CN106650755A CN106650755A CN201611221227.4A CN201611221227A CN106650755A CN 106650755 A CN106650755 A CN 106650755A CN 201611221227 A CN201611221227 A CN 201611221227A CN 106650755 A CN106650755 A CN 106650755A
- Authority
- CN
- China
- Prior art keywords
- image
- color
- feature
- target
- coarse positioning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
-
- 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
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
Abstract
The invention belongs to a feature extraction method based on color features. According to the method, a rough positioning process comprises the steps of carrying out downsampling on a collected image, wherein the sample times is four times; converting the image from an RGB color space to an LAB space and removing an L channel; establishing a color feature base, collecting and storing target color information and carrying out clustering analysis on colors in the image through a K-means algorithm; and matching a target image with the color feature base, determining an optimal matching position, realizing rough positioning and the like. The method provided by the invention comprises an improved color image feature matching method which combines image rough positioning and precise positioning. According to a pyramid downsampling method in the rough positioning, the resolution of the image is reduced by 4 times, while the color information in the image is reserved, and the processing rate of the color image is greatly improved; the previous defect that the processing speed of the color image is improved; and the processing timeliness of the color image is realized.
Description
Technical field
The invention belongs to a kind of feature extracting method of image processing field.
Background technology
Feature extraction is first had in the committed step in image processing process, and graphical analyses and image recognition processes
Problems faced.Conventional characteristics of image has color characteristic, textural characteristics, shape facility, spatial relation characteristics.Color is used as coloured silk
One key character of color image, has vital status in computer vision and area of pattern recognition.With cromogram
As the increase of the reduction, the continuous improvement of computer process ability and coloured image application of acquisition cost, color is used as feature
Become the focus of image processing field.Feature extracting method in conventional gray level image is quite ripe, and for example, yardstick is not
Become feature descriptor (Scale Invariant Feature Descriptor, SIFT) algorithm and be widely used in characteristic point
Extract, which has rotation and scale invariability, but generally only more sensitive to strong texture and constant high brightness two dimensional image, for
Brightness flop and weak texture object can substantially reduce effect, and real-time is not high.
As coloured image is there is provided the information than gray level image more horn of plenty, therefore can be entered by the use of color as feature
Row global registration realizes coarse positioning, then arranges search window essence search is carried out to target and realize fine positioning so that ensureing accurate
Property while improve real-time, the present invention will have broader practice prospect in image characteristics extraction field.
The content of the invention
The invention aims to overcome traditional feature extraction that shape, angle point etc. are carried out on gray-scale maps, and carry
Take effect not accurate enough, it is impossible to the problems such as meeting requirement of real-time, there is provided a kind of utilization has more rich information than gray-scale maps
Coloured image, carries out coarse positioning using color as global characteristics, then fine positioning is carried out to local feature region, realizes high accuracy, real
When the feature extracting method based on color characteristic.
The object of the present invention is achieved like this：
(1) coarse positioning process is comprised the steps of：
(1.1) image to collecting carry out it is down-sampled, sampling multiple be 4 times；
(1.2) image is transformed into into LAB spaces by RGB color, and removes L * channel；
(1.3) color characteristic storehouse is set up, target color information is gathered and is preserved, K-means is utilized to the color in image
Algorithm carries out cluster analyses；
(1.4) target image is matched with color characteristic storehouse, determines best match position, realizes coarse positioning；
(2) it is accurately positioned process to comprise the steps of：
(2.1) area-of-interest of the target being partitioned into during obtaining coarse positioning；
(2.2) in the window for clustering the centrally disposed 10*10 of color of object of above-mentioned zone, carry out smart search；
(2.3) characteristic point is extracted by the search of search window；
(2.4) characteristic point extracted is matched, is realized fine positioning.
The beneficial effects of the present invention is：The method of the invention includes that image coarse positioning is mutually tied by one kind with being accurately positioned
The improvement coloured image feature matching method of conjunction.The down-sampled method of pyramid in coarse positioning causes the resolution of image to reduce
4 times, but the colouring information in image all also retains, and Color Image Processing speed is but greatly improved, and improves conventional coloured image
The slow shortcoming of processing speed, realizes the real-time of Color Image Processing；Adopt in coarse positioning by image by RGB color
LAB spaces are transformed into, and remove the method for L * channel and reduce dependence of the coloured image to brightness, compared gray level image and process normal
Because the change of brightness affects treatment effect to have more stability；The 10*10 search windows arranged during fine positioning are also improved
The precision of Search Results.The characteristics of the method for the invention has real-time, good stability, high precision, is a kind of new
Effective color image processing method.
Description of the drawings
Fig. 1 is the method for the invention flow chart.
Fig. 2 is the basic framework figure of rough localization method.
Fig. 3 is the basic framework figure of accurate positioning method.
Specific embodiment
The present invention is further described with reference to the accompanying drawings and examples.
The technical solution adopted for the present invention to solve the technical problems is：The simple and clear introduction of method flow diagram shown in Fig. 1
Whole characteristic extraction procedure：Camera captures image with the frame rate collection image of 10fps in the video of real-time Transmission,
Color characteristic is extracted on image carries out color matching, and this process is coarse positioning；Coarse positioning process is substantially partitioned into target place
Area-of-interest, then in the region target characteristic point is extracted, the characteristic point extracted is matched with source images, this mistake
Journey is to be accurately positioned.Coarse positioning is combined with fine positioning and realizes the main-process stream of whole feature extraction.
Wherein, coarse positioning process refinement is comprised the steps of, as shown in Figure 2：
Step one：Image to collecting carry out it is down-sampled, sampling multiple be 4 times.
Step 2：Image is transformed into into LAB spaces by RGB color, and removes L * channel, brightness is reduced to color
The impact of feature.
Step 3：Color characteristic storehouse is set up, target color information is gathered and is preserved, K- is utilized to the color in image
Means algorithms carry out cluster analyses.
Step 4：Target image is matched with color characteristic storehouse, determines best match position, realizes coarse positioning.
It is accurately positioned process refinement to comprise the steps of, as shown in Figure 3：
Step one：The area-of-interest of the target being partitioned into during obtaining coarse positioning.
Step 2：In the window of the centrally disposed 10*10 of cluster color of object of above-mentioned zone, smart search is carried out.
Step 3：Characteristic point is extracted by the search of search window.
Step 4：The characteristic point extracted is matched, fine positioning is realized.
Camera captures image with the frame rate collection image of 10fps in the video of real-time Transmission.Image is divided into source images
And target image, image of the source images as collection color characteristic；Target image is matched with source images for real time imaging.
Color characteristic is extracted on image carries out color matching, and this process becomes coarse positioning；Coarse positioning process is substantially partitioned into target place
Area-of-interest, then in the region target characteristic point is extracted, the characteristic point extracted is matched with source images, this
Process is to be accurately positioned.Coarse positioning is combined with fine positioning and realizes the main-process stream of whole feature extraction.
Wherein, coarse positioning process refinement is comprised the steps of, as shown in Figure 2：
Step one：The image resolution ratio of collection is 1116*1956, in order to meet the requirement of real-time output, image is carried out
Pyramid is down-sampled, and sampling multiple is 4 times.Image resolution ratio after down-sampled is changed into the color in 279*489, but image
Information does not change, and the speed of image procossing is but improved.
Step 2：Image is transformed into into Lab space by RGB color, Lab color spaces selection three is mutually perpendicular
Coordinate axess, the existing lightness of L axis bodies are in vain in top, black in bottom；Colored characteristic is expressed together with b axles by a axles, red
Change is embodied with a axles positive direction, and the change of green is then that negative direction embodies；The change of yellow is embodied with b axles positive direction, blue
Change negative direction.Remove L * channel, only retain the passage of two expression colouring informations of a, b, brightness is reduced to color characteristic
Affect.
Step 3：Color characteristic storehouse is set up, target color information is extracted repeatedly in source images and is preserved, waited and target
Image is matched.As color of object there are k kinds, that is, k kind color templates are extracted, while K- is utilized to the color template in image
Means algorithms carry out cluster analyses.Should be noted that during extraction and color of object is extracted into complete, it is to avoid because the disappearance of colouring information causes
Matching error.
Step 4：Target image is matched with the color characteristic storehouse of source images, is determined best match position, and is gathered
Class center.As color of object there are k kinds, that is, k+1 cluster centre is obtained, coarse positioning is realized.
It is accurately positioned process refinement to comprise the steps of, as shown in Figure 3：
Step one：The area-of-interest and cluster centre of the target being partitioned into during obtaining coarse positioning.
Step 2：In the window of the centrally disposed 10*10 of cluster color of object of above-mentioned zone, smart search is carried out.
Step 3：Characteristic point is extracted by the search of search window, characteristic point includes angle point, central point, focus point etc..
Step 4：The characteristic point extracted is matched with the characteristic point in source images, fine positioning is realized.
Claims (1)
1. the feature extracting method based on color characteristic, it is characterised in that comprise the steps：
(1) coarse positioning process is comprised the steps of：
(1.1) image to collecting carry out it is down-sampled, sampling multiple be 4 times；
(1.2) image is transformed into into LAB spaces by RGB color, and removes L * channel；
(1.3) color characteristic storehouse is set up, target color information is gathered and is preserved, K-means algorithms are utilized to the color in image
Carry out cluster analyses；
(1.4) target image is matched with color characteristic storehouse, determines best match position, realizes coarse positioning；
(2) it is accurately positioned process to comprise the steps of：
(2.1) area-of-interest of the target being partitioned into during obtaining coarse positioning；
(2.2) in the window for clustering the centrally disposed 10*10 of color of object of above-mentioned zone, carry out smart search；
(2.3) characteristic point is extracted by the search of search window；
(2.4) characteristic point extracted is matched, is realized fine positioning.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611221227.4A CN106650755A (en) | 2016-12-26 | 2016-12-26 | Feature extraction method based on color feature |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611221227.4A CN106650755A (en) | 2016-12-26 | 2016-12-26 | Feature extraction method based on color feature |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106650755A true CN106650755A (en) | 2017-05-10 |
Family
ID=58831370
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611221227.4A Pending CN106650755A (en) | 2016-12-26 | 2016-12-26 | Feature extraction method based on color feature |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106650755A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110852319A (en) * | 2019-11-08 | 2020-02-28 | 深圳市深视创新科技有限公司 | Rapid universal roi matching method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1570969A (en) * | 2003-07-23 | 2005-01-26 | 西北工业大学 | An image retrieval method based on marked interest point |
| CN101105836A (en) * | 2007-07-06 | 2008-01-16 | 哈尔滨工程大学 | Infrared image real-time target identification and tracking device for movement background, and the method |
| CN102609723A (en) * | 2012-02-08 | 2012-07-25 | 清华大学 | Image classification based method and device for automatically segmenting videos |
| WO2013164043A1 (en) * | 2012-05-03 | 2013-11-07 | Thomson Licensing | Method and system for determining a color mapping model able to transform colors of a first view into colors of at least one second view |
| CN105139017A (en) * | 2015-08-27 | 2015-12-09 | 重庆理工大学 | License plate positioning algorithm fusing affine invariant corner feature and visual color feature |
-
2016
- 2016-12-26 CN CN201611221227.4A patent/CN106650755A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1570969A (en) * | 2003-07-23 | 2005-01-26 | 西北工业大学 | An image retrieval method based on marked interest point |
| CN101105836A (en) * | 2007-07-06 | 2008-01-16 | 哈尔滨工程大学 | Infrared image real-time target identification and tracking device for movement background, and the method |
| CN102609723A (en) * | 2012-02-08 | 2012-07-25 | 清华大学 | Image classification based method and device for automatically segmenting videos |
| WO2013164043A1 (en) * | 2012-05-03 | 2013-11-07 | Thomson Licensing | Method and system for determining a color mapping model able to transform colors of a first view into colors of at least one second view |
| CN105139017A (en) * | 2015-08-27 | 2015-12-09 | 重庆理工大学 | License plate positioning algorithm fusing affine invariant corner feature and visual color feature |
Non-Patent Citations (2)
| Title |
|---|
| 王招娣: ""基于SURF和颜色特征的图像匹配算法研究"", 《万方学位论文》 * |
| 贲志伟等: ""基于改进的K均值聚类算法提取彩色图像有意义区域"", 《计算机应用与软件》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110852319A (en) * | 2019-11-08 | 2020-02-28 | 深圳市深视创新科技有限公司 | Rapid universal roi matching method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102704215B (en) | Automatic cutting method of embroidery cloth based on combination of DST file parsing and machine vision | |
| CN104951940B (en) | A kind of mobile payment verification method based on personal recognition | |
| CN109344701A (en) | A kind of dynamic gesture identification method based on Kinect | |
| CN104167003A (en) | Method for fast registering remote-sensing image | |
| CN104778470B (en) | Text detection based on component tree and Hough forest and recognition methods | |
| CN106203420B (en) | A kind of bayonet vehicle color identification method | |
| CN109409208A (en) | A kind of vehicle characteristics extraction and matching process based on video | |
| CN104598907B (en) | Lteral data extracting method in a kind of image based on stroke width figure | |
| CN103473551A (en) | Station logo recognition method and system based on SIFT operators | |
| CN102194108A (en) | Smiley face expression recognition method based on clustering linear discriminant analysis of feature selection | |
| CN110032932B (en) | Human body posture identification method based on video processing and decision tree set threshold | |
| CN106874884A (en) | Human body recognition methods again based on position segmentation | |
| CN112906550B (en) | Static gesture recognition method based on watershed transformation | |
| CN107545550B (en) | Cell image color cast correction method | |
| CN104463138A (en) | Text positioning method and system based on visual structure attribute | |
| CN102147867A (en) | Method for identifying traditional Chinese painting images and calligraphy images based on subject | |
| CN107203757B (en) | Building extracting method based on binary features classifier | |
| CN108073940A (en) | A kind of method of 3D object instance object detections in unstructured moving grids | |
| CN106780428A (en) | A kind of number of chips detection method and system based on colour recognition | |
| CN104268509B (en) | The method and system of dump truck car plate detection | |
| CN108022245A (en) | Photovoltaic panel template automatic generation method based on upper thread primitive correlation model | |
| CN107145888A (en) | Video caption real time translating method | |
| CN106650755A (en) | Feature extraction method based on color feature | |
| CN113538585A (en) | High-precision multi-target intelligent identification, positioning and tracking method and system based on unmanned aerial vehicle | |
| CN108596262A (en) | A kind of method and system carrying out plant disease spot classification based on computer vision |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170510 |