CN103679179A - High-speed local relevance calculation method and device - Google Patents
High-speed local relevance calculation method and device Download PDFInfo
- Publication number
- CN103679179A CN103679179A CN201210344228.3A CN201210344228A CN103679179A CN 103679179 A CN103679179 A CN 103679179A CN 201210344228 A CN201210344228 A CN 201210344228A CN 103679179 A CN103679179 A CN 103679179A
- Authority
- CN
- China
- Prior art keywords
- correlation
- overlay area
- coordinate
- data
- segmentation
- 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
Images
Abstract
The invention provides a high-speed local relevance calculation method and device. The method is used for sectionally extracting two frame images, obtaining a sectional relevance matrix and calculating an extreme value and a coordinate of the matrix. The method adapts to processing of an image obtained by moving of an acquisition object on an acquisition window or an image obtained by moving the acquisition window, such as splicing of images acquired by fingerprint scraping sensors. The extreme value and the coordinate extracted by adopting the scheme can be applied to navigation, track and speed tracking, image detection, relevance calculation and the like. The effects of conducting an identification process in accurate and diversified manners and better recovering a motion curve of the acquisition object can be achieved; moreover, the image detection process and the image splicing effect are more accurate, and the phenomena of remarkable image dislocation, image deformation, short images, image trailing, background image splicing and the like are corrected.
Description
Technical field
The present invention relates to field of image recognition, relate in particular to high speed local correlations computing method and device.
Background technology
Image Mosaics technology can solve due to image acquisition window visual angle and big or small restriction, can not produce the problem of very large picture.So-called Image Mosaics technology exactly by two or two above adjacent have partly overlapping image carry out seamless spliced, generate one large-scale and have a technology of the high-definition picture at wider visual angle.Image Mosaics technology is applied to aviation field the earliest, has now been widely used in the fields such as digital video, performance analysis, fingerprint reconstruct.
The precondition of Image Mosaics is that between adjacent image, some is logically identical, and must have certain intersection.The splicing most important of image is to select a robust and image registration methods efficiently.Image registration methods is exactly to determine the overlapping degree of adjacent image on width and height.This overlapping degree is called the correlativity of two two field pictures.
Existing image correlation computing method exist various defects: as calculated amount is very large, computing velocity is slow, precision is low, cost is high, cannot adapt to images with large data volume splicing.These a lot of links when for back-end processing all can cause serious impact, and effect is also had a greatly reduced quality.As: the correlation data calculating, during for image detection, discrimination power is low, insensitive; The shortcomings such as while being used for navigating, it is not accurate that direction detects, and the poor and response speed of recognition effect is slow.And have a greatly reduced quality in the cost performance of product.Therefore the method that needs an energy also to have higher price-performance ratio under comprehensive above-mentioned condition solves these problems.
Summary of the invention
For solving the problems of the technologies described above, this programme provides a kind of high speed local correlations computing method and device.Be applicable to two or two above adjacent and there is partly overlapping image correlation calculating, and image without spin.With the mode of ASIC realized at a high speed, low-cost, calculated amount is little and regulatable correlation calculations, makes the correlation data calculating can be applied to preferably the aspects such as Image Mosaics, image detection and navigation.
Refer to high speed local correlations computing method, it is characterized in that, comprise the steps:
S1, carries out segmentation to the selected overlay area under a kind of coverage condition of two two field pictures; Selected overlay area is divided into two or more at acquisition target on the main moving direction with respect to acquisition window;
S2, carries out correlation calculations to the data in segmentation, obtains the correlation data of every section of region under current coverage condition;
S3, generates correlation matrix, and the correlation data according to every section of region under all coverage conditions generates the segmentation correlation matrix of every section;
S4, obtains maximal value, minimum value and coordinate thereof in the segmentation correlation matrix of every section.
The matrix that described correlation matrix is mapped to for the displacement coordinate that the correlation data obtaining in step S2 under all coverage conditions is moved with two two field pictures, the number of described segmentation correlation matrix is identical with the hop count that selected overlay area is divided on the main moving direction with respect to acquisition window at acquisition target; Can also obtain overall relevance matrix, described overall relevance matrix is comprised of the correlation data that in whole selected overlay area, all data calculate; Described coordinate is rectangular coordinate system, the corresponding coordinate of each coverage condition, the coverage condition respective coordinates initial point that front and back two two field pictures overlap completely.
Described two two field pictures are respectively template frame and present frame, and described template frame and present frame can exchange.
The production method of the overlay area in described step S1 is on the basis overlapping completely at template frame and present frame two two field pictures, and fixed form frame is motionless, and present frame moves up and down and produces an overlay area within the scope of two field picture length and width; Move and take Pixel-level as least unit each time.
The transverse and longitudinal both direction that selected overlay area in described step S1 is chosen can be continuous or discontinuous, select region with behavior unit, can be divided into and be more than or equal to a region in the horizontal, select region with the unit of classifying as, can be divided into and be more than or equal to a region in the vertical;
The corresponding correlation data of each coverage condition of described two two field pictures, described correlation data computing method are for asking squared difference by the data of selected overlay area one to one in selected overlay area in two two field pictures, obtain squared difference value one to one, each squared difference value adds up and obtains the correlation data under current coverage condition.
When described maximal value or the minimum value value corresponding with the centre coordinate of correlation matrix equates, centre coordinate replaces corresponding extreme value coordinate; Correlation data when described centre coordinate is template frame and the covering of present frame non-displacement.
High speed local correlations calculation element, is characterized in that, comprises segmentation module, correlation calculations module, correlation matrix generation module, extreme value coordinate generation module;
Described segmentation module carries out segmentation to the selected overlay area under a kind of coverage condition of two two field pictures; Selected overlay area is divided into two or more at acquisition target on the main moving direction with respect to acquisition window;
Described correlation calculations module is connected with segmentation module, for the data of segmentation are carried out to correlation calculations, obtains the correlation data of every section of region under current coverage condition;
Described correlation matrix generation module is connected with correlation calculations module, for according to every section of region, the correlation data under all coverage conditions obtains the segmentation correlation matrix of every section;
Described extreme value coordinate generation module is connected with correlation matrix generation module, for obtaining maximal value, minimum value and the coordinate thereof of the segmentation correlation matrix of every section.
The matrix that described correlation matrix is mapped to for the displacement coordinate that the correlation data obtaining in correlation calculations module under all coverage conditions is moved with two two field pictures, the number of described segmentation correlation matrix is identical with the hop count that selected overlay area is divided on the main moving direction with respect to acquisition window at acquisition target; Can also obtain overall relevance matrix, described overall relevance matrix is comprised of the correlation data that in whole selected overlay area, all data calculate; Described coordinate is rectangular coordinate system, the corresponding coordinate of each coverage condition, the coverage condition respective coordinates initial point that front and back two two field pictures overlap completely.
Described two two field pictures are respectively template frame and present frame, and described template frame and present frame can exchange.
The production method of the overlay area in described segmentation module is on the basis overlapping completely at template frame and present frame two two field pictures, and fixed form frame is motionless, and present frame moves up and down and produces an overlay area within the scope of two field picture length and width; Move and take Pixel-level as least unit each time.
The transverse and longitudinal both direction that the selected overlay area of described segmentation module is chosen can be continuous or discontinuous, select region with behavior unit, can be divided into and be more than or equal to a region in the horizontal, select region with the unit of classifying as, can be divided into and be more than or equal to a region in the vertical;
The corresponding correlation data of each coverage condition of described two two field pictures, described correlation data computing method are for asking squared difference by the data of selected overlay area one to one in selected overlay area in two two field pictures, obtain squared difference value one to one, each squared difference value adds up and obtains the correlation data under current coverage condition.
When described maximal value or the minimum value value corresponding with the centre coordinate of correlation matrix equates, centre coordinate replaces corresponding extreme value coordinate; Correlation data when described centre coordinate is template frame and the covering of present frame non-displacement.
The present invention, to two two field picture stage extractions, obtains segmentation correlation matrix and calculates its extreme value and coordinate.Be adapted to gather the processing of the consecutive image that the thing mobile consecutive image obtaining or mobile collection window on acquisition window obtain, the splicing of the consecutive image collecting as fingerprint scraping sensor.The extreme value and the coordinate thereof that adopt this programme to extract, can apply to navigation, track and speed tracking, image detection, correlation calculations etc.Can reach identification precision, variation, recover preferably the movement locus of acquisition target; And making image detection more accurate, stitching image effect is more accurate, corrects and has the phenomenons such as obvious image offset, image deformation, short image, streaking, background picture mosaic.
Accompanying drawing explanation
Fig. 1 is process flow diagram of the present invention;
Fig. 2 is the module diagram of high speed local correlations calculation element;
Fig. 3 is that schematic diagram is chosen in overlay area; Wherein Fig. 3-1 is that two two field picture overlay area parts are chosen schematic diagram, and Fig. 3-2 are that limiting case schematic diagram is chosen in two two field picture overlay areas, and Fig. 3-3 are that the discontinuous schematic diagram of transverse and longitudinal two direction is chosen in two two field picture overlay areas;
Fig. 4 is correlation matrix piecemeal schematic diagram;
Fig. 5 is the corresponding coverage condition schematic diagram of correlation matrix;
Fig. 6 is the fingerprint image that scraping sensor gathers;
Fig. 7 is that the fingerprint image of Fig. 6 is through the spliced image of this programme;
Fig. 8 solves and leaks the fingerprint image contrast figure that spells phenomenon with this programme; Wherein Fig. 8-1 is this programme splicing result, and Fig. 8-2 are for having brachydactylia line and leaking the fingerprint image of spelling;
Fig. 9 is the fingerprint image contrast figure of many spellings phenomenon of case solution; Fig. 9-1 is for utilizing the fingerprint image of this programme splicing, and for there are the images of spelling phenomenon Fig. 9-2 more;
The fingerprint image contrast figure that Figure 10 is the splicing inconsistent phenomenon of solution; Wherein Figure 10-1 is this programme splicing result, and Figure 10-2 are for there being the fingerprint image of splicing inconsistent phenomenon.
Embodiment
This programme solves that existing correlation calculations method calculated amount is large, computing velocity is slow, precision is low, high in cost of production technical matters.
This programme is realized the relevant coordinate of stage extraction and extreme value by the mode of ASIC, the max min of two two field picture correlation matrixes before and after obtaining, and coordinate, several groups of extreme values extracting and coordinate thereof, realized at a high speed, low-cost, calculated amount is little and regulatable correlation calculations, makes the correlation data calculating can be applied to preferably the aspects such as Image Mosaics, image detection and navigation.
Below with reference to Fig. 1-Figure 10, the preferred embodiments of the present invention are described, specific embodiment described herein only, in order to explain the present invention, is not intended to limit the present invention.
The process flow diagram of high speed local correlations computing method is as shown in Figure 1:
S1, carries out segmentation to the selected overlay area under a kind of coverage condition of two two field pictures; Selected overlay area is divided into two or more at acquisition target on the main moving direction with respect to acquisition window;
Two two field pictures before and after extracting, in the present embodiment, the image of extraction is two two field pictures that acquisition target arrives with respect to acquisition window mobile collection.Two two field pictures are set as respectively to template frame T and present frame C, and template frame T and present frame C are any two field picture of extracted two two field pictures.Template frame T is that motionless frame of reference in above-mentioned two two field pictures, and present frame C is with reference to regular another two field picture moving up and down of template frame T in above-mentioned two two field pictures.
On the basis overlapping completely at template frame T and present frame C two two field pictures, T is motionless for fixed form frame, the regular present frame C that moves up and down within the scope of two field picture length and width, and two frames produce the relative displacement between an overlay area M and two frames.As shown in Figure 3, it is capable that template frame T and present frame C comprise respectively r, and overlay area is the represented region of M, and template frame T and present frame C displacement are in the horizontal direction w pixel, and displacement is h pixel in the vertical direction.
As shown in Fig. 3-1, the present embodiment is selected M1, two overlay areas of M2 in the M of overlay area, wherein selected two frames corresponding to overlay area M1 are the first template frame region T1 and the first present frame region C1, and the first template frame region T1 and the first present frame region C1 overlap completely; Selected two frames corresponding to overlay area M2 are the second template frame region T2 and the second present frame region C2, and the second template frame region T2 and the second present frame region C2 overlap completely.The effective length of selected overlay area M1 and selected overlay area M2 is len_w, the significant height of selected overlay area M1 is len_h1, selected overlay area M2 significant height is len_h2, and two significant height len_h1, len_h2 and effective width len_w are adjustable, be adjusted to limiting case as shown in Fig. 3-2, under limiting case, selected overlay area summation is whole overlay area M.
Two two field picture overlay area M choose transverse and longitudinal both direction can discontinuous (referring to Fig. 3-3), select region in the horizontal, can be divided into and be more than or equal to a region, as Fig. 3-2 Zhong Yi behavior unit has divided two regions with behavior unit.In like manner, select region in the vertical, with the unit of classifying as, can be divided into and be more than or equal to a region.
As shown in Fig. 4-2, selected overlay area M1 and selected overlay area M2 are divided into several sections at acquisition target on the main moving direction with respect to acquisition window, hop count is two or more, and each section of overlay area of dividing can be discontinuous on horizontal, vertical both direction, and the present embodiment will select overlay area M1 and selected overlay area M2 is divided into respectively tri-sections of ar1, ar2, ar3.
S2, carries out correlation calculations to the data in segmentation, obtains the correlation data of every section of region under current coverage condition;
Template frame T and present frame C two two field pictures have correlation data under every kind of coverage condition, respectively the data of selected overlay area are carried out to correlation operation under different coverage conditions, draw the correlation data under current coverage condition.
Choose the part or all of data of overlay area under current coverage condition, the template frame T choosing and present frame C two two field picture overlay area data are for corresponding one by one.As shown in Fig. 3-1, what the first template frame region T1 and the second template frame region T2 chose is data corresponding in template frame T, and what the first present frame region C1 and the second present frame region C2 chose is data corresponding in present frame C.The column region of peek is adjustable, also can adjust the line number that participates in correlation operation, while choosing data, can jump and select according to the row, column of frame image data, and that can also need to be interrupted according to difference fetches data, and each zone length not necessarily equates.
Data one to one in two two field pictures that take out are asked to squared difference, obtain squared difference value one to one, then each squared difference value is added up and obtains the correlation data under coverage condition.Referring to formula:
Wherein: m=len_h1+b
n=len_w+a
q=len_w+c
p=len_h2+d
In above-mentioned formula, the coordinate (a, b) of corresponding templates frame, (c, d), the coordinate of corresponding present frame is (a-w, b+h), (c-w, d+h), slice_a and slice_b are respectively template frame T and coordinate corresponding to present frame C.Cor is the correlation data under current a kind of coverage condition.The overlay area data that traversal slice_a and slice_b choose, correspondence is asked squared difference, then each squared difference is added up and obtains the correlation data cor under a kind of coverage condition.Above-mentioned correlation data, when generating correlation matrix, need be done normalized.
S3, generates correlation matrix, and the correlation data according to every section of region under all coverage conditions obtains the segmentation correlation matrix of every section;
Present frame C be take template frame T as with reference to moving up and down, each coverage condition correspondence calculates a correlation data according to step S2, it is segmentation correlation matrix that the correlation data of every section of selected overlay area that step S2 is obtained under all coverage conditions be take the matrix that displacement coordinate that two frames move is mapped to, and each coordinate of correlation matrix represents that two two field pictures are with the coverage condition of this coordinate offset.As shown in Figure 5, described coordinate is rectangular coordinate system, and the coverage condition that two two field pictures of take overlap is completely true origin, i.e. the displacement of the horizontal and vertical direction between two frames.
As only ar1 area data carried out respectively under each coverage condition to correlation operation in Fig. 4-2, obtain ar1 area data correlation matrix.In like manner obtain ar2, ar3 area data correlation matrix, three correlation matrixes of gained are referred to as segmentation correlation matrix, segmentation correlation matrix number is identical with the hop count that overlay area data are divided on the main moving direction with respect to acquisition window at acquisition target, as overlay area being divided in Fig. 4-2 to three sections, has three segmentation correlation matrixes.
The matrix that the correlation data that in whole selected overlay area, all data calculate forms is overall relevance matrix.
S4, obtains maximal value, minimum value and coordinate thereof in the segmentation correlation matrix of every section.
Each in each segmentation correlation matrix of repeating query is worth respectively, obtains maximal value, minimum value and the coordinate thereof of each segmentation correlation matrix.When the value of maximal value or minimum value (when template frame T with present frame C non-displacement the cover correlation data) corresponding with its centre coordinate equates, with centre coordinate, replace corresponding extreme value coordinate, off-centring is paid the utmost attention to.
The present invention also comprises high speed local correlations calculation element (referring to Fig. 2), comprising: segmentation module, correlation calculations module, correlation matrix generation module, extreme value coordinate generation module;
Segmentation module carries out segmentation to the selected overlay area under a kind of coverage condition of two two field pictures; Selected overlay area is divided into two or more at acquisition target on the main moving direction with respect to acquisition window.
Two two field pictures before and after segmentation module extracts, in the present embodiment, the image of extraction is two two field pictures that acquisition target arrives with respect to acquisition window mobile collection.Two two field pictures are set as respectively to template frame T and present frame C, and template frame T and present frame C are any two field picture of extracted two two field pictures.Template frame T is that motionless frame of reference in above-mentioned two two field pictures, and present frame C is with reference to regular another two field picture moving up and down of template frame T in above-mentioned two two field pictures.
On the basis overlapping completely at template frame T and present frame C two two field pictures, T is motionless for fixed form frame, the regular present frame C that moves up and down within the scope of two field picture length and width, and two frames produce the relative displacement between an overlay area M and two frames.As shown in Figure 3, it is capable that template frame T and present frame C comprise respectively r, and overlay area is the represented region of M, and template frame T and present frame C displacement are in the horizontal direction w pixel, and displacement is h pixel in the vertical direction.
As shown in Fig. 3-1, the present embodiment is selected M1, two overlay areas of M2 in the M of overlay area, wherein selected two frames corresponding to overlay area M1 are the first template frame region T1 and the first present frame region C1, and the first template frame region T1 and the first present frame region C1 overlap completely; Selected two frames corresponding to overlay area M2 are the second template frame region T2 and the second present frame region C2, and the second template frame region T2 and the second present frame region C2 overlap completely.The effective length of selected overlay area M1 and selected overlay area M2 is len_w, the significant height of selected overlay area M1 is len_h1, selected overlay area M2 significant height is len_h2, and two significant height len_h1, len_h2 and effective width len_w are adjustable, be adjusted to limiting case as shown in Fig. 3-2, under limiting case, selected overlay area summation is whole overlay area M.
Two two field picture overlay area M choose transverse and longitudinal both direction can discontinuous (referring to Fig. 3-3), select region in the horizontal, can be divided into and be more than or equal to a region, as Fig. 3-2 Zhong Yi behavior unit has divided two regions with behavior unit.In like manner, select region in the vertical, with the unit of classifying as, can be divided into and be more than or equal to a region.
As shown in Fig. 4-2, selected overlay area M1 and selected overlay area M2 are divided into several sections at acquisition target on the main moving direction with respect to acquisition window, hop count is two or more, and each section of overlay area of dividing can be discontinuous on horizontal, vertical both direction, and the present embodiment will select overlay area M1 and selected overlay area M2 is divided into respectively tri-sections of ar1, ar2, ar3.
Correlation calculations module is connected with segmentation module, and the data in segmentation are carried out to correlation calculations, obtains the correlation data of every section of region under current coverage condition;
Template frame T and present frame C two two field pictures have correlation data under every kind of coverage condition, respectively the data of selected overlay area are carried out under different coverage conditions correlation operation, draw the correlation data under current coverage condition.
Choose the part or all of data of overlay area under current coverage condition, the template frame T choosing and present frame C two two field picture overlay area data are for corresponding one by one.As shown in Fig. 3-1, what the first template frame region T1 and the second template frame region T2 chose is data corresponding in template frame T,
What the first present frame region C1 and the second present frame region C2 chose is data corresponding in present frame C.The column region of peek is adjustable, also can adjust the line number that participates in correlation operation, while choosing data, can jump and select according to the row, column of frame image data, and that can also need to be interrupted according to difference fetches data, and each zone length not necessarily equates.
Data one to one in two two field pictures that take out are asked to squared difference, obtain squared difference value one to one, then each squared difference value is added up and obtains the correlation data under coverage condition.Referring to formula:
Wherein: m=len_h1+b
n=len_w+a
q=len_w+c
p=len_h2+d
In above-mentioned formula, the coordinate (a, b) of corresponding templates frame, (c, d), the coordinate of corresponding present frame is (a-w, b+h), (c-w, d+h), lice_a and slice_b are respectively template frame T and coordinate corresponding to present frame C.Cor is the correlation data under current a kind of coverage condition.The overlay area data that traversal slice_a and slice_b choose, correspondence is asked squared difference, then each squared difference is added up and obtains the correlation data cor under a kind of coverage condition.
Correlation matrix generation module is connected with correlation calculations module, and for generating correlation matrix, the correlation data according to every section of region under all coverage conditions obtains the segmentation correlation matrix of every section;
Present frame C be take template frame T as with reference to moving up and down, the corresponding correlation calculations module of each coverage condition calculates a correlation data, it is segmentation correlation matrix that the correlation data of every section of selected overlay area that correlation calculations module is obtained under all coverage conditions be take the matrix that displacement coordinate that two frames move is mapped to, and each coordinate of correlation matrix represents that two two field pictures are with the coverage condition of this coordinate offset.As figure
Shown in 5, described coordinate is rectangular coordinate system, and the coverage condition that two two field pictures of take overlap is completely true origin, i.e. the displacement of the horizontal and vertical direction between two frames.
As only ar1 area data carried out respectively under each coverage condition to correlation operation in Fig. 4-2, obtain ar1 area data correlation matrix.In like manner obtain ar2, ar3 area data correlation matrix, three correlation matrixes of gained are referred to as segmentation correlation matrix, segmentation correlation matrix number is identical with the hop count that overlay area data are divided on the main moving direction with respect to acquisition window at acquisition target, as overlay area being divided in Fig. 4-2 to three sections, has three segmentation correlation matrixes.
By the correlation data that in whole selected overlay area, all data calculate, form overall relevance matrix.
Extreme value coordinate generation module is connected with correlation matrix generation module, for obtaining maximal value, minimum value and the coordinate thereof of the segmentation correlation matrix of every section.
Each in each segmentation correlation matrix of repeating query is worth respectively, obtains maximal value, minimum value and the coordinate thereof of each segmentation correlation matrix.When the value of maximal value or minimum value (when template frame T with present frame C non-displacement the cover correlation data) corresponding with its centre coordinate equates, with centre coordinate, replace corresponding extreme value coordinate, off-centring is paid the utmost attention to.
Several groups of extreme values and the coordinate thereof of said extracted can apply to navigation, motion track and speed tracking, image detection, two field picture correlation operation etc.While making to obtain correlation data navigation, can reach gesture identification precision, variation, recover preferably the movement locus of acquisition target.And making image detection more accurate, stitching image effect is more accurate, corrects and has obvious image offset, image deformation, short image, streaking, background picture mosaic
Etc. phenomenon (referring to Fig. 6, Fig. 7, Fig. 8, Fig. 9, Figure 10).
Those skilled in the art is not departing under the condition of the definite the spirit and scope of the present invention of claims, can also carry out various modifications to above content.Therefore scope of the present invention is not limited in above explanation, but determined by the scope of claims.
Claims (10)
1. high speed local correlations computing method, is characterized in that, comprise the steps:
S1, carries out segmentation to the selected overlay area under a kind of coverage condition of two two field pictures; Selected overlay area is divided into two or more at acquisition target on the main moving direction with respect to acquisition window;
S2, carries out correlation calculations to the data in segmentation, obtains the correlation data of every section of region under current coverage condition;
S3, generates correlation matrix, and the correlation data according to every section of region under all coverage conditions obtains the segmentation correlation matrix of every section;
S4, obtains maximal value, minimum value and coordinate thereof in the segmentation correlation matrix of every section.
2. high speed local correlations computing method according to claim 1, it is characterized in that, the matrix that described correlation matrix is mapped to for the displacement coordinate that the correlation data obtaining in step S2 under all coverage conditions is moved with two two field pictures, the number of described segmentation correlation matrix is identical with the hop count that selected overlay area is divided on the main moving direction with respect to acquisition window at acquisition target; Can also obtain overall relevance matrix, described overall relevance matrix is comprised of the correlation data that in whole selected overlay area, all data calculate; The coordinate of described displacement coordinate is rectangular coordinate system.
3. high speed local correlations computing method according to claim 1, is characterized in that, described two two field pictures are respectively template frame and present frame, and described template frame and present frame can exchange.
4. high speed local correlations computing method according to claim 3, it is characterized in that, the production method of the overlay area in described step S1 is on the basis overlapping completely at template frame and present frame two two field pictures, fixed form frame is motionless, and present frame moves up and down and produces an overlay area within the scope of two field picture length and width;
The transverse and longitudinal both direction that selected overlay area in described step S1 is chosen can be continuous or discontinuous, select region with behavior unit, can be divided into and be more than or equal to a region in the horizontal, select region with the unit of classifying as, can be divided into and be more than or equal to a region in the vertical;
The corresponding correlation data of each coverage condition of described two two field pictures, described correlation calculations method is for asking squared difference by the data of selected overlay area one to one in selected overlay area in two two field pictures, obtain squared difference value one to one, each squared difference value adds up and obtains the correlation data under current coverage condition.
5. high speed local correlations computing method according to claim 1, is characterized in that, when described maximal value or the minimum value value corresponding with the centre coordinate of correlation matrix equates, centre coordinate replaces corresponding extreme value coordinate; Correlation data when described centre coordinate is template frame and the covering of present frame non-displacement.
6. high speed local correlations calculation element, is characterized in that, comprises segmentation module, correlation calculations module, correlation matrix generation module, extreme value coordinate generation module;
Described segmentation module carries out segmentation to the selected overlay area under a kind of coverage condition of two two field pictures; Selected overlay area is divided into two or more at acquisition target on the main moving direction with respect to acquisition window;
Described correlation calculations module is connected with segmentation module, for the data of segmentation are carried out to correlation calculations, obtains the correlation data of every section of region under current coverage condition;
Described correlation matrix generation module is connected with correlation calculations module, for according to every section of region, the correlation data under all coverage conditions obtains the segmentation correlation matrix of every section;
Described extreme value coordinate generation module is connected with correlation matrix generation module, for obtaining maximal value, minimum value and the coordinate thereof of the segmentation correlation matrix of every section.
7. high speed local correlations calculation element according to claim 6, it is characterized in that, the matrix that described correlation matrix is mapped to for the displacement coordinate that the correlation data obtaining in correlation calculations module under all coverage conditions is moved with two two field pictures, the number of described segmentation correlation matrix is identical with the hop count that selected overlay area is divided on the main moving direction with respect to acquisition window at acquisition target; Can also obtain overall relevance matrix, described overall relevance matrix is comprised of the correlation data that in whole selected overlay area, all data calculate; The coordinate of described displacement coordinate is rectangular coordinate system.
8. high speed local correlations calculation element according to claim 6, is characterized in that, described two two field pictures are respectively template frame and present frame, and described template frame and present frame can exchange.
9. high speed local correlations calculation element according to claim 8, it is characterized in that, the production method of the overlay area in described segmentation module is on the basis overlapping completely at template frame and present frame two two field pictures, fixed form frame is motionless, and present frame moves up and down and produces an overlay area within the scope of two field picture length and width;
The transverse and longitudinal both direction that the selected overlay area of described segmentation module is chosen can be continuous or discontinuous, select region with behavior unit, can be divided into and be more than or equal to a region in the horizontal, select region with the unit of classifying as, can be divided into and be more than or equal to a region in the vertical;
The corresponding correlation data of each coverage condition of described two two field pictures, described correlation calculations method is for asking squared difference by the data of selected overlay area one to one in selected overlay area in two two field pictures, obtain squared difference value one to one, each squared difference value adds up and obtains the correlation data under current coverage condition.
10. high speed local correlations calculation element according to claim 6, is characterized in that, when described maximal value or the minimum value value corresponding with the centre coordinate of correlation matrix equates, centre coordinate replaces corresponding extreme value coordinate; Correlation data when described centre coordinate is template frame and the covering of present frame non-displacement.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210344228.3A CN103679179A (en) | 2012-09-18 | 2012-09-18 | High-speed local relevance calculation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210344228.3A CN103679179A (en) | 2012-09-18 | 2012-09-18 | High-speed local relevance calculation method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103679179A true CN103679179A (en) | 2014-03-26 |
Family
ID=50316668
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210344228.3A Pending CN103679179A (en) | 2012-09-18 | 2012-09-18 | High-speed local relevance calculation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103679179A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5890808A (en) * | 1996-01-05 | 1999-04-06 | Mcdonnell Douglas Corporation | Image processing method and apparatus for correlating a test image with a template |
| CN101086766A (en) * | 2007-06-04 | 2007-12-12 | 北京航空航天大学 | Fingerprint image frame sequence combination method based on wave form match |
| CN101533470A (en) * | 2009-04-16 | 2009-09-16 | 成都方程式电子有限公司 | Method for scraping sensor image split joint and device thereof |
| CN102609937A (en) * | 2012-01-12 | 2012-07-25 | 中国人民解放军国防科学技术大学 | Hardware accelerator oriented image matching method |
-
2012
- 2012-09-18 CN CN201210344228.3A patent/CN103679179A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5890808A (en) * | 1996-01-05 | 1999-04-06 | Mcdonnell Douglas Corporation | Image processing method and apparatus for correlating a test image with a template |
| CN101086766A (en) * | 2007-06-04 | 2007-12-12 | 北京航空航天大学 | Fingerprint image frame sequence combination method based on wave form match |
| CN101533470A (en) * | 2009-04-16 | 2009-09-16 | 成都方程式电子有限公司 | Method for scraping sensor image split joint and device thereof |
| CN102609937A (en) * | 2012-01-12 | 2012-07-25 | 中国人民解放军国防科学技术大学 | Hardware accelerator oriented image matching method |
Non-Patent Citations (2)
| Title |
|---|
| 张强: ""图像匹配算法研究"", 《中国优秀硕士学位论文全文数据库（信息科技辑）》 * |
| 胡璟璟: ""复杂场景下目标跟踪的多模板匹配算法研究"", 《中国优秀硕士学位论文全文数据库（信息科技辑）》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103400129A (en) | Target tracking method based on frequency domain saliency | |
| CN107146239B (en) | Satellite video moving target detection method and system | |
| CN101137003B (en) | Gray associated analysis based sub-pixel fringe extracting method | |
| CN102426647B (en) | Station identification method and device | |
| CN103559237A (en) | Semi-automatic image annotation sample generating method based on target tracking | |
| EP1901228A1 (en) | Image processor | |
| CN103413276A (en) | Depth enhancing method based on texture distribution characteristics | |
| CN105447828B (en) | The one-view image deblurring method of one-dimensional deconvolution is carried out along motion blur path | |
| CN103955888A (en) | High-definition video image mosaic method and device based on SIFT | |
| CN103440664A (en) | Method, system and computing device for generating high-resolution depth map | |
| CN103440667A (en) | Automatic device for stably tracing moving targets under shielding states | |
| CN104200492A (en) | Automatic detecting and tracking method for aerial video target based on trajectory constraint | |
| CN104463873A (en) | Image target repositioning method based on local uniform scaling | |
| CN103914829B (en) | Method for detecting edge of noisy image | |
| CN101276416A (en) | Text tracking and multi-frame reinforcing method in video | |
| CN103226824A (en) | Video retargeting system for maintaining visual saliency | |
| CN105303566B (en) | A kind of SAR image azimuth of target method of estimation cut based on objective contour | |
| CN104050674B (en) | Salient region detection method and device | |
| CN111914596B (en) | Lane line detection method, device, system and storage medium | |
| CN101572820B (en) | Preprocessing method of video signal in detection process of moving target | |
| CN110322479B (en) | Dual-core KCF target tracking method based on space-time significance | |
| CN115830064B (en) | Weak and small target tracking method and device based on infrared pulse signals | |
| CN103679179A (en) | High-speed local relevance calculation method and device | |
| CN104484997B (en) | A kind of regional movement traffic flow detecting method based on video image technology | |
| CN111062968A (en) | Robot path skeleton extraction method based on edge scanning and centerline extraction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20140326 |
|
| WD01 | Invention patent application deemed withdrawn after publication |