CN106023181B - A kind of local line segment irrelevance feature matching method based on printed matter - Google Patents
A kind of local line segment irrelevance feature matching method based on printed matter Download PDFInfo
- Publication number
- CN106023181B CN106023181B CN201610322170.0A CN201610322170A CN106023181B CN 106023181 B CN106023181 B CN 106023181B CN 201610322170 A CN201610322170 A CN 201610322170A CN 106023181 B CN106023181 B CN 106023181B
- Authority
- CN
- China
- Prior art keywords
- distance
- straight line
- feature
- printed matter
- location point
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
Landscapes
- Image Analysis (AREA)
Abstract
The present invention discloses a kind of local line segment irrelevance feature matching method based on printed matter, the straightway irrelevance feature of printed matter is extracted first, then length being overlapped using minimum and symbol variance rate carrying out characteristic matching, the matching relationship of printed matter part is determined finally by the matching result of straightway.The present invention solves the matching problem between different length feature, and avoiding can not be matched and range error problem as caused by straightway location point inaccuracy.
Description
Technical field
The present invention relates to image processing techniques more particularly to a kind of local line segment irrelevance characteristic matchings based on printed matter
Method.
Background technique
The feature extraction of image and characteristic matching are always the emphasis and hot spot of field of image processing, are image registration, figure
As one of committed steps of related fieldss such as fusion.The feature of image mainly includes point feature, line feature and region feature.Image
With being broadly divided into based on feature and based on the two major classes of gray scale.
Currently, the feature extraction and matching algorithm based on SIFT is applied the most extensive in feature-based matching scheme.
It can preferably handle translation, rotation and dimensional variation, but for the printed matter part less for point feature, matching effect
And it is unsatisfactory.
Summary of the invention
The technical problem to be solved in the present invention is that for the defects in the prior art, providing a kind of office based on printed matter
Portion's line segment irrelevance feature matching method.
The technical solution adopted by the present invention to solve the technical problems is: a kind of local line segment irrelevance based on printed matter
Feature matching method, comprising the following steps:
1) straightway irrelevance feature is locally carried out to the identical image of reference printed matter and printed matter to be matched respectively to mention
It takes;Obtain multiple feature vectors with reference to printed matter image localWith printing to be matched
Multiple feature vectors of product image local
2) feature vector is chosen:
Step 2.1: distinguishing selected characteristic vector D from set R and set MiWith D 'j;
Step 2.2: determining feature vector DiWith D 'jMinimum be overlapped length LMin;
Since image passes through standardization, the scale of feature is identical.Known features vector DiWith D 'jLength point
It Wei not LiWith L 'jIf LiWith L 'jIt differs within 30%, then LMinComputation rule it is as follows:
LMin=0.9*min (Li,L′j),
If LiWith L 'j30% is differed by more than, return step 2.1 reselects feature vector;
3) feature vector D is calculatediWith D 'jThe distance of intersection;
If feature vector DiWith D 'jThe length of intersection is L, it is clear that L > Lmin;Calculate feature vector DiWith D 'jIt is overlapped
The different component number Num of symbol in respective components in part, then feature vector DiWith D 'jBetween distance be d (Di,D′j)=
Minimum value in Num/L;
4) matching judgment: if distance is less than set distance threshold value, then it is assumed that two line matchs, on the contrary it mismatches;
5) step 2) is repeated to the combination for 4), traversing whole feature vectors pair;
6) matching relationship between printed matter part is determined according to the matching result between line segment character pair;
The matching characteristic determined by step 5) is that element number is respectively in NumMatch, set R and set M to number
NumRAnd NumM, then with Ratio=NumMatch/ (NumR*NumM) size judge part whether match;If Ratio is greater than
0.02 local matching, on the contrary part mismatches.
According to the above scheme, the line segment irrelevance feature extracting method of print characteristics is directed in the step 1), including following
Step:
1) image procossing: acquiring the image of printed matter, converts standard picture for printed matter image;
2) edge detection: edge detection is carried out to image using Canny edge detection operator, obtains the marginal position of image
Point;
3) straight-line detection: straight-line detection is carried out to the marginal position point detected using Hough transform principle, is obtained a plurality of
Straight line;
4) location point is sorted out: distance and the projected position point that each marginal position point arrives straight line are calculated, it is nearest according to distance
Location point is assigned on each straight line by principle;
It is specific as follows: to set up two threshold values, the minimum distance d1 and maximum distance d2 of location point to straight line;When location point arrives
When the distance of arbitrary line is both greater than maximum distance d2, then the location point is not belonging to any straight line;When location point is to a plurality of
When the distance of straight line is both less than minimum distance d1, then the location point belongs to a plurality of straight line that distance is less than minimum distance d1;Remaining
Location point belongs to apart from nearest straight line;
5) projected position of location point on straight line is clustered, determines the straightway number on straight line;
6) it according to the projected position of location point on straight line on each straightway, is ranked up, determines according to rectilinear direction
The corresponding location point of two endpoints of straightway;
Further include following pre-treatment step before step 6): the location point on straightway being screened, deletion is greater than
The location point that 1.5 times of average distance, the average distance is all location points to the distance apart from nearest straight line, if location point
It is less than d1 to the distance apart from nearest straight line, then distance is denoted as d1;
Further include following pre-treatment step before step 6): delete position point number is less than the straightway of preset value.
7) feature vector is formed: specific steps are as follows:
7.1) to the distance of the location point on every straightway to straight line, arrangement is carried out according to rectilinear direction and forms original spy
Vector is levied, the value of each component of feature vector is distance value of the corresponding position point to straight line;
7.2) sign symbol of each component of vector is determined according to the positional relationship of location point and straight line, specific rules are such as
Under:
One terminal A of selected straightway, the intersection point of any position point B, location point B to straight line are C, and by vector
With vectorIt is expanded into three-dimensional vector, then vectorWith vectorThe symbol of the third dimension component of multiplication cross result is location point B
The symbol of respective components;
8) normalized, specific steps are as follows:
8.1) equidistant interpolation is carried out to the original feature vector that step 7) obtains, the feature vector after obtaining interpolation;
8.2) the mean value dMean of each component in interpolation feature vector is calculated;
8.3) interpolation feature vector components makes are subtracted into mean value dMean, obtains normalized feature vector.
According to the above scheme, set distance threshold value is 0.2 in the step 4).
The principle of the method for the present invention is: using the intrinsic straightway irrelevance feature of printed matter, passing through the minimum between feature
The symbol difference being overlapped between length and character pair component determines the matching relationship between printed matter image.
The beneficial effect comprise that: it is solved between the different feature of length by the minimum method for being overlapped length
With problem, while avoiding can not matching problem as caused by straightway endpoint location inaccuracy.The present invention passes through corresponding special
The symbol difference between component is levied as the distance between feature, avoids the distance as caused by the location point inaccuracy of edge detection
Error.Accuracy is matched in small scale experiments, between matching characteristic and mismatches recognition correct rate between feature close to 100%.
Detailed description of the invention
Present invention will be further explained below with reference to the attached drawings and examples, in attached drawing:
Fig. 1 is the flow chart of the embodiment of the present invention.
Fig. 2 is reference printed matter and printed matter feature extraction schematic diagram to be matched in the embodiment of the present invention;
Fig. 3 is the characteristic matching schematic diagram in the embodiment of the present invention.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to embodiments, to the present invention
It is further elaborated.It should be appreciated that described herein, specific examples are only used to explain the present invention, is not used to limit
The fixed present invention.
In embodiment, respectively to reference printed matter and printed matter one (actual match) to be matched and with reference to printed matter with to
Explanation is compared with printed matter two (practical to mismatch).As shown in Figure 1, a kind of part based on printed matter provided by the invention
Line segment irrelevance feature matching method, comprising the following steps:
Step 1: the image local of reference printed matter, printed matter to be matched one and printed matter to be matched two being carried out respectively straight
Line irrelevance feature extraction;
According to the standardization to reference printed matter and printed matter to be matched respectively of the irrelevance feature extracting method of straightway
Image local extracts feature, obtains multiple feature vectors with reference to printed matter image localMultiple feature vectors of one image local of printed matter to be matchedWith multiple feature vectors of two image local of printed matter to be matchedAs shown in Fig. 2, wherein Fig. 2 (a) is the standardized images office with reference to printed matter
Portion, Fig. 2 (d) and Fig. 2 (g) they are respectively that the standardized images of printed matter one and printed matter to be matched two to be matched are local, Fig. 2 (b),
Fig. 2 (e) and Fig. 2 (h) is edge detection results, and Fig. 2 (c), Fig. 2 (f) and Fig. 2 (i) are the straight-line detection result of Hough transform.
After straightway cluster and length filtration processing, the straightway less than 100 of counting is not involved in characteristic matching, feature in set R
Vector number is 4, and the middle feature vector number of set M ' is 6, and feature vector number is 5 in set M ".
Step 2: characteristic matching being carried out to the feature of extraction, as shown in Figure 3;
Due to the influence of various factors, may result in the inaccurate of straightway endpoint, therefore in matching, need to do into
The processing of one step, specific steps are as follows:
Step 2.1: from set R and set M ' difference selected characteristic vector DiWith D 'j;
Step 2.2: determining feature vector DiWith D 'jMinimum be overlapped length LMin;
Since image passes through standardization, the scale of feature is identical.Known features vector DiWith D 'jLength point
It Wei not LiWith L 'jIf LiWith L 'jIt differs within 30%, then LMinComputation rule it is as follows:
LMin=0.9*min (Li,L′j),
Otherwise 2.1 are entered step, feature vector is reselected.
Step 2.3: calculating feature vector DiWith D 'jThe distance of intersection;
Assuming that feature vector DiWith D 'jThe length of intersection is L, it is clear that L >=LMin.Calculate feature vector DiWith D 'j
The different component number Num of symbol in respective components in intersection, then feature vector DiWith D 'jBetween distance be d (Di,D′j)
Minimum value in=Num/L.
Step 2.4: if distance is less than a certain threshold value, then it is assumed that two line matchs, on the contrary it mismatches.
It finds in an embodiment of the present invention, as feature vector DiWith D 'jBetween distance be generally less than d (Di,D′j) < 0.2
When, feature vector DiWith D 'jMatching, on the contrary then feature vector DiWith D 'jIt mismatches.
Step 2.5: repeating step 2.1-2.4, calculate whole feature vectors pair.
Step 3: the matching relationship between printed matter part is determined according to the matching result between line segment character pair;
The matching characteristic determined by step 2 is that the middle element number of NumMatch, set R and set M ' is respectively to number
NumRAnd NumM, then with Ratio=NumMatch/ (NumR*NumM) size judge part whether match.If Ratio is greater than
0.02 local matching, on the contrary part mismatches.
As shown in table 1, table 2, with reference to appearance of the minimum range less than 0.2 between printed matter and printed matter to be matched one 13
A, ratio reaches 0.54, remote former greater than preset threshold 0.02;And refer to minimum range between printed matter and printed matter to be matched two
Basic both greater than 0.3, do not occur less than 0.2.It can be seen that matching process proposed by the invention is when identifying irrelevance feature
All there is higher accuracy to matching characteristic and mismatch feature.
Table 1
Table 2
It should be understood that for those of ordinary skills, it can be modified or changed according to the above description,
And all these modifications and variations should all belong to the protection domain of appended claims of the present invention.
Claims (2)
1. a kind of local line segment irrelevance feature matching method based on printed matter, which comprises the following steps:
1) feature extraction of straightway irrelevance is locally carried out to the identical image of reference printed matter and printed matter to be matched respectively;?
To multiple feature vectors of reference printed matter image localWith printed matter figure to be matched
As multiple feature vectors of part
2) feature vector is chosen:
Step 2.1) distinguishes selected characteristic vector D from set R and set MiWith D 'j;
Step 2.2) determines feature vector DiWith D 'jMinimum be overlapped length LMin;
Since image passes through standardization, the scale of feature is identical, it is known that feature vector DiWith D 'jLength be respectively
LiWith L 'jIf LiWith L 'jIt differs within 30%, then LMinComputation rule it is as follows:
LMin=0.9*min (Li,L′j),
If LiWith L 'j30% is differed by more than, return step 2.1 reselects feature vector;
3) feature vector D is calculatediWith D 'jThe distance of intersection;
If feature vector DiWith D 'jThe length of intersection is L, it is clear that L > Lmin;Calculate feature vector DiWith D 'jIn intersection
The different component number Num of symbol in respective components, then feature vector DiWith D 'jBetween distance be d (Di,D′jIn)=Num/L
Minimum value;
4) matching judgment: if distance is less than set distance threshold value, then it is assumed that two line matchs, on the contrary it mismatches;
5) step 2) is repeated to the combination for 4), traversing whole feature vectors pair;
6) matching relationship between printed matter part is determined according to the matching result between line segment character pair;It is specific as follows:
The matching characteristic determined by step 5) is that element number is respectively Num in NumMatch, set R and set M to numberRWith
NumM, then with Ratio=NumMatch/ (NumR*NumM) size judge part whether match;The office if Ratio is greater than 0.02
Portion's matching, on the contrary part mismatches;
In the step 1), feature extraction uses following methods:
1.1) image procossing: acquiring the image of printed matter, converts standard picture for printed matter image;
1.2) edge detection: edge detection is carried out to image using Canny edge detection operator, obtains the marginal position of image
Point;
1.3) straight-line detection: carrying out straight-line detection to the marginal position point detected using Hough transform principle, obtains a plurality of straight
Line;
1.4) location point sort out: calculate each marginal position point arrive straight line distance and projected position point, according to distance recently original
Then location point is assigned on each straight line;
It is specific as follows: to set up two threshold values, the minimum distance d1 and maximum distance d2 of location point to straight line;When location point is to arbitrarily
When the distance of straight line is both greater than maximum distance d2, then the location point is not belonging to any straight line;When location point to a plurality of straight line
Distance when being both less than minimum distance d1, then the location point belongs to a plurality of straight line that distance is less than minimum distance d1;Remaining position
Point belongs to apart from nearest straight line;
1.5) projected position of location point on straight line is clustered, determines the straightway number on straight line;
1.6) it according to the projected position of location point on straight line on each straightway, is ranked up, determines straight according to rectilinear direction
The corresponding location point of two endpoints of line segment;
Further include following pre-treatment step before step 1.6): the location point on straightway is screened, deletes to be greater than and put down
The location point of equal 1.5 times of distance, the average distance is all location points to the distance apart from nearest straight line, if location point arrives
Distance apart from nearest straight line is less than d1, then distance is denoted as d1;
Further include following pre-treatment step before step 1.6): delete position point number is less than the straightway of preset value;
1.7) feature vector is formed: specific steps are as follows:
1.7.1) to the distance of the location point on every straightway to straight line, arrangement is carried out according to rectilinear direction and forms primitive character
Vector, the value of each component of feature vector are distance value of the corresponding position point to straight line;
1.7.2 the sign symbol of each component of vector) is determined according to the positional relationship of location point and straight line, specific rules are as follows:
One terminal A of selected straightway, the intersection point of any position point B, location point B to straight line are C, and by vectorWith to
AmountIt is expanded into three-dimensional vector, then vectorWith vectorThe symbol of the third dimension component of multiplication cross result is B pairs of location point
Answer the symbol of component;
1.8) normalized, specific steps are as follows:
1.8.1 equidistant interpolation) is carried out to the original feature vector that step 7) obtains, the feature vector after obtaining interpolation;
1.8.2 the mean value dMean of each component in interpolation feature vector) is calculated;
1.8.3 interpolation feature vector components makes) are subtracted into mean value dMean, obtain normalized feature vector.
2. part line segment irrelevance feature matching method according to claim 1, which is characterized in that set in the step 4)
Set a distance threshold value is 0.2.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610322170.0A CN106023181B (en) | 2016-05-16 | 2016-05-16 | A kind of local line segment irrelevance feature matching method based on printed matter |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610322170.0A CN106023181B (en) | 2016-05-16 | 2016-05-16 | A kind of local line segment irrelevance feature matching method based on printed matter |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106023181A CN106023181A (en) | 2016-10-12 |
CN106023181B true CN106023181B (en) | 2019-03-08 |
Family
ID=57097306
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610322170.0A Active CN106023181B (en) | 2016-05-16 | 2016-05-16 | A kind of local line segment irrelevance feature matching method based on printed matter |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106023181B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113920121B (en) * | 2021-12-14 | 2022-03-15 | 武汉亚美高工艺制品有限公司 | Artwork printing streak detection method and system based on image processing |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101315698A (en) * | 2008-06-25 | 2008-12-03 | 中国人民解放军国防科学技术大学 | Characteristic matching method based on straight line characteristic image registration |
CN104820983A (en) * | 2015-04-23 | 2015-08-05 | 清华大学 | Image matching method |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR100370220B1 (en) * | 2000-12-27 | 2003-02-12 | 삼성전자 주식회사 | Image matching method based on straight line |
-
2016
- 2016-05-16 CN CN201610322170.0A patent/CN106023181B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101315698A (en) * | 2008-06-25 | 2008-12-03 | 中国人民解放军国防科学技术大学 | Characteristic matching method based on straight line characteristic image registration |
CN104820983A (en) * | 2015-04-23 | 2015-08-05 | 清华大学 | Image matching method |
Non-Patent Citations (3)
Title |
---|
MSLD:A robust discriptor for line matching;Zhiheng Wang 等;《Pattern recognition》;20090531;第42卷(第5期);941-953 |
基于Hough变换的直线提取与匹配;刘春阁;《中国优秀硕士学位论文全文数据库 信息科技辑》;20120215(第02期);I140-814 |
序列图像中的直线特征提取与匹配;陈志良;《中国优秀硕士学位论文全文数据库 信息科技辑》;20130115(第01期);I138-1596 |
Also Published As
Publication number | Publication date |
---|---|
CN106023181A (en) | 2016-10-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Shivakumara et al. | Multioriented video scene text detection through Bayesian classification and boundary growing | |
Ren et al. | Figure/ground assignment in natural images | |
CN113344857B (en) | Defect detection network training method, defect detection method and storage medium | |
CN107247950A (en) | A kind of ID Card Image text recognition method based on machine learning | |
CN110189344A (en) | A kind of line segment extraction method in power knife switch image | |
Forczmański et al. | Stamps detection and classification using simple features ensemble | |
CN108038458B (en) | Method for automatically acquiring outdoor scene text in video based on characteristic abstract diagram | |
CN110675442B (en) | Local stereo matching method and system combined with target recognition technology | |
CN106778777B (en) | Vehicle matching method and system | |
CN109635679B (en) | Real-time target paper positioning and loop line identification method | |
CN106023181B (en) | A kind of local line segment irrelevance feature matching method based on printed matter | |
CN105844235B (en) | The complex environment method for detecting human face of view-based access control model conspicuousness | |
Hossen et al. | Vehicle license plate detection and tilt correction based on HSI color model and SUSAN corner detector | |
CN111767909B (en) | Character recognition method and device and computer readable storage medium | |
CN114926635B (en) | Target segmentation method in multi-focus image combined with deep learning method | |
CN107122714B (en) | Real-time pedestrian detection method based on edge constraint | |
CN106022335B (en) | A kind of line segment irrelevance feature extracting method for print characteristics | |
CN107122783B (en) | Method for quickly identifying assembly connector based on angular point detection | |
CN105550706A (en) | Method of recognizing 2D image and 3D image | |
Leifman et al. | Pattern-driven colorization of 3d surfaces | |
CN109636844B (en) | Complex desktop point cloud segmentation method based on 3D bilateral symmetry | |
Sayahi et al. | An intelligent and robust multi-oriented image scene text detection | |
Gaceb et al. | A new mixed binarization method used in a real time application of automatic business document and postal mail sorting. | |
Yasukochi et al. | A recognition method of restricted hand shapes in still image and moving image as a man-machine interface | |
Too et al. | Extraction of scene text information from video |
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 |