CN102945368B - Method for positioning and identifying laser character of beer bottle cap - Google Patents

Method for positioning and identifying laser character of beer bottle cap Download PDF

Info

Publication number
CN102945368B
CN102945368B CN201210395549.6A CN201210395549A CN102945368B CN 102945368 B CN102945368 B CN 102945368B CN 201210395549 A CN201210395549 A CN 201210395549A CN 102945368 B CN102945368 B CN 102945368B
Authority
CN
China
Prior art keywords
character
image
finger url
bottle cap
region
Prior art date
Application number
CN201210395549.6A
Other languages
Chinese (zh)
Other versions
CN102945368A (en
Inventor
武吉梅
李晶
Original Assignee
西安理工大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 西安理工大学 filed Critical 西安理工大学
Priority to CN201210395549.6A priority Critical patent/CN102945368B/en
Publication of CN102945368A publication Critical patent/CN102945368A/en
Application granted granted Critical
Publication of CN102945368B publication Critical patent/CN102945368B/en

Links

Abstract

The invention discloses a method for positioning and identifying a laser character of a beer bottle cap, which comprises the following steps: step 1, character positioning which comprises coarse positioning and fine positioning, that is, carrying out binaryzation, corrosion expansion, edge detection, contour tracing and Hough transform processing on a positioning character image, searching an angle between the straight lines on which two edges of the positioned character A are positioned, calculating a rotating angle of the character, searching coordinates of the center of the bottle cap, translating the center of the bottle cap to the center of an original bottle cap image, aligning the bottle cap image according to the calculated rotating angle of the character and repositioning the character; and step 2, character division and identification, wherein the character identification comprises the steps of extracting a 13 characteristic, extracting a protection statistical characteristic, extracting a coarse grid characteristic and finally carrying out matched identification on the character on the basis of the three character characteristics by utilizing a template matching method. The method disclosed by the invention aims at the character image of the bottle cap, which has a complex background and nonuniform illumination and implements the positioning on the laser character of the beer bottle cap, which rotates at a random angle.

Description

The positioning identifying method of beer bottle cap laser character
Technical field
The invention belongs to electronic supervision code character recognition technologies field, relate to a kind of positioning identifying method of beer bottle cap laser character.
Background technology
In recent years, the research of the electronic supervision code character recognition in Internet of Things printing packaging field achieves a large amount of achievements in research, and these achievements are widely used in the occasion such as Car license recognition, printing packaging Symbol recognition.Under current character recognition condition, because the picture quality be identified is generally better, target area contrast is large, and target is static, and only there will be low-angle rotation, and noise is little, and recognition accuracy is high.
The location of beer bottle cap laser character realizes by images match.Beer bottle cap laser character is made up of two parts, i.e. the numerical character composition of changeless character " A " (being called as finger URL) and five variations.Due to the character that finger URL is constant, slightly mated by correlation coefficient process by the template of finger URL, find the position of finger URL, then essence coupling is carried out to finger URL image, to find image rotation angle, image of becoming a full member, thus navigate to concrete numerical character.
But in the identification of beer bottle cap laser character, bottle cap image gathers in production line motion process, thus character picture there will not be on fixing position, and character picture has and rotates at any angle, brings difficulty to character locating and identification.
Therefore, the character locating recognition methods of Effect of Anti Arbitrary Rotation, the management problems reviewed for the electronic monitoring solved in the circulation of similar Beer Brewage and logistics is of great immediate significance, and is that character recognition technologies moves towards the practical technical matters that must solve.
Summary of the invention
The object of this invention is to provide a kind of positioning identifying method of beer bottle cap laser character, solve in prior art when beer bottle cap Arbitrary Rotation, the impalpable problem of the character on bottle cap.
The technical solution adopted in the present invention is, a kind of positioning identifying method of beer bottle cap laser character, implements according to following steps:
Step 1, character locating, comprise coarse positioning and fine positioning,
1.1) Image semantic classification
Collect the beer bottle cap original image that resolution is 640 × 480 on a production line, at this, Image semantic classification refers to and carries out medium filtering to image,
1.2) coarse positioning
The resolution of locator template image is reduced, the locator template image of low resolution is got with picture centre is the center of circle, radius is the pixel in the circle of r, if total H pixel, by the method connected by column, H pixel is formed a H dimensional vector, put into the first row of array b, with step delta θ rotary template, form the secondary series of b by postrotational H pixel, rotate until turn over 360 degree successively, the array b of final generation H × (2 π/Δ θ) size
Original bottle cap image is reduced resolution, at bottle cap intra-zone from top to bottom, from left and right search, the pixel in circle that radius is r is got at point (x, the y) place on low resolution bottle cap image, and use the same method generation H n dimensional vector n a,
Vector a and array b are relevant by column, record related coefficient and positional information now,
Move to next position, generate new a, the step of double counting related coefficient, until searched for pixels all in bottle cap image range, the related coefficient of position sorted, the position that maximum correlation coefficient occurs is exactly the position of finger URL,
1.3) fine positioning
Intercept step 1.2) middle image of locating the finger URL region obtained, picture size is 56 × 56, first first binary conversion treatment is carried out to finger URL image by OTSU method, add up the number of black pixel in now finger URL binary image, when pixel number black after binaryzation is in [200,900] scope, the binaryzation effect of finger URL " A " is best, if stain number is not within the scope of this, then need to re-start binaryzation to finger URL image
Count the number of black pixel point on finger URL binary image, thus finger URL image is expanded or corrosion treatment, when black pixel point number is greater than 580, need to carry out corrosion treatment to image, when black pixel point number is less than 500, need to carry out expansion process to finger URL image
Horizontal projection method and vertical projection method is adopted to find the border of finger URL to remove unnecessary noise spot,
Gauss-Laplace operator is adopted to carry out rim detection to finger URL image;
The Contour extraction of finger URL edge-detected image based on 8 connected regions is processed;
Two straight lines the longest and horizontal angle in finger URL image is found, i.e. the angle of place, finger URL A two limits straight line, in order to the anglec of rotation of calculating character by Hough transformation;
Obtained the angle of place, finger URL A both sides straight line by Hough transformation, calculate the angle of the center line of both sides angle, be considered as having become a full member to finger URL time vertical when center line rotates, the anglec of rotation now calculated is exactly the anglec of rotation of character; But, no matter finger URL A is anti-just, the angle on two limits exported is all the same, the anglec of rotation only calculated according to a kind of situation carries out angularity correction to bottle cap image, bottle cap image after the correction likely obtained is down, judge the original anglec of rotation to be added the situation that calculating angle need adjust 180 degree, again rotates;
By the angle calculated, rotation correction is done to original bottle cap image, interpolation calculation has been carried out in the process of rotation correction, thus after correcting, the position of finger URL changes, it is no longer the position of coarse positioning, need again to position finger URL, concrete steps are as follows: 1. to the finger URL region binaryzation that coarse positioning finds; 2. by the center of bottle cap center translation to original bottle cap image; 3. the part of now picture centre is rotated according to the angle calculated before; 4. because finger URL region is binarized, the pixel that a large amount of gray-scale value is 0 and 255 should be there is in the region comprising finger URL, finger URL is searched within the scope of the left half-image of the bottle cap image after rotation correction, if pixel value is that total number of the point of 0 or 255 is greater than the threshold value 130 preset in the larger context, this region is exactly the region at finger URL place
Character picture, just after finger URL, intercepts character picture, just completes the operation of character locating;
Step 2, Character segmentation and identification
2.1) Character segmentation
Before Character segmentation first to character picture medium filtering to weaken noise,
Intercept image-region in a big way from the left side of character picture, first character is included in the inside, to this region binaryzation, finds the border of character, be partitioned into first character through follow-up projection denoising; Again with the right margin of first character for left benchmark, intercept region in a big way to the right, second character is included in the inside, again character boundary is found to this region binaryzation, projection, be partitioned into second character, by that analogy, use the same method and the character of 5 in character picture split successively;
To the single character normalization be partitioned into, character all unification be 24 × 40 sizes;
2.2) character recognition
Extract 13 features of single character, projection statistical nature and thick meshed feature, the feature of single character with 0 to 9 template characteristic mate, identify character,
Character to be identified is 10 numerals of ocr-a font, and the single character selecting stroke full, effective from a large amount of single characters is as template, and character boundary is the size after normalization,
Character feature extracts and comprises the following steps:
2.2.1) 13 features are extracted;
2.2.2) projection statistical nature is extracted, to the scanning that character picture to be identified carries out line by line from left to right, count the number of often row black pixel point, then scanning is by column carried out from top to bottom, count the number of the black pixel point often arranged, using the result that the counts proper vector as character;
2.2.3) extract thick meshed feature, normalization Digital Character Image to be identified according to 4 × 4 piecemeals, count the number of the black pixel point in each macrolattice, using the result that the counts proper vector as character;
After obtaining three kinds of above-mentioned features, just obtain the feature of normalization character picture, finally on the basis of these three kinds of character features, utilize template matching method to carry out match cognization to character.
The invention has the beneficial effects as follows, the identification of Internet of Things printing packaging electronic supervision code character locating, for the bottle cap character picture of background complexity, uneven illumination, effectively split character, can resistant to arbitrary angle rotate, little on image definition impact, ensure that the accurate location of character, achieve the beer bottle cap laser character location of Arbitrary Rotation.
Accompanying drawing explanation
Fig. 1 is the original bottle cap image in the inventive method;
Fig. 2 is the locator template image in the inventive method;
Fig. 3 is the coarse positioning image in the inventive method;
Fig. 4 is the finger URL image in the inventive method, and Fig. 4 a is wherein the finger URL image that need corrode, and Fig. 4 b is the finger URL image that need expand, and Fig. 4 c is the image after Fig. 4 a corrodes, and Fig. 4 d is the image after Fig. 4 b expands;
Fig. 5 is the finger URL corrosion expansion plans picture in the inventive method, and Fig. 5 a is wherein finger URL corrosion image, and Fig. 5 b is horizontal projection image, and Fig. 5 c is vertical projection image, and Fig. 5 d is that finger URL cuts image;
Fig. 6 is the finger URL projected image in the inventive method, and Fig. 6 a is wherein image before rim detection, and Fig. 6 b is edge-detected image;
Fig. 7 is the finger URL edge-detected image in the inventive method, and Fig. 7 a is wherein image before Contour extraction, and Fig. 7 b is Contour extraction image;
Fig. 8 is the finger URL Contour extraction image in the inventive method;
Fig. 9 is the finger URL Hough transformation image in the inventive method, Fig. 9 a is wherein the Hough transformation image of finger URL one, Fig. 9 b is the rotation correction image of finger URL one, Fig. 9 c is ambiguity problem image, Fig. 9 d is the Hough transformation image of finger URL two, Fig. 9 e is the rotation correction image of finger URL two, and Fig. 9 f is the image of rotation correction again of finger URL image two;
Figure 10 is the anglec of rotation computed image in the inventive method, and Figure 10 a is wherein finger URL region binary image, and Figure 10 b is bottle cap displacement images, and Figure 10 c is bottle cap rotation correction image, and Figure 10 d is character picture;
Figure 11 is that the finger URL in the inventive method reorientates image;
Figure 12 is the single character binary image in the inventive method;
Figure 13 is the single character projected image in the inventive method, and Figure 13 a is wherein the character picture be partitioned into, and Figure 13 b is character filtering image.
Embodiment
The principle of work of the positioning identifying method of beer bottle cap laser character of the present invention is: beer bottle cap laser code is made up of two parts, the numerical character of special character " A " (being called as finger URL) that namely specify and five variations, therefore all character locating comprise coarse positioning and fine positioning two parts, coarse positioning navigates to the position of finger URL, and fine positioning finds image rotation angle by finger URL image, image is become a full member and navigates to character.The information of locator template records template when rotating with a fixed step size and rotates to the half-tone information of all angles place locator template, makes and mates, thus can realize the character recognition of Arbitrary Rotation of it and bottle cap image to be matched.
The principle of work of coarse positioning is: the resolution of bottle cap image and locator template image is reduced, rotates locator template image, the template gray information of each angle stored in array with certain step-length; The bottle cap image of low resolution calculates the related coefficient of half-tone information in bottle cap around each position and locator template information, and the maximum position of related coefficient is exactly the region at finger URL place; The coordinate of finger URL on former figure is calculated according to the coordinate of finger URL in low resolution.
The principle of work of fine positioning is: the image intercepting finger URL region, to finger URL image binaryzation, corrosion expansion, rim detection, Contour extraction and Hough transformation, finds the angle that character rotates, and this angle is exactly the angle that image needs to rotate; Find the coordinate at bottle cap center in image, by the center of bottle cap center translation to image, image is become a full member, and is repositioned onto character, is presented on interface by character picture, thus achieves the character locating of Arbitrary Rotation.
The positioning identifying method of beer bottle cap laser character of the present invention, implement according to following steps:
Step 1, character locating, comprise coarse positioning and fine positioning.
1.1) Image semantic classification
Collect the beer bottle cap original image that resolution is 640 × 480 on a production line, Image semantic classification here refers to and carries out medium filtering to image,
Fig. 1 is original bottle cap image, and Fig. 2 is locator template image.
1.2) coarse positioning
The resolution of original bottle cap image and locator template image is reduced, rotates locator template image, each postrotational template gray information stored in array with certain step-length.The bottle cap image of low resolution calculates the related coefficient of half-tone information in bottle cap around each position and locator template image information, the maximum position of related coefficient is exactly the region at finger URL place, calculate the coordinate of finger URL on original bottle cap image again according to the coordinate of finger URL in low resolution, detailed process is as follows:
The resolution of locator template image is reduced, the locator template image of low resolution is got with picture centre is the center of circle, radius is the pixel in the circle of r, if total H pixel, by the method connected by column, H pixel is formed a H dimensional vector, put into the first row of array b, with step delta θ rotary template, form the secondary series of b by postrotational H pixel, rotate until turn over 360 degree successively, the array b of final generation H × (2 π/Δ θ) size
Original bottle cap image is reduced resolution, at bottle cap intra-zone from top to bottom, from left and right search, the pixel in circle that radius is r is got at point (x, the y) place on low resolution bottle cap image, and use the same method generation H n dimensional vector n a,
Vector a and array b are relevant by column, record related coefficient and positional information now,
Move to next position, generate new a, the step of double counting related coefficient, until searched for pixels all in bottle cap image range, the related coefficient of position sorted, the position that maximum correlation coefficient occurs is exactly the position of finger URL,
Fig. 3 is the image that coarse positioning obtains.
1.3) fine positioning
Intercept the image of finger URL region, binaryzation, corrosion expansion, rim detection, Contour extraction and Hough transformation process are carried out to finger URL image, finds the angle of the straight line at the place, two limits of finger URL " A ", calculate the angle that character rotates; Find the coordinate at bottle cap center, by the center of bottle cap center translation to original bottle cap image, become a full member by bottle cap image by the character anglec of rotation calculated, be repositioned onto character, detailed process is as follows:
Intercept step 1.2) middle image of locating the finger URL region obtained, picture size is 56 × 56, first first binary conversion treatment is carried out to finger URL image by OTSU method, add up the number of black pixel in now finger URL binary image, when pixel number black after binaryzation is [200,900] time in scope, the binaryzation effect of finger URL " A " is best, if stain number does not need to re-start binaryzation to finger URL image within the scope of this.
Count the number of black pixel point on finger URL binary image, thus finger URL image is expanded or corrosion treatment.When black pixel point number is greater than 580, need to carry out corrosion treatment to image, when black pixel point number is less than 500, need to carry out expansion process to finger URL image.
Reference Fig. 4 is finger URL corrosion expansion plans picture, and Fig. 4 a is wherein the finger URL image that need corrode, and Fig. 4 b is the finger URL image that need expand, and Fig. 4 c is the image after Fig. 4 a corrodes, and Fig. 4 d is the image after Fig. 4 b expands.
Horizontal projection method and vertical projection method is adopted to find the border of finger URL to remove unnecessary noise spot, sometimes can not remove out the noise (as shown in the black picture element part of Fig. 5 a upper right corner) on four angles of finger URL image simply by virtue of horizontal projection and vertical projection completely, just need to eliminate this noise like by the pixel on four angles is directly bleached; ;
Reference Fig. 5 is that finger URL projection cuts image, and Fig. 5 a is wherein finger URL corrosion image, and Fig. 5 b is horizontal projection image, and Fig. 5 c is vertical projection image, and Fig. 5 d is that finger URL cuts image.
Gauss-Laplace operator is adopted to carry out rim detection to finger URL image;
Reference Fig. 6 is finger URL edge-detected image, and Fig. 6 a is wherein image before rim detection, and Fig. 6 b is edge-detected image.
The Contour extraction of finger URL edge-detected image based on 8 connected regions is processed;
Reference Fig. 7 is finger URL Contour extraction image, and Fig. 7 a is wherein image before Contour extraction, and Fig. 7 b is Contour extraction image.
Two straight lines the longest and horizontal angle in finger URL image is found, i.e. the angle of place, finger URL A two limits straight line, in order to the anglec of rotation of calculating character by Hough transformation;
Fig. 8 is finger URL Hough transformation image, and be finger URL one Hough transformation image with reference to Fig. 9 a, Fig. 9 b is finger URL one rotation correction image.
Obtained the angle of place, finger URL A both sides straight line by Hough transformation, calculate the angle of the center line of both sides angle, be considered as having become a full member to finger URL time vertical when center line rotates, the anglec of rotation now calculated is exactly the anglec of rotation of character.But, no matter finger URL A is anti-just, the angle on two limits exported is all the same (being ambiguity problem image as is shown in fig. 9 c), the anglec of rotation only calculated according to a kind of situation carries out angularity correction to bottle cap image, bottle cap image after the correction likely obtained is down, judge the situation that calculating angle need adjust, the original anglec of rotation is added 180 degree, again rotate, reference, Fig. 9 d is finger URL two Hough transformation image, and Fig. 9 e is finger URL two rotation correction image, and Fig. 9 f is finger URL image two rotation correction image again.
By the angle calculated, rotation correction is done to original bottle cap image, interpolation calculation has been carried out in the process of rotation correction, thus after correcting, the position of finger URL changes, it is no longer the position of coarse positioning, need again to position finger URL, concrete steps are as follows: 1. to the finger URL region binaryzation (being finger URL region binary image as shown in Figure 10 a) that coarse positioning finds, 2. by the center (be as shown in fig. lob bottle cap displacement images) of bottle cap center translation to original bottle cap image, 3. the part of now picture centre is rotated (being bottle cap rotation correction image as shown in figure l oc) according to the angle calculated before, 4. because finger URL region is binarized, the pixel that a large amount of gray-scale value is 0 and 255 should be there is in the region comprising finger URL, finger URL is searched within the scope of the left half-image of the bottle cap image (as shown in figure l oc) after rotation correction, if pixel value is that total number of the point of 0 or 255 is greater than the threshold value (preferably the numerical value of 130) preset in the larger context, this region is exactly the region at finger URL place, character picture is just after finger URL, character picture is intercepted, the operation just completing character locating is character picture as shown in fig. 10d.
Step 2, Character segmentation and identification
2.1) Character segmentation
The region comprising first character is intercepted from the character picture (as shown in fig. 10d) containing five characters navigated to, binaryzation is carried out to it, single character boundary is found by projection, be partitioned into single character, use the same method after first character and intercept out 4 remaining characters, to the single character filtering be partitioned into and normalization, concrete steps are as follows:
Because the former bottle cap image in character locating process there occurs rotation, this process can produce certain noise, therefore, before Character segmentation first to character picture medium filtering to weaken noise,
First intercept image-region in a big way from the left side of character picture, first character is included in the inside, to this region binaryzation (as shown in figure 11), finds the border of character, be partitioned into first character through follow-up projection denoising; Again with the right margin of first character for left benchmark, intercept region in a big way to the right, second character is included in the inside, again character boundary is found to this region binaryzation, projection, be partitioned into second character, by that analogy, use the same method and the character of 5 in character picture split successively.
Figure 11 is single character binary image.
Because the regional compare of binaryzation is above large, needing to reduce the scope, tentatively determining the region (as shown in figure 12) at single character place by projecting to the single character binary image intercepted.
Figure 12 is single character projected image.
Or due to problems such as bottle cap background complexity, uneven illuminations, sometimes carry out binaryzation in the larger context, its treatment effect is not still very desirable, needs to reduce the scope again binaryzation to weaken the interference of neighboring pixel again as far as possible; The region at single character place has tentatively been determined, to this zonule again binaryzation by projection; Again projection is re-started to this zonule, remove noise spot and determine the border of single character.
Or the character picture be partitioned into after binaryzation is the character picture be partitioned into as depicted in fig. 13 a, burr is many sometimes, and needing to carry out medium filtering process and round and smooth being convenient to of stroke is identified, is character filtering image as illustrated in fig. 13b.
So far, to the single character normalization be partitioned into, character all unification be 24 × 40 sizes.
2.2) character recognition
Extract 13 features of single character, projection statistical nature and thick meshed feature, the feature of single character with 0 to 9 template characteristic mate, identify character,
Character to be identified is 10 numerals of ocr-a font, and the single character selecting stroke full, effective from a large amount of single characters is as template, and character boundary is the size after normalization.
Character feature extracts and comprises the following steps:
2.2.1) 13 features are extracted
13 described feature extractions, be a kind of method had compared with strong adaptability, concrete steps are as follows:
First character is divided into 8 regions fifty-fifty, count the number of the black pixel point in these 8 regions respectively, so just obtain 8 features, and these features are in turn denoted as I (1) ~ I (8) according to order from top to bottom, from left to right, then black pixel point number is counted on two row in the middle of vertical direction and horizontal direction middle two rows as 4 features, be denoted as I (9) ~ I (12) successively, finally count 13rd feature of number as character of all black pixel points in character picture;
2.2.2) projection statistical nature is extracted
To the scanning that character picture to be identified carries out line by line from left to right, count the number of often row black pixel point, then carry out scanning by column from top to bottom, count the number of the black pixel point often arranged, using the result that the counts proper vector as character;
2.2.3) extract thick meshed feature
Normalization Digital Character Image to be identified according to 4 × 4 piecemeals, count the number of the black pixel point in each macrolattice, using the result that the counts proper vector as character;
After obtaining three kinds of above-mentioned features, just obtain the feature of normalization character picture, finally on the basis of these three kinds of character features, utilize template matching method to carry out match cognization to character.

Claims (3)

1. a positioning identifying method for beer bottle cap laser character, is characterized in that, implements according to following steps:
Step 1, character locating, comprise coarse positioning and fine positioning,
1.1) Image semantic classification
Collect the beer bottle cap original image that resolution is 640 × 480 on a production line, this original image is carried out to the process of medium filtering;
1.2) coarse positioning
The resolution of locator template image is reduced, the locator template image of low resolution is got with picture centre is the center of circle, radius is the pixel in the circle of r, if total H pixel, by the method connected by column, H pixel is formed a H dimensional vector, put into the first row of array b, with step delta θ rotary template, form the secondary series of b by postrotational H pixel, rotate until turn over 360 degree successively, the array b of final generation H × (2 π/Δ θ) size
Original bottle cap image is reduced resolution, at bottle cap intra-zone from top to bottom, from left and right search, the pixel in circle that radius is r is got at point (x, the y) place on low resolution bottle cap image, and use the same method generation H n dimensional vector n a,
Vector a and array b are relevant by column, record related coefficient and positional information now,
Move to next position, generate new a, the step of double counting related coefficient, until searched for pixels all in bottle cap image range, the related coefficient of position sorted, the position that maximum correlation coefficient occurs is exactly the position of finger URL;
1.3) fine positioning
Intercept step 1.2) middle image of locating the finger URL region obtained, picture size is set to 56 × 56, first carries out binary conversion treatment by OTSU method to finger URL image,
Add up the number of black pixel in now finger URL binary image, when pixel number black after binaryzation is in [200,900] scope, the binaryzation effect of finger URL " A " is best, if stain number is not within the scope of this, then need to re-start binaryzation to finger URL image
Count the number of black pixel point on finger URL binary image, thus finger URL image is expanded or corrosion treatment, when black pixel point number is greater than 580, need to carry out corrosion treatment to image, when black pixel point number is less than 500, need to carry out expansion process to finger URL image
Horizontal projection method and vertical projection method is adopted to find the border of finger URL to remove unnecessary noise spot,
Gauss-Laplace operator is adopted to carry out rim detection to finger URL image;
The Contour extraction of finger URL edge-detected image based on 8 connected regions is processed;
Two straight lines the longest and horizontal angle in finger URL image is found, i.e. the angle of place, finger URL A two limits straight line, in order to the anglec of rotation of calculating character by Hough transformation;
Obtained the angle of place, finger URL A both sides straight line by Hough transformation, calculate the angle of the center line of both sides angle, be considered as having become a full member to finger URL time vertical when center line rotates, the anglec of rotation now calculated is exactly the anglec of rotation of character; But, no matter finger URL A is anti-just, the angle on two limits exported is all the same, the anglec of rotation only calculated according to a kind of situation carries out angularity correction to bottle cap image, bottle cap image after the correction likely obtained is down, judge the original anglec of rotation to be added the situation that calculating angle need adjust 180 degree, again rotates;
By the angle calculated, rotation correction is done to original bottle cap image, interpolation calculation has been carried out in the process of rotation correction, thus after correcting, the position of finger URL changes, it is no longer the position of coarse positioning, need again to position finger URL, concrete steps are as follows: 1. to the finger URL region binaryzation that coarse positioning finds; 2. by the center of bottle cap center translation to original bottle cap image; 3. the part of now picture centre is rotated according to the angle calculated before; 4. because finger URL region is binarized, the pixel that a large amount of gray-scale value is 0 and 255 is there is in the region comprising finger URL, finger URL is searched within the scope of the left half-image of the bottle cap image after rotation correction, if this pixel values in regions is total number of the point of 0 or 255 be greater than the threshold value preset, this region is exactly the region at finger URL place, character picture, just after finger URL, intercepts character picture, namely completes the operation of character locating;
Step 2, Character segmentation and identification
2.1) Character segmentation
First medium filtering is carried out to weaken noise to character picture before Character segmentation,
Intercept an image-region from the left side of character picture, first character is included in the inside, to this region binaryzation, finds the border of character, be partitioned into first character through follow-up projection denoising; Again with the right margin of first character for left benchmark, intercept another image-region to the right, second character is included in the inside, again character boundary is found to this region binaryzation, projection, be partitioned into second character, by that analogy, use the same method and the character of 5 in character picture split successively;
To the single character normalization be partitioned into, character all unification be 24 × 40 sizes;
2.2) character recognition
Extract 13 features of single character, projection statistical nature and thick meshed feature, the feature of single character with 0 to 9 template characteristic mate, identify character,
Character to be identified is 10 numerals of ocr-a font, and the single character selecting stroke full, effective from a large amount of single characters is as template, and character boundary is the size after normalization,
Character feature extracts and comprises the following steps:
2.2.1) 13 features are extracted;
2.2.2) projection statistical nature is extracted, to the scanning that character picture to be identified carries out line by line from left to right, count the number of often row black pixel point, then scanning is by column carried out from top to bottom, count the number of the black pixel point often arranged, using the result that the counts proper vector as character;
2.2.3) extract thick meshed feature, normalization Digital Character Image to be identified according to 4 × 4 piecemeals, count the number of the black pixel point in each macrolattice, using the result that the counts proper vector as character;
After obtaining three kinds of above-mentioned features, just obtain the feature of normalization character picture, finally on the basis of these three kinds of character features, utilize template matching method to carry out match cognization to character.
2. the positioning identifying method of beer bottle cap laser character according to claim 1, it is characterized in that: described step 2.1) in, the regional compare of the binaryzation obtained due to previous step is large, needs reduce the scope, and tentatively determine the region at single character place by projecting to the single character binary image intercepted;
The region at single character place has tentatively been determined, to this zonule again binaryzation by projection; Again projection is re-started to this zonule, remove noise spot and determine the border of single character;
Carry out medium filtering process.
3. the positioning identifying method of beer bottle cap laser character according to claim 1, is characterized in that: described step 2.2.1) in 13 feature extractions, concrete steps are as follows:
First character is divided into 8 regions fifty-fifty, count the number of the black pixel point in these 8 regions respectively, so just obtain 8 features, and these features are in turn denoted as I (1) ~ I (8) according to order from top to bottom, from left to right
Then black pixel point number is counted on two row in the middle of vertical direction and horizontal direction middle two rows as 4 features, be denoted as I (9) ~ I (12) successively, finally count 13rd feature of number as character of all black pixel points in character picture.
CN201210395549.6A 2012-10-17 2012-10-17 Method for positioning and identifying laser character of beer bottle cap CN102945368B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201210395549.6A CN102945368B (en) 2012-10-17 2012-10-17 Method for positioning and identifying laser character of beer bottle cap

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201210395549.6A CN102945368B (en) 2012-10-17 2012-10-17 Method for positioning and identifying laser character of beer bottle cap

Publications (2)

Publication Number Publication Date
CN102945368A CN102945368A (en) 2013-02-27
CN102945368B true CN102945368B (en) 2015-04-22

Family

ID=47728307

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201210395549.6A CN102945368B (en) 2012-10-17 2012-10-17 Method for positioning and identifying laser character of beer bottle cap

Country Status (1)

Country Link
CN (1) CN102945368B (en)

Families Citing this family (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105844271A (en) * 2015-01-13 2016-08-10 征图新视(江苏)科技有限公司 High-speed variable number character automatic extraction detection method and system
CN104751489A (en) * 2015-04-09 2015-07-01 苏州阔地网络科技有限公司 Grid-based relay tracking method and device in online class
CN104820827A (en) * 2015-04-28 2015-08-05 电子科技大学 Method for recognizing punctiform characters on surfaces of cables
CN105004323B (en) * 2015-07-03 2016-08-17 长沙理工大学 A kind of cap like wine bottle cap rotation angle measurement based on machine vision and modification method
CN105447469A (en) * 2015-12-01 2016-03-30 天津普达软件技术有限公司 Bottle cover character spray-printing detection method for mineral spring water bottle
CN105787936A (en) * 2016-02-25 2016-07-20 天津普达软件技术有限公司 Method for determining position of bottle cap
CN106251341B (en) * 2016-07-22 2019-12-24 凌云光技术集团有限责任公司 Printing quality detection method
CN106204631B (en) * 2016-08-23 2019-01-15 苏州德创测控科技有限公司 Plug-hole detection method based on frequency domain
CN106583271A (en) * 2016-12-28 2017-04-26 天津普达软件技术有限公司 Method for removing defective packing box with first line of ink-jet printed characters lacked
CN106874909B (en) * 2017-01-18 2019-10-11 深圳怡化电脑股份有限公司 A kind of recognition methods of image character and its device
CN107292840B (en) * 2017-06-07 2020-01-17 上海智臻智能网络科技股份有限公司 Image restoration method and device, computer-readable storage medium and terminal
CN107247057B (en) * 2017-06-09 2020-02-28 青岛软控机电工程有限公司 Method for detecting defects of production date on surface of bottle cap
CN107330881B (en) * 2017-06-30 2020-11-06 燕山大学 Method for identifying wheel hub valve hole
CN110753503A (en) * 2017-11-15 2020-02-04 深圳市柔宇科技有限公司 Control method of intelligent pillow, intelligent pillow and computer readable storage medium
CN108985305B (en) * 2018-06-26 2020-08-25 武汉科技大学 Laser etching industrial detonator coded image positioning and correcting method

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102609699A (en) * 2012-02-17 2012-07-25 重庆大学 Device and method for recognizing number of cast workpiece scanned by laser
CN102622783A (en) * 2012-03-29 2012-08-01 北京速通科技有限公司 Multilane free-flow electronic toll collection method based on accurate position match
CN102637275A (en) * 2012-02-28 2012-08-15 刘杨志 Safe shopping method and system based on Internet of things

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP3844594B2 (en) * 1998-03-31 2006-11-15 株式会社リコー Industrial character recognition method

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102609699A (en) * 2012-02-17 2012-07-25 重庆大学 Device and method for recognizing number of cast workpiece scanned by laser
CN102637275A (en) * 2012-02-28 2012-08-15 刘杨志 Safe shopping method and system based on Internet of things
CN102622783A (en) * 2012-03-29 2012-08-01 北京速通科技有限公司 Multilane free-flow electronic toll collection method based on accurate position match

Also Published As

Publication number Publication date
CN102945368A (en) 2013-02-27

Similar Documents

Publication Publication Date Title
Ju et al. Depth saliency based on anisotropic center-surround difference
Teoh et al. Symmetry-based monocular vehicle detection system
Broggi Robust real-time lane and road detection in critical shadow conditions
US8867790B2 (en) Object detection device, object detection method, and program
Liow et al. Use of shadows for extracting buildings in aerial images
CN104636701B (en) A kind of laser two-dimensional code recognition methods based on image repair
CN1255767C (en) Production of reconfiguration image data according to partial data moving distance and angle
CN102790841B (en) Method of detecting and correcting digital images of books in the book spine area
Kluge Extracting road curvature and orientation from image edge points without perceptual grouping into features
CN100351039C (en) Precisive measurement of static knife profile
Kang et al. Road lane segmentation using dynamic programming for active safety vehicles
CN103093191B (en) A kind of three dimensional point cloud is in conjunction with the object identification method of digital image data
CN102096795B (en) Method for recognizing worn two-dimensional barcode image
CN103047943B (en) Based on the door skin geomery detection method of single projection coded structured light
Cheng et al. Building boundary extraction from high resolution imagery and lidar data
KR20130030220A (en) Fast obstacle detection
CN102501140B (en) Method for positioning and monitoring wear of ball nose end mill cutter
CN102930268B (en) A kind of for polluting and the accurate positioning method of DataMatrix code in various visual angles situation
US10102408B2 (en) Quick response code feature detection method and system
CN107609451A (en) A kind of high-precision vision localization method and system based on Quick Response Code
CN102222229B (en) Method for preprocessing finger vein images
TWI455034B (en) Barcode recognion method and a computer product thereof
Mishra et al. Segmenting “simple” objects using RGB-D
CN107045629B (en) Multi-lane line detection method
CN103793712A (en) Image recognition method and system based on edge geometric features

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant