CN107301429B - License plate similar character recognition method based on local position value scoring - Google Patents
License plate similar character recognition method based on local position value scoring Download PDFInfo
- Publication number
- CN107301429B CN107301429B CN201710500232.7A CN201710500232A CN107301429B CN 107301429 B CN107301429 B CN 107301429B CN 201710500232 A CN201710500232 A CN 201710500232A CN 107301429 B CN107301429 B CN 107301429B
- Authority
- CN
- China
- Prior art keywords
- character
- template
- matrix
- license plate
- detected
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/63—Scene text, e.g. street names
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/625—License plates
Abstract
The invention discloses a license plate similar character recognition method based on local position value scoring, which is characterized in that on the premise of template matching, after a similar template character group is determined according to a template matching result, a recognition sequence, a recognition layer number and a maximum local position value difference matrix of each layer are determined through the similar template character group, the local position value scoring of a character to be detected and each similar template character is matched, the category of the layer of similar character group to which the character to be detected belongs after template matching is judged layer by layer according to the height of the scoring, and the character with higher template matching error recognition rate is subjected to secondary recognition, so that the purpose of improving the license plate recognition rate is achieved. The method is simple and practical, can be applied to portable hardware systems with general processing speed, and has obvious improvement on the recognition rate and recognition rate of the license plate compared with a local HOG characteristic recognition method.
Description
Technical Field
The invention belongs to the technical field of pattern recognition and artificial intelligence, and particularly relates to a design of a license plate similar character recognition method based on local position value scoring.
Background
The license plate recognition technology is an application of a computer video image recognition technology in vehicle license plate recognition, and refers to a technology for extracting a moving vehicle license plate from a complex background and detecting and recognizing the license plate. The whole process comprises photographing, image preprocessing, license plate region extraction, character segmentation, character feature recognition and the like.
At present, many character recognition methods exist at home and abroad, for example: template matching is combined with local HOG features, a neural network character recognition method, a Support Vector Machine (SVM) -based recognition method and the like. In the template matching identification method, characters to be detected are compared with a standard template, and the characters with the highest similarity degree with the template are judged as template characters, so that the aim of identifying the license plate is fulfilled. The template matching method directly carries out the 'AND' operation on the image matrixes of the characters to be detected and the template characters under the condition of converting the image matrixes into the same size, and the proportion of the image matrixes and the template character matrixes is obtained, so that the people with large proportion win the game. The template matching recognition algorithm is simple in principle and easy to implement, so that the template matching recognition algorithm is widely applied, but in practical application, the overall difference of a plurality of characters is small, and the characters are greatly different only in local details, so that the template matching method is easy to misjudge, and recognition errors are caused.
The HOG feature, namely the directional gradient histogram, is firstly proposed by French researchers Dalal in 2005 CVPR, HOG + SVM has a good effect on pedestrian detection, a local HOG feature algorithm utilizes the gradient direction and the size of each pixel point of a local feature region to count the direction into a plurality of regions, the gradient size is used as a weight projected in the region to be accumulated, the weight accumulated sum of each region is obtained, and finally the Euclidean distance between a character to be detected and a template character is calculated to perform character recognition.
The neural network character recognition method is based on an error back propagation algorithm or BP algorithm, is essentially a mathematical model for neural network learning, can achieve higher fault-tolerant rate in character recognition by using BP neural network training for character feature selection and extraction, but has high operation complexity and long time consumption in training samples, and is difficult to meet the requirements of recognition rate and low cost.
The support vector machine is an emerging machine learning method, which is developed on the basis of a statistical theory, can reduce the size of a training sample space as much as possible and improve the learning capacity of a machine on the basis of the principle of minimizing structural risk, has obvious advantages compared with a neural network learning method, but has resistance which hinders the development of the neural network learning method due to the fact that the training sample time is too long and the classification rate is slow.
The vehicle license plate recognition methods are combined, a method of combining template matching with local HOG features is adopted to achieve high recognition rate of similar characters, the method is simple relative to a neural network and a support vector machine and easy to achieve, and the recognition rate of the method is not as good as that of a machine learning method after a large number of samples are trained.
Disclosure of Invention
The invention aims to solve the problem that the recognition rate of similar characters is low in the existing license plate character recognition method, and provides a license plate similar character recognition method based on local position value scoring, so that the purpose of detecting a license plate at a higher speed and a higher recognition rate is achieved.
The technical scheme of the invention is as follows: a license plate similar character recognition method based on local position value scoring comprises the following steps:
s1, performing template matching on the character to be detected and a standard template in a database to obtain a template character group similar to the character to be detected;
s2, defining the sequence and the layer number of layer-by-layer segmentation according to the number and the characteristics of the template character groups;
s3, searching a maximum local position value difference matrix of the character to be detected and the template character group, and defining the maximum local position value difference matrix as an M matrix;
s4, segmenting the template character group layer by layer according to the M matrix, and discarding characters which do not meet the conditions;
and S5, judging whether the template character group only contains one character, if so, finishing the identification, otherwise, returning to the step S3.
Wherein, step S4 specifically includes the following sub-steps:
s41, establishing the same rectangular coordinate system for M matrixes in the same area of the character to be detected and the template character group respectively;
s42, dividing the M matrix area into intervals with the same angle;
s43, calculating the score of the character to be detected in each angle interval matching template character;
and S44, calculating the total scores of all the angle intervals, judging the character to be detected as the type of the template character with the highest score, and discarding other template characters which do not meet the conditions.
Wherein, step S43 specifically includes the following sub-steps:
s431, respectively calculating the position value of each angle interval of the character to be measured according to the formula (1):
in the formula XkDenotes thetakTo thetak+1The sum of the position values of the characters to be detected in the interval range, M represents the number of pixel points 1 in the interval range, and xi,yiRespectively representing the horizontal and vertical coordinates of the point of which the ith pixel is 1;
s432, respectively calculating the position value of each angle interval of the template character according to the formula (2):
in the formulaDenotes thetakTo thetak+1The sum of the position values of the template characters in the interval range;
s433, calculating similarity scores of the character to be detected and the template character in each angle interval according to a formula (3):
in the formula SkDenotes thetakTo thetak+1And (5) scoring the similarity between the character to be detected and the template character within the interval range.
The invention has the beneficial effects that: the invention mainly solves the dilemma that the license plate recognition rate and the recognition rate cannot be simultaneously met in the license plate recognition process based on the local difference of similar license plate characters. The invention improves the recognition rate of the license plate characters under the condition of reducing the complexity of recognizing the similar characters of the license plate, and simultaneously compared with a license plate character recognition method based on an artificial neural network, the invention can achieve very high recognition rate of the similar characters by only needing one standard template library without a large number of training samples for machine learning, so that the recognition of the license plate is simple, convenient and efficient, and the invention is suitable for portable hardware systems with low operation rate, such as a portable license plate recognition instrument and the like.
Furthermore, the character to be detected and the template character group are binary images, a '1' represents white, a '0' represents black in numerical value, and the M matrix is two or two matrixes with the most obvious local black-white difference and the smallest range.
The beneficial effects of the further scheme are as follows: according to the characteristics of Chinese license plate characters, letters and numbers, the M matrix of the similar characters selects one part as "0" and the other part as "1" as much as possible, namely, the binary image shows that the part in the M matrix area is completely black, and the part in the M matrix area is more in white area, so that the score difference can be enlarged, and the recognition rate is improved. Meanwhile, the area of the M matrix is as small as possible, so that the identification rate is improved.
Further, the specific method for establishing the rectangular coordinate system for the M matrix in step S41 is as follows:
and taking the first pixel point at the upper left corner of the M matrix as an origin, wherein the first row of the M matrix is an X-axis positive half shaft, and the first column is a Y-axis negative half shaft, so that the M matrix is positioned in the fourth quadrant of the rectangular coordinate system.
Step S42 specifically includes: the M matrix is divided into 6 sections of the same angle, each section having an angle of 15 °, with the origin as the starting point and the side of the M matrix other than the X, Y axis as the end point.
The beneficial effects of the further scheme are as follows: through experimental tests, the M matrix is divided into 6 angle intervals, each interval is 15 degrees, and the highest recognition rate can be achieved. If the angle of each interval is too large, the number of the intervals is reduced, the identification basis is insufficient, and the identification rate is reduced; if the angle of each interval is too small, the algorithm is sensitive to the inclination of the character, and the inclined character recognition rate is also reduced.
Further, in step S433, if the similarity score S in a certain angle intervalkWhen the calculation is negative, recording S in the angle interval k0; if X is in a certain angle intervalkAndwhen all are 0, recording S in the angle intervalk=10。
The beneficial effects of the further scheme are as follows: since the similarity score S is calculatedkMay sometimes appear SkIs a negative number or XkAndboth 0' S result in the case where the denominator in equation (3) is 0, so S for these special caseskAnd (5) assigning, so that the accuracy of similarity score calculation is improved, and the recognition rate of license plate characters is improved.
Drawings
Fig. 1 is a flowchart illustrating a license plate similar character recognition method based on a local position value score according to an embodiment of the present invention.
Fig. 2 is a schematic diagram illustrating an embodiment of dividing an M matrix area into the same angle intervals.
Fig. 3 is a schematic diagram of a specific layer-by-layer segmentation template character group according to an embodiment of the present invention.
Detailed Description
Exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It is to be understood that the embodiments shown and described in the drawings are merely exemplary and are intended to illustrate the principles and spirit of the invention, not to limit the scope of the invention.
The embodiment of the invention provides a license plate similar character recognition method based on local position value scoring, which comprises the following steps of:
and S1, performing template matching on the character to be detected and a standard template in the database to obtain a template character group similar to the character to be detected.
In the embodiment of the invention, the character to be detected and the template character group are binary images, and the value of 1 represents white and 0 represents black.
And S2, defining the sequence and the layer number of the layer-by-layer segmentation through the number and the characteristics of the template character groups.
S3, finding the maximum local position value difference matrix of the character to be detected and the template character group, and defining the maximum local position value difference matrix as an M matrix.
In the embodiment of the invention, the M matrix is defined as a local matrix area with the same maximum position value difference of two or two similar characters with the same size, and the position value is defined as a coordinate value of a pixel point of the image matrix in a rectangular coordinate system. According to the characteristics of Chinese license plate characters, letters and numbers, the selection principle of the M matrix is as follows:
(1) one side of the M matrix is selected to be 0 and the other side is selected to be 1 (the two-value graph shows that the other side of the M matrix is completely black and the other side of the M matrix is more white), so that the score difference can be enlarged, and the recognition rate is improved.
(2) The area of the M matrix is as small as possible to improve the identification rate.
And S4, segmenting the template character group layer by layer according to the M matrix, discarding characters which do not meet the conditions, wherein the sequence and the number of layers of segmentation layer by layer are already defined in the step S2.
Step S4 specifically includes the following substeps:
and S41, establishing the same rectangular coordinate system for the M matrixes in the same area of the character group to be detected and the template character group respectively. The specific method for establishing the rectangular coordinate system comprises the following steps:
and taking the first pixel point at the upper left corner of the M matrix as an origin, wherein the first row of the M matrix is an X-axis positive half shaft, and the first column is a Y-axis negative half shaft, so that the M matrix is positioned in the fourth quadrant of the rectangular coordinate system.
S42, taking the origin as the starting point, taking the side of the M matrix other than X, Y as the end point, and clockwise dividing the M matrix into 6 sections with the same angle with the X axis in the forward direction, wherein the angle of each section is 15 °, as shown in fig. 2.
In fig. 2, numeral 1 indicates a maximum difference matrix (M matrix) of local position values of an extracted character "8" compared with a character "B", 2 indicates that a rectangular coordinate system is established for the M matrix, 3 indicates the number of rows and columns of the M matrix, 4 indicates that the M matrix is divided into 6 identical angle intervals, and 5 indicates that each interval is 15 °.
According to the experimental effect, 6 parameters can be set, which are respectively: the method comprises the following steps of template character numbering, character numbering to be detected, local initial row coordinates of template characters and a character matrix to be detected, local initial column coordinates of the template characters and the character matrix to be detected, M matrix row numbers and M matrix column numbers. Experiments show that the influence of the division of the M matrix angle interval on the algorithm identification rate is small, the time consumption difference is about 1ms, and the identification rate and the division of the M matrix angle interval have a great relationship. If the angle of each interval is too large, the number of the intervals is reduced, and the identification rate is reduced because the identification basis is insufficient; if the angle of each interval is too small, the algorithm is sensitive to the inclination of the character, and the inclined character recognition rate is also reduced. The table 1 shows the relationship between the number of the M matrix sections of the 20 random license plates and the algorithm recognition rate and recognition rate. As can be seen from the data in table 1, the M matrix is divided into 6 angular intervals, each 15 °, which achieves the highest recognition rate.
TABLE 1
And S43, calculating the score of the character to be detected in each angle interval matching the template character.
Step S43 specifically includes the following substeps:
s431, respectively calculating the position value of each angle interval of the character to be measured according to the formula (1):
in the formula XkDenotes thetakTo thetak+1The sum of the position values of the characters to be detected in the interval range, M represents the number of pixel points 1 in the interval range, and xi,yiRespectively, the abscissa and ordinate of the point where the ith pixel is 1.
S432, respectively calculating the position value of each angle interval of the template character according to the formula (2):
in the formulaDenotes thetakTo thetak+1And the sum of the position values of the template characters in the interval range.
S433, calculating similarity scores of the character to be detected and the template character in each angle interval according to a formula (3):
in the formula SkDenotes thetakTo thetak+1And (5) scoring the similarity between the character to be detected and the template character within the interval range.
S44, calculating the total score of all angle intervals according to the formula (4):
and judging the type of the character to be detected as the template character with the highest score, and discarding other template characters which do not meet the conditions.
And S5, judging whether the template character group only contains one character, if so, finishing the identification, otherwise, returning to the step S3.
Taking the character to be tested as "8" as an example, the template character groups similar to the character to be tested are obtained as "5", "R", "8" and "B" through template matching. As shown in fig. 3, the binary images of the character to be detected and the template characters "5", "R", "8", and "B" are adjusted to the same size, and according to the number and features of the template character groups, the recognition of the similar characters is divided into three layers, and the recognition sequence is:
the first layer is that the similar character group is divided into two sub-character groups, which are respectively a character '5' and characters 'R', '8' and 'B', and M matrixes (regions I) of the character to be detected and each similar character are simultaneously extracted, the position value of the M matrix of the character to be detected and the position value of the M matrix of each template character are calculated, the score of the character to be detected matching each M matrix of the template character is respectively calculated, the scores of the character to be detected matching 4 template character position values are compared, if the score of the matched character '5' is the highest, the character to be detected is identified as the character '5', otherwise, the second layer of identification is carried out, at this time, the characters 'R', '8' and 'B' are left in the similar template character group, and the character.
and on the second layer, the similar character group is divided into two sub-character groups, namely a character R, a character 8 and a character B, M matrixes (regions) of the character to be detected and template characters are extracted simultaneously, the method on the first layer is repeated, the position value scores of the matched 3 template characters of the character to be detected are compared, if the score of the matched character R is the highest, the character to be detected is identified as the character R, otherwise, the third layer of identification is carried out, at the moment, the characters 8 and the characters B are left in the similar template character group, and the character R is abandoned.
and in the third layer, the similar character group is divided into two sub-character groups, namely a character 8 and a character B, the M matrixes (the region III) of the character to be detected and the similar characters are simultaneously extracted, the method of the first layer is repeated, the position value scores of the matched 2 template characters of the character to be detected are compared, if the score of the matched character 8 is high, the character to be detected is recognized as the character 8, otherwise, the character to be detected is recognized as the character B, and the recognition is finished.
The method is based on local differences of similar license plate characters, and mainly solves the dilemma that the license plate recognition rate and the recognition rate cannot be simultaneously met in the license plate recognition process. 728 characters of 104 automobile license plates temporarily parked in a street are taken as experimental samples, the method is compared with a license plate recognition method combining template matching and local HOG characteristics, a comparison result of recognition rates is given in table 2, and a comparison result of recognition rates is given in table 3.
TABLE 2
TABLE 3
As can be seen from the experimental data in tables 2 and 3, compared with the template matching and local HOG characteristics, the recognition rate of the invention is improved by 1.9 percentage points, the time consumption for recognizing a single character is reduced by 108.25ms, and the purpose of improving the recognition rate of the license plate characters is achieved under the condition of reducing the complexity of recognizing the similar characters of the license plate.
It will be appreciated by those of ordinary skill in the art that the embodiments described herein are intended to assist the reader in understanding the principles of the invention and are to be construed as being without limitation to such specifically recited embodiments and examples. Those skilled in the art can make various other specific changes and combinations based on the teachings of the present invention without departing from the spirit of the invention, and these changes and combinations are within the scope of the invention.
Claims (8)
1. A license plate similar character recognition method based on local position value scoring is characterized by comprising the following steps:
s1, performing template matching on the character to be detected and a standard template in a database to obtain a template character group similar to the character to be detected;
s2, defining the sequence and the layer number of layer-by-layer segmentation according to the number and the characteristics of the template character groups;
s3, searching a maximum local position value difference matrix of the character to be detected and the template character group, and defining the maximum local position value difference matrix as an M matrix;
s4, segmenting the template character group layer by layer according to the M matrix, and discarding characters which do not meet the conditions;
s5, judging whether the template character group only contains one character, if yes, finishing the identification, otherwise, returning to the step S3;
the step S4 specifically includes the following sub-steps:
s41, establishing the same rectangular coordinate system for M matrixes in the same area of the character to be detected and the template character group respectively;
s42, dividing the M matrix area into intervals with the same angle;
s43, calculating the score of the character to be detected in each angle interval matching template character;
s44, calculating the total scores of all angle intervals, judging the character to be detected as the type of the template character with the highest score, and discarding other template characters which do not meet the conditions;
the step S43 specifically includes the following sub-steps:
s431, respectively calculating the position value of each angle interval of the character to be measured according to the formula (1):
in the formula XkDenotes thetakTo thetak+1The sum of the position values of the characters to be detected in the interval range, M represents the number of pixel points 1 in the interval range, and xi,yiRespectively representing the horizontal and vertical coordinates of the point of which the ith pixel is 1;
s432, respectively calculating the position value of each angle interval of the template character according to the formula (2):
in the formulaDenotes thetakTo thetak+1The sum of the position values of the template characters in the interval range;
s433, calculating similarity scores of the character to be detected and the template character in each angle interval according to a formula (3):
in the formula SkDenotes thetakTo thetak+1And (5) scoring the similarity between the character to be detected and the template character within the interval range.
2. The license plate similar character recognition method of claim 1, wherein the to-be-detected character and the template character group are binary images, and numerically, 1 represents white and 0 represents black.
3. The license plate similar character recognition method of claim 1, wherein the M matrix is two or more types of matrices with the most obvious local black-white difference and the smallest range.
4. The license plate similar character recognition method of claim 1, wherein the specific method for establishing the rectangular coordinate system for the M matrix in step S41 is as follows:
and taking the first pixel point at the upper left corner of the M matrix as an origin, wherein the first row of the M matrix is an X-axis positive half shaft, and the first column is a Y-axis negative half shaft, so that the M matrix is positioned in the fourth quadrant of the rectangular coordinate system.
5. The license plate similar character recognition method according to claim 4, wherein the step S42 specifically comprises:
the M matrix is divided into 6 sections of the same angle, each section having an angle of 15 °, with the origin as the starting point and the side of the M matrix other than the X, Y axis as the end point.
6. The license plate similar character recognition method of claim 1, wherein in the step S433, if the similarity score in a certain angle interval is S, the similarity score is SkWhen the calculation is negative, recording S in the angle intervalk=0。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710500232.7A CN107301429B (en) | 2017-06-27 | 2017-06-27 | License plate similar character recognition method based on local position value scoring |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710500232.7A CN107301429B (en) | 2017-06-27 | 2017-06-27 | License plate similar character recognition method based on local position value scoring |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107301429A CN107301429A (en) | 2017-10-27 |
CN107301429B true CN107301429B (en) | 2020-05-19 |
Family
ID=60135082
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710500232.7A Active CN107301429B (en) | 2017-06-27 | 2017-06-27 | License plate similar character recognition method based on local position value scoring |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107301429B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109086738B (en) * | 2018-08-23 | 2021-05-11 | 深圳市同维通信技术有限公司 | Character recognition method and device based on template matching |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103065137A (en) * | 2012-12-30 | 2013-04-24 | 四川九洲电器集团有限责任公司 | License plate character recognition method |
CN106529592A (en) * | 2016-11-07 | 2017-03-22 | 湖南源信光电科技有限公司 | License plate recognition method based on mixed feature and gray projection |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9536315B2 (en) * | 2015-01-13 | 2017-01-03 | Xerox Corporation | Annotation free license plate recognition method and system |
-
2017
- 2017-06-27 CN CN201710500232.7A patent/CN107301429B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103065137A (en) * | 2012-12-30 | 2013-04-24 | 四川九洲电器集团有限责任公司 | License plate character recognition method |
CN106529592A (en) * | 2016-11-07 | 2017-03-22 | 湖南源信光电科技有限公司 | License plate recognition method based on mixed feature and gray projection |
Non-Patent Citations (1)
Title |
---|
"基于模板匹配的局部HOG特征的车牌识别算法";高聪 等;《计算机系统应用》;20170115;第26卷(第1期);第3-4节 * |
Also Published As
Publication number | Publication date |
---|---|
CN107301429A (en) | 2017-10-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111401257B (en) | Face recognition method based on cosine loss under non-constraint condition | |
CN108681693B (en) | License plate recognition method based on trusted area | |
CN108121991B (en) | Deep learning ship target detection method based on edge candidate region extraction | |
CN109389121B (en) | Nameplate identification method and system based on deep learning | |
CN106682598A (en) | Multi-pose facial feature point detection method based on cascade regression | |
CN110675437A (en) | Image matching method based on improved GMS-ORB characteristics and storage medium | |
CN104809481A (en) | Natural scene text detection method based on adaptive color clustering | |
CN109325507B (en) | Image classification method and system combining super-pixel saliency features and HOG features | |
CN110472625B (en) | Chinese chess piece visual identification method based on Fourier descriptor | |
CN106503694B (en) | Digit recognition method based on eight neighborhood feature | |
CN111259880A (en) | Electric power operation ticket character recognition method based on convolutional neural network | |
WO2015131468A1 (en) | Method and system for estimating fingerprint pose | |
CN110378239A (en) | A kind of real-time traffic marker detection method based on deep learning | |
CN103886619A (en) | Multi-scale superpixel-fused target tracking method | |
CN110852311A (en) | Three-dimensional human hand key point positioning method and device | |
CN110197113B (en) | Face detection method of high-precision anchor point matching strategy | |
CN108664975A (en) | A kind of hand-written Letter Identification Method of Uighur, system and electronic equipment | |
CN107610177A (en) | A kind of method and apparatus that characteristic point is determined in synchronous superposition | |
CN107122775A (en) | A kind of Android mobile phone identity card character identifying method of feature based matching | |
CN111091124A (en) | Spine character recognition method | |
JPH0695192A (en) | Device for image recognition and method for inputting image | |
CN109727363A (en) | A method of identifying amount in Chinese character in bill | |
CN110443184A (en) | ID card information extracting method, device and computer storage medium | |
CN111339932A (en) | Palm print image preprocessing method and system | |
CN107301429B (en) | License plate similar character recognition method based on local position value scoring |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |