CN109190625A - A kind of container number identification method of wide-angle perspective distortion - Google Patents
A kind of container number identification method of wide-angle perspective distortion Download PDFInfo
- Publication number
- CN109190625A CN109190625A CN201810739072.6A CN201810739072A CN109190625A CN 109190625 A CN109190625 A CN 109190625A CN 201810739072 A CN201810739072 A CN 201810739072A CN 109190625 A CN109190625 A CN 109190625A
- Authority
- CN
- China
- Prior art keywords
- container
- character
- perspective distortion
- wide
- subset
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 21
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 8
- 238000003066 decision tree Methods 0.000 claims abstract description 5
- 238000003754 machining Methods 0.000 claims description 8
- 238000012549 training Methods 0.000 claims description 8
- 238000012360 testing method Methods 0.000 claims description 5
- 230000009466 transformation Effects 0.000 claims description 5
- 230000002708 enhancing effect Effects 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 4
- 238000012937 correction Methods 0.000 claims description 2
- 238000001514 detection method Methods 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 229910052742 iron Inorganic materials 0.000 description 1
- JEIPFZHSYJVQDO-UHFFFAOYSA-N iron(III) oxide Inorganic materials O=[Fe]O[Fe]=O JEIPFZHSYJVQDO-UHFFFAOYSA-N 0.000 description 1
- 238000007639 printing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/148—Segmentation of character regions
- G06V30/153—Segmentation of character regions using recognition of characters or words
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/24—Aligning, centring, orientation detection or correction of the image
- G06V10/245—Aligning, centring, orientation detection or correction of the image by locating a pattern; Special marks for positioning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Evolutionary Biology (AREA)
- Character Discrimination (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of container number identification methods of wide-angle perspective distortion, comprising the following steps: 1) carries out perspective transform pretreatment to container representation；2) character locating and identification of depth convolutional neural networks are constructed；3) case number (CN) identification is carried out based on cascade decision tree.Compared with prior art, the present invention has the advantages that identify that there are the container numbers of obvious perspective distortion in high precision.
Description
Technical field
The present invention relates to container terminal loading and unloading technical field of operation, more particularly, to a kind of collection of wide-angle perspective distortion
Vanning number identification method.
Background technique
Container number identification is the key technology that automation is realized in container terminal loading and unloading operation.Based on computer
The character identifying method of vision, because it is not necessarily to increase additional attachments to container, as long as obtaining the image comprising container number
It can be realized, therefore have become the main way of modern port use.
Case number (CN) identification based on computer vision mainly has the difficult point of following 3 aspects: (1) case number (CN) spread pattern is various
(it is horizontally-arranged it is single, it is horizontally-arranged it is multiple rows of, vertical setting of types is single, vertical setting of types is multiple rows of), (2) case face, which is easy to appear, to be stained interference (unrelated printing, case number (CN) is de-
Fall, iron rust etc.), interference caused by (3) severe natural weather, night work light filling etc., therefore, opposite Car license recognition is swept
Character recognition is retouched, container number identifies that difficulty is higher.
Nowadays, Container shipping case number (CN) identification technology generallys use image preprocessing, case number (CN) positioning, case number (CN) Character segmentation, case number (CN) word
Symbol 4 steps of identification improve picture quality in favor of the processing in later period, then according to image using various filtering algorithms first
The low-level features such as edge feature position the region where container number within the scope of entire image, finally in this regional scope
Interior progress Character segmentation and character recognition.When wide-angle perspective distortion is not present in container representation, this technical solution can be taken
Good accuracy of identification is obtained, but when container representation is there are when wide-angle perspective distortion, this technical solution is in practical applications
Just fail.
Summary of the invention
It is an object of the present invention to overcome the above-mentioned drawbacks of the prior art and provide a kind of wide-angle perspectives to become
The container number identification method of shape.
The purpose of the present invention can be achieved through the following technical solutions:
A kind of container number identification method of wide-angle perspective distortion, comprising the following steps:
1) perspective transform pretreatment is carried out to container representation；
2) character locating and identification of depth convolutional neural networks are constructed；
3) case number (CN) identification is carried out based on cascade decision tree.
The step 1) specifically includes the following steps:
11) a pair of image shot from different perspectives for same container of selection is concentrated in container representation, wherein one
For there are the image of wide-angle perspective distortion, an image for not perspective distortion, selected on two images respectively to
Few 5 pairs of same places, and equation is established to each pair of same place, 9 perspective transform parameters are calculated using least square adjustment；
12) Perspective transformation model obtained according to resolving, to all container representations concentration, there are the images of perspective distortion
It is corrected.
In the step 11), each pair of same place establishes 2 equations, specifically:
Wherein, (u, v) is there are the coordinate of target point in wide-angle perspective distortion image, and (x, y) is no perspective distortion
Image in correspond to the coordinate of target point, a11、a12、a21、a22、a31、a32、a13、a23、a33For perspective transform parameter.
The step 2) specifically includes the following steps:
21) container representation after correcting perspective transform zooms to fixed size, marks the class of container number by hand
Not and boundary rectangle frame, and after small perspective transform progress sample enhancing is added, training dataset, test data set are classified as
And validation data set；
22) depth convolutional neural networks arameter optimization is obtained to identify 10 comprising " 0 "-" 9 " using training dataset
The character machining network of a number, 26 of " A "-" Z " capitalization English characters and background totally 37 classes；
23) simultaneously location character, and output set C=is identified on container representation using trained character machining network
{ck}K=1....K, wherein ck={ bk,ck,sk, b is the rectangle frame for surrounding character, and c is character class, and s is score, and K is candidate word
The number of symbol.
The step 3) identifies container number using three-level decision rule, specifically includes the following steps:
31) according to score threshold and the ratio of width to height threshold value preliminary screening character targets rectangle frame, it is most left to search container representation
The letter of top is successively searched to the right after first member as case number (CN) is put into subset C', and interval is less than character pitch threshold
The member of value is put into subset C'；
32) all members in subset C' are fitted into straight line α, traverses the member c' in subset C'k, as member c'k
It when being greater than the threshold value of setting with a distance from straight line α, is then rejected from subset C', and character line β is fitted according to the member of C' again；
33) if subset C' is unsatisfactory for container number check code rule, will be less than in complete or collected works C apart from character line β distance
Subset C' is added in all characters of threshold value, guarantees that the character number in subset C' is no less than 11, it is straight to traverse all combinations
Meet container number check code rule to 11 case number (CN)s, exports this result as container number.
The priori conditions of the corresponding container number of three-level decision rule and container representation include:
A. before the character of container number 4 be letter, latter 7 be number；
B. letter is in the left side of container representation, and number is on the right side or lower section of letter；
C. character is adjacent two-by-two, and standoff distance is in a certain range, and left and right character possesses similar height and width is high
Than；
D.11 bit number forms character line feature, and the central point of character is capable of forming straight line；
E. container code rule are as follows: the 11st is to check character, and can be calculated by preceding 10 characters
Compared with prior art, the invention has the following advantages that
One, perspective transform correction is added in the present invention in image preprocessing, can identify that there are wide-angle perspective distortions
Container number.
Two, the sequence that case number (CN) positions character recognition again is first changed to the case number (CN) identification again of first character machining by workflow, not only
It can use deep neural network algorithm of target detection and improve character locating and accuracy of identification, and in the identification of subsequent case number (CN)
More higher level knowledges in container number coding rule can be added, effectively improve container number accuracy of identification.
Detailed description of the invention
Fig. 1 is flow chart of the method for the present invention.
Specific embodiment
The present invention is described in detail with specific embodiment below in conjunction with the accompanying drawings.
The present invention provides a kind of container number identification method of wide-angle perspective distortion, and this method can be divided into three portions
Point: 1) perspective image preconditioning；2) character locating and identification based on depth convolutional neural networks；3) based on cascade decision
The case number (CN) of tree identifies.Flow chart is as shown in Fig. 1.
First part: perspective image preconditioning, steps are as follows:
(1) representational a pair of of picture is picked out in image set, selects at least 5 pairs of same places, building perspective by hand
Equation of transformation resolves 9 perspective transform parameters；
(2) all container representations are corrected using Perspective transformation model obtained above.
Second part, character locating and identification based on depth convolutional neural networks, steps are as follows:
(1) container representation after correcting perspective transform zooms to fixed size, marks the class of container number by hand
Other and boundary rectangle frame, and small perspective transform is added and carries out sample enhancing；
(2) using classical deep neural network target detection model, pass through arameter optimization (Fine using training sample
Tuning) obtain capable of identifying the word comprising " 0 "-" 9 " 10 numbers, " A "-" Z " 26 capitalization English characters and background totally 37 classes
Symbol detection network；
(3) simultaneously location character, output set C=are identified on container representation with trained character machining network
{ck}K=1....K, wherein ck={ bk,ck,sk, b indicates to surround character rectangle frame, and c indicates character class, and s indicates score, and K is to wait
The number of word selection symbol.
Part III, the case number (CN) identification based on cascade decision tree, steps are as follows:
(1) rule of thumb score threshold and the ratio of width to height threshold value, preliminary screening character targets rectangle frame then look for image most
The letter of upper left is put into C', then successively finds to the right as first member of case number (CN), and interval is less than reasonable threshold value
Member is put into C'；
(2) straight line α is fitted with all members in C', traverses the member c in C'kIf it is with a distance from straight line α
It greater than certain threshold value, is then rejected from subset C', straight line β can be fitted from the member of C' again；
(3) if C' is unsatisfactory for " container number check code " rule, will be less than in complete or collected works C apart from character line β distance
C' is added in all characters of threshold value, guarantees that the character number in C' is no less than 11, traverses all assembled schemes until 11 casees
Number by the inspection of coding rule, this result is exported as container number.
Embodiment:
In many practical applications, the camera of storage yard is mounted on suspender, i.e. the camera oblique upper that is located at container,
Therefore there are the perspective distortions of wide-angle for the imaging of case number (CN).If directly carrying out case number (CN) identification, existing recognition methods is unable to satisfy
Actual requirement.For this there are the container representation of wide-angle perspective distortion, we have proposed a kind of new case number (CN) identification sides
Method.
(1) perspective transform pre-processes
A pair of image shot from different perspectives for same container of selection is concentrated in container representation, wherein one is
There are the image of wide-angle perspective distortion, an image for no obvious perspective distortion, if (u, v) is target in deformation pattern
The coordinate of point, (x, y) is the coordinate of corresponding points in no obvious deformation pattern, and at least 5 pairs of same places are selected on two images,
As shown in formula (1), each pair of same place establishes 2 equations, at least establishes 10 equations, is calculated using least square adjustment
9 perspective transform parameter a out11,a12,a13,a21,a22,a23,a31,a32,a33；
12) Perspective transformation model obtained according to resolving, to all container representations concentration, there are the images of perspective distortion
It is corrected.(2) character locating and identification based on deep neural network
Container representation after perspective transform is corrected is compressed to 224 × 224, mark by hand container number classification and
Boundary rectangle frame, and small perspective transform is added and carries out sample enhancing, 2100 training samples are obtained, in total 23100 words
Symbol.Training dataset, test data set and validation data set are splitted data into according to 60%, 20%, 20% ratio.
Character locating and identification are carried out using Faster R-CNN target detection model, due to the character on container representation
Target is smaller, we select to gather around the ZF network architecture there are five convolutional layer.It obtains to know by arameter optimization using training sample
Not Bao Han " 0 "-" 9 " 10 numbers, " A "-" Z " 26 capitalization English characters and background totally 37 classes character machining network.
Simultaneously location character, output set C=are identified on a width container representation with trained character machining network
{ck}K=1 ..., K, wherein ck={ bk,ck,sk, b indicates that character area-encasing rectangle frame, c indicate character class, and s indicates score, and K is
The number of candidate characters.
(3) based on the case number (CN) identification of cascade decision tree
The priori knowledge of container number and container representation includes:
A. first 4 are letters, and latter 7 are numbers.
B. letter is in the left side of image, and number is on the right side of letter or lower section.
C. character is adjacent two-by-two, and in a certain range, left and right character possesses similar height and the ratio of width to height to standoff distance.
D.11 bit number forms character line feature, and the central point of character can substantially form straight line.
E. the 11st word can be calculated by first 10 it is found that the 11st is to check character by container code rule
Symbol.
According to above-mentioned knowledge, the three-level decision rule that we design are as follows:
1) tri- rule of a, b, c, first rule of thumb score threshold and the ratio of width to height threshold value, preliminary screening character targets are based on
Rectangle frame then looks for the upper leftmost letter of image, as first member of case number (CN), is put into C', then successively seeks to the right
It looks for, the member that interval is less than reasonable threshold value is put into C'；
2) rule-based d fits straight line α with all members in C', traverses the member c' in C'kIf it
It is greater than certain threshold value with a distance from straight line α, then is rejected from subset C', straight line β can be fitted from the member of C' again；
3) rule-based e, if C' be unsatisfactory for " container number check code " rule, by complete or collected works C apart from character line β
C' is added in all characters that distance is less than threshold value, guarantees that the character number in C' is no less than 11, it is straight to traverse all assembled schemes
To 11 case number (CN)s by the inspection of coding rule, this result is exported as container number.
(4) evaluation of result
When 11 dimension character all correct identification in piece image, case number (CN) just calculates correct identification, uses according to this standard
It is proposed that the accuracy that is obtained on test set of number identification method be 97%.
Claims (6)
1. a kind of container number identification method of wide-angle perspective distortion, which comprises the following steps:
1) perspective transform pretreatment is carried out to container representation；
2) character locating and identification of depth convolutional neural networks are constructed；
3) case number (CN) identification is carried out based on cascade decision tree.
2. a kind of container number identification method of wide-angle perspective distortion according to claim 1, which is characterized in that institute
The step 1) stated specifically includes the following steps:
11) a pair of image shot from different perspectives for same container of selection is concentrated in container representation, wherein one is
There are the image of wide-angle perspective distortion, an image for not perspective distortion selects at least 5 pairs on two images respectively
Same place, and equation is established to each pair of same place, 9 perspective transform parameters are calculated using least square adjustment；
12) Perspective transformation model obtained according to resolving, to all container representations concentration, there are the progress of the image of perspective distortion
Correction.
3. a kind of container number identification method of wide-angle perspective distortion according to claim 2, which is characterized in that institute
In the step 11) stated, each pair of same place establishes 2 equations, specifically:
Wherein, (u, v) is there are the coordinate of target point in wide-angle perspective distortion image, and (x, y) is the figure of not perspective distortion
The coordinate of target point, a are corresponded to as in11、a12、a21、a22、a31、a32、a13、a23、a33For perspective transform parameter.
4. a kind of container number identification method of wide-angle perspective distortion according to claim 1, which is characterized in that institute
The step 2) stated specifically includes the following steps:
21) container representation after correcting perspective transform zooms to fixed size, mark by hand container number classification and
Boundary rectangle frame, and be added after small perspective transform carries out sample enhancing, it is classified as training dataset, test data set and tests
Demonstrate,prove data set；
22) depth convolutional neural networks arameter optimization is obtained to identify 10 numbers comprising " 0 "-" 9 " using training dataset
The character machining network of word, 26 of " A "-" Z " capitalization English characters and background totally 37 classes；
23) simultaneously location character, and output set C=is identified on container representation using trained character machining network
{ck}K=1....K, wherein ck={ bk,ck,sk, b is the rectangle frame for surrounding character, and c is character class, and s is score, and K is candidate word
The number of symbol.
5. a kind of container number identification method of wide-angle perspective distortion according to claim 1, which is characterized in that institute
The step 3) stated identifies container number using three-level decision rule, specifically includes the following steps:
31) according to score threshold and the ratio of width to height threshold value preliminary screening character targets rectangle frame, container representation most upper left side is searched
Letter successively search to the right after first member as case number (CN) is put into subset C', interval is less than character pitch threshold value
Member is put into subset C'；
32) all members in subset C' are fitted into straight line α, traverses the member c' in subset C'k, as member c'kFrom straight
When line α distance is greater than the threshold value of setting, then rejected from subset C', and character line β is fitted according to the member of C' again；
If 33) subset C' is unsatisfactory for container number check code rule, threshold value will be less than apart from character line β distance in complete or collected works C
All characters subset C' is added, guarantee that the character number in subset C' is no less than 11, traverse all combinations until 11
Position case number (CN) meets container number check code rule, exports this result as container number.
6. a kind of container number identification method of wide-angle perspective distortion according to claim 5, which is characterized in that institute
The priori conditions of the corresponding container number of three-level decision rule and container representation stated include:
A. before the character of container number 4 be letter, latter 7 be number；
B. letter is in the left side of container representation, and number is on the right side or lower section of letter；
C. character is adjacent two-by-two, and standoff distance is in a certain range, and left and right character possesses similar height and the ratio of width to height；
D.11 bit number forms character line feature, and the central point of character is capable of forming straight line；
E. container code rule are as follows: the 11st is to check character, and can be calculated by preceding 10 characters.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810739072.6A CN109190625B (en) | 2018-07-06 | 2018-07-06 | Large-angle perspective deformation container number identification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810739072.6A CN109190625B (en) | 2018-07-06 | 2018-07-06 | Large-angle perspective deformation container number identification method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109190625A true CN109190625A (en) | 2019-01-11 |
| CN109190625B CN109190625B (en) | 2021-09-03 |
Family
ID=64936297
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810739072.6A Active CN109190625B (en) | 2018-07-06 | 2018-07-06 | Large-angle perspective deformation container number identification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109190625B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109941885A (en) * | 2019-03-07 | 2019-06-28 | 无锡顶视科技有限公司 | A kind of container number candid photograph and identification device and its method based on telescopic arm |
| CN110598697A (en) * | 2019-08-23 | 2019-12-20 | 上海撬动网络科技有限公司 | Container number positioning method based on thickness character positioning |
| CN110659634A (en) * | 2019-08-23 | 2020-01-07 | 上海撬动网络科技有限公司 | Container number positioning method based on color positioning and character segmentation |
| CN111161227A (en) * | 2019-12-20 | 2020-05-15 | 成都数之联科技有限公司 | Target positioning method and system based on deep neural network |
| CN111291748A (en) * | 2020-01-15 | 2020-06-16 | 广州玖峰信息科技有限公司 | Cascade distributed artificial intelligence case number identification system |
| CN111414844A (en) * | 2020-03-17 | 2020-07-14 | 北京航天自动控制研究所 | Container number identification method based on convolution cyclic neural network |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060153455A1 (en) * | 2001-08-02 | 2006-07-13 | Paceco Corp. | Method and apparatus of automated optical container code recognition with positional identification for a transfer container crane |
| CN102024144A (en) * | 2010-11-23 | 2011-04-20 | 上海海事大学 | Container number identification method |
| US20150032449A1 (en) * | 2013-07-26 | 2015-01-29 | Nuance Communications, Inc. | Method and Apparatus for Using Convolutional Neural Networks in Speech Recognition |
| CN105184275A (en) * | 2015-09-21 | 2015-12-23 | 北京中科虹霸科技有限公司 | Infrared local face key point selecting and obtaining method based on binary decision tree |
| CN105701490A (en) * | 2016-02-24 | 2016-06-22 | 上海海事大学 | Container number adaptive positioning method based on image entropy |
-
2018
- 2018-07-06 CN CN201810739072.6A patent/CN109190625B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060153455A1 (en) * | 2001-08-02 | 2006-07-13 | Paceco Corp. | Method and apparatus of automated optical container code recognition with positional identification for a transfer container crane |
| CN102024144A (en) * | 2010-11-23 | 2011-04-20 | 上海海事大学 | Container number identification method |
| US20150032449A1 (en) * | 2013-07-26 | 2015-01-29 | Nuance Communications, Inc. | Method and Apparatus for Using Convolutional Neural Networks in Speech Recognition |
| CN105184275A (en) * | 2015-09-21 | 2015-12-23 | 北京中科虹霸科技有限公司 | Infrared local face key point selecting and obtaining method based on binary decision tree |
| CN105701490A (en) * | 2016-02-24 | 2016-06-22 | 上海海事大学 | Container number adaptive positioning method based on image entropy |
Non-Patent Citations (1)
| Title |
|---|
| GUO-RONG HUANG ET AL.: "Study on content-independent feature matching systems for container images", 《 2013 INTERNATIONAL CONFERENCE ON SYSTEM SCIENCE AND ENGINEERING (ICSSE)》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109941885A (en) * | 2019-03-07 | 2019-06-28 | 无锡顶视科技有限公司 | A kind of container number candid photograph and identification device and its method based on telescopic arm |
| CN109941885B (en) * | 2019-03-07 | 2024-07-26 | 上海安维尔信息科技股份有限公司 | Container number snapshot and recognition device and method based on telescopic arm |
| CN110598697A (en) * | 2019-08-23 | 2019-12-20 | 上海撬动网络科技有限公司 | Container number positioning method based on thickness character positioning |
| CN110659634A (en) * | 2019-08-23 | 2020-01-07 | 上海撬动网络科技有限公司 | Container number positioning method based on color positioning and character segmentation |
| CN111161227A (en) * | 2019-12-20 | 2020-05-15 | 成都数之联科技有限公司 | Target positioning method and system based on deep neural network |
| CN111291748A (en) * | 2020-01-15 | 2020-06-16 | 广州玖峰信息科技有限公司 | Cascade distributed artificial intelligence case number identification system |
| CN111414844A (en) * | 2020-03-17 | 2020-07-14 | 北京航天自动控制研究所 | Container number identification method based on convolution cyclic neural network |
| CN111414844B (en) * | 2020-03-17 | 2023-08-29 | 北京航天自动控制研究所 | Container number identification method based on convolutional neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109190625B (en) | 2021-09-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109190625A (en) | A kind of container number identification method of wide-angle perspective distortion | |
| CN111325203B (en) | American license plate recognition method and system based on image correction | |
| CN110543878B (en) | Pointer instrument reading identification method based on neural network | |
| CN113160192B (en) | Visual sense-based snow pressing vehicle appearance defect detection method and device under complex background | |
| CN113591967B (en) | Image processing method, device, equipment and computer storage medium | |
| CN104615986B (en) | The method that pedestrian detection is carried out to the video image of scene changes using multi-detector | |
| CN109543753B (en) | License plate recognition method based on self-adaptive fuzzy repair mechanism | |
| CN105608456A (en) | Multi-directional text detection method based on full convolution network | |
| CN113449594A (en) | Multilayer network combined remote sensing image ground semantic segmentation and area calculation method | |
| CN105809121A (en) | Multi-characteristic synergic traffic sign detection and identification method | |
| CN102629322B (en) | Character feature extraction method based on stroke shape of boundary point and application thereof | |
| CN106446896A (en) | Character segmentation method and device and electronic equipment | |
| CN113888536B (en) | Printed matter double image detection method and system based on computer vision | |
| CN112949338A (en) | Two-dimensional bar code accurate positioning method combining deep learning and Hough transformation | |
| CN110599463B (en) | Tongue image detection and positioning algorithm based on lightweight cascade neural network | |
| CN108509950B (en) | Railway contact net support number plate detection and identification method based on probability feature weighted fusion | |
| CN111369526B (en) | Multi-type old bridge crack identification method based on semi-supervised deep learning | |
| CN114387592B (en) | Character positioning and identifying method under complex background | |
| CN111965197A (en) | Defect classification method based on multi-feature fusion | |
| CN111488911A (en) | Image entity extraction method based on Mask R-CNN and GAN | |
| CN113705673A (en) | Character detection method, device, equipment and storage medium | |
| CN111950556A (en) | License plate printing quality detection method based on deep learning | |
| CN112132798B (en) | Method for detecting complex background PCB mark point image based on Mini ARU-Net network | |
| KR101849933B1 (en) | Apparatus of Covariance-matching-based Model for musical-symbol recognition | |
| CN109829511B (en) | Texture classification-based method for detecting cloud layer area in downward-looking infrared image |
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 | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20220113 Address after: 908, block a, floor 8, No. 116, Zizhuyuan Road, Haidian District, Beijing 100089 Patentee after: ZHONGZI DATA CO.,LTD. Address before: 200092 Siping Road 1239, Shanghai, Yangpu District Patentee before: TONGJI University |
|
| TR01 | Transfer of patent right |