CN110378332A - A kind of container terminal case number (CN) and Train number recognition method and system - Google Patents
A kind of container terminal case number (CN) and Train number recognition method and system Download PDFInfo
- Publication number
- CN110378332A CN110378332A CN201910514876.0A CN201910514876A CN110378332A CN 110378332 A CN110378332 A CN 110378332A CN 201910514876 A CN201910514876 A CN 201910514876A CN 110378332 A CN110378332 A CN 110378332A
- Authority
- CN
- China
- Prior art keywords
- picture
- case
- image
- character
- train number
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
- 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
- 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of container terminal case number (CN) and Train number recognition method, which comprises obtains picture to be processed；The picture to be processed is scanned, character zone is obtained；The coordinate of character in the picture is detected, character area image is split；Gray proces are carried out to the character area image；Image grayscale is modified, and character after binary conversion treatment is cut is carried out to revised image；Character after the cutting is handled using neural network detection model, obtains prediction result.Using the embodiment of the present invention, the discrimination and accuracy rate of container terminal case number (CN) and license number are improved.
Description
Technical field
The present invention relates to marking of cars detection technique fields, more particularly to a kind of container terminal case number (CN) and Train number recognition
Method and system.
Background technique
Container number and license number be current harbour, in logistics transportation operation for tracking the important symbol of container,
It is played a crucial role in the situations such as harbour road junction and harbour facilities handling.In the prior art, case number (CN) and license number detection with
Identification is faced with complex background interference, text blur degradation, uncertain illumination, the diversity of font, vertical case number (CN), inclination
Etc. numerous challenges.The system that traditional OCR uses a set of pre-fixed rule, often to the shooting matter of picture when case number (CN) identifies
Amount and shooting angle have higher requirement, and recognizer degraded performance, efficiency are insufficient, nothing limited for the extraction of identification feature
Method identifies container number or license number well.This allows for dock operation personnel and needs that the time is spent to carry out handmarking's case
Number and license number, to reduce the production efficiency of harbour.
As it can be seen that in the prior art, by handmarking's case number (CN) and license number, leading to the problem that production efficiency is lower.
Summary of the invention
In view of the foregoing deficiencies of prior art, the purpose of the present invention is to provide a kind of container terminal case number (CN) and vehicles
Number recognition methods and system, improve the discrimination and accuracy rate of container terminal case number (CN) and license number.
In order to achieve the above objects and other related objects, the present invention provides a kind of container terminal case number (CN) and Train number recognition side
Method, which comprises
Obtain picture to be processed；
The picture to be processed is scanned, character zone is obtained；
The coordinate of character in the picture is detected, character area image is split；
Gray proces are carried out to the character area image；
Image grayscale is modified, and character after binary conversion treatment is cut is carried out to revised image；
Character after the cutting is handled using neural network detection model, obtains prediction result.
In a kind of implementation of the invention, the method also includes:
Case number (CN) coding formulas Y=X × 2 are setA-1, wherein A is digit value range, and X is number or the corresponding number of letter
Word code, and Y is digit code value；
Check bit is set and verifies formula
Formula is verified according to the case number (CN) coding formulas and the check bit, obtains identifying code；
According to the prediction result and the identifying code, determine whether prediction result is correct.
In a kind of implementation of the invention, the step of acquisition picture to be processed, includes:
According to preset time period, the corresponding video frame images of video in the preset time period are obtained；
Video frame images are determined as picture to be processed.
In a kind of implementation of the invention, described the step of gray proces are carried out to each video frame figure, comprising:
By gray value and image RGB conversion formula, gray proces are carried out to image；
Wherein, the conversion formula are as follows:
I=0.299 × R+0.587 × G+0.114 × B
I is gray value, and RGB is respectively the red of image, green and blue component color component value；
Image grayscale is modified by image smoothing, removes noise signal.
It is described that the picture to be processed is scanned in a kind of implementation of the invention, obtain the step of character zone
Suddenly, comprising:
Using the method for progressive scan, license plate is positioned and is divided；
It is past from a left side according to the character stroke change frequency on license plate using the strategy progressively scanned since image base
It scans pixel-by-pixel on the right side；
Reach in the stroke change frequency of judgement scanning to target line in some specific width and presets value and write down the row
Number and the stroke area beginning and end；
According to the Aspect Ratio feature precomputation license plate area of license plate, and in the region there is the row of similar features to reach and set
When determining range, determine that the region is license number region.
In a kind of implementation of the invention, it is described using neural network detection model to character after the cutting at
The step of reason, acquisition prediction result, comprising:
Neural network convolutional layer extracts characteristic sequence from the image of each input automatically, and uses recirculating network, is used for
Each frame of the characteristic sequence of convolutional layer output is predicted；
Convolution layer assembly is constructed by convolutional layer and maximum pond layer, extracts the convolution characteristic pattern of input picture；
Convert the image into the convolution eigenmatrix of particular size, and when using each column of Feature Mapping as being used as one
Between piece be input in LSTM；
Every frame predictive conversion that RNN is done is followed by classification function at sequence label, each timeslice of LSTM
Softmax exports posterior probability matrix, obtains the classification of each column output character.
The invention discloses a kind of container terminal case number (CN) and car number identification systems, comprising: server-side and client；
The server-side includes: case number (CN) identification service node, picture sample study node, picture servers node and data
Library node composition；
Wherein, case number (CN) identification service node, which provides, receives terminal uploading pictures, push picture sample training result to terminal,
More new demand servicing is pushed to terminal service；Server-side and terminal node network interaction agreement use HTTP mode, and service end node will
According to configuration information, the update status of data, to determine the frequency of propelling data to terminal；Server-side uses the more examples of nginx+
Deployment way；The picture sample study node, which provides, receives the container picture that picture servers transmit, and carries out to picture
Study and training obtain identification model, and model file is pushed to case number (CN) identification service node；The picture servers are for depositing
The container picture that terminal uploads is stored up, picture servers use multiple instances deployment mode；The database node provides system
Parameter configuration, data storage service；
The client includes: that case Train number recognition service node and case Train number recognition are formed using end；
Wherein, case Train number recognition service node is by the service of case Train number recognition, case Train number recognition engine, picture synchronization, data
Update four component compositions；The case Train number recognition service is as service entrance, for providing FUZZY H TTP service, for specific
User calls, and the relevant parameter for needing to need when the picture identified and identification is provided when calling；The case Train number recognition
Engine is the real time picture identification engine installed in bridging device terminal, provides application by providing HTTP for upper-level system；
When case Train number recognition client is called, after the completion of the picture to be identified is identified, what which asynchronous can identify this
Picture is uploaded in the picture storage server in background server, is learnt for picture sample study node to do to the sample；
Data more New Parent provides server-side and issues using update.
As described above, a kind of container terminal case number (CN) provided in an embodiment of the present invention and Train number recognition method and system, mention
For utilizing neural network algorithm, the operating mode of cerebral neuron is simulated, the edge for carrying out object first with low layer neuron is special
Sign is extracted, and in the basic configuration goods local shape using middle layer neuron identification object, is finally carried out using high-rise neuron
Case number (CN) and Train number recognition.Effectively raise the discrimination and accuracy rate of container terminal case number (CN) and license number.
Detailed description of the invention
Fig. 1 is a kind of flow diagram provided in an embodiment of the present invention；
Fig. 2 is the first effect diagram provided in an embodiment of the present invention；
Fig. 3 is second of effect diagram provided in an embodiment of the present invention；
Fig. 4 is the third effect diagram provided in an embodiment of the present invention；
Fig. 5 is the 4th kind of effect diagram provided in an embodiment of the present invention；
Fig. 6 is the 5th kind of effect diagram provided in an embodiment of the present invention；
Fig. 7 is the 6th kind of effect diagram provided in an embodiment of the present invention；
Fig. 8 is the 7th kind of effect diagram provided in an embodiment of the present invention；
Fig. 9 is the 8th kind of effect diagram provided in an embodiment of the present invention.
Specific embodiment
Illustrate embodiments of the present invention below by way of specific specific example, those skilled in the art can be by this specification
Other advantages and efficacy of the present invention can be easily understood for disclosed content.The present invention can also pass through in addition different specific realities
The mode of applying is embodied or practiced, the various details in this specification can also based on different viewpoints and application, without departing from
Various modifications or alterations are carried out under spirit of the invention.
Please refer to Fig. 1-9.It should be noted that only the invention is illustrated in a schematic way for diagram provided in the present embodiment
Basic conception, only shown in schema then with related component in the present invention rather than component count, shape when according to actual implementation
Shape and size are drawn, when actual implementation kenel, quantity and the ratio of each component can arbitrarily change for one kind, and its component cloth
Office's kenel may also be increasingly complex.
As shown in Figure 1, embodiment provides a kind of container terminal case number (CN) and Train number recognition method, the side when present invention
Method includes:
S101 obtains picture to be processed.
It is understood that obtaining to the container and truck figure during dock work, truck transport being collected into etc.
Piece.Specifically, since an identification image may obtain the photo of multiple shootings, and a face (two sides) often has multiple
Picture.This kind of picture is spliced into a complete picture for by the way of panoramic mosaic by the present invention.
S102 is scanned the picture to be processed, obtains character zone.
The characteristics of according to China's license plate, using the method for progressive scan, license plate is positioned and is divided.
S103, the coordinate of detection character in the picture, character area image is split.
It is past from a left side according to the character stroke change frequency on license plate using the strategy progressively scanned since image base
It scans pixel-by-pixel on the right side；It presets value α when certain row stroke change frequency reaches in some specific width and writes down the line number and the pen
The beginning and end in partition domain；Further according to the Aspect Ratio feature precomputation license plate area of license plate, when the region has similar spy
The row of sign, which reaches setting range β then, indicates that the region is license number region；Normalized through on the license plate of over-segmentation Chinese character,
English and Arabic numerals.Noise reduction and Shape correction are carried out to Chinese character.
S104 carries out gray proces to the character area image.
Of the invention obtains container picture or video to be identified in the specific implementation, passing through, for processing speed and identification
The considerations of effect, carries out gray proces to image.The corresponding relationship of gray value and image RGB are I=0.299 × R+0.587 × G
+ 0.114 × B, wherein R, G, B are 3 color component values of color image.Adaptive Thresholding is recycled to carry out gray level image
Binary conversion treatment.Since neural network requires the discrete value of input 0 or 1, binary conversion treatment is carried out to image, while removing image
Neutralize the noise of background similar gray value.
Container terminal case number (CN) and Train number recognition OCR deep learning system of the invention is adopted as shown in Fig. 2, including one
The case number (CN) and Train number recognition OCR deep learning system established to the key technology based on neural network, wherein utilize described
Deep learning method to be collected into dock work, truck transport etc. during container and Jia Ka picture carry out feature mention
It takes, merge, text filed detection and localization finally carries out identification verification and output to text.
Container picture or video to be identified are obtained, picture (video) each region will be scanned, literal field is detected
The coordinate of domain in the picture, character area image is split, and is used for next text detection.As shown in Figure 3, Figure 4, side
Frame is the image-region detected.Detection module is built using neural network, and model mainly consists of three parts, and feature is taken out
It takes, Fusion Features, positions coordinate measurement.Feature extraction uses convolutional neural networks, and Fusion Features use deconvolution+up-sampling,
It positions coordinate measurement and uses full Connection Neural Network.By the coordinate navigated to, character area to be identified is cut out, such as Fig. 5 institute
Show.Using picture as the input of Text region module, neural network is built using CNN+RNN+CTC, is known by end-to-end training
Other random length word content, corresponding text is as a result in each picture of final output.The first picture for such as inputting Fig. 5 will
Return to 561209 as a result.The regular correction verification module verification for being carried out example 4 is returned the result, container number verification rule is met
Data will export.
S105 is modified image grayscale, and carries out character after binary conversion treatment is cut to revised image.
Image grayscale is modified by image smoothing, eliminates or reduce to the greatest extent noise.Due to pictograph master
Low frequency range is appeared in, and the noise of image and false contouring tend to occur at high frequency region, it can it is realized and is schemed by low frequency filtering
As smooth.
Detection model is built using neural network.Case number (CN) region is positioned based on semantic segmentation algorithm.Model mainly includes spy
Sign extraction, Fusion Features and positioning coordinate measurement；Picture (video) each region is scanned, detection character area is being schemed
Coordinate as in, character area image is split；According to coordinate of the text in picture, text to be identified is cut out.
Random length text is identified by end-to-end training using identification model.
S106 is handled character after the cutting using neural network detection model, obtains prediction result.
BP neural network identification process:
In the bottom of neural network, convolutional layer extracts characteristic sequence from the image of each input automatically.In convolutional network
On, a recirculating network is constructed, each frame of the characteristic sequence for exporting to convolutional layer is predicted.Using turning for top
Every frame prediction of circulation layer is converted sequence label by record layer.Assuming that input picture size is (32,100,3), the image referred to
It is all [Height, Width, Channel]
(1) convolutional layer CNN characteristic sequence extracts
Convolution layer assembly is constructed with maximum pond layer by using the convolutional layer in the CNN model of standard, extracts input
The image that size is (32,100,3) is converted (1,25,512) size by the Convolutional feature maps of image
Convolution eigenmatrix.Each column of Feature map are used as to be input in LSTM as a timeslice.If feature map
Size is m*T (m=512, T=25 in Fig. 6).All since t=1, i.e. 1≤t≤T is defined as time series t in step 2It is wherein each to be classified as
(2) circulation layer sequence labelling
Here circulation layer is the two-way LSTM network of a deep layer, and stack shape deep layer is utilized on the basis of convolution feature
Bi-directional configuration continues to extract word sequence feature.Such as Fig. 6, since the defeated feature map of convolutional layer is that (1,25,512) is big
It is small, (there are 25 time inputs, each input x for RNN maximum time length T=25tThere is D=512).
(3) it transcribes, every frame predictive conversion that RNN is done is at sequence label.Each timeslice of LSTM is followed by
Softmax, output y is a posterior probability matrix, is defined as y=(y1, y2..., yT), wherein each be classified as
Wherein n represents the character set length for needing to identify, andArgmax0 behaviour is carried out to each column of y
Make, can be obtained the classification of each column output character.
The detection process of railway carriage text includes:
Complex background interference, text blur degradation, unpredictable is faced with for container terminal case number (CN) and Train number recognition
Illumination, the factor of many factors such as font multiplicity is thought, case number (CN) is vertical, inclination influence, using CTPN in natural scene
Text is detected, and has stronger robustness.
(1) it uses VGG16 to extract feature as base net, obtains the feature of conv5_3 as feature map, greatly
Small is W × H × C；
(2) sliding window is then done on this feature map, window size is 3 × 3.I.e. each window can obtain one
A length is the feature vector of 3 × 3 × C.This feature vector will be used to predict offset distance between 10 anchor,
That is each window center can predict 10 text propsoal.
(3) feature (W*3*3*C) of the corresponding 3*3*C of all windows of every a line is input in RNN (BLSTM), is obtained
To the output of W*256；
(4) W*256 of RNN is input to the fc layer of 512 dimensions；
(5) fc layers of feature are input to three classification or return in layer.2k scores indicates the classification letter of k anchor
Breath (be character or be not character).2k vertical coordinate and third k side-refinement is used to return k
The location information of a anchor.2k vertical coordinate is because anchor's is that (y is sat for the height of center
Mark) and height two of rectangle frame values indicate, so one with 2k output.This part k side-refinement is main
Two endpoints for refine line of text, expression be each proposal horizontal translation amount.Return the box come out such as
The red elongate rectangular of those in figure, their width is certain.The text proposal of dense prediction is obtained, uses one
The non-maxima suppression algorithm of a standard filters out extra box.
(6) with simple line of text construction algorithm, proposal (the elongated square in upper figure for the text that classification is obtained
Shape) it is merged into line of text.
Whether alarm according to the rules plus dangerous material mark for container.It, can be right after client connects TOS system
Whether each container note is dangerous material case.If what is recorded in system is dangerous material case, but after passing through identification container picture,
It was found that being identified without dangerous material, then alarm.Further, it is also possible in conjunction with business datum (TOS operation management system), further progress industry
Business management case number (CN) and license number verification, such as by STOWAGE PLAN case, operation technique, truck state (loaded vehicle empty wagons etc.) verifies case number (CN) and vehicle
Number, while also may determine that and checking possible manual work maloperation.
The embodiment of the present invention utilizes neural network algorithm, simulates the operating mode of cerebral neuron, first with low layer nerve
Member carries out the Edge Gradient Feature of object, in the basic configuration goods local shape using middle layer neuron identification object, finally
Case number (CN) and Train number recognition are carried out using high-rise neuron.
The present invention compares with prior art, and effect is actively apparent.The present invention effectively raises container code
The discrimination and accuracy rate of head case number and license number.
In a kind of implementation of the invention, further includes: the then output for meeting verification rule does not meet the defeated of verification rule
Prompt information out.Container number verification rule is as follows: the corresponding number 0-9 of the digital 0-9 of case number (CN), and letter corresponds to A=10B
=12C=13D=14E=15F=16G=17H=18I=19J=20K=21L=23M=24N=25O=26P=27Q=
28R=29S=30T=31U=32V=34W=35X=36Y=37Z=38.Case number (CN) coding formulas Y=X × 2A-1Wherein A is
Digit value range is (0,10), and X is number or the corresponding digital code of letter, and Y is digit code value.Check bitSuch as when identifying that case number (CN) is the container of CBHU3202732, then E=4061%11=2, just
It is the 11st of this case number (CN).The case number (CN) is exported when calculated check code is consistent with the check code identified, is otherwise refused
Output.Therefore, can determine whether prediction result is correct according to the prediction result and the identifying code.
Case identification in left and right is generally used for the case where two chests identify together.As shown in fig. 7, requiring to do under the scene
Distinguished to good corresponding 6 bit digital of 4 English of the case number (CN) detected from the background and 1 bit check position by Picture Coordinate positioning and
Splicing.Splice box refers to the case where a pile empty van is put together.The identification of splice box with left and right case, by Picture Coordinate position into
Row is distinguished and splicing case number (CN) exports result queue.
Illustratively, left and right case and splice box (case group) identification function.Case identification in left and right is generally used for two containers one
Play the scene of identification.4 English of case number (CN) for accomplishing to detect from the background and corresponding 6 bit digital and 1 bit check are required under the scene
Position is distinguished and is spliced by Picture Coordinate positioning.Splice box refers to the case where a pile empty van is put together.For splice box
Identification with left and right case identify, by Picture Coordinate positioning distinguish and splice case number (CN) output result queue.
Further present invention mainly solves the identification of single case, the identification of more casees, weather colour of sky bring unfavorable factor image more
Head, which is shot in the technological difficulties such as the picture splicing of unified container and case number (CN) and Train number recognition, there is multiple containers or license number
The ballot technology interfered and the alarm technique whether there is or not danger signal.Whether warning function is primarily referred to as to container according to rule
Surely it alarms plus risk identification.Whether after client connects TOS system, can record to each container is dangerous material case.
If what is recorded in system is dangerous material case, but marked by discovery after identification container picture without dangerous material, then alarms, such as scheme
Shown in 8.
The invention discloses a kind of container terminal case number (CN) and car number identification systems, including system to be divided into server-side and industry control
Generator terminal (client).As shown in figure 9, server-side learns node, picture servers section by case number (CN) identification service node, picture sample
Point and database node composition.Wherein case number (CN) identification service node, which provides, receives terminal uploading pictures, push picture sample training
As a result to terminal, push more new demand servicing to terminal etc. is serviced.Server-side and terminal node network interaction agreement use HTTP mode,
Servicing end node will be according to configuration information, the update status of data, to determine the frequency of propelling data to terminal.Server-side uses
The deployment way of the more examples of nginx+ guarantees that system high efficiency is available.Picture sample learns node and provides reception picture servers biography
The container picture come learn to picture and training obtains identification model, and model file is pushed to case number (CN) identification service
Node.Picture servers are used to store the container picture of terminal upload, and picture servers use multiple instances deployment mode, guarantee
The High Availabitity of service, it is ensured that data are not lost.Database node provides the parameter configuration of system, the service such as data storage.
Client is mainly made of case Train number recognition service node and case Train number recognition using end.Wherein, case Train number recognition
Service node is made of four components such as the service of case Train number recognition, case Train number recognition engine, synchronous, the data updates of picture.Boxcar
Number identification service is used as service entrance, provides FUZZY H TTP service, calls for specific user, provides when calling and need to identify
Picture and the relevant parameter that needs of when identification；Case Train number recognition engine is the installation in bridging device terminal (client)
Real time picture identify engine.Application is provided by providing HTTP for upper-level system；When case Train number recognition client is called,
After the completion of the picture to be identified is identified, the component can the asynchronous picture for identifying this be uploaded to the picture in background server
In storage server, learn for picture sample study node to be done to the sample；Data more New Parent offer server-side, which issues, answers
With update, parameter adjustment, the data such as picture learning outcome are used for client.Client is facilitated to carry out automatic updating, deployment
Operation.The use end of case Train number recognition is the operation program (calling recognizer by interface) of client oneself.
The above-described embodiments merely illustrate the principles and effects of the present invention, and is not intended to limit the present invention.It is any ripe
The personage for knowing this technology all without departing from the spirit and scope of the present invention, carries out modifications and changes to above-described embodiment.Cause
This, institute is complete without departing from the spirit and technical ideas disclosed in the present invention by those of ordinary skill in the art such as
At all effect modifications and changes, should be covered by the claims of the present invention.
Claims (7)
1. a kind of container terminal case number (CN) and Train number recognition method, which is characterized in that the described method includes:
Obtain picture to be processed；
The picture to be processed is scanned, character zone is obtained；
The coordinate of character in the picture is detected, character area image is split；
Gray proces are carried out to the character area image；
Image grayscale is modified, and character after binary conversion treatment is cut is carried out to revised image；
Character after the cutting is handled using neural network detection model, obtains prediction result.
2. a kind of container terminal case number (CN) according to claim 1 and Train number recognition method, which is characterized in that the method
Further include:
Case number (CN) coding formulas Y=X × 2 are setA-1, wherein A is digit value range, and X is number or letter corresponding digital generation
Code, and Y is digit code value；
Check bit is set and verifies formula
Formula is verified according to the case number (CN) coding formulas and the check bit, obtains identifying code；
According to the prediction result and the identifying code, determine whether prediction result is correct.
3. a kind of container terminal case number (CN) according to claim 1 and Train number recognition method, which is characterized in that the acquisition
The step of picture to be processed includes:
According to preset time period, the corresponding video frame images of video in the preset time period are obtained；
Video frame images are determined as picture to be processed.
4. a kind of container terminal case number (CN) according to claim 1 and Train number recognition method, which is characterized in that described to every
The step of one video frame figure carries out gray proces, comprising:
By gray value and image RGB conversion formula, gray proces are carried out to image；
Wherein, the conversion formula are as follows:
I=0.299 × R+0.587 × G+0.114 × B
I is gray value, and RGB is respectively the red of image, green and blue component color component value；
Image grayscale is modified by image smoothing, removes noise signal.
5. a kind of container terminal case number (CN) according to claim 1 and Train number recognition method, which is characterized in that described to institute
The step of stating picture to be processed to be scanned, obtaining character zone, comprising:
Using the method for progressive scan, license plate is positioned and is divided；
Using since image base progressively scan strategy, according to the character stroke change frequency on license plate, from left to right by
Picture element scan；
Judgement scanning to target line stroke change frequency some specific width reach preset value write down the line number with
The beginning and end of the stroke area；
According to the Aspect Ratio feature precomputation license plate area of license plate, and in the region there is the row of similar features to reach setting model
When enclosing, determine that the region is license number region.
6. a kind of container terminal case number (CN) according to claim 1 and Train number recognition method, which is characterized in that the use
The step of neural network detection model handles character after the cutting, obtains prediction result, comprising:
Neural network convolutional layer extracts characteristic sequence from the image of each input automatically, and uses recirculating network, for volume
Each frame of the characteristic sequence of lamination output is predicted；
Convolution layer assembly is constructed by convolutional layer and maximum pond layer, extracts the convolution characteristic pattern of input picture；
The convolution eigenmatrix of particular size is converted the image into, and using each column of Feature Mapping as a timeslice
It is input in LSTM；
Every frame predictive conversion that RNN is done is followed by classification function softmax at sequence label, each timeslice of LSTM,
Posterior probability matrix is exported, the classification of each column output character is obtained.
7. a kind of container terminal case number (CN) and car number identification system, which is characterized in that the system comprises: server-side and client
End；
The server-side includes: case number (CN) identification service node, picture sample study node, picture servers node and database section
Point composition；
Wherein, case number (CN) identification service node, which provides, receives terminal uploading pictures, push picture sample training result to terminal, push
More new demand servicing is to terminal service；Server-side and terminal node network interaction agreement use HTTP mode, service end node for basis
Configuration information, the update status of data, to determine the frequency of propelling data to terminal；Server-side uses the portion of the more examples of nginx+
Management side formula；The picture sample study node, which provides, receives the container picture that picture servers transmit, and learns to picture
Identification model is obtained with training, model file is pushed to case number (CN) identification service node；The picture servers are for storing end
The container picture uploaded is held, picture servers use multiple instances deployment mode；The database node provides the parameter of system
Configuration, data storage service；
The client includes: that case Train number recognition service node and case Train number recognition are formed using end；
Wherein, case Train number recognition service node is updated by the service of case Train number recognition, case Train number recognition engine, picture synchronization, data
Four component compositions；The case Train number recognition service is as service entrance, for providing FUZZY H TTP service, for specifically using
Person calls, and the relevant parameter for needing to need when the picture identified and identification is provided when calling；The case Train number recognition engine
It is the real time picture identification engine installed in bridging device terminal, provides application by providing HTTP for upper-level system；Work as case
When Train number recognition client is called, after the completion of the picture to be identified is identified, which can the asynchronous picture for identifying this
It is uploaded in the picture storage server in background server, learns for picture sample study node to be done to the sample；Data
More New Parent provides server-side and issues using update.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910514876.0A CN110378332A (en) | 2019-06-14 | 2019-06-14 | A kind of container terminal case number (CN) and Train number recognition method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910514876.0A CN110378332A (en) | 2019-06-14 | 2019-06-14 | A kind of container terminal case number (CN) and Train number recognition method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110378332A true CN110378332A (en) | 2019-10-25 |
Family
ID=68250321
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910514876.0A Pending CN110378332A (en) | 2019-06-14 | 2019-06-14 | A kind of container terminal case number (CN) and Train number recognition method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110378332A (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110909745A (en) * | 2019-12-24 | 2020-03-24 | 新疆九洲熏蒸消毒有限责任公司 | Train disinfection channel identification system |
| CN110991279A (en) * | 2019-11-20 | 2020-04-10 | 北京灵伴未来科技有限公司 | Document image analysis and recognition method and system |
| CN111039186A (en) * | 2019-12-18 | 2020-04-21 | 中国铁路上海局集团有限公司 | Method for identifying license plate number and vehicle number of container gantry crane |
| CN111144309A (en) * | 2019-12-27 | 2020-05-12 | 苏州海管家物流科技有限公司 | Automatic identification method for container door image |
| 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 |
| CN111833302A (en) * | 2020-06-05 | 2020-10-27 | 中铁武汉勘察设计研究院有限公司 | Car number and case number identification association system |
| CN111860690A (en) * | 2020-07-31 | 2020-10-30 | 河北交投智能交通技术有限责任公司 | Dangerous goods vehicle detection and identification method based on deep learning |
| CN112070087A (en) * | 2020-09-14 | 2020-12-11 | 成都主导软件技术有限公司 | Train number identification method and device with end position and readable storage medium |
| CN112209197A (en) * | 2020-10-30 | 2021-01-12 | 深兰人工智能芯片研究院（江苏）有限公司 | Elevator fault detection method and device |
| CN112580738A (en) * | 2020-12-25 | 2021-03-30 | 特赞（上海）信息科技有限公司 | AttentionOCR text recognition method and device based on improvement |
| CN112651952A (en) * | 2020-12-31 | 2021-04-13 | 上海振华重工(集团)股份有限公司 | Container number and vehicle head top number identification method and system for container terminal |
| CN113011423A (en) * | 2019-12-21 | 2021-06-22 | 北京师范大学珠海分校 | Text line structure optimization calculation method based on CTPN system and application thereof |
| CN113378646A (en) * | 2021-05-18 | 2021-09-10 | 上海平奥供应链管理有限公司 | Freight train information identification system and identification method |
| CN113873442A (en) * | 2021-09-08 | 2021-12-31 | 宁波大榭招商国际码头有限公司 | External hub card positioning method |
| CN113971811A (en) * | 2021-11-16 | 2022-01-25 | 北京国泰星云科技有限公司 | Intelligent container feature identification method based on machine vision and deep learning |
| CN114299408A (en) * | 2021-12-31 | 2022-04-08 | 哪吒港航智慧科技(上海)有限公司 | Identification method and system for tire crane number box number, storage medium and terminal |
| CN117037158A (en) * | 2023-10-09 | 2023-11-10 | 之江实验室 | Urban brain cloud edge cooperative computing method and device based on video semantic driving |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101398894A (en) * | 2008-06-17 | 2009-04-01 | 浙江师范大学 | Automobile license plate automatic recognition method and implementing device thereof |
| CN102024144A (en) * | 2010-11-23 | 2011-04-20 | 上海海事大学 | Container number identification method |
| CN103473121A (en) * | 2013-08-20 | 2013-12-25 | 西安电子科技大学 | Mass image parallel processing method based on cloud computing platform |
| CN105554052A (en) * | 2015-09-25 | 2016-05-04 | 深圳市美贝壳科技有限公司 | Server photograph information remote synchronous modification method |
| CN105677763A (en) * | 2015-12-29 | 2016-06-15 | 华南理工大学 | Image quality evaluating system based on Hadoop |
| CN106203239A (en) * | 2015-05-04 | 2016-12-07 | 杭州海康威视数字技术股份有限公司 | Information processing method, device and system for container tally |
| CN106446891A (en) * | 2016-08-29 | 2017-02-22 | 昆山邦泰汽车零部件制造有限公司 | Automatic identification method of motor vehicle license plate |
| CN109447078A (en) * | 2018-10-23 | 2019-03-08 | 四川大学 | A kind of detection recognition method of natural scene image sensitivity text |
-
2019
- 2019-06-14 CN CN201910514876.0A patent/CN110378332A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101398894A (en) * | 2008-06-17 | 2009-04-01 | 浙江师范大学 | Automobile license plate automatic recognition method and implementing device thereof |
| CN102024144A (en) * | 2010-11-23 | 2011-04-20 | 上海海事大学 | Container number identification method |
| CN103473121A (en) * | 2013-08-20 | 2013-12-25 | 西安电子科技大学 | Mass image parallel processing method based on cloud computing platform |
| CN106203239A (en) * | 2015-05-04 | 2016-12-07 | 杭州海康威视数字技术股份有限公司 | Information processing method, device and system for container tally |
| CN105554052A (en) * | 2015-09-25 | 2016-05-04 | 深圳市美贝壳科技有限公司 | Server photograph information remote synchronous modification method |
| CN105677763A (en) * | 2015-12-29 | 2016-06-15 | 华南理工大学 | Image quality evaluating system based on Hadoop |
| CN106446891A (en) * | 2016-08-29 | 2017-02-22 | 昆山邦泰汽车零部件制造有限公司 | Automatic identification method of motor vehicle license plate |
| CN109447078A (en) * | 2018-10-23 | 2019-03-08 | 四川大学 | A kind of detection recognition method of natural scene image sensitivity text |
Non-Patent Citations (1)
| Title |
|---|
| 赵显劲等: ""集装箱材料与配件手册"", 《中国铁道出版社》 * |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110991279A (en) * | 2019-11-20 | 2020-04-10 | 北京灵伴未来科技有限公司 | Document image analysis and recognition method and system |
| CN110991279B (en) * | 2019-11-20 | 2023-08-22 | 北京灵伴未来科技有限公司 | Document Image Analysis and Recognition Method and System |
| CN111039186A (en) * | 2019-12-18 | 2020-04-21 | 中国铁路上海局集团有限公司 | Method for identifying license plate number and vehicle number of container gantry crane |
| CN113011423A (en) * | 2019-12-21 | 2021-06-22 | 北京师范大学珠海分校 | Text line structure optimization calculation method based on CTPN system and application thereof |
| CN110909745A (en) * | 2019-12-24 | 2020-03-24 | 新疆九洲熏蒸消毒有限责任公司 | Train disinfection channel identification system |
| CN111144309A (en) * | 2019-12-27 | 2020-05-12 | 苏州海管家物流科技有限公司 | Automatic identification method for container door image |
| CN111144309B (en) * | 2019-12-27 | 2023-06-20 | 苏州海管家物流科技有限公司 | Automatic identification method for container door image |
| 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 |
| CN111833302A (en) * | 2020-06-05 | 2020-10-27 | 中铁武汉勘察设计研究院有限公司 | Car number and case number identification association system |
| CN111860690A (en) * | 2020-07-31 | 2020-10-30 | 河北交投智能交通技术有限责任公司 | Dangerous goods vehicle detection and identification method based on deep learning |
| CN111860690B (en) * | 2020-07-31 | 2024-05-28 | 河北交投智能科技股份有限公司 | Dangerous goods vehicle detection and identification method based on deep learning |
| CN112070087B (en) * | 2020-09-14 | 2023-06-02 | 成都主导软件技术有限公司 | Train number identification method and device with end bit and readable storage medium |
| CN112070087A (en) * | 2020-09-14 | 2020-12-11 | 成都主导软件技术有限公司 | Train number identification method and device with end position and readable storage medium |
| CN112209197A (en) * | 2020-10-30 | 2021-01-12 | 深兰人工智能芯片研究院（江苏）有限公司 | Elevator fault detection method and device |
| CN112580738A (en) * | 2020-12-25 | 2021-03-30 | 特赞（上海）信息科技有限公司 | AttentionOCR text recognition method and device based on improvement |
| CN112651952A (en) * | 2020-12-31 | 2021-04-13 | 上海振华重工(集团)股份有限公司 | Container number and vehicle head top number identification method and system for container terminal |
| CN113378646A (en) * | 2021-05-18 | 2021-09-10 | 上海平奥供应链管理有限公司 | Freight train information identification system and identification method |
| CN113873442A (en) * | 2021-09-08 | 2021-12-31 | 宁波大榭招商国际码头有限公司 | External hub card positioning method |
| CN113971811A (en) * | 2021-11-16 | 2022-01-25 | 北京国泰星云科技有限公司 | Intelligent container feature identification method based on machine vision and deep learning |
| CN113971811B (en) * | 2021-11-16 | 2023-06-06 | 北京国泰星云科技有限公司 | Container characteristic intelligent recognition method based on machine vision and deep learning |
| CN114299408A (en) * | 2021-12-31 | 2022-04-08 | 哪吒港航智慧科技(上海)有限公司 | Identification method and system for tire crane number box number, storage medium and terminal |
| CN117037158A (en) * | 2023-10-09 | 2023-11-10 | 之江实验室 | Urban brain cloud edge cooperative computing method and device based on video semantic driving |
| CN117037158B (en) * | 2023-10-09 | 2024-01-09 | 之江实验室 | Urban brain cloud edge cooperative computing method and device based on video semantic driving |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110378332A (en) | A kind of container terminal case number (CN) and Train number recognition method and system | |
| CN109886210A (en) | A kind of traffic image recognition methods, device, computer equipment and medium | |
| CN106897681B (en) | Remote sensing image contrast analysis method and system | |
| CN108765325A (en) | A kind of small drone Restoration method of blurred image | |
| CN104463241A (en) | Vehicle type recognition method in intelligent transportation monitoring system | |
| CN110555464A (en) | Vehicle color identification method based on deep learning model | |
| US11495022B2 (en) | Method for recognizing an object of a mobile unit | |
| CN111967313A (en) | Unmanned aerial vehicle image annotation method assisted by deep learning target detection algorithm | |
| CN111027538A (en) | Container detection method based on instance segmentation model | |
| CN108648210B (en) | Rapid multi-target detection method and device under static complex scene | |
| CN110991447A (en) | Train number accurate positioning and identification method based on deep learning | |
| Xu et al. | Convolutional neural network based traffic sign recognition system | |
| CN114519819B (en) | Remote sensing image target detection method based on global context awareness | |
| CN115601741A (en) | Non-motor vehicle retrograde detection incremental learning and license plate recognition method | |
| CN112053407B (en) | Automatic lane line detection method based on AI technology in traffic law enforcement image | |
| Wang et al. | Multiscale traffic sign detection method in complex environment based on YOLOv4 | |
| CN108509826A (en) | A kind of roads recognition method and its system of remote sensing image | |
| CN112001873B (en) | Data generation method based on container X-ray image | |
| CN109284752A (en) | A kind of rapid detection method of vehicle | |
| CN111091165A (en) | Fine-grained identification method of image | |
| CN115359306A (en) | Intelligent identification method and system for high-definition images of railway freight inspection | |
| CN113343977B (en) | Multipath automatic identification method for container terminal truck collection license plate | |
| CN114926456A (en) | Rail foreign matter detection method based on semi-automatic labeling and improved deep learning | |
| CN114998689A (en) | Generation method of track data set, track identification method and system | |
| Xu et al. | SPNet: Superpixel pyramid network for scene parsing |
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 |