CN105844266A - Occlusion and alteration-preventing license plate recognition system and method - Google Patents
Occlusion and alteration-preventing license plate recognition system and method Download PDFInfo
- Publication number
- CN105844266A CN105844266A CN201610406626.1A CN201610406626A CN105844266A CN 105844266 A CN105844266 A CN 105844266A CN 201610406626 A CN201610406626 A CN 201610406626A CN 105844266 A CN105844266 A CN 105844266A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- unit
- character
- license plate
- car
- 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
- 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
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/017—Detecting movement of traffic to be counted or controlled identifying vehicles
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/04—Detecting movement of traffic to be counted or controlled using optical or ultrasonic detectors
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/042—Detecting movement of traffic to be counted or controlled using inductive or magnetic detectors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/08—Detecting or categorising vehicles
-
- 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
- G06V30/28—Character recognition specially adapted to the type of the alphabet, e.g. Latin alphabet
- G06V30/293—Character recognition specially adapted to the type of the alphabet, e.g. Latin alphabet of characters other than Kanji, Hiragana or Katakana
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses an occlusion and alteration-preventing license plate recognition system and method. According to the occlusion and alteration-preventing license plate recognition system and method, in license plate recognition, judgment on omission and incompletion is carried out; the number and locations of missing characters are outputted; possible characters of incomplete characters are rendered, and the characters are sequenced according to possibility; a vehicle model recognition sub module and a vehicle color recognition sub module are started; license plate number information which is most matched with above information is extracted from an external data database according to the above information; and an obtained possible license plate number is outputted. With the system and method adopted, judgment and processing are performed on completely occluded and altered characters and partially occluded and altered characters, so that the accuracy of license plate recognition can be improved; and a visible light active light source, a visible light camera, a near-infrared active light source and a near-infrared camera are arranged in an image acquisition module, so that images with higher accuracy and higher clarity in passive and active imaging can be obtained, and a license plate number region can be displayed in a highlighted manner, and high-quality images can be provided for the improvement of the accuracy of the license plate recognition.
Description
Technical field
The present invention relates to intelligent transportation field, be specifically related to the anti-Vehicle License Plate Recognition System blocked and alter and method.
Background technology
Car plate is that vehicle is specifically numbered, there is character arrangements feature, the information having affiliated vehicle is contained in car plate color and characters on license plate color, content, position, and this information includes the information such as type and affiliated province, city, the county of affiliated vehicle, by the owner of car plate this vehicle of lockable.Car license recognition can be applicable to the occasions such as parking lot fee collection management, traffic flux measurement, vehicle location, hypervelocity automatic monitoring, toll station.
Life has a large amount of false-trademark, pseudo-board, the phenomenon of deck, affect the interests of car owner, and existing Vehicle License Plate Recognition System and method are generally only capable of judging that characters on license plate is the most complete, there may be false-trademark, pseudo-board and the phenomenon of deck, but the scope of characters on license plate can not be reduced further, providing the license plate number that this car plate is possible, the accuracy causing Car license recognition is low, not exclusively can be suitably used for the occasions such as parking lot fee collection management, traffic flux measurement, vehicle location, hypervelocity automatic monitoring, toll station.
Summary of the invention
The present invention provides the anti-Vehicle License Plate Recognition System blocked and alter and method, solves the problem that Car license recognition accuracy rate is low.
The present invention solves the problems referred to above by the following technical programs:
The anti-Vehicle License Plate Recognition System blocked and alter, same as the prior art is, including vehicle detection module, image capture module and controller module, the outfan of vehicle detection module is connected with the input of image capture module, the outfan of image capture module is connected with controller module, unlike the prior art, described controller module includes Car license recognition submodule, vehicle cab recognition submodule and car colour recognition submodule；Described Car license recognition submodule is made up of Character segmentation unit, judging unit, screening unit, sequencing unit and output unit；The input of Character segmentation unit is connected with the outfan of image capture module, and the outfan of Character segmentation unit is connected with the input of judging unit；No. one outfan of judging unit is connected with the input of screening unit, it is judged that another road outfan of unit is connected with a road input of output unit；The outfan of screening unit is connected with the input of sequencing unit；No. one outfan of sequencing unit is connected with the control end of vehicle cab recognition submodule, and another road outfan of sequencing unit is connected with the control end of car colour recognition submodule, and another road outfan of sequencing unit is connected with another road input of output end unit；The outfan of vehicle cab recognition submodule is connected with another road input of output unit；The outfan of car colour recognition submodule is connected with the another road input of output unit；Output unit is also connected with external data base；Vehicle cab recognition submodule with car colour recognition submodule respectively outfan with image information collecting module be connected.
Further, described vehicle detection module is video detector.
Further, described image information collecting system includes near-infrared video camera, near-infrared active light source, visible light camera and visible ray active light source；Near-infrared video camera is horizontally or vertically connected installation, near-infrared active light source and near-infrared camera horizon or right angle setting with visible light camera, it is seen that light video camera and visible ray active light source are horizontally or vertically installed；Having vehicle daytime through out-of-date, near-infrared active light source starts, and near-infrared camera operation is in Active Imaging state, it is seen that light active light source is closed, it is seen that light camera operation is in imaging and passive imaging state；Having vehicle evening through out-of-date, near-infrared active light source and near-infrared video camera are turned off, it is seen that light active light source works, it is seen that light camera operation is in Active Imaging state.
Further, the character in car plate part in image is split by image information collecting system output image to Character segmentation unit, Character segmentation unit, inputs segmentation content to judging unit；Judging unit first determines whether whether character quantity meets national regulation, if any gaps and omissions, illustrate car plate has character changed by screening completely, quantity and the position of the character changed by screening completely are inputted to output unit, as without gaps and omissions, then judges to remain whether character exists incompleteness, as without incomplete, then by inputting to output unit without incomplete character, if any incompleteness, then the incomplete quantity of character, position and stroke are moved towards input to screening unit；Screening unit is according to all possible character of stroke trend estimate of incomplete character, and inputs characters into sequencing unit；Sequencing unit carries out probability size sequence to prediction character, inputs the prediction character being ranked up by probability size to output unit, starts vehicle cab recognition submodule and car colour recognition submodule simultaneously；Vehicle cab recognition submodule identifies the vehicle of this vehicle, inputs vehicle result to output unit；Car colour recognition submodule identifies the color of vehicle, inputs car color to output unit；Output unit is according to the position at the character place changed by screening completely, the incomplete character of nothing identified and position, each prediction character of incomplete character and position, the vehicle of vehicle and color, the license plate number perfectly correlated with above-mentioned information information of vehicles is extracted from external data base, compare with above-mentioned license plate number, verify, export most probable license plate number.
Further, the identification step of described vehicle cab recognition module is:
1) image is split and pretreatment, by image cropping specification, then carry out rim detection and morphological operation；
2) image after morphological operation is carried out shape and texture feature extraction, area pixel number, average, standard deviation, third moment, entropy form 5 dimensional feature vectors, carry out HOG feature extraction；The HOG characteristic vector of images all in training sample being constituted matrix, carries out PCA dimensionality reduction, the dimension after dimensionality reduction combines above-mentioned 5 dimensions, forms new feature vector, uses LDA algorithm to calculate scatter matrix S in its classwAnd scatter matrix S between classb, take matrix Sw -1SbFront 4 characteristic vectors as the projection matrix of LDA proper subspace；
3) determine that parameter combines, use Radial basis kernel function, construct three one-against-one devices, and construct directed acyclic graph, carry out vehicle classification by the method for decision-making directed acyclic graph.
Further, the identification step of described car colour recognition module is:
1) determine that the parameter in network is: input layer node number, the number of output layer node, initial weight, threshold value, the hidden layer number of plies, each hidden layer node number, learning rate, error function type and excitation function type；
2) input layer number is 3；Output layer node number is 9,；In network, excitation function is Sigmoid type action function；Error function is error of sum square function；Initial weight is the different little random number between-1 to 1；Threshold value is given by user；The hidden layer number of plies is 1；Each hidden layer node number is 5；Trend according to error change dynamically changes learning rate, for adaptively changing learning rate；
3) it is attached power correction according to the gradient direction before the moment and current gradient descent direction.
Further, step 2) described in the method for adaptively changing learning rate be: initial learn rate is set, if error increases after an iteration, then learning rate is multiplied by a constant less than 1, if error reduces after an iteration, then learning coefficient are multiplied by a constant more than 1.
Advantages of the present invention with effect is:
1, when carrying out Car license recognition, carry out gaps and omissions and incomplete judgement respectively, gaps and omissions character is exported its quantity and position, incomplete character is provided possible character, and arrange according to probability size, and start vehicle cab recognition submodule and car colour recognition submodule, according to above-mentioned information, the license plate number information mated the most with above-mentioned information is extracted, each possible license plate number output that will obtain from external data base；The method all carries out judging and processing for hiding completely to change and partly hide the character changed, and improves the accuracy rate of Car license recognition；
2, visible ray active light source, visible light camera, near-infrared active light source and near-infrared video camera are set in image capture module, near-infrared video camera is horizontally or vertically connected installation with visible light camera, near-infrared active light source and near-infrared camera horizon or right angle setting, visible light camera and visible ray active light source are horizontally or vertically installed, definition, accuracy higher image can be obtained when obtaining active imaging and passive imaging, the license plate number region being especially highlighted, provides qualitative picture for improving the accuracy rate of Car license recognition.
Accompanying drawing explanation
Fig. 1 is present configuration theory diagram.
Detailed description of the invention
Below in conjunction with embodiment, the invention will be further described, but the invention is not limited in these embodiments.
The anti-Vehicle License Plate Recognition System blocked and alter, same as the prior art is, all include vehicle detection module, image capture module and controller module, the outfan of vehicle detection module is connected with the input of image capture module, the outfan of image capture module is connected with controller module, unlike the prior art, all it is correspondingly improved in each module.
Vehicle detection module uses one or more in ElectroMagnetic Induction coil, ultrasound wave, laser, dynamic weighing and video image, for detecting whether there is vehicle pass-through, testing result is sensitive, just can avoid judging by accident, just be avoided that image capture module error starting.
The method of electromagnetic induction coil detection vehicle is: lay electromagnetic induction coil under highway, passing to high frequency electric, when vehicle passes through, the metal contained by vehicle produces eddy current in electromagnetic induction coil, electromagnetic induction coil inductance value is reduced, thus obtains traffic flow signal.The method of above-mentioned detection vehicle, when being used for detecting the traffic parameter such as traffic flow, occupation rate, accuracy rate is higher, has an advantage in that low cost, weatherproof, and shortcoming is that mobility is poor, be easily subject to damage, the life-span is short.
The method of ultrasound examination vehicle is: launched high frequency waves by supersonic generator, moving vehicle return with the frequency of change, record frequecy characteristic by transducer, thus carry out vehicle detection.The method of above-mentioned detection vehicle, it is adaptable to the vehicle detection of highway, have an advantage in that length in service life, removable, assumes convenient, shortcoming is that reflected signal instability, accuracy of detection are poor and the most affected by environment.
The method of laser detection vehicle is: with pulse laser as medium, carry out detecting distance information by the reflection results in Laser Measurement face, obtains the three-dimensional profile shape of vehicle according to different distance, carries out vehicle detection and vehicle cab recognition., there is signal disturbing between similar laser detector in the method for above-mentioned detection vehicle, it is adaptable in highway and Fare Collection System, but equipment cost is too high, can produce reception error in data, reduces the reliability of system, but also pollutes the environment.
The method of dynamic weighing detection vehicle is: when vehicle passes through, detector stress produces deformation, vehicle is detected by back information, measure the axle weight of vehicle, wheelbase, gross weight, speed etc., and by the vehicle classification table pre-established, automatically identifying vehicle, have an advantage in that the consuming time is few, efficiency is high, its shortcoming is that equipment installs complexity, the life-span is short, precision is the most impacted.
The method of Video Detection vehicle is: video detector has detected when vehicle passes through, information gathering is carried out by collecting vehicle information module, controller analysis video image, from video flowing, image is extracted by photographic head, effective information is extracted from image, vehicle detection is carried out according to information, have an advantage in that simple installation, removable, detection range big, obtain contain much information, noiseless between pollution-free, video detector, its shortcoming is that requirement of real-time is high, and under complex background, the accuracy rate of vehicle detection is difficult to reach practical degree.
User can according to use occasion and actual demand, select in above-mentioned 5 kinds of vehicle detection modes one or more.The present invention is preferably, the method selecting Video Detection vehicle, i.e. vehicle detection module is video detector, and the outfan of video detector is connected with the input of image capture module, by control signal input to image capture module, control startup and the closedown of image capture module.
Collecting vehicle information module includes near-infrared video camera, near-infrared active light source, visible light camera and visible ray active light source composition.Near-infrared video camera is horizontally or vertically connected installation, near-infrared active light source and near-infrared camera horizon or right angle setting with visible light camera, it is seen that light video camera and visible ray active light source are horizontally or vertically installed.Having vehicle daytime through out-of-date, near-infrared active light source starts, and near-infrared camera operation is in Active Imaging state, it is seen that light active light source is closed, it is seen that light camera operation is in imaging and passive imaging state；Having vehicle evening through out-of-date, near-infrared active light source and near-infrared video camera are turned off, it is seen that light active light source works, it is seen that light camera operation is in Active Imaging state.The mode of operation switching of active imaging and passive imaging, the when of contributing to carrying out License Plate with evening by day, gets highlighted car plate position.Active Imaging can overcome external illumination effect, it is achieved being highlighted of license plate area in image, vehicle body, other backgrounds and screening change the low bright display of material.Video detector has detected that vehicle passes through, signal will be triggered send to collecting vehicle information module, start visible light camera and/or near-infrared video camera, during daytime, what collecting vehicle information module exported is a group of near-infrared, visible images of close visual field, during evening, the output of collecting vehicle information module is a secondary visible images.
The secondary visible images of above-mentioned one group of near-infrared, visible images or one is delivered to Car license recognition submodule, by Car license recognition submodule, acquired image is carried out Car license recognition, as required, also acquired image is inputted to vehicle cab recognition submodule and car colour recognition submodule, carry out vehicle and the identification of car color.Having in actual life to hide and change the phenomenon of car plate, certain character portion is changed by screening or certain character is changed by screening completely and the present invention is directed to the screening of single character and change situation and process；When car plate occurs that multiple screening changes phenomenon, relevant character is identified by the present invention one by one.
Controller module includes Car license recognition submodule, vehicle cab recognition submodule and car colour recognition submodule；Car license recognition submodule is made up of Character segmentation unit, judging unit, screening unit, sequencing unit and output unit；The input of Character segmentation unit is connected with the outfan of image capture module, and the outfan of Character segmentation unit is connected with the input of judging unit；No. one outfan of judging unit is connected with the input of screening unit, it is judged that another road outfan of unit is connected with a road input of output unit；The outfan of screening unit is connected with the input of sequencing unit；No. one outfan of sequencing unit is connected with the control end of vehicle cab recognition submodule, and another road outfan of sequencing unit is connected with the control end of car colour recognition submodule, and another road outfan of sequencing unit is connected with another road input of output end unit；The outfan of vehicle cab recognition submodule is connected with another road input of output unit；The outfan of car colour recognition submodule is connected with the another road input of output unit；Output unit is also connected with external data base；Vehicle cab recognition submodule with car colour recognition submodule respectively outfan with image information collecting module be connected.
Image inputs to Character segmentation unit, and the character in car plate part in image is split by Character segmentation unit, inputs segmentation content to judging unit；Judging unit first determines whether whether character quantity meets national regulation, if any gaps and omissions, illustrate car plate has character changed by screening completely, quantity and the position of the character changed by screening completely are inputted to output unit, as without gaps and omissions, then judges to remain whether character exists incompleteness, as without incomplete, then by inputting to output unit without incomplete character, if any incompleteness, then the incomplete quantity of character, position and stroke are moved towards input to screening unit；Screening unit is according to all possible character of stroke trend estimate of incomplete character, and inputs characters into sequencing unit；Sequencing unit carries out probability size sequence to prediction character, inputs the prediction character being ranked up by probability size to output unit, starts vehicle cab recognition submodule and car colour recognition submodule simultaneously；Vehicle cab recognition submodule identifies the vehicle of this vehicle, inputs vehicle result to output unit；Car colour recognition submodule identifies the color of vehicle, inputs car color to output unit；Output unit is according to the position at the character place changed by screening completely, the incomplete character of nothing identified and position, each prediction character of incomplete character and position, the vehicle of vehicle and color, the license plate number perfectly correlated with above-mentioned information information of vehicles is extracted from external data base, compare with above-mentioned license plate number, verify, export most probable license plate number.
Vehicle cab recognition submodule obtains after vehicle image, carries out vehicle cab recognition according to the following steps:
1) image is split and pretreatment, by image cropping specification, then carry out rim detection and morphological operation；
2) image after morphological operation is carried out shape and texture feature extraction, area pixel number, average, standard deviation, third moment, entropy form 5 dimensional feature vectors, carry out HOG feature extraction；The HOG characteristic vector of images all in training sample being constituted matrix, carries out PCA dimensionality reduction, the dimension after dimensionality reduction combines above-mentioned 5 dimensions, forms new feature vector, uses LDA algorithm to calculate scatter matrix S in its classwAnd scatter matrix S between classb, take matrix Sw -1SbFront 4 characteristic vectors as the projection matrix of LDA proper subspace；
3) determine that parameter combines, use Radial basis kernel function, construct three one-against-one devices, and construct directed acyclic graph, carry out vehicle classification by the method for decision-making directed acyclic graph.
Car colour recognition submodule obtains after vehicle pictures, carries out car colour recognition by following steps:
1) parameter in network is determined: input layer node number, the number of output layer node, initial weight, threshold value, the hidden layer number of plies, each hidden layer node number, learning rate, error function type and excitation function type；
2) input layer number is 3, respectively R, G, B feature；Output layer node number is 9, respectively red, green, blue, Huang, orange, silver, black, white and grey；In network, excitation function is Sigmoid type action function；Error function is error of sum square function；Initial weight is the different little random number between-1 to 1, and little random number is used for ensureing that network will not enter local minimum point in advance because weights are excessive；Threshold value is given by user；The hidden layer number of plies is 1；Choosing of each hidden layer node number, if number of network nodes is very few, unit networks, by not setting up judgement circle of complexity, makes network training the most out, or the network of training is the strongst, the sample not having before can not identifying, poor fault tolerance and number of network nodes is too much, the complexity of network can be increased, cause learning time long, also can reduce the Generalization Ability of network, and train and be easily trapped into local minimum point and can not get globe optimum, node in hidden layer optimal in the present invention is 5；The increase of learning rate can make error increase, and makes iterations increase, and affects convergence rate, and the present invention dynamically changes learning rate, i.e. adaptively changing learning rate according to the trend of error change；
3) if only adjusting by the gradient descent direction of moment error, and do not account for the gradient direction before the moment, shake during being easily caused training, convergence is slowly, for accelerating convergence and preventing concussion, when connection weight being corrected every time, by a certain percentage plus front once learn time correcting value, i.e. increase momentum arithmetic, during (n+1)th iteration, the adjustment amount of weights is relevant to during nth iteration, owing between each sample of sample cluster, dependency is very strong, the learning outcome of previous sample, used by next sample, so can accelerate convergence rate.
The method of above-mentioned adaptively changing learning rate: first set an initial learn rate, if error increases after an iteration, then learning rate is multiplied by a constant less than 1, if error reduces after an iteration, then learning coefficient are multiplied by a constant more than 1, the most neither increase too many amount of calculation, make again learning rate reasonably be adjusted.Here, not regularized learning algorithm rate after a sample learning, but regularized learning algorithm rate again after the study once of whole training sample database, i.e. use batch processing mode regularized learning algorithm rate.
In the present invention, vehicle detection module uses video detector, is used for starting image capture module and carries out image acquisition；Image capture module is provided with visible ray active light source, visible light camera, near-infrared active light source and near-infrared video camera, for obtaining active imaging and passive imaging by day, obtaining Active Imaging at night, no matter how environment changes, and all can get the most highlighted vehicle image；Vehicle image is separately input into Car license recognition submodule, when character each in license plate number all can identify, the most directly export this license plate number, otherwise start vehicle cab recognition submodule and car colour recognition submodule carries out vehicle and car colour recognition, stroke trend, vehicle and the color of the character that bound fraction is blocked, from external data base, transfer the information of vehicles coincideing the most with features described above, all possible license board information is exported, thus improve Car license recognition accuracy.
Claims (7)
- The most anti-Vehicle License Plate Recognition System blocked and alter, including vehicle detection module, image acquisition mould Block and controller module, the outfan of vehicle detection module is connected with the input of image capture module, The outfan of image capture module is connected with controller module, it is characterised in that:Described controller module includes Car license recognition submodule, vehicle cab recognition submodule and car colour recognition Submodule；Described Car license recognition submodule is by Character segmentation unit, judging unit, screening unit, sequence list Unit and output unit form；The outfan phase of the input of Character segmentation unit and image capture module Even, the outfan of Character segmentation unit is connected with the input of judging unit；One tunnel of judging unit is defeated Go out end to be connected with the input of screening unit, it is judged that another road outfan of unit and the one of output unit Road input is connected；The outfan of screening unit is connected with the input of sequencing unit；Sequencing unit No. one outfan is connected with the control end of vehicle cab recognition submodule, another road outfan of sequencing unit with The control end of car colour recognition submodule is connected, another road outfan of sequencing unit and output end unit Another road input be connected；The outfan of vehicle cab recognition submodule inputs with another road of output unit End is connected；The outfan of car colour recognition submodule is connected with the another road input of output unit；Defeated Go out unit to be also connected with external data base；Vehicle cab recognition submodule and car colour recognition submodule respectively with The outfan of image information collecting module is connected.
- The anti-Vehicle License Plate Recognition System blocked and alter the most according to claim 1, its feature exists In: described vehicle detection module is video detector.
- The anti-Vehicle License Plate Recognition System blocked and alter the most according to claim 1, its feature exists In:Described image information collecting system includes near-infrared video camera, near-infrared active light source, visible ray Video camera and visible ray active light source；Near-infrared video camera is horizontally or vertically connected installation, near-infrared active light source with visible light camera With near-infrared camera horizon or right angle setting, it is seen that light video camera and visible ray active light source level or Right angle setting；Having vehicle daytime through out-of-date, near-infrared active light source starts, and near-infrared camera operation is actively Image formation state, it is seen that light active light source is closed, it is seen that light camera operation is in imaging and passive imaging state；Having vehicle evening through out-of-date, near-infrared active light source and near-infrared video camera are turned off, it is seen that light Active light source works, it is seen that light camera operation is in Active Imaging state.
- 4. based on the anti-Car license recognition system blocked and alter described in any one in claim 1-3 The recognition methods of system, it is characterised in that be accomplished by:Image information collecting system output image is to Character segmentation unit, and Character segmentation unit is by image Character in car plate part is split, and inputs segmentation content to judging unit；Judging unit first determines whether whether character quantity meets national regulation, if any gaps and omissions, car plate is described In have character to be changed by screening completely, the quantity of the character changed by screening completely and position are inputted to exporting list Unit, as without gaps and omissions, then judges to remain whether character exists incompleteness, as incomplete in nothing, then by without incompleteness Character inputs to output unit, if any incompleteness, then by the incomplete quantity of character, position and stroke trend Input is to screening unit；Screening unit is according to all possible character of stroke trend estimate of incomplete character and character is defeated Enter to sequencing unit；Sequencing unit carries out probability size sequence to prediction character, will be ranked up by probability size Prediction character input to output unit, start vehicle cab recognition submodule and car colour recognition submodule simultaneously Block；Vehicle cab recognition submodule identifies the vehicle of this vehicle, inputs vehicle result to output unit；Car colour recognition submodule identifies the color of vehicle, inputs car color to output unit；Output unit according to the position at the character place changed by screening completely, identify without incomplete character and Its position, each prediction character of incomplete character and position, the vehicle of vehicle and color, from outward Portion data base extracts the license plate number perfectly correlated with above-mentioned information information of vehicles, enters with above-mentioned license plate number Row comparison, verification, export most probable license plate number.
- The anti-licence plate recognition method blocking and altering the most according to claim 4, its feature exists In, the identification step of described vehicle cab recognition module is:1) image is split and pretreatment, by image cropping specification, then carry out rim detection And morphological operation；2) image after morphological operation is carried out shape and texture feature extraction, by area pixel number, Average, standard deviation, third moment, entropy form 5 dimensional feature vectors, carry out HOG feature extraction；Will In training sample, the HOG characteristic vector of all images constitutes matrix, carries out PCA dimensionality reduction, after dimensionality reduction Dimension combine above-mentioned 5 dimensions, form new feature vector, use LDA algorithm to calculate in its class and spread Matrix SwAnd scatter matrix S between classb, take matrix Sw -1SbFront 4 characteristic vectors as LDA feature The projection matrix of subspace；3) determine that parameter combines, use Radial basis kernel function, construct three one-against-one devices, and Structure directed acyclic graph, carries out vehicle classification by the method for decision-making directed acyclic graph.
- The anti-licence plate recognition method blocking and altering the most according to claim 5, its feature exists In, the identification step of described car colour recognition module is:1) determine that the parameter in network is: input layer node number, the number of output layer node, just Beginning weights, threshold value, the hidden layer number of plies, each hidden layer node number, learning rate, error function class Type and excitation function type；2) input layer number is 3；Output layer node number is 9,；Excitation function in network For Sigmoid type action function；Error function is error of sum square function；Initial weight arrives for-1 Different little random number between 1；Threshold value is given by user；The hidden layer number of plies is 1；Each hidden layer is tied Point number is 5；Trend according to error change dynamically changes learning rate, learns for adaptively changing Rate；3) it is attached weighing school according to the gradient direction before the moment and current gradient descent direction Just.
- The anti-licence plate recognition method blocking and altering the most according to claim 6, its feature exists In:Step 2) described in the method for adaptively changing learning rate be: initial learn rate is set, if After an iteration, error increases, then learning rate is multiplied by a constant less than 1, if an iteration Rear error reduces, then learning coefficient are multiplied by a constant more than 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610406626.1A CN105844266A (en) | 2016-06-12 | 2016-06-12 | Occlusion and alteration-preventing license plate recognition system and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610406626.1A CN105844266A (en) | 2016-06-12 | 2016-06-12 | Occlusion and alteration-preventing license plate recognition system and method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105844266A true CN105844266A (en) | 2016-08-10 |
Family
ID=56575755
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610406626.1A Pending CN105844266A (en) | 2016-06-12 | 2016-06-12 | Occlusion and alteration-preventing license plate recognition system and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105844266A (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106331656A (en) * | 2016-10-26 | 2017-01-11 | 桂林理工大学 | License plate assisted positioning and tracking device with visible light communication vehicle lamp |
| CN107292291A (en) * | 2017-07-19 | 2017-10-24 | 北京智芯原动科技有限公司 | A kind of vehicle identification method and system |
| CN107977596A (en) * | 2016-10-25 | 2018-05-01 | 杭州海康威视数字技术股份有限公司 | A kind of car plate state identification method and device |
| CN108182808A (en) * | 2017-12-28 | 2018-06-19 | 广东傲智创新科技有限公司 | A kind of movable type car plate data collecting system |
| CN108205897A (en) * | 2016-12-16 | 2018-06-26 | 北京迪科达科技有限公司 | A kind of car plate analysis and identification system of vehicle |
| CN110335473A (en) * | 2019-08-14 | 2019-10-15 | 中国联合网络通信集团有限公司 | Block the personal identification method and system of license plate vehicle |
| CN110472633A (en) * | 2019-08-15 | 2019-11-19 | 南京拓控信息科技股份有限公司 | A kind of detection of train license number and recognition methods based on deep learning |
| CN111144390A (en) * | 2019-12-20 | 2020-05-12 | 武汉瑞纳捷电子技术有限公司 | Missing character searching method and system in license plate recognition and electronic equipment |
| US10691968B2 (en) | 2018-02-08 | 2020-06-23 | Genetec Inc. | Systems and methods for locating a retroreflective object in a digital image |
| CN111460186A (en) * | 2020-03-31 | 2020-07-28 | 河北工业大学 | Method for establishing database containing vehicle visible light images and infrared images |
| CN113674262A (en) * | 2021-08-26 | 2021-11-19 | 安徽天诚交通工程有限公司 | Vehicle trajectory tracking method and system |
| US11188776B2 (en) | 2019-10-26 | 2021-11-30 | Genetec Inc. | Automated license plate recognition system and related method |
| CN115116152A (en) * | 2022-06-21 | 2022-09-27 | 江苏重行信息科技有限公司 | ZigBee-based intelligent parking lot charging management system |
Citations (4)
| 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 |
| CN102156862A (en) * | 2011-05-06 | 2011-08-17 | 杨星 | License plate recognition system and license plate recognition method preventing blocking and altering |
| CN104408931A (en) * | 2014-10-29 | 2015-03-11 | 合肥指南针电子科技有限责任公司 | Incomplete sign license plate identification system and method |
| CN105185123A (en) * | 2015-10-08 | 2015-12-23 | 南京信息工程大学 | Fake plate vehicle recognition system and method |
-
2016
- 2016-06-12 CN CN201610406626.1A patent/CN105844266A/en active Pending
Patent Citations (4)
| 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 |
| CN102156862A (en) * | 2011-05-06 | 2011-08-17 | 杨星 | License plate recognition system and license plate recognition method preventing blocking and altering |
| CN104408931A (en) * | 2014-10-29 | 2015-03-11 | 合肥指南针电子科技有限责任公司 | Incomplete sign license plate identification system and method |
| CN105185123A (en) * | 2015-10-08 | 2015-12-23 | 南京信息工程大学 | Fake plate vehicle recognition system and method |
Non-Patent Citations (2)
| Title |
|---|
| 杨丹: "基于BP神经网络的汽车颜色识别", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 陈明明: "视频中的车辆检测及车型识别研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107977596A (en) * | 2016-10-25 | 2018-05-01 | 杭州海康威视数字技术股份有限公司 | A kind of car plate state identification method and device |
| CN106331656A (en) * | 2016-10-26 | 2017-01-11 | 桂林理工大学 | License plate assisted positioning and tracking device with visible light communication vehicle lamp |
| CN106331656B (en) * | 2016-10-26 | 2023-08-15 | 桂林理工大学 | License plate-assisted visible light communication car lamp positioning and tracking device |
| CN108205897A (en) * | 2016-12-16 | 2018-06-26 | 北京迪科达科技有限公司 | A kind of car plate analysis and identification system of vehicle |
| CN107292291A (en) * | 2017-07-19 | 2017-10-24 | 北京智芯原动科技有限公司 | A kind of vehicle identification method and system |
| CN107292291B (en) * | 2017-07-19 | 2020-04-03 | 北京智芯原动科技有限公司 | Vehicle identification method and system |
| CN108182808A (en) * | 2017-12-28 | 2018-06-19 | 广东傲智创新科技有限公司 | A kind of movable type car plate data collecting system |
| US11367267B2 (en) | 2018-02-08 | 2022-06-21 | Genetec Inc. | Systems and methods for locating a retroreflective object in a digital image |
| US11830256B2 (en) | 2018-02-08 | 2023-11-28 | Genetec Inc. | Systems and methods for locating a retroreflective object in a digital image |
| US10691968B2 (en) | 2018-02-08 | 2020-06-23 | Genetec Inc. | Systems and methods for locating a retroreflective object in a digital image |
| CN110335473A (en) * | 2019-08-14 | 2019-10-15 | 中国联合网络通信集团有限公司 | Block the personal identification method and system of license plate vehicle |
| CN110335473B (en) * | 2019-08-14 | 2021-05-11 | 中国联合网络通信集团有限公司 | Identity recognition method and system for vehicle with license plate covered |
| CN110472633A (en) * | 2019-08-15 | 2019-11-19 | 南京拓控信息科技股份有限公司 | A kind of detection of train license number and recognition methods based on deep learning |
| US11188776B2 (en) | 2019-10-26 | 2021-11-30 | Genetec Inc. | Automated license plate recognition system and related method |
| US12067743B2 (en) | 2019-10-26 | 2024-08-20 | Genetec Inc. | Automated license plate recognition system and related method |
| US12125234B2 (en) | 2019-10-26 | 2024-10-22 | Genetec Inc. | Automated license plate recognition system and related method |
| CN111144390A (en) * | 2019-12-20 | 2020-05-12 | 武汉瑞纳捷电子技术有限公司 | Missing character searching method and system in license plate recognition and electronic equipment |
| CN111144390B (en) * | 2019-12-20 | 2023-10-27 | 武汉瑞纳捷半导体有限公司 | Missing character searching method and system in license plate recognition and electronic equipment |
| CN111460186B (en) * | 2020-03-31 | 2022-04-08 | 河北工业大学 | Method for establishing database containing vehicle visible light images and infrared images |
| CN111460186A (en) * | 2020-03-31 | 2020-07-28 | 河北工业大学 | Method for establishing database containing vehicle visible light images and infrared images |
| CN113674262A (en) * | 2021-08-26 | 2021-11-19 | 安徽天诚交通工程有限公司 | Vehicle trajectory tracking method and system |
| CN115116152A (en) * | 2022-06-21 | 2022-09-27 | 江苏重行信息科技有限公司 | ZigBee-based intelligent parking lot charging management system |
| CN115116152B (en) * | 2022-06-21 | 2023-09-05 | 肇庆市停容易停车管理有限公司 | Intelligent parking lot charging management system based on ZigBee |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105844266A (en) | Occlusion and alteration-preventing license plate recognition system and method | |
| CN113111974B (en) | Vision-laser radar fusion method and system based on depth canonical correlation analysis | |
| Chadwick et al. | Distant vehicle detection using radar and vision | |
| CN112380952B (en) | Power equipment infrared image real-time detection and identification method based on artificial intelligence | |
| CN107918753B (en) | Processing Method of Point-clouds and device | |
| CN109948661A (en) | A kind of 3D vehicle checking method based on Multi-sensor Fusion | |
| US9552524B2 (en) | System and method for detecting seat belt violations from front view vehicle images | |
| CN111079640B (en) | Vehicle type identification method and system based on automatic amplification sample | |
| CN108983219A (en) | A kind of image information of traffic scene and the fusion method and system of radar information | |
| CN111274976A (en) | Lane detection method and system based on multi-level fusion of vision and laser radar | |
| CN105574552A (en) | Vehicle ranging and collision early warning method based on monocular vision | |
| IL271568B1 (en) | Rare instance classifiers | |
| CN103902976A (en) | Pedestrian detection method based on infrared image | |
| CN108875754B (en) | Vehicle re-identification method based on multi-depth feature fusion network | |
| CN107985189A (en) | Towards driver's lane change Deep Early Warning method under scorch environment | |
| CN109815884A (en) | Unsafe driving behavioral value method and device based on deep learning | |
| CN111523352A (en) | Method for intelligently and rapidly identifying illegal modified vehicle and monitoring system thereof | |
| CN115690714A (en) | Multi-scale road target detection method based on area focusing | |
| CN106295605A (en) | Traffic lights detection and recognition methods | |
| CN111523401B (en) | Method for recognizing vehicle type | |
| CN112949510A (en) | Human detection method based on fast R-CNN thermal infrared image | |
| CN111898427A (en) | Multispectral pedestrian detection method based on feature fusion deep neural network | |
| Shafie et al. | Smart video surveillance system for vehicle detection and traffic flow control | |
| Balci et al. | Front-View Vehicle Make and Model Recognition on Night-Time NIR Camera Images | |
| Amirgaliyev et al. | License plate verification method for automatic license plate recognition systems |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20160810 |
|
| WD01 | Invention patent application deemed withdrawn after publication |