CN111325209B - License plate recognition method and system - Google Patents
License plate recognition method and system Download PDFInfo
- Publication number
- CN111325209B CN111325209B CN201811529296.0A CN201811529296A CN111325209B CN 111325209 B CN111325209 B CN 111325209B CN 201811529296 A CN201811529296 A CN 201811529296A CN 111325209 B CN111325209 B CN 111325209B
- Authority
- CN
- China
- Prior art keywords
- license plate
- plate image
- character
- image
- standardized
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 67
- 238000012545 processing Methods 0.000 claims abstract description 40
- 238000010606 normalization Methods 0.000 claims abstract description 14
- 230000008569 process Effects 0.000 claims description 16
- 238000011049 filling Methods 0.000 claims description 13
- 238000013507 mapping Methods 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 3
- 238000011068 loading method Methods 0.000 abstract description 11
- 238000004422 calculation algorithm Methods 0.000 abstract description 3
- 238000012544 monitoring process Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 13
- 238000004590 computer program Methods 0.000 description 8
- 238000005286 illumination Methods 0.000 description 7
- 230000008901 benefit Effects 0.000 description 6
- 239000003086 colorant Substances 0.000 description 5
- 230000004913 activation Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000006386 memory function Effects 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000012015 optical character recognition Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Abstract
The invention relates to a license plate recognition method and system. The method comprises the following steps: receiving an original license plate image and processing the original license plate image to obtain a standardized license plate image; performing character positioning and judgment on the standardized license plate image based on a single detector, and performing normalization processing on the standardized license plate image by adopting a background expansion method to obtain a target license plate image; and inputting DeepOCR a model to identify the target license plate image, and obtaining target license plate character information. The invention is convenient to overcome the disadvantages of the traditional license plate recognition algorithm in the license plate recognition of the loading and unloading port, greatly improves the recognition rate of the license plate recognition, enables the loading and unloading port under the complex background to use the intelligent license plate recognition system to manage the vehicle, and greatly improves the monitoring and management capability of the vehicle in the loading and unloading port.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a license plate recognition method and system.
Background
The traditional license plate recognition system utilizes traditional image processing methods, such as image projection, morphology, gradient and the like, to extract image features, so that character segmentation is completed, and meanwhile, the characters are recognized by utilizing the features and traditional machine learning methods. Aiming at the freight loading and unloading port wagon license plate, the method has the advantages of larger disadvantages, firstly, difficult characteristic selection, larger selected workload and unstable result. The reason is that the condition of the loading and unloading port is complex and various, and the manually selected characteristics are difficult to adapt to various states. Second, the lack of data driving, insufficient generalization capability, and inability to accommodate more complex backgrounds.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide a license plate recognition method and system.
According to an aspect of the present invention, there is provided a license plate recognition method including the steps of:
Receiving an original license plate image and processing the original license plate image to obtain a standardized license plate image;
performing character positioning and judgment on the standardized license plate image based on a single detector, and performing normalization processing on the standardized license plate image by adopting a background expansion method to obtain a target license plate image;
And inputting DeepOCR a model to identify the target license plate image, and obtaining target license plate character information.
Preferably, the receiving the original license plate image and processing the original license plate image to obtain a standardized license plate image includes:
Converting the color space of the original license plate image;
And collecting and setting background image data of the original license plate image.
Preferably, the character positioning and judgment are performed on the standardized license plate image by using a trained single-time detector, which comprises the following steps:
Performing character positioning and judgment on the standardized license plate image based on a two-class method;
And judging whether the standardized license plate image is a character or not based on a single detector.
Preferably, inputting DeepOCR a model to identify the target license plate image, and obtaining target license plate character information includes:
And extracting character features of the target license plate image based on DeepOCR models, acquiring character features, processing the character features, and acquiring the target license plate information.
Preferably, the normalization processing is performed on the standardized license plate image by adopting a background expansion method to obtain a target license plate image, which comprises the following steps:
Extracting characters by a threshold method;
Setting a background filled pixel value, and filling an original license plate image;
Mapping characters to the target license plate image.
According to another aspect of the present invention, there is provided a license plate recognition system including:
The image processing unit is configured to receive the original license plate image and process the original license plate image to obtain a standardized license plate image;
The character positioning unit is configured to perform character positioning and judgment on the standardized license plate image based on the single detector, and perform normalization processing on the standardized license plate image by adopting a background expansion method to obtain a target license plate image;
And the character recognition unit is configured to input DeepOCR a model to recognize the target license plate image and obtain target license plate character information.
Preferably, the image processing unit includes:
A first image processing subunit configured to convert a color space of an original license plate image;
And the second image processing subunit is configured to acquire and set background image data of the original license plate image.
Preferably, the character positioning unit includes:
The second classification subunit is configured to perform character positioning and judgment on the standardized license plate image based on a second classification method;
and the character judging subunit is configured to judge whether the standardized license plate image is a character or not based on the single detector.
Preferably, the character recognition unit includes:
And the character feature processing subunit is configured to extract character features of the target license plate image based on the DeepOCR model, acquire character features, process the character features and acquire the target license plate information.
Preferably, the character positioning unit includes:
a character extraction module configured to extract characters using a threshold method;
The image filling module is configured to set a background filling pixel value and fill an original license plate image;
and the character mapping module is configured to map characters to the target license plate image.
Compared with the prior art, the invention has the following beneficial effects:
1. according to the license plate recognition method, an original license plate image is received and processed to obtain a standardized license plate image; performing character positioning and judgment on the standardized license plate image based on a single detector, and performing normalization processing on the standardized license plate image by adopting a background expansion method to obtain a target license plate image; and inputting DeepOCR a model to identify the target license plate image, and obtaining target license plate character information.
The method has the advantages that special design is not needed for the network, the existing available data is fully utilized, the good effect can be achieved by adopting a general network structure, the advantages of the deep convolution network are fully exerted, and the method has the advantages of being simple in design, good in robustness, high in detection accuracy and low in omission ratio. Therefore, the method can better overcome the disadvantages of the traditional license plate recognition algorithm in the license plate recognition of the loading and unloading port, greatly improve the recognition rate of the license plate recognition, and enable the loading and unloading port under a complex background to also use the intelligent license plate recognition system to manage the vehicle, thereby greatly improving the monitoring and management capability of the vehicle in the loading and unloading port.
2. The invention discloses a license plate recognition system, which comprises an image processing unit, a license plate recognition unit and a license plate recognition unit, wherein the image processing unit is configured to receive an original license plate image and process the original license plate image to obtain a standardized license plate image; the character positioning unit is configured to perform character positioning and judgment on the standardized license plate image based on the single detector, and perform normalization processing on the standardized license plate image by adopting a background expansion method to obtain a target license plate image; and the character recognition unit is configured to input DeepOCR a model to recognize the target license plate image and obtain target license plate character information.
The units cooperate with each other, so that the recognition rate of license plates is improved, and the loading and unloading ports under a complex background can also use the intelligent license plate recognition system to manage vehicles, so that the improvement of the vehicle management and monitoring capability in the loading and unloading ports is facilitated.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a block flow diagram of the training process and the operation process of the present invention.
Detailed Description
For a better understanding of the technical solution of the present invention, the present invention will be further described with reference to the following specific examples and the accompanying drawings.
Embodiment one:
the license plate recognition method of the embodiment comprises the following steps:
S1, receiving an original license plate image and processing the original license plate image to obtain a standardized license plate image;
Wherein S1 comprises:
Converting the color space of the original license plate image;
And collecting and setting background image data of the original license plate image.
Specifically, the original photographed image is an image directly obtained after photographing by using the photographing device, and may be an image photographed under a fixed application scene, or may be an image photographed under a natural application scene, generally an RGB image, that is, a true color image. According to the principle of three primary colors, each color can be composed of three primary colors of red, green and blue according to different proportions.
The basic storage structure of image data is a state in which RGB (red, green and blue) is arranged in accordance with RGB, and RGB cannot express colors well in various states, so that the color space of an image is converted into HSV (chromaticity, saturation, purity) space, specifically as follows: let (r, g, b) be the red, green and blue coordinates of one color, respectively, their values being real numbers between 0 and 1. Let max be the maximum of r, g and b. Let min be equal to the minimum of these values. To find the (h, s, v) value in HSV space, where h ε [0,360 ] is the hue angle of the angle and s, v ε [0,1] is the saturation and brightness, calculated as:
ν＝max
Wherein: yellow and blue HSV channels.
Blue: h-190, 245, S-0.35,1, V-0.3,1
Yellow: h-28, 55, S-0.35,1, V-0.3,1 ]
On the H channel, yellow and blue can be distinguished to a large extent, relatively stable.
Under the condition of dazzling light, only the value of the V channel relatively changes greatly under the same color, so that the license plate type can be well distinguished in the space based on HSV.
According to the type of the license plate, an image blurring technology is utilized, a kernel size (100 x 100) average filtering mode is utilized to solve the background, the image after the average value is differentiated from the original image, gradient stretching is conducted according to the type of the color, and finally a standardized license plate image is formed. The image gray morphology suppresses the foreground signal, highlights the background signal, and simultaneously smoothes the background signal by using an average filter, then performs subtraction by using the smoothed background signal and the original image, finally obtains the foreground signal of the image, and then normalizes the gray scale of the image to [0,255] by using gray scale normalization, and finally obtains the standardized image.
In the RGB model, if r=g=b, the color represents a gray color, wherein a value of r=g=b is called a gray value, and a process of converting the color into gray is called a graying process. The standardized license plate image at this time is that the character is the prospect that the pixel gray value is higher, and the background is blackish, and the size is 600x300, but the image noise is more. About 3000 pictures of the type are collected, meanwhile, manual labeling is carried out on the pictures, the label of the character is target, and finally, data are manufactured into a specific database.
The image needs to be normalized while the algorithm is running. The reason is that two points exist, firstly, the license plate occupies a small area of the whole image (the length of the whole image is 1280 x 960, the size of the license plate is 120 x 64), the resolution of the license plate is too low, and finally the character definition of the license plate is influenced. Secondly, license plates of the loading and unloading ports are classified according to the colors of the ground color and the characters, and the license plates are approximately classified into black ground white characters, yellow ground black characters, green ground black characters and blue ground white characters, and besides, the reflective colors are not less than 128 due to the influence of external illumination. The ground color of each state directly affects the contrast of the character, thereby affecting the quality of character segmentation.
S2, carrying out character positioning and judgment on the standardized license plate image based on a single detector, and carrying out normalization processing on the standardized license plate image by adopting a background expansion method to obtain a target license plate image;
Wherein S2 includes:
Performing character positioning and judgment on the standardized license plate image based on a two-class method;
And judging whether the standardized license plate image is a character or not based on a single detector.
Wherein S2 further comprises:
Extracting characters by a threshold method;
Setting a background filled pixel value, and filling an original license plate image;
Mapping characters to the target license plate image.
Specifically, the normalization of the image size is performed by using a background filling method, wherein the background filling refers to that after characters are extracted in a threshold mode, the average value of all pixels of the background is used as a pixel value of the background filling, and finally the pixel value is used for filling in four directions up, down, left and right until the size of a picture is consistent with the set size, and finally the foreground, namely the characters, is mapped back into the image through affine transformation, so that normalization is finally formed. Compared with the traditional interpolation mode, the method has the greatest advantage that the original character morphological characteristics are saved, and the images with the same size are obtained.
A single detector (Single Shot MultiBox Detector, hereinafter referred to as SSD model) is a model that implements target detection and identification using a single deep neural network model. In this embodiment, the SSD model needs to be trained before use. The SSD model adopts VGG-16 as a basic network, an auxiliary structure is added to form a deep learning model to position characters, and the problem of character positioning is simplified into simple two-class problems (one is a character, and the other is a background).
For character recognition and positioning tasks, it is currently more common practice to train the positioning and recognition of characters in the same task, which makes SSD networks perform both recognition and positioning tasks on a multi-class (about 35 classes) basis, and the method is typically based on color images. Because the external environment changes more, the neural network has more tolerance and more information difficulty, and finally the convergence accuracy is also influenced. In this way, the information of the image is reduced, and a standardized image is formed. Meanwhile, only the SSD is needed to complete the task of positioning the characters, so that the SSD network only needs to recognize what is the character, what is the non-character, and no concern is given to which number or character the character is. At the moment, SSD only needs to complete one classification problem, so that the precision is guaranteed, and the speed is guaranteed.
When the SSD judges whether the character is the character, the ability of the neural network to extract the characteristic is utilized, the characteristic of the character is formed into a multidimensional vector, the softmax is utilized for normalization, and the cross entropy loss function is utilized to measure whether the target is the character, wherein the mathematical expression of the softmax is as follows:
The cross entropy is expressed mathematically as follows:
the SSD model is characterized by extracting the normalized image containing the characters and the background through VGG-16, because the image has other illumination interference signals besides the characters, the illumination interference signals are supposed to belong to foreground signals and belong to one category with the characters of the image from the design thought of the normalized image, but for the characters, the illumination interference signals belong to background signals, so that contradiction occurs when the foreground and the background are defined. Meanwhile, the shapes of the same character are different from one another, if the SSD model is required to distinguish all signal differences, such as the difference between an illumination disturbance signal and a character in the foreground, the difference of each different character is insufficient in classification precision, the difference between the illumination disturbance signal and the character signal is far away through observation, VGG-16 is easier to learn the difference, so that the character and non-character (such as illumination disturbance) characteristics are extracted only by VGG-16, then whether the characteristics are similar to artificial classification is evaluated through a cross entropy function, and finally, character classification is converted into a two-class structure.
And for judging each character, after the image is manually calibrated in advance, pixels such as coordinate points (x, y) belong to the characters, and the like, an external rectangular frame containing all the pixels of the characters is obtained, then the normalized image is subjected to convolution operation by using convolution checks of different sizes, a feature map is obtained, the feature map is converted into a feature vector, the feature vector is then used for regressing an external rectangle, finally, the distance between the generated external rectangle and the artificially marked rectangle is evaluated by using the Euclidean distance, and when the two distances are in a proper range, the pixels contained in the external rectangle are the pixels of the characters. Thereby judging the position of the character.
The two-classification method can neglect the differences of the characters, can well overcome the problem of unbalanced character data distribution, and can accurately position each character finally.
And S3, inputting DeepOCR a model to identify the target license plate image, and obtaining target license plate character information.
Wherein S3 includes:
And extracting character features of the target license plate image based on DeepOCR models, acquiring character features, processing the character features, and acquiring the target license plate information.
Because the SSD model is a deep neural network model and is used for license plate character recognition, the process is complicated, and therefore a DeepOCR model is adopted for recognizing license plate images. And selecting DeepOCR networks for character classification on the acquired character images with the same size. The network has the advantages of less network parameters, high calculation speed and high accuracy.
DeepOCR is fully referred to as: deep convolution network for Optical Character Recognition is a shallow convolutional neural network, which is composed of a 7-layer network structure, and a first layer, a data layer, mainly loads character data and performs character preprocessing, wherein the preprocessing comprises scaling and mirror projection of characters. The second to sixth layers are combined layers of a convolution layer and an activation function, 5 layers are used, and characters are mainly extracted, wherein all the characters are obtained through network automatic learning and data self-driving. The last layer is a loss layer, the learned characteristics are input into a softmax function and a cross entropy function to judge the characters, and finally, the character recognition task is completed through iterative correction.
DeepOCR is a small neural network composed of three convolution layers, two pooling layers, two activation functions and a loss function, wherein the three convolution layers mainly extract edge features of each character, and then the pooling layers mainly reduce dimensions of images, so that the convolution layers can extract high-dimensional features. And the activation function mainly improves the nonlinear fitting capacity of the network, so that the network adapts to the characteristics of different characters and expresses the characteristics. The loss function is mainly used for evaluating the gap between character features generated by the network and character features which are classified artificially, and is an optimization target of the network. Through the small network, the network can learn the characteristics of each character by using a gradient descent method, and finally the network can have the character classification capability through evaluation of a loss layer.
As shown in fig. 2, deepOCR need to be trained prior to use. In the training process, the iteration condition is also needed to be judged, if the iteration condition is not met, and the iteration number is larger than 10000, the SSD is needed to be used for positioning the character again, and the iteration number is increased by one until the condition is met.
The running process is comparable to the training process, except that the transition from database input to real-time data stream input is merely omitted.
A license plate recognition system of the present embodiment includes:
The image processing unit is configured to receive the original license plate image and process the original license plate image to obtain a standardized license plate image;
The character positioning unit is configured to perform character positioning and judgment on the standardized license plate image based on the single detector, and perform normalization processing on the standardized license plate image by adopting a background expansion method to obtain a target license plate image;
And the character recognition unit is configured to input DeepOCR a model to recognize the target license plate image and obtain target license plate character information.
Characterized in that the image processing unit comprises:
A first image processing subunit configured to convert a color space of an original license plate image;
And the second image processing subunit is configured to acquire and set background image data of the original license plate image.
Preferably, the character positioning unit includes:
The second classification subunit is configured to perform character positioning and judgment on the standardized license plate image based on a second classification method;
and the character judging subunit is configured to judge whether the standardized license plate image is a character or not based on the single detector.
Preferably, the character recognition unit includes:
And the character feature processing subunit is configured to extract character features of the target license plate image based on the DeepOCR model, acquire character features, process the character features and acquire the target license plate information.
Preferably, the character positioning unit includes:
a character extraction module configured to extract characters using a threshold method;
The image filling module is configured to set a background filling pixel value and fill an original license plate image;
and the character mapping module is configured to map characters to the target license plate image.
The embodiment also provides an apparatus, including:
One or more processors;
A data storage for acquiring and storing data and one or more programs;
the input/output device is used for realizing the input/output function of the equipment;
The one or more programs, when executed by the one or more processors, cause the one or more processors to perform the methods described above.
The device is a device capable of automatically performing numerical calculation and/or information processing according to a preset or stored instruction, and its hardware includes, but is not limited to, a microprocessor, an Application SPECIFIC INTEGRATED Circuit (ASIC), a Programmable gate array (Field-Programmable GATE ARRAY, FPGA), a digital Processor (DIGITAL SIGNAL Processor, DSP), an embedded device, and the like.
The device may be any electronic product that can interact with a user in a human-computer manner, such as a Personal computer, tablet computer, smart phone, personal digital assistant (Personal DIGITAL ASSISTANT, PDA), game console, interactive internet protocol television (Internet Protocol Television, IPTV), smart wearable device, etc. The device may be a server including, but not limited to, a single web server, a server group of multiple web servers, or a Cloud Computing (Cloud Computing) based Cloud consisting of a large number of hosts or web servers, where Cloud Computing is one of distributed Computing, one super virtual computer consisting of a group of loosely coupled computer sets. The network in which the device is located includes, but is not limited to, the internet, a wide area network, a metropolitan area network, a local area network, a virtual private network (Virtual Private Network, VPN), and the like.
The memory is used for storing programs and various data of the license plate recognition method, and realizes high-speed and automatic access of the programs or the data in the running process of the equipment. The memory may be an external storage device and/or an internal storage means of the device.
Further, the Memory may be a circuit with a Memory function such as RAM (Random-Access Memory), FIFO (FIRST IN FIRST Out), etc. without physical form in the integrated circuit, or the Memory may be a Memory device with physical form such as a Memory bank, TF card (Trans-FLASH CARD), etc.
The processor may be a central processing unit (CPU, central Processing Unit). The CPU is a very large scale integrated circuit, and is an operation Core (Core) and a Control Core (Control Unit) of the device. The processor may execute an operating system of the device and various installed applications, program codes, etc., for example, execute the operating system in each module or unit in the license plate recognition system and various installed applications, program codes, etc., so as to implement the license plate recognition method.
The input-output device is mainly used for realizing input-output functions of the apparatus, such as receiving and transmitting input digital or character information, or displaying information input by a user or information provided to the user, and various menus of the apparatus.
The present embodiment provides a computer-readable storage medium storing a computer program which, when executed by a processor, implements the above-described method.
The modules/units integrated with the device may be stored in a computer readable storage medium if implemented in the form of software functional units and sold or used as a stand alone product.
Based on such understanding, the present invention may implement all or part of the flow of the method of the above embodiment, or may be implemented by a computer program to instruct related hardware, where the computer program may be stored in a computer readable storage medium, and when the computer program is executed by a processor, the computer program may implement the steps of each of the method embodiments described above. Wherein the computer program comprises computer program code which may be in source code form, object code form, executable file or some intermediate form etc. The computer readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), an electrical carrier signal, a telecommunications signal, a software distribution medium, and so forth. It should be noted that the computer readable medium contains content that can be appropriately scaled according to the requirements of jurisdictions in which such content is subject to legislation and patent practice, such as in certain jurisdictions in which such content is subject to legislation and patent practice, the computer readable medium does not include electrical carrier signals and telecommunication signals.
In this embodiment, it should be understood that the disclosed method and system may be implemented in other manners. For example, the above-described embodiment of the apparatus is merely illustrative, and for example, the division of the service modules is merely a logic function division, and there may be other division manners in actual implementation.
The modules described as separate components may or may not be physically separate, and components shown as modules may or may not be physical units, may be located in one place, or may be distributed over multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each service module in this embodiment may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units can be realized in a form of hardware or a form of hardware and a form of software functional modules.
The above description is only illustrative of the preferred embodiments of the present application and of the principles of the technology employed. It will be appreciated by persons skilled in the art that the scope of the application referred to in the present application is not limited to the specific combinations of the technical features described above, but also covers other technical features formed by any combination of the technical features described above or their equivalents without departing from the inventive concept. Such as the features described above, have similar functions to those disclosed in the present application (but are not limited to).
Claims (4)
1. The license plate recognition method is characterized by comprising the following steps of:
Receiving an original license plate image and processing the original license plate image to obtain a standardized license plate image;
performing character positioning and judgment on the standardized license plate image based on a single detector, and performing normalization processing on the standardized license plate image by adopting a background expansion method to obtain a target license plate image;
Inputting DeepOCR a model to identify the target license plate image, and obtaining target license plate character information;
The method for receiving the original license plate image and processing the original license plate image to obtain the standardized license plate image comprises the following steps:
Converting the color space of the original license plate image;
collecting and setting background image data of an original license plate image;
Performing character positioning and judgment on the standardized license plate image based on a single detector, including:
Performing character positioning and judgment on the standardized license plate image based on a two-class method;
judging whether the standardized license plate image is a character or not based on a single detector;
the normalized license plate image is normalized by adopting a background expansion method to obtain a target license plate image, and the method comprises the following steps:
Extracting characters by a threshold method;
Setting a background filled pixel value, and filling an original license plate image;
Mapping characters to the target license plate image.
2. The license plate recognition method according to claim 1, wherein the step of inputting DeepOCR a model to recognize the target license plate image to obtain target license plate character information includes:
And extracting character features of the target license plate image based on DeepOCR models, acquiring character features, processing the character features, and acquiring the target license plate character information.
3. A license plate recognition system, comprising:
The image processing unit is configured to receive the original license plate image and process the original license plate image to obtain a standardized license plate image;
The character positioning unit is configured to perform character positioning and judgment on the standardized license plate image based on the single detector, and perform normalization processing on the standardized license plate image by adopting a background expansion method to obtain a target license plate image;
the character recognition unit is configured to input DeepOCR a model to recognize the target license plate image and obtain target license plate character information;
Wherein the image processing unit includes:
A first image processing subunit configured to convert a color space of an original license plate image;
the second image processing subunit is configured to collect and set background image data of the original license plate image;
the character positioning unit includes:
The second classification subunit is configured to perform character positioning and judgment on the standardized license plate image based on a second classification method;
A character judging subunit configured to judge whether the standardized license plate image is a character based on a single detector;
a character extraction module configured to extract characters using a threshold method;
The image filling module is configured to set a background filling pixel value and fill an original license plate image;
and the character mapping module is configured to map characters to the target license plate image.
4. The license plate recognition system according to claim 3, wherein the character recognition unit includes:
And the character feature processing subunit is configured to extract character features of the target license plate image based on the DeepOCR model, acquire character features, process the character features and acquire the target license plate character information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811529296.0A CN111325209B (en) | 2018-12-14 | License plate recognition method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811529296.0A CN111325209B (en) | 2018-12-14 | License plate recognition method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111325209A CN111325209A (en) | 2020-06-23 |
| CN111325209B true CN111325209B (en) | 2024-06-07 |
Family
ID=
Citations (6)
| 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 |
| CN107273896A (en) * | 2017-06-15 | 2017-10-20 | 浙江南自智能科技股份有限公司 | A kind of car plate detection recognition methods based on image recognition |
| CN107679531A (en) * | 2017-06-23 | 2018-02-09 | 平安科技（深圳）有限公司 | Licence plate recognition method, device, equipment and storage medium based on deep learning |
| CN108052937A (en) * | 2017-12-28 | 2018-05-18 | 百度在线网络技术（北京）有限公司 | Based on Weakly supervised character machining device training method, device, system and medium |
| CN108416348A (en) * | 2018-01-29 | 2018-08-17 | 重庆邮电大学 | Plate location recognition method based on support vector machines and convolutional neural networks |
| CN108830192A (en) * | 2018-05-31 | 2018-11-16 | 珠海亿智电子科技有限公司 | Vehicle and detection method of license plate under vehicle environment based on deep learning |
Patent Citations (6)
| 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 |
| CN107273896A (en) * | 2017-06-15 | 2017-10-20 | 浙江南自智能科技股份有限公司 | A kind of car plate detection recognition methods based on image recognition |
| CN107679531A (en) * | 2017-06-23 | 2018-02-09 | 平安科技（深圳）有限公司 | Licence plate recognition method, device, equipment and storage medium based on deep learning |
| CN108052937A (en) * | 2017-12-28 | 2018-05-18 | 百度在线网络技术（北京）有限公司 | Based on Weakly supervised character machining device training method, device, system and medium |
| CN108416348A (en) * | 2018-01-29 | 2018-08-17 | 重庆邮电大学 | Plate location recognition method based on support vector machines and convolutional neural networks |
| CN108830192A (en) * | 2018-05-31 | 2018-11-16 | 珠海亿智电子科技有限公司 | Vehicle and detection method of license plate under vehicle environment based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108717524B (en) | Gesture recognition system based on double-camera mobile phone and artificial intelligence system | |
| CN111488756B (en) | Face recognition-based living body detection method, electronic device, and storage medium | |
| CN109344701A (en) | A kind of dynamic gesture identification method based on Kinect | |
| CN103413147B (en) | A kind of licence plate recognition method and system | |
| WO2017092431A1 (en) | Human hand detection method and device based on skin colour | |
| CN111275696B (en) | Medical image processing method, image processing method and device | |
| CN111079764B (en) | Low-illumination license plate image recognition method and device based on deep learning | |
| CN108090511B (en) | Image classification method and device, electronic equipment and readable storage medium | |
| CN106951869B (en) | A kind of living body verification method and equipment | |
| CN110569782A (en) | Target detection method based on deep learning | |
| CN109977834B (en) | Method and device for segmenting human hand and interactive object from depth image | |
| CN105678318A (en) | Traffic label matching method and apparatus | |
| CN116681636B (en) | Light infrared and visible light image fusion method based on convolutional neural network | |
| CN105184771A (en) | Adaptive moving target detection system and detection method | |
| CN110298829A (en) | A kind of lingual diagnosis method, apparatus, system, computer equipment and storage medium | |
| CN111709305B (en) | Face age identification method based on local image block | |
| CN116052090A (en) | Image quality evaluation method, model training method, device, equipment and medium | |
| Tan et al. | Image haze removal based on superpixels and Markov random field | |
| CN108711160A (en) | A kind of Target Segmentation method based on HSI enhancement models | |
| CN111027564A (en) | Low-illumination imaging license plate recognition method and device based on deep learning integration | |
| CN113128428B (en) | Depth map prediction-based in vivo detection method and related equipment | |
| CN106980371A (en) | It is a kind of based on the mobile augmented reality exchange method for closing on heterogeneous distributed structure | |
| CN110910497B (en) | Method and system for realizing augmented reality map | |
| CN113191189A (en) | Face living body detection method, terminal device and computer readable storage medium | |
| CN111325209B (en) | License plate recognition method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |