KR101935010B1 - Apparatus and method for recognizing license plate of car based on image - Google Patents

Abstract

The present invention relates to an image-based license plate recognition device and method, and more specifically, to an image-based license plate recognition device and method which can identify a license plate of a vehicle from an image received through a camera to recognize the vehicle number while tracking a vehicle changing the lane based on an image to stably recognize the license plate, thereby increasing accuracy of vehicle monitoring. The present invention can identify a license plate of a vehicle-related object detected from an image while tracking the object, thereby easily identifying a vehicle number of the vehicle changing the lane or traveling across a plurality of lanes in an area to be monitored by the camera, unlike in a conventional technique by which a vehicle number of a vehicle changing the lane or traveling across a plurality of lanes in an area to be monitored (or to be captured) by the camera cannot be identified by operating a camera fixed and focused at a specific point by lane constituting a road.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention [0002] The present invention relates to an image-

The present invention relates to an image-based license plate recognition apparatus and method, and more particularly, to a vehicle-license plate recognition apparatus and method for recognizing a license plate number of a vehicle from a video image received through a camera, And more particularly, to a device and method for recognizing an image based license plate which can improve the accuracy of vehicle monitoring.

On the present road, a camera is installed for crime prevention and speeding prevention, and a vehicle monitoring system is provided for identifying a license plate of a vehicle through an image received from the camera and measuring the speed of the vehicle.

However, in the conventional vehicle monitoring system, when the vehicle is monitored for a road composed of a plurality of lanes, the camera is fixedly operated so that the lens of the camera is focused at a specific point for each lane, There is a problem in that it is difficult to identify the vehicle license plate or to detect the speed of the vehicle in the case of a vehicle that runs over a plurality of lanes.

In addition, since the existing vehicle monitoring system recognizes the license plates based on the image analysis, it is necessary to construct a device for analyzing the image separately. Most of such devices are expensive and bulky, which makes it difficult to manage them.

In addition, the conventional vehicle monitoring system frequently fails to recognize the vehicle number in the image analysis process for the images in which the license plate of the vehicle is not clearly displayed due to external environmental influences such as illumination, There is a problem of poor accuracy.

As described above, the existing vehicle monitoring system has a problem that the accuracy and reliability of vehicle tracking are low and the management efficiency is low due to the inefficient configuration in which the cameras are opposed to each other for each lane.

Korean Patent Publication No. 10-1999-0070562

In order to solve the above-described problems, the present invention operates a camera in a state of being fixedly focused at a specific point for each lane constituting a road, thereby changing a lane in a surveillance target area of a camera or identifying a license plate of a vehicle traveling on a plurality of lanes And it is an object of the present invention to solve the problem of the prior art which can not be done.

The present invention also provides a front-end-based hardware configuration capable of performing image analysis on a license plate through a lightened neural network in the camera itself, thereby ensuring ease and economy of the system configuration related to license plate recognition of a vehicle .

A method for recognizing a license plate of an image-based license plate recognition apparatus including a control unit configured in a camera unit according to an embodiment of the present invention includes a detecting step of detecting a vehicle-related object from an image received from the camera unit, A license plate identification step of identifying a license plate area in the object detected in each of a plurality of frames constituting an image while the control unit tracks the object; and a control step of analyzing the license plate area through a lightened neural network excluding the floating- The control unit compares the vehicle numbers generated for each of the plurality of frames through the analysis step, and transmits the vehicle numbers through a vote to the vehicle numbers It consists of the most recognized characters per constituent digit And a recognition step of providing a final vehicle number.

In an embodiment of the present invention, in the analysis step, the control unit identifies an arrangement pattern of a character area based on an area where characters are detected in the license plate area, and compares the arrangement pattern with a plurality of preset reference arrangement patterns for each license plate type Identifying a license plate type corresponding to the license plate area and dividing the license plate area into a plurality of different character areas based on a preset reference pattern corresponding to the identified license plate type; And generating the vehicle number by identifying the character by character area through the neural network in each of the character areas.

As an example related to the present invention, the step of identifying the character may include a step of recognizing the character, wherein the control unit determines the size of the character area, the space between the character areas, the ratio of the character area , The reference layout pattern, and the order of sorting by character area, with the plurality of character areas, and dividing the number plate area into a plurality of different character areas. can do.

In one embodiment of the present invention, in the recognition step, the control unit may vote for a vehicle number calculated for each frame on a per-digit basis according to the number plate type, select a character having the highest recognition count for each digit, And a step of calculating a difference between the first and second threshold values.

In one embodiment of the present invention, the controller calculates the velocity of the object on the basis of the movement distance of the object in each of the plurality of frames and the time interval between the plurality of frames after the recognition step, And generating object information including the final vehicle number calculated for the object and the speed of the object.

In one embodiment of the present invention, after the recognition step, the controller transmits the object information to a predetermined electric signboard device when the object information is generated, so that the vehicle number and the speed according to the object information are displayed through the electric signboard device The method comprising the steps of:

In an image-based license plate recognition apparatus including a control unit configured in a camera unit according to an exemplary embodiment of the present invention, the control unit includes an object detection unit that detects a vehicle-related object from an image received from the camera unit, A number plate extracting unit for identifying a number plate region in the object detected in each of a plurality of frames constituting the image through the tracking unit; Analyzing the license plate area through a lightened neural network excluding a floating point calculation to generate a car number according to identification of each character displayed in the license plate area, The vehicle numbers are compared with each other, and the vehicle numbers And a recognition unit for providing a final vehicle number composed of the most recognized characters for each of the constituent locations.

In an embodiment of the present invention, the recognition unit may extend the zoom of the camera unit under predetermined conditions to control the camera unit to photograph an image of the vehicle at a long distance.

Unlike the prior art in which the camera is operated in a state of being fixedly focused at a specific point for each lane constituting the road and the license plate of the vehicle traveling on a plurality of lanes can not be identified, Related objects are detected and the license plate of the object is identified to change the lane in the monitored area of the camera or to easily identify the vehicle number of the vehicle traveling in a plurality of lanes.

In addition, unlike the conventional art in which characters of a license plate appearing in an image of a camera are partially lost due to an external environment, image quality is deteriorated, license plate identification fails, or license plate recognition errors occur frequently in the camera itself, The vehicle identification number for the same vehicle in each of the frames of the object is used to calculate the final vehicle number through a voting procedure so that even though the vehicle number is not correctly identified in some frames, The accuracy and reliability of recognition of the car number can be greatly increased.

In addition, the present invention supports a front-end-based hardware configuration that enables image analysis of a license plate through a lightened neural network in the camera itself, thereby ensuring high performance with low hardware, It is not necessary to configure a separate server or device, and it is possible to greatly reduce system construction cost in system configuration for license plate recognition in addition to convenience of installation and management.

1 is a configuration diagram of an image-based license plate recognition apparatus according to an embodiment of the present invention;
2 is a configuration diagram of a control unit according to an embodiment of the present invention;
FIG. 3 is a diagram illustrating an operation example of vehicle-related object tracking and license plate area detection of an image-based license plate recognition apparatus according to an embodiment of the present invention;
FIG. 4 and FIG. 5 are views illustrating a process of dividing a character region in a license plate area of an image-based license plate recognition apparatus according to an embodiment of the present invention. FIG.
FIG. 6 and FIG. 7 illustrate operation examples of a vehicle number calculation process of an image-based license plate recognition apparatus according to an embodiment of the present invention;
FIG. 8 is a diagram illustrating an example of providing a vehicle speed and a vehicle number through a display board of an image-based license plate recognition apparatus according to an embodiment of the present invention; FIG.
9 is a flowchart illustrating a method of recognizing an image based license plate according to an exemplary embodiment of the present invention.

Hereinafter, detailed embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram of an image-based license plate recognizing apparatus according to an embodiment of the present invention. As shown in FIG. 1, a camera unit 1 for providing images of vehicles 10a, 10b and 10c, And a control unit 100 for recognizing objects related to the vehicles 10a, 10b and 10c through image analysis of the image received from the unit 1 and recognizing the license plate A in the object.

At this time, the image-based license plate recognition apparatus may be configured by the camera unit 1 itself, and the control unit 100 may be formed inside the camera unit 1. [

The camera unit 1 may be configured to photograph the vehicles 10a, 10b, and 10c in operation and may be configured to be installed on a support or the like so as to face the traveling direction of the vehicles 10a, 10b, and 10c.

The control unit 100 may be configured within the camera unit 1 to receive images of the license plates A of the vehicles 10a, 10b, 10c through the images provided from the camera unit 1, The control unit 100, which is an analysis related module, can be configured integrally with the camera unit 1. [

In addition, the camera may include a storage unit 160 for storing images photographed through the camera unit 1, and the storage unit 160 may be managed by the control unit 100 And may be included in the control unit 100.

At this time, the control unit 100 executes the overall control function of the camera unit 1 using the program and data stored in the storage unit 160. [ The control unit 100 may include a RAM, a ROM, a CPU, a GPU, and a bus, and the RAM, the ROM, the CPU, and the GPU may be connected to each other via a bus. The CPU accesses the storage unit 160 and performs booting using an O / S (Operating System) stored in the storage unit 160. The CPU accesses various programs stored in the storage unit 160, Can be used to perform various operations described below.

In addition, the camera unit 1 may photograph the vehicles 10a, 10b, and 10c traveling in a plurality of lanes, generate an image thereof, and provide the images to the controller 100. [

According to the above-mentioned configuration, the present invention detects a vehicle-related object through an image within a shooting range of the camera unit 1, tracks the object, identifies the license plate through image analysis, Algorithm to perform image analysis without requiring a complicated hardware configuration as in the prior art, so that the controller 100 (FIG. 1) performing image analysis without having to construct a separate image analysis apparatus separate from the camera unit 1, Is built in the camera unit 1 so that it can easily recognize the license plate while tracking the vehicle while changing the lane or easily tracking the vehicle running on a plurality of lanes and recognizing the license plate through the license plate By providing information on the car number, it is possible to guarantee the accuracy and reliability of license plate identification.

Based on the above-described configuration, an image-based license plate recognition apparatus according to an embodiment of the present invention will be described in detail with reference to the drawings.

2 is a block diagram illustrating a configuration of a control unit 100 for recognizing a license plate and recognizing a vehicle number through image analysis of an image received from the camera unit 1 in an image-based license plate recognition apparatus according to an embodiment of the present invention. .

The control unit 100 may include an object detection unit 110, a tracking unit 120, a license plate extraction unit 130, an analysis unit 140, and a recognition unit 150 have.

First, the object detecting unit 110 may detect an object related to the vehicle 10 from the image received from the camera unit 1. [

To this end, the object detection unit 110 may include object feature information including features of the object related to the vehicle 10 to identify the object related to the vehicle 10 in the image, Setting information including calibration information for mutual mapping of the three-dimensional space corresponding to the real space in the real space and the image photographed through the camera unit 1 may be set in advance.

Accordingly, the object detecting unit 110 may filter objects other than the vehicle 10 from the image based on the setting information, and may detect the vehicle 10 related object.

3, the tracking unit 120 interlocks with the object detecting unit 110 to detect the same object as the object for each frame constituting the image with respect to the object detected from the image And the tracking unit 120 can easily track the same object through a Kalman filter.

At this time, the tracking unit 120 can continuously track the object to be tracked until the object disappears from the image.

In addition, the license plate extracting unit 130 can identify and extract the license plate area A from the objects detected in each of the plurality of frames through the tracking unit 120.

4, the license plate extracting unit 130 extracts a histogram by analyzing the brightness of all the pixels constituting the frame, and performs a histogram stretching (Histogram Stretching) Information can be corrected.

Next, the license plate extracting unit 130 binarizes the brightness of surrounding pixels using a moving average binarization technique that enables extraction of edge information of the horizontal and vertical edges, It can be generated as a binary image.

In addition, the license plate extracting unit 130 finds an edge having a pattern similar to a character in the binarized frame, and searches for a license plate area in the frame based on preset feature setting information such as a license plate type standard, aspect ratio, edge density, Can be identified and detected.

The analysis unit 140 may divide the license plate area identified by the license plate extraction unit 130 into a plurality of different character areas, and convert the plurality of different character areas into light weight The user can identify each character displayed in the license plate area and generate a vehicle number.

5 and 6, the analyzer 140, in cooperation with the license plate extractor 130, extracts characters (e.g., characters) from the license plate area of the object identified in the frame, Can be divided.

For this, the analysis unit 140 may divide the individual characters appearing in the license plate area using a blob analysis technique.

For example, when the detection of the license plate area is completed through the license plate extracting unit 130, the analysis unit 140 divides the license plate into a plurality of individual character zones. A letter zone means a separate zone in which each letter can be included.

The analysis unit 140 analyzes the brightness of pixels in the license plate area to extract a histogram and extracts separated histogram regions having non-zero values from the coordinate values of the separated histogram regions, It is possible to create individual zone lines and set and divide the character zones corresponding to the plurality of characters appearing in the license plate zone.

When the division into each character region is completed, the analysis unit 140 extracts one minimum bounding box including one character in each character region.

At this time, only a part of a character may appear as a result of noise, shading, non-uniform lighting, and a result of image processing up to the middle, or one character may appear as a separated piece, and foreground blobs Or a background pixel may be extracted as foreground pixels and foreground pixels may appear as background pixels. Therefore, in order to improve the accuracy, it is desirable to perform pixel correction for correcting the error. To this end, the analysis unit 140 may perform a morphology operation.

In addition, the analysis unit 140 extracts a blob according to a character zone by performing a labeling operation when the correction operation through the morphology operation is completed. Here, a blob means a set of connected pixels having the same characteristics or properties, and the blob extraction can use the Connected Component Labeling (CCL) technique.

When the blob is extracted as described above, the analyzer 140 may complete the processing for the final character division by dividing the minimum rectangular area including the blob into character areas.

At this time, the analyzer 140 may divide the license plate area into a plurality of different character areas, and then identify the arrangement pattern of the plurality of different character areas.

In addition, the analyzer 140 may compare the arrangement pattern with a plurality of different license plate types, and identify the license plate type corresponding to the arrangement pattern.

For example, the analyzer 140 compares a reference layout pattern previously set for each of a plurality of different license plate types with a layout pattern identified corresponding to the plurality of different character areas, and obtains a reference layout pattern most similar to the layout pattern It is possible to identify the excluded reference arrangement pattern and identify the license plate type in which the identified reference arrangement pattern corresponding to the arrangement pattern is set.

Accordingly, the analyzing unit 140 analyzes the size of the character area, the space between the character areas, the ratio of the character area, the reference arrangement pattern, and the character area, which correspond to the license plate type identified on the basis of the arrangement pattern And comparing the division reference information including the sort order and the like with the plurality of character regions to determine a plurality of remaining character regions excluding a character region which is different from the division reference information by a predetermined reference value or more, You can decide.

5 (a), the analyzer 140 determines the arrangement pattern determined on the basis of the plurality of divided character regions corresponding to the license plate area, and a plurality of preset reference arrangement patterns for each license plate type And determines that the layout pattern corresponds to the first plate type having one line of the reference layout pattern and that a difference from the reference layout pattern of the first plate type among the plurality of character areas set in the layout pattern The final character area can be set by excluding the minimum rectangular block of the character area.

5 (b), when the arrangement pattern corresponds to the second number plate type having the reference arrangement pattern of two lines, the analyzing section 140 selects one of the plurality of character areas The final character area can be set by excluding the minimum rectangular block of the character area in which a difference from the reference layout pattern of the second license plate type is generated.

6, the license plate extracting unit 130 extracts the object detected while tracking the object through the object detecting unit 110 and the tracking unit 120 from the image received from the camera unit 1, The license plate area detected through the above-described configuration may be set and the image of the license plate area may be stored in the storage unit 160 provided in the camera unit 1 and connected to the controller 100.

In addition, the analysis unit 140 may pre-set a weighted neural network deep learning algorithm except for the floating point calculation, and learn the license plate area set in the image through the deep learning algorithm, .

At this time, the neural network may be configured as an Optimized Wide Residual Network.

In addition, the analysis unit 140 may analyze the reference feature information including various feature-specific parameters such as the size of the image on the image, the arrangement pattern of the character region, the character type (font), the outline, Lt; / RTI >

At this time, the reference feature information is set as an operation reference of a plurality of layers (for example, an input layer, one or more hidden layers and an output layer) constituting the neural network-related deep learning algorithm The analysis unit 140 may continuously update the reference feature information through repeated learning on images continuously received from the camera unit 1 and accumulated in the storage unit 160, The deep-running algorithm can be optimized.

The analysis unit 140 may further include a weight sharing technique for calculating a fixed-point format excluding floating-point arithmetic operations and reducing an amount of computation between nodes constituting the layer, and a weight sharing technique for eliminating unnecessary connections between the plurality of layers. A lightweight deep-running algorithm employing a Huffman compression scheme for pruning and memory optimization can be set.

According to the above-described configuration, the analyzer 140 divides the license plate area identified in the frame into a plurality of different character areas with respect to the frame provided from the license plate extracting unit 130, as described above And the characters of each of the plurality of character regions can be identified through the neural network-related deep learning algorithm.

In addition, the analyzer 140 may identify the license plate type corresponding to the license plate area in cooperation with the license plate extraction unit 130, and may identify the license plate type corresponding to the identified license plate type, It is possible to generate the vehicle number (candidate vehicle number) by sorting the characters identified in each of the plurality of character areas.

In addition, the analyzer 140 detects a license plate area of a specific object for each frame by tracing a specific object for each of a plurality of different frames constituting an image in cooperation with the license plate extractor 130, The license plate area can be divided into a plurality of character areas as described above, and the vehicle number (candidate car number) can be generated for each frame through the neural network.

In addition, the analysis unit 140 may provide a plurality of vehicle numbers (candidate vehicle numbers) corresponding to a plurality of different frames to the recognition unit 150. [

As described above, the present invention can identify the vehicle number while tracking the same specific object in each of the frames, but characters among the plurality of frames can be differently identified according to the image quality or the influence of the external environment. A method unique to the present invention for providing an accurate vehicle number of the vehicle will be described with reference to FIG.

As shown in the figure, the recognition unit 150 may receive the vehicle number generated through the analysis of each of a plurality of frames including a specific object from the analysis unit 140 as a candidate vehicle number.

At this time, the object detecting unit 110 may assign a unique identification number corresponding to a specific object detected in the image, and may assign the same unique identification number to the same object.

The number plate extracting unit 130 extracts a number plate region from the object detected by the object detecting unit 110 in each of a plurality of frames constituting an image in cooperation with the object detecting unit 110 and the tracking unit 120, The unique identification number may be set in the license plate area and provided to the analysis unit 140.

 In addition, the analyzing unit 140, in cooperation with the license plate extracting unit 130, matches the unique identification number assigned to the specific object each time a candidate vehicle number is generated corresponding to the specific object in each of a plurality of frames, (150).

Accordingly, the recognition unit 150 sequentially receives a plurality of candidate vehicle numbers generated by tracking the specific object based on the image from the analysis unit 140, and acquires a plurality of candidate vehicle numbers corresponding to the specific object The storage unit 160 may store a plurality of candidate vehicle numbers corresponding to a specific object.

In addition, the recognition unit 150 may classify each of the plurality of candidate vehicle numbers corresponding to the specific object stored in the storage unit 160 as the sorting for sorting according to the license plate type identified corresponding to the license plate area of the specific object (Or the number of recognitions) for each of at least one character in each place by comparing the same place in the order, and then counting the number of duplicates (or the number of recognizations) by voting ), And the final vehicle number can be calculated by combining the characters calculated through the above-mentioned voting procedure for each place.

Accordingly, the recognition unit 150 can generate and provide the final vehicle number of the vehicle corresponding to the specific object.

Meanwhile, in the above-described configuration, the tracking unit 120 may calculate a velocity of the specific object by tracking a specific object in cooperation with the object detection unit 110, and may calculate the velocity of the specific object, And provides the information to the license plate extracting unit 130.

For example, the tracking unit 120 may calculate the velocity of the object based on a movement distance of the specific object in each of the plurality of frames and a time interval between the plurality of frames.

Accordingly, the analysis unit 140 may provide the plurality of candidate vehicle numbers to the recognition unit 150 in correspondence with the object information received from the license plate extraction unit 130, and the recognition unit 150 May calculate the final vehicle number through the voting procedure for the plurality of candidate vehicle numbers, and then update the object information by replacing the unique identification number included in the object information with the final vehicle number.

Accordingly, the recognition unit 150 can generate object information including a final vehicle number and a speed of a specific object.

The camera unit 1 or the control unit 100 may be provided with a communication unit for communication through a communication network with a predetermined external display device 200. The control unit 100 controls the recognition unit 150 And transmits the object information including the final vehicle number generated through the communication unit to the external electric sign board device 200 through the communication unit.

At this time, the communication unit can be applied to various widely known wired / wireless communication systems.

8, the external electric sign board device 200 receives the object information from the communication unit configured in the camera unit 1, and displays the final vehicle number and the speed according to the object information have.

Also, the control unit 100 may transmit the object information to an external server through the communication unit, thereby enabling a user of the external server to obtain information about the vehicle traveling on the road, , It is possible to provide a crime prevention function for controlling an overspeed vehicle or for tracking an escape vehicle or a crime vehicle.

At this time, the control unit 100 visually sets (or indexes) the final vehicle number to the object detected in the image or each of the plurality of frames based on the object information, And may be transmitted to the external server through the communication unit.

In addition, the controller 100 may transmit only the final vehicle number excluding the speed-related information to the external server.

Meanwhile, the recognition unit 150 may control the camera unit 1 to enlarge the zoom of the camera unit 1 under a predetermined condition and photograph an image of the vehicle at a long distance.

For example, the recognition unit 150 may extend the zoom of the camera unit 1 when a speed over a predetermined reference value is detected through a predetermined time period during which the overspeed is likely to occur or in conjunction with the tracking unit 120 , It is possible to control the camera unit to capture an image of a remote vehicle.

Accordingly, the controller 100 can detect an object by receiving an image of a target surveillance area wider than before the zoom of the camera unit 1 from the camera unit 1, It is possible to improve the problem that the possibility of failing to recognize the number plate of the overspeed vehicle is high and the range of the surveillance area is expanded through zooming of the camera section so that a sufficient frame number So that the license plate of the speeding vehicle can be easily recognized.

According to the above-described configuration, the present invention can be applied to a vehicle in which a camera is operated in a state of being fixedly focused at a specific point for each lane constituting a road, thereby changing a lane in a surveillance target area (or a shooting target area) Unlike the conventional technique in which the license plate of the vehicle can not be identified, the vehicle-related object detected in the image is tracked, and the license plate of the object is identified to change the lane in the monitored area of the camera, And so on.

In addition, unlike the conventional art in which characters of a license plate appearing in an image of a camera are partially lost due to an external environment, image quality is deteriorated, license plate identification fails, or license plate recognition errors occur frequently in the camera itself, The vehicle identification number for the same vehicle in each of the frames of the object is used to calculate the final vehicle number through a voting procedure so that even though the vehicle number is not correctly identified in some frames, The accuracy and reliability of recognition of the car number can be greatly increased.

In addition, the present invention supports a front end based hardware configuration that enables image analysis on a license plate through a neural network that is lightened by the camera itself, thereby ensuring high performance with low hardware, It is not necessary to construct a separate server or device for image analysis, so that it is possible to greatly reduce the installation cost of the system configuration for the license plate recognition as well as the installation and management convenience.

9 is a flowchart of a method of recognizing an image-based license plate according to an embodiment of the present invention. As shown in FIG. 9, the control unit 100 can detect a vehicle-related object in the image received from the camera unit 1 (S1).

In addition, the controller 100 may identify the license plate area in the object detected in each of a plurality of frames constituting the image while tracking the object (S2).

In step S3, the controller 100 analyzes the license plate area through the lightened neural network except for the floating point operation, and generates a car number according to the identification of each character in the license plate area.

Next, the control unit 100 compares the vehicle numbers generated for each of the plurality of frames through the analysis step, and obtains a final vehicle number composed of the most recognized characters for each place constituting the vehicle number (S4).

The various devices and components described herein may be implemented by hardware circuitry (e.g., CMOS-based logic circuitry), firmware, software, or a combination thereof. For example, it can be implemented utilizing transistors, logic gates, and electronic circuits in the form of various electrical structures.

It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit or essential characteristics thereof. Therefore, the embodiments disclosed in the present invention are intended to illustrate rather than limit the scope of the present invention, and the scope of the technical idea of the present invention is not limited by these embodiments. The scope of protection of the present invention should be construed according to the following claims, and all technical ideas within the scope of equivalents should be construed as falling within the scope of the present invention.

1: camera unit 100:
110: Object detection unit 120:
130: number plate extracting unit 140:
150:

Claims (8)

A license plate recognition method of an image-based license plate recognition apparatus including a control unit configured in a camera unit,
Detecting a vehicle-related object from the image received from the camera unit;
A license plate identification step of identifying the license plate area in the object detected in each of a plurality of frames constituting an image while the control unit tracks the object;
Analyzing the license plate area through the lightened neural network except for the floating point operation to generate a car number according to identification of each character displayed in the license plate area; And
The control unit compares the vehicle numbers generated for each of the plurality of frames through the analysis step, and provides a final vehicle number composed of the most recognized characters for each place constituting the vehicle number through a vote
Based vehicle license plate identification method.
The method according to claim 1,
The analysis step
The control unit identifies the layout pattern of the character area based on the area where the characters are detected in the number plate area, and compares the arrangement pattern with a plurality of preset reference layout patterns for each license plate type to identify the license plate type corresponding to the license plate area ; And
Wherein the control unit divides and sets the license plate area into a plurality of different character areas based on a preset reference arrangement pattern corresponding to the identified number plate type, And generating a vehicle number
Further comprising the steps of: receiving an image of the vehicle;
The method of claim 2,
The step of identifying the character
Wherein the control unit includes at least one of a size of the character area, a space between the character areas, a ratio of the character area, a reference arrangement pattern, and a sort order for each character area corresponding to the license plate type identified on the basis of the arrangement pattern And dividing the license plate area into a plurality of different character areas by comparing the divided reference information with the plurality of character areas.
The method of claim 2,
The recognition step
Further comprising the step of the control unit performing a vote for each vehicle number calculated for each frame according to the number according to the license plate type to select a character having the highest recognition frequency for each digit to calculate a final vehicle number Image based license plate recognition method.
The method according to claim 1,
After the recognizing step
The control unit calculates the speed of the object on the basis of the movement distance of the object in each of the plurality of frames and the time interval between the plurality of frames, Wherein the step of generating the object-based license plate identification information comprises the step of generating object information including the velocity.
The method of claim 5,
After the recognizing step
Further comprising the step of the controller transmitting the object information to the electric display panel device set in advance when the object information is generated so that the vehicle number and the speed according to the object information are displayed through the electric display panel device License plate recognition method.
An image-based license plate recognizing apparatus comprising a control unit configured in a camera unit,
The control unit
An object detection unit for detecting a vehicle-related object from the image received from the camera unit;
A tracking unit for tracking the object detected in the image in cooperation with the object detecting unit;
A license plate extracting unit for identifying a license plate area in the object detected in each of a plurality of frames constituting the image through the tracing unit;
An analysis unit for analyzing the license plate area identified through the license plate extraction unit through a lightened neural network except for floating point calculation to generate a car number according to identification of each character displayed in the license plate area; And
A recognition unit for comparing the vehicle numbers generated for each of the plurality of frames through the analysis unit and providing a final vehicle number composed of the most recognized characters for each place constituting the vehicle number through a vote,
Based license plate recognition device.
The method of claim 7,
Wherein the recognition unit extends the zoom of the camera unit under predetermined conditions to control the camera unit to photograph an image of the vehicle at a long distance.

Images