KR101996101B1 - A smart light-car recognition system, and stand-alone based smart car-number recogniton sytem therefor - Google Patents

Abstract

An intelligent light vehicle discrimination system for achieving the object of the present invention analyzes a vehicle image using an intelligent algorithm without employing any sensors and discriminates whether the vehicle is light or small according to the result, And the maintenance cost can be greatly reduced. Further, the intelligent light vehicle discrimination system according to the present invention can discriminate whether a vehicle is a light vehicle by analyzing a vehicle image input from an image camera that is used in another system. In addition, a parking control system, a road control system, System, vehicle black box, mobile terminal, and the like. Therefore, the intelligent light vehicle discrimination system according to the present invention has developed a highly discriminating light vehicle discrimination method in terms of purpose, structure and effect, thereby realizing high accuracy, high reliability, real time response, installation place independence, manufacturing cost / installation / maintenance It is possible to reduce the cost, so it has secured the high performance and price competitiveness compared to the existing one. It is expected to be widely applied in the fields requiring discrimination of light vehicles in the future.
In order to achieve another object of the present invention, a standalone-based intelligent car number recognition system has organically combined the intelligent car discrimination algorithm and the car number recognition system. In particular, the vehicle number recognition system according to the present invention detects the vehicle license plate contour line more quickly than the conventional vehicle license plate line by scanning from the lower line to the upper side of the area based on the vehicle contour area detected by the miniscule discrimination algorithm. (FIRM-WARE) -based firmware that does not require an OS (Operating System), because it can recognize the vehicle number by extracting (detecting) and matching the feature vector of the vehicle, and simplifying unnecessary processing steps and logic, , The program capacity can be reduced while the function calling stage can be reduced and the loading time can be shortened, thereby improving the processing speed compared with the performance. Therefore, the intelligent light vehicle discrimination and car number recognition system according to the present invention reduces unnecessary processing steps compared with the prior art and adopts an intelligent algorithm to further improve the accuracy and processing speed, and can reduce the size, weight, stabilization, , Compatibility, cost reduction (manufacturing / operation cost / maintenance cost), ease of installation and maintenance, and also achieved performance and price competitiveness compared to the existing one.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention [0001] The present invention relates to an intelligent vehicle identification system and a stand-alone intelligent vehicle identification system using the intelligent vehicle identification system.

The present invention relates to an intelligent light vehicle discrimination system and a standalone-based vehicle number recognition system employing the intelligent light vehicle discrimination system. More particularly, the present invention relates to an intelligent light vehicle discrimination system for rapidly and accurately discriminating whether a vehicle is light or small by analyzing a vehicle image using an intelligent algorithm And a stand-alone based intelligent car number recognition system that organically combines the car discrimination system.

Generally, vehicles are divided into small cars, small cars, mid-size cars and large cars according to size and displacement. A minicar is notified of a vehicle with a displacement of 1000cc or less, a full width (width) of 1.6 meters or less, a total length (length) of 3.6 meters or less, and a height of 2.0 meters or less. In light of the government policy, there is a 50% discount on parking fees for public parking lots, congestion tolls for motorway tolls on tunnels and tunnels, and so on. In the public parking lot, highway toll gate and tunnel gate, Respectively.

Conventionally, there are four methods for discriminating light vehicles. First, information such as the type of vehicle is registered at the time of membership entry, and the high-pass management system reads the vehicle information from the high-pass card information at the time of entering the vehicle and directly discriminates whether or not the vehicle is a light vehicle. The second method is to embed a loop sensor in the lower part of the vehicle access road or to embed wheels or weight sensors horizontally in the direction of the vehicle in advance to detect the vehicle width or vehicle length. Third, the width of the vehicle is detected by using an ultrasonic sensor (or an infrared sensor) on the front side of the access road. Fourth, a plurality of pairs of optical sensors having a light emitting portion and a light receiving portion are disposed on both sides of an entrance through which a vehicle passes, thereby detecting the vehicle width.

However, the above-mentioned first method can not cover all the vehicles at the same time because it is useless in a place not equipped with a high-pass management system or a vehicle not using a high-pass.

In the second method, since a separate sensor is required to be laid under the road, a large installation cost is required and a large number of malfunctions occur in the sensor current characteristic value due to aging or abnormal weather (cold or hot), and maintenance is complicated .

In the third method, the more the distance between the ultrasonic sensor and the vehicle is, the more the error is increased and the accuracy is lowered. Due to the characteristics of the ultrasonic sensor, the error of the infrared sensor is higher than that of the ultrasonic sensor, There is still a problem that it is difficult to practical use.

As shown in FIG. 1, the fourth method disclosed in Patent Registration No. 10-0895844 (Light Vehicle Identification Apparatus and Method) requires that the sensors be arranged at very regular intervals in the vertical direction on both sides of the vehicle access road, There is a problem that the maintenance is complicated, the construction cost is large, the error occurs according to the interval between the neighboring sensors when the vehicle enters the skew, and the maintenance is very inconvenient.

Meanwhile, a car number recognition system using a camera has recently attracted a great interest in the fields of a parking control system, a road control system, and a vehicle surveillance system. A vehicle number recognition system using a camera basically includes a general-purpose PC equipped with a predetermined image camera and an operating system (OS), and is used by incorporating car number recognition software in a general-purpose PC. The vehicle number recognition software based on the operating system (OS) detects the license plate area from the vehicle image of the input camera, extracts characteristics of the car number in letters and numbers in the detected license plate area, It is common to RECOGNITON the car number.

Here, performance of recognition rate (accuracy) and recognition speed (real-time processing capability) is the biggest issue. Particularly, the existing car number recognition system significantly decreases the recognition rate of the car number when the car number plate is backlit, when the vehicle is inclined, or when the car is at night or overcast. In order to solve these problems, a vehicle number identification image camera is provided, and the image input from the camera is preprocessed and / or post-processed to correct sharpness correction and screen inclination, or to detect a license plate area there have been newly proposed methods employing various algorithms in the detection step and the vehicle number recognition step.

However, such a car number recognition system improves the car number recognition rate by employing artificial intelligence learning or a complex mathematical algorithm, but the complexity of the calculation becomes complicated due to the complexity of the calculation, which makes it difficult to respond in real time. On the other hand, if a high-performance CPU and a highly integrated image processing board are additionally mounted to improve the operation speed, there arises a problem that the complexity of the system and the manufacturing cost increase greatly.

Especially, general-purpose PC-based devices equipped with operating system (OS) and car number recognition software (SW) are large in housing size, vulnerable to external shocks, The cost of purchasing software is so high that it is difficult to hire small- and medium-sized businesses or small businesses.

Recently, there is an increasing demand for an embedded system which is easy to install and maintain in various mobile environments and special sites. However, the embedded-based system still has a problem in the commercialization due to the problem of the component cost compared to the hardware performance described above. An embedded system that performs a specific function is generally used for a purpose of reducing the manufacturing cost by employing a hardware specification lower than that of a general-purpose PC, and programming and embedding only a specific function in a large amount. However, in embedded systems, as in the case of PCs, an operating system such as MS Windows or Linux is installed to run the software. Therefore, in an embedded system in which the specification is dropped due to operation of the corresponding module in each operating system loading and function processing step, The limit is significantly reduced in speed. Specifically, the car number recognition program must call the operating system in order to call the corresponding device manager for a specific function module. Therefore, the step becomes more complicated and the number of clocks increases accordingly, so that the system-on- The lower the SoC specification, the slower the processing speed. In addition, since the complex operation of the embedded system constituted by the intermediate and low-end specifications is limited, the license plate area detection and the car number recognition algorithm are also very limited, and in reality, satisfying the performance required in the market (real- It is a difficult situation.

To solve this problem, it is necessary to purchase a system-on-chip (SoC) with an expensive integrated circuit (VLSI; CPU + memory + cache memory + I / O interface) or to manufacture a customized system- The limit is clear to ensure competitiveness. Therefore, embedded-based vehicle identification system that satisfies both performance and price competitiveness will take a considerable amount of time. In order to realize an embedded-based vehicle identification system that satisfies performance and price competitiveness at the same time with a system-on-a-chip (SoC) of a normal low-end specification, it is necessary to change the idea from a development mindset or control logic configuration Is required.

Patent Registration No. 10-0895844 (Miniature Discriminating Apparatus and Method, Announcement on May 9, 2009)

It is an object of the present invention to provide an intelligent light vehicle discrimination system which automatically discriminates whether a light vehicle is very fast or accurately by using an intelligent new concept algorithm without employing existing sensors .

It is another object of the present invention to solve the above-mentioned problems, and it is another object of the present invention to solve the above problems by simultaneously combining the intelligent light vehicle discrimination algorithm with a predetermined vehicle number recognition algorithm and a special programming technique to simultaneously improve a vehicle number recognition speed and a vehicle number recognition rate, (STAND-ALONE) based intelligent car number recognition system that is easy to install, maintain, and maintain.

According to another aspect of the present invention, there is provided an intelligent light vehicle discrimination system, comprising: a memory table for storing size information of a vehicle license plate by type; (A) and a longitudinal length (b) of the vehicle license plate contour are detected from a predetermined vehicle image, and the ratio of the length to the length of the vehicle license plate contour line and the license plate size information for each type of the memory table And a discriminating unit for discriminating whether the vehicle is a light vehicle or not based on the detected result.

Here, the judging unit judges whether or not at least one of the ratio of the length (a) to the length (b) of the outline of the license plate of the vehicle image and the actual width A1 of the license plate of the vehicle image The actual transverse length A1 of the license plate corresponds to the ratio of the transverse length (a) to the transverse length (b) of the vehicle license plate contour of the vehicle image, when it is matched with the license plate size type of the large- Is compared with the ratio of the actual transverse length (A) to the actual transverse length (B) of the license plate according to the type of the memory table, and is obtained from the license plate size information of the type matched with each other.

Here, the discrimination unit further detects the vehicle contour line from the predetermined vehicle image, calculates the actual vehicle width Y of the vehicle on the basis of Equation 1 below, compares the calculation result with the lightness reference value (total width of 1.6 meters or less) Thereby discriminating whether or not the vehicle is a light vehicle.

Where Y is the actual vehicle width of the photographed vehicle, y is the vehicle outline vehicle width (lateral length) of the image, and A1 is the vehicle width Is the actual transverse length of the license plate, and a is the transverse length of the license plate of the image. A1 compares the ratio of the length (a) to the length (b) of the license plate contour in the vehicle image with the ratio of the actual width (A) to the actual length (B) of the license plate for each type of memory table And can be immediately confirmed from the license plate size information of the type matched to each other.

According to another aspect of the present invention, there is provided a standalone-based intelligent car number recognition system comprising: an input interface unit for inputting a predetermined car image; A first memory table for storing the size information of the license plate by type; A second memory table for storing a predetermined type of vehicle number feature vector; The vehicle contour line and the license plate contour line are detected from the vehicle image, the actual vehicle width Y of the vehicle is calculated on the basis of Equation 1 below, and the calculation result is compared with the light criterion value (full width of 1.6 meters or less) And extracts the feature vector of the vehicle number in the outline area of the detected license plate and compares the feature vector with the vehicle number feature vector set in the second memory table to recognize the vehicle number of the vehicle Car identification and car number recognition unit; And a management event control unit for controlling each configuration of the system as a whole and performing predetermined management events preset based on light vehicle discrimination information and vehicle number recognition information.

(1), where Y is the actual vehicle width (lateral length) of the photographed vehicle, y is the contour vehicle width (lateral length) of the vehicle image , A1 is an actual transverse length of the photographed vehicle license plate, a is a vehicle license plate contour transverse length of the vehicle image, A1 is a transverse length (a) of the license plate contour of the vehicle image, Ratio can be confirmed directly from the license plate size information of the type that is matched with the ratio of the actual transverse length (A) to the actual longitudinal length (B) of the license plate for each type classified in the first memory table.

Here, a predetermined output (OUTPUT) interface for externally transmitting at least one of the lightness difference discrimination information, the car number recognition information, and the control signal according to the execution of the predetermined management event may be organically provided have.

Here, it is more preferable that the outline of the license plate is scanned sequentially from the lower line to the upper direction of the vehicle outline area and detected.

Here, it is preferable that the car number recognition is performed by detecting a feature vector of a car number composed of Hangul characters or numbers (a combination of the number of turns, desirable.

Here, it is preferable that the algorithm (logic) performed in the light vehicle discrimination, the vehicle number recognition unit, and the management control unit is programmed based on firmware without an operating system (OS).

Here, the intelligent car number recognition system according to the present invention is constituted by a stand-alone (STAND-ALONE) -based terminal built-in system that performs self-complete identification of a light vehicle and a car number recognition algorithm, thereby achieving miniaturization, , Mobility, scalability, site-specific installation, and maintenance are easily combined organically.

An intelligent light vehicle discrimination system for achieving the object of the present invention analyzes a vehicle image using an intelligent algorithm without employing any sensors and discriminates whether the vehicle is light or small according to the result, And the maintenance cost can be greatly reduced. Further, the intelligent light vehicle discrimination system according to the present invention can discriminate whether a vehicle is a light vehicle by analyzing a vehicle image input from an image camera that is used in another system. In addition, a parking control system, a road control system, System, vehicle black box, mobile terminal, and the like. Therefore, the intelligent light vehicle discrimination system according to the present invention has developed a highly discriminating light vehicle discrimination method in terms of purpose, structure and effect, thereby realizing high accuracy, high reliability, real time response, installation place independence, manufacturing cost / installation / maintenance It is possible to reduce the cost, so it has secured the high performance and price competitiveness compared to the existing one. It is expected to be widely applied in the fields requiring discrimination of light vehicles in the future.

In order to achieve another object of the present invention, a standalone-based intelligent car number recognition system has organically combined the intelligent car discrimination algorithm and the car number recognition system. In particular, the vehicle number recognition system according to the present invention detects the vehicle license plate contour line more quickly than the conventional vehicle license plate line by scanning from the lower line to the upper side of the area based on the vehicle contour area detected by the miniscule discrimination algorithm. (FIRM-WARE) -based firmware that does not require an OS (Operating System), because it can recognize the vehicle number by extracting (detecting) and matching the feature vector of the vehicle, and simplifying unnecessary processing steps and logic, , The program capacity can be reduced while the function calling stage can be reduced and the loading time can be shortened, thereby improving the processing speed compared with the performance. Therefore, the intelligent light vehicle discrimination and car number recognition system according to the present invention reduces unnecessary processing steps compared with the prior art and adopts an intelligent algorithm to further improve the accuracy and processing speed, and can reduce the size, weight, stabilization, , Compatibility, cost reduction (manufacturing / operation cost / maintenance cost), ease of installation and maintenance, and also achieved performance and price competitiveness compared to the existing one.

FIG. 1 is a schematic diagram showing a light-vehicle discrimination method according to a conventional system,
FIG. 2 is a conceptual diagram for explaining the core principle of the intelligent light vehicle discrimination algorithm and the system according to the present invention,
FIG. 3 is a configuration diagram showing a schematic configuration of an intelligent light vehicle discrimination system according to a preferred embodiment of the present invention,
FIG. 4 is a flowchart schematically showing an intelligent light-discernment discrimination algorithm applied to the light-discernment discrimination system of FIG. 3 according to a preferred embodiment of the present invention,
FIG. 5 is a block diagram schematically illustrating a stand-alone-based intelligent car number recognition system according to an embodiment of the present invention.
FIG. 6 is a flowchart schematically illustrating an intelligent light vehicle discrimination and car number recognition algorithm employed in the standalone-based intelligent car number recognition system of FIG. 6 to accomplish another object of the present invention.

Since 1973, 1996, and 2004, the Korean license plate has undergone major changes in terms of combination style, color, etc. since the liberation, and the large frame, which has been maintained since April 15, 1973, , "Use of Korean characters", and "Last four-digit serial number - 1000 to 9999". The license plates are divided into 12 kinds including the temporary license plates, and the license plate sizes are used in duplicate.

If the reclassification is based only on the <horizontal-vertical> size of the vehicle license plate, the license plate sizes of all the vehicles currently operating in Korea are classified into one of the five types as shown in Table 1 below.

According to the size classification of the license plate according to the present invention By size Size (mm) (Width - Height) Aspect ratio Applicable vehicle Type 1 520 - 110 4.73 New front Type 2 335 - 155 2.16 New rear (optional) Type 3 335 - 170 1.97 Spherical general Type 4 440 - 200 2.20 Large non-business use Type 5 440 - 220 2.00 Large Business

According to the Ministry of Land Transportation Statistics (as of the end of September 2014), 86.6% of all registered vehicles (18.73030 units, excluding business use) replaced the front plate of the vehicle with the new type plate of Type 1, only 13.4% It is equipped with a license plate, but it is expected that most of the registration will be canceled within the next 10 years.

As can be seen from Table 1, the license plates of all domestic vehicles are reclassified into a total of five sizes (four in total based on the front surface of the vehicle). In view of this point, the inventor of the present invention has developed an intelligent We have invented the submarine discrimination algorithm and the submarine discrimination system.

An intelligent light vehicle discrimination system according to the present invention is an algorithm and system for discriminating whether a vehicle is light or not by a simple, fast, and accurate method using an existing car number identification camera, The inventor of the present invention originally developed it after a long research and effort.

The intelligent vehicle discrimination system according to the present invention is a key discrimination algorithm and a system of the present invention which are a combination of the following five features newly discovered and proposed.

FIG. 2 is a conceptual diagram for explaining the core principle of the intelligent light vehicle discrimination algorithm and the system according to the present invention, and the concept of the concept.

First, as shown in FIG. 2, the size of the vehicle in the image taken by the camera and the actual size of the vehicle are directly proportional to each other.

Secondly, the ratio of 'the length (a): length (b)' of the license plate in the vehicle image corresponds to 'the length A1 of the actual license plate of the vehicle: the length of the actual license plate B' , The vehicle license plate photographed by the camera can be specified by any one of the types shown in Table 1, so that the actual vehicle license plate lateral length A1 of the photographed vehicle can be immediately confirmed with reference to Table 1 It is a point.

Third, the ratio of the actual vehicle width (Y) of the vehicle to the vehicle width (y) in the vehicle image and the actual lateral length (A1) of the license plate of the vehicle are: The actual vehicle width Y of the vehicle can be calculated very simply and accurately, as shown in FIG. 2, by using the fact that the ratio of the length (a) 'of the vehicle to the width (a) In this case, a and b are obtained by measuring the length and the length of the license plate in the vehicle image with a program, and A1 is confirmed immediately in the second step.

Fourthly, comparing the actual vehicle width Y of the corresponding vehicle calculated in the third step with 1.6 m, which is the reference value of the full width of the vehicle (vehicle width), it is possible to determine very accurately whether the vehicle is light or not.

Fifthly, in the above Table 1, the type 4 (for large-scale non-business use) and type 5 (for large-sized business) is the license plate size limited to the use of large vehicles, 5, it can be immediately determined that the vehicle is not a light vehicle. Also, if the actual transverse length A1 of the vehicle license plate obtained from the second characteristic is matched with the type 4 or 5 type, the vehicle can be immediately determined that the vehicle is not a light vehicle. Therefore, in this case, the third and fourth steps are omitted and the discrimination speed is faster.

Therefore, the intelligent light vehicle discrimination system according to the present invention excludes measurement errors of the external sensors, thereby remarkably improving the accuracy, and secondly, there is no need to install a separate sensor as in the conventional method, thereby reducing manufacturing cost and installation / maintenance cost , It is very easy to maintain and repair. Therefore, it can be widely applied to a parking control system using a camera, a road control system, and a vehicle sensing system, and has a very accurate discrimination accuracy and real time processing. It is evaluated as a highly innovative and unique intelligent light vehicle discrimination system.

 DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a configuration and operation of an intelligent light vehicle discrimination system and a standalone-based intelligent vehicle identification system according to a preferred embodiment of the present invention will be described in detail with reference to the accompanying drawings.

3 is a block diagram showing a schematic configuration of an intelligent light vehicle discrimination system according to a preferred embodiment of the present invention. The light vehicle discrimination system according to the embodiment of FIG. 3 may include an image camera and a light vehicle discriminating means.

The image camera of FIG. 3 is located on the front side of an entering vehicle and photographs a vehicle entering or passing through it, and is not an essential component of the intelligent vehicle discrimination system according to the present invention. Accordingly, the image camera may further include the image camera, or may be a video camera used for parking control, road control, or vehicle monitoring. The intelligent light vehicle discrimination system of the present invention can discriminate whether a light vehicle is a car or a vehicle by receiving a photographed image or a vehicle image photographed by another device.

3 includes an input end, a memory table, a discriminating portion, and an output end. When the captured vehicle image is received through the input terminal, the determination unit intelligently analyzes the vehicle image and determines whether the vehicle is light or not by referring to the memory table. The light vehicle discriminating means may further include a predetermined output terminal or a wire / wireless communication module when necessary, so that the light vehicle discrimination result of the vehicle can be transmitted to another device.

3, the actual transverse length (a) and the actual transverse length (b) of the license plate and the length ratio a / b of the transverse length and longitudinal length of the license plate are corresponded to each other for each license plate type Lt; / RTI &gt; In this case, the memory table may have a DB (lookup table) for each size type of the license plate as one of the following three types.

First, as shown in Table 2 below, the actual width (A), the actual length (B), and the ratio (A / B or B / A) Can be stored. In this case, since it is not necessary to separately calculate the above-mentioned ratio while including the basic size information of the vehicle, the execution speed is higher, which is preferable.

According to the present invention, the size classification DB size
By type License plate
Actual width (mm) (A) License plate
Actual length (mm) (B) Width: aspect ratio (A / B) Type 1 520 110 4.73 Type 2 335 155 2.16 Type 3 335 170 1.97 Type 4 440 200 2.20 Type 5 440 220 2.00

Second, only actual length (A) and actual length (B) of the license plate can be stored for each size type, as shown in Table 3 below. In this case, it is inconvenient for the minuscule discrimination system to separately calculate the ratio at the time of discriminating whether or not the vehicle is a light vehicle.

According to the present invention, the size classification DB size
By type License plate
Actual width (mm) (A) License plate
Actual length (mm) (B) Type 1 520 110 Type 2 335 155 Type 3 335 170 Type 4 440 200 Type 5 440 220

Thirdly, as shown in Table 4 below, only the actual transverse length (A) of the license plate and the ratio (A / B or B / A) of transverse length to transverse length may be stored for each size type, There is an advantage that it is not necessary to separately calculate the ratio at the time of discriminating whether the vehicle is a light vehicle or not.

According to the present invention, the size classification DB size
By type License plate
Actual width (mm) (A) License plate
Actual width (A): Length (B) Ratio (A / B) Type 1 520 4.73 Type 2 335 2.16 Type 3 335 1.97 Type 4 440 2.20 Type 5 440 2.00

As described above, it is possible to store the size information of the memory table, all of which are within the scope of the present invention, and it is also possible to add additional information such as a vehicle map in addition to the item. Also, it is obvious that the memory table according to the present invention may be embedded in the system or provided outside the system, and may be connected and referenced using a wired / wireless communication network. It will be appreciated by those skilled in the art that reclassifying the size of the corresponding license plate, and thus reconfiguring the memory table, even if the legislation associated with the license plate changes in the future, falls within the scope of the present invention .

Next, the operation of the system configuration shown in Fig. 3 will be described in detail with reference to Fig. FIG. 4 is a flowchart schematically showing an intelligent submarine discrimination algorithm applied to the submarine discrimination system of FIG. 3 according to a preferred embodiment of the present invention.

Referring to FIG. 4, first, the camera captures an image of a target vehicle entering or passing through (step 300). The photographed vehicle image is input to the light-vehicle discriminating means (step 310). The light-vehicle discrimination means detects the contour of the vehicle from the input vehicle image and measures the lateral length (y) of the vehicle contour (step 320). Here, a brief binarization can be performed on the inputted vehicle image to detect the contour of the vehicle.

Next, the light-vehicle discriminating means detects the license plate contour line on the basis of the detected vehicle contour line (step 330). Here, it is general that most of the vehicle license plates attached to the front of the vehicle are disposed in a balanced manner in the right and left direction with respect to the center of the vehicle in the horizontal direction, and at the same time, they are significantly lowered downwardly from the center of the vehicle vertical direction. Conventionally, when extracting a license plate, the image was scanned from the upper line to the lower side of the image screen, and from left to right by a predetermined pixel unit or block unit. Therefore, there is a problem that a full screen must be scanned, and a considerable delay time is generated by scanning the license plate from a position far from the position. However, according to the present invention, the search speed and processing efficiency of the contour line of the license plate are compared with each other by sequentially scanning the vehicle line from the lower line to the upper line of the contour area on the basis of the detected vehicle contour line and from the center line of the vehicle contour area Improvement. In addition, for the image of the inner area of the license plate contour line to detect the contour of the license plate, the image can be binarized by further subdividing the contour edge and the vehicle number area inside thereof from the step 320.

If the license plate contour is detected in step 330, the light-vehicle discrimination means calculates a / b (or b / a) which is the ratio of the length (a) to the length (b) of the license plate corresponding to the vehicle image contour .

In step 340, the light-vehicle discriminating means compares the calculated ratio a / b (or b / a) with the five types of vehicle license plate size types classified in the memory table, The actual transverse length A1 is checked. For example, if the ratio is matched to 4.73, it corresponds to a new type plate of Type 1, and it can be immediately confirmed that the actual width A1 of the license plate is 520 mm.

 In step 350, when the ratio a / b obtained in step 330 matches the corresponding ratio value of type 4 or type 5 of Table 2a to Table 2c in step 340, It is determined that the vehicle is not a light vehicle and the process goes to step 380 to wait for a subsequent management event.

In step 350, when the actual transverse length A1 of the vehicle license plate obtained in step 340 is approximately 440 mm (YES), the light vehicle discriminating means determines that the vehicle is not a light vehicle And proceeds to step 390 to wait for a subsequent management event. For reference, the term &quot; judgment &quot; used in the flowcharts of FIGS. 4 and 6 is limited to the same meaning as described above.

If it is determined in step 350 that the vehicle does not match type 4 or type 5 of Tables 2, 3, 4, the vehicle is sure to be not a large vehicle once,

In step 360, the actual vehicle width Y (lateral length of the vehicle contour) of the vehicle is calculated according to the following equation (1).

(1) &quot; Y = (A1 / a) xy '

For reference, since the actual vehicle size and the photographed vehicle image are directly proportional to each other, 'Y / A = y / a' is established, and the equation (1) is derived therefrom.

Here, Y is the actual vehicle width of the photographed vehicle,

Here, y is a vehicle contour vehicle width (transverse length) of the image (measured at S320)

Here, A1 is the actual transverse length (calculated in S340) of the photographed license plate,

Here, a is the length of the vehicle license plate contour of the image (measured at S330).

When the actual vehicle width Y of the vehicle is calculated in step 360, the process proceeds to step 370. If the calculated actual vehicle width Y is 1.6 meters or less, it is determined that the vehicle is a light vehicle (step 380a) If the Y value exceeds 1.6 meters, it is determined that the vehicle is not a light vehicle (step 380b).

When the final discrimination is completed, the light-amount discrimination unit returns to the initial state of step 300 and waits, and on the other hand, when the management event corresponding to the result of discriminating the lightness difference is set in advance, .

The major management event expected in the light car discrimination system is to control the parking fee at 50% discount in conjunction with the parking charge settlement system at the paid public parking place. In the highway or pay tunnel, In the field of vehicle surveillance, it is possible to provide parking discrimination information to the corresponding system by linking with the surveillance system.

As described above, the intelligent parking discrimination system according to the present invention can intelligently analyze the vehicle image even without installing the existing sensor at all, thereby accurately and quickly discriminating whether the vehicle is light or small.

FIG. 5 is a block diagram schematically illustrating an embodiment of a standalone-based intelligent car number recognition system for achieving another object of the present invention. 5, when the vehicle enters the vehicle (step 500), the image camera captures the image (step 510), and a stand-alone (STAND-ALONE) based intelligent vehicle number recognition system according to the preferred embodiment of the present invention The control unit 520 analyzes the input vehicle image to recognize whether the vehicle is a light vehicle or not and recognizes the vehicle number.

The car number recognition system 520 includes a power supply SMPS, an INPUT interface, a minor discrimination and vehicle number recognition unit, an event management control unit, a first memory table, a second memory table, an OUTPUT interface, And a communication module.

The power supply (SMPS) receives AC or DC power and supplies the driving voltage to each configuration of the system. The input interface receives a vehicle image from a camera or an external device. The light vehicle discrimination and the vehicle number recognizing unit, which are composed of a predetermined CPU, a microprocessor, a RAM, a cache memory, or a system-on-chip (SoC), intelligently analyze inputted vehicle images to discriminate whether or not the vehicle is light, Fast and accurate recognition.

The CPU, the RAM, the cache memory or the SoC constituting the light-vehicle discrimination and the car number recognizing unit according to the present invention are configured with a mid-price specification that is not a high performance but a price competitiveness. We implemented an intelligent algorithm to make it possible.

The first memory table contains license plate size information for each license plate size type in Tables 2, 3, and 4 above. The second memory table stores the feature vector information of Korean characters or numbers to be referred to when recognizing the car number. The output interface and the communication module enable data transmission and communication with an external device, and predetermined wired and wireless networks and communication protocols can be employed. For example, RS232C, 485 communication standard and LAN can be selectively configured in the case of wired, and wireless communication (WIFI), RF, Bluetooth, Zigbee .

The event management control unit can control the overall configuration of the system and issue event-specific follow-up control signals customized to the purpose and installation place. For example, if it is determined that the vehicle is a light vehicle, the event management control unit may control the charge control system so that the vehicle can be discounted by 50% in a pay parking lot or on the road. On the other hand, if the vehicle number is recognized, the event management control unit can further check whether the vehicle is a registered member and turn on / off the breaker 530. Alternatively, the event management controller can control the vehicle number recognition information of the vehicle to an external control server 540, or may control to display the vehicle number on a predetermined indicator screen next to the breaker or gate.

The breaker 530 is not an essential component of the car number identification system according to the present invention but can be interlocked with the car number recognition system by wire or wireless using the output interface and / ), Rear gate (entrance), entrance gate of toll road or pay tunnel, and can be used to control access vehicle. Here, the breaker can be automatically turned on and off according to the application set by the administrator, such as whether the member is registered or the fare is settled. If necessary, the manager can manually turn on / off the controller.

The control server 540 is not an indispensable component of the vehicle identification system according to the present invention, but can be interlocked with the vehicle identification system by wire or wireless using the output interface and / or the predetermined communication module, , Roads and tunnels gates, vehicle surveillance systems (which can intelligently detect when a license plate of a registered vehicle is attached to another vehicle or other vehicle when theft, fines and delinquency are monitored) It can be used for management purposes.

FIG. 6 is a flowchart schematically illustrating an intelligent light vehicle discrimination and vehicle number recognition process and an algorithm of the vehicle number recognition system of FIG. 5 according to an embodiment of the present invention.

Referring to FIG. 6, in a preferred embodiment of the present invention, control logic for intelligently analyzing input vehicle images to discriminate whether a vehicle is a light vehicle or not, and control logic for further analyzing the vehicle image to recognize a vehicle number are organically combined Respectively.

The control algorithm of FIG. 6 in S600 to S690 is identical to the algorithm described in FIG. 4, and detailed description thereof will be omitted.

Next, the algorithm for further analyzing the same vehicle image and recognizing the vehicle number shares the result of the miniscule discrimination algorithm of steps 600 to 630 shown in FIG. 6, and then proceeds to step 710.

If the outline of the vehicle number is detected in step 630, then in step 710, the area inside the outline of the license plate is binarized to make the contrast of the vehicle number higher. Then, in step 720, the feature vectors of the vehicle numbers of numbers and Hangul characters are extracted (detected) from the binned license plate area images.

Here, the vehicle number feature vector is decomposed into the initial, neutral, and Arabic numerals of the Hangeul character, and the feature vector is detected for each disassembled individual unit. The Arabic numerals on the license plate consist of 0, 1, 2, 3, 4, 5, 6, 7, 8, and 9, and the Hangul characters used in car numbers are also very limited. Therefore, according to the present invention, by appropriately combining the number of inflection points, the angle of bend, and the direction of bending of individual units of numbers and letters, the feature vectors of the car number having discriminative power can be distinguished and classified by their types. This method has higher recognition accuracy and faster processing speed than the template matching method which recognizes characters according to the degree of overlapping by overlapping a predetermined template with a character or a figure of a figure (contour). In the present invention, it is possible to minimize the program capacity, improve the processing speed, and improve the vehicle number recognition rate by directly programming the feature vector-based recognition algorithm with a firmware method in which an operating system (OS) is unnecessary.

In step 730, the feature vector of the vehicle number detected in the step is compared with the feature vectors of each type set in the second memory table to recognize the corresponding vehicle number. Here, the feature vector matching is selectively applied to the installation location and use of the car number recognition system, by dividing the feature vector by the car number including the 4-digit Arabic numerals, 6 Arabic numerals, and Arabic numerals.

If the vehicle number is recognized, the process proceeds to the subsequent management event step of step 740 to wait for a predetermined management event, and returns to the initial mode (step 600 or 610) to wait for the next vehicle image input. Here, the predetermined management event can be selectively set according to the application and location to which the vehicle identification system according to the present invention is applied. For example, by interacting with an external device, it is possible to inquire whether the vehicle is registered as a member, Inquiry, breaker on / off control, vehicle number display control on the display, transmission control of the recognized car number information to the control server, and the like.

As shown in FIG. 6, in the system for simultaneously performing the miniscule discrimination and the car number recognition, since the steps 300 to 330 of FIG. 6 are organically shared, the processing speed of the algorithm and the system The efficiency is improved and the detection speed of the license plate is further improved by detecting the license plate contour in the detected vehicle contour area instead of the entire screen. Particularly, at the time of detection of the license plate area and detection of the car number feature vector in the area, it is possible to perform the search from the lower line of the detected license plate contour area upwards, so that the search (SCAN) Improvement can be doubled.

In addition, the neural network learning used in existing car number recognition systems and the car number recognition algorithms using complex mathematical expressions are very slow in terms of performance compared with performance, so they are practically used in a stand-alone system employing a system-on- However, in the present invention, in combination with the light-difference discrimination algorithm, the vehicle license plate area is quickly searched for, the car number feature vector is detected to recognize the car number, and the car number is embedded in the firmware. Based on Intelligent Vehicle Number Identification System that can improve the recognition rate and real time response even on the system - on - chip basis of the specification.

In addition, firmware-based programming that does not require an operating system (OS) is used in the design of an individual device manager module in a small-sized embedded system. However, even those having ordinary skill in the art, It will be appreciated by those skilled in the art that implementing software on a firmware basis is not easy and can only be accomplished if the expertise gained in hardware and software and the intelligent algorithms according to the present invention are organically coupled.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments. It will be understood that various modifications may be made without departing from the scope of the present invention.

100, 500: Entry or passing vehicle
200, 510: image camera 300: light vehicle discrimination means
520: Standalone-based intelligent vehicle identification system
530: breaker 540: control server

Claims (2)

A system for recognizing a light vehicle of a vehicle photographed using a camera and recognizing a vehicle number of the vehicle,
An input interface unit for receiving a vehicle image from the camera;
A first memory table for storing a ratio (A1 / B) of an actual transverse length (A1) to an actual transverse length (B) of the license plate size and an actual transverse length (A1)
A car number characteristic in which the combination of the number of break points, the angle of bend, and the direction of bending of the disassembled individual unit is decomposed into the initial and neutral and arabic numbers of the Korean character corresponding to the car number, A second memory table for storing the vector classified by the type of the Hangul character and the Arabic numeral of the car number;
Detecting a vehicle contour line from the input vehicle image and sequentially scanning the vehicle from the lower line to the upper line of the vehicle contour area to detect the license plate area and calculating the actual vehicle width Y of the vehicle according to the following equation And compares the calculated result with the reference value (1.6 meters or less) to determine whether the vehicle is a light vehicle or not,
The vehicle is scanned sequentially from the lower line to the upper side of the detected license plate area and scanned left and right with respect to the center line of the license plate area to separate the initial characters of the Korean character corresponding to the car number into neutral and arabic digits A car number identification function for recognizing the car number of the vehicle by comparing the car number feature vector extracted by analyzing the number of break points, the angle of tilt and the direction of tilt with the car number feature vector stored in the second memory table, ; And
And a management event control unit for controlling each configuration of the system and performing a predetermined management event preset based on light vehicle discrimination information and vehicle number recognition information,
Equation 1 is "Y = (A1 / a) xy"
Here, Y is an actual vehicle width (lateral length) of the vehicle,
(A / b) of the length (a) to the length (b) of the vehicle license plate outline of the vehicle image as the actual width of the vehicle license plate of the vehicle, (A1 / B) of the actual transverse length (A1) to the actual transverse length (B) stored in the vehicle license plate type information,
Here, a is a vehicle license plate contour length of the vehicle image,
Here, y is a lateral length (vehicle width) of the vehicle image,
Wherein the discrimination and control logic performed in the light vehicle discrimination, the vehicle number recognizing unit, and the management event control unit is programmed based on firmware that does not require an operating system (OS). delete

Images