KR102584182B1 - High-speed distributed vehicle number recognition system using cloud computing technology - Google Patents

Abstract

The present invention includes a video capture unit that captures images of running vehicles and transmits the captured images, receives images captured from each video capture unit in the field through a wired or wireless network, and extracts a detection area from the received images. A cloud server including a vehicle detection unit that detects a vehicle in the detection area, a license plate recognition unit that extracts a recognition area from the image when the vehicle is detected by the vehicle detection unit and recognizes the vehicle license plate in the recognition area, and server information of the cloud server It is possible to provide a high-speed distributed license plate recognition system using cloud computing technology that includes a database that stores vehicle information. According to the above, by centralizing vehicle detection and recognition functions on a cloud server, the performance of the server can be utilized to the fullest, and the cost of system construction can be reduced by making the field system lightweight, and at the same time, maintenance costs can be reduced by being free from frequent errors in the field system. Not only that, but also system integrated configuration management is possible.

Description

{High-speed distributed vehicle number recognition system using cloud computing technology}

The present invention relates to a high-speed distributed license plate recognition system using cloud computing technology. More specifically, by centralizing the vehicle detection and recognition functions of the field system on a server, the performance of the server can be maximized and the field system can be made lighter. This is about a high-speed distributed license plate recognition system using cloud computing technology that can reduce the cost of building a vehicle information collection system and at the same time significantly reduce maintenance costs by being free from frequent errors in the field system.

In general, a vehicle license plate is a plate that is attached to the front or rear of a vehicle and has standardized characters (including letters, numbers, and symbols) written on it in a specific pattern. This license plate includes the vehicle's name, manufacturer, production date, engine displacement, engine type, etc. You can check basic information, and you can also check important information such as accident history, maintenance history, and vehicle registration status while driving.

Meanwhile, vehicle license plate recognition technology is a technology that acquires a vehicle's license plate from an image and recognizes the number through an inherent algorithm, and its scope of use is diversifying as a foundation technology for the social safety net.

Recently, thanks to the increase in the market size of ITS (intelligent transport systems) and the development of ITS, license plate recognition systems are being used in illegal parking enforcement devices, vehicle speed monitoring devices, building vehicle access systems, and CCTV security systems. In consideration of the processing capacity and speed of the central server, this license plate recognition system detects the vehicle and recognizes the license plate number in each field system and transmits it to the central server for management.

However, since this conventional license plate recognition system consists of detecting and recognizing vehicles at each site and then transmitting them to the central server, it is difficult to manage from the slowdown in processing speed due to concentration of data on the central server and frequent errors in the field system. There was a weak problem.

In addition, this conventional license plate recognition system has a problem in that the reliability of vehicle detection and recognition is lowered, as the camera is installed outside and is sensitive to camera shake caused by wind or rain, and a lot of noise is generated as a result.

Republic of Korea Patent Publication No. 10-2272279

By centralizing vehicle detection and recognition functions on a cloud server, the present invention is freed from frequent errors in field systems, reducing maintenance costs, making full use of the server's performance, and improving data processing speed through an efficient vehicle detection algorithm. The purpose is to provide a high-speed distributed license plate recognition system using cloud computing technology that can lighten the field system and reduce the cost of system construction.

In addition, the present invention can effectively reduce image noise when detecting a vehicle, thereby improving the edge search effect, and can improve processing speed and reliability of vehicle detection and vehicle recognition by using cloud computing technology to improve the reliability of vehicle detection and recognition. The purpose is to provide a distributed license plate recognition system.

The present invention includes an image capture unit that captures images of driving vehicles and transmits the captured images; A vehicle detection unit that receives the images captured from each of the video capture units in the field through a wired or wireless network, extracts a detection area from the received images, and detects the vehicle in the detection area, and the vehicle is detected by the vehicle detection unit. When detected, a cloud server including a license plate recognition unit that extracts a recognition area from the image and recognizes a license plate in the recognition area; A database that stores server information and vehicle information of the cloud server; wherein the vehicle detection unit includes a resolution change unit that reduces the image by changing the resolution of the image to a first setting value to increase detection and recognition speed; , a detection area extraction unit that extracts the detection area by cutting the image whose size has been reduced due to change in resolution into a strip of the entire width of the image and a set height at a random position, and a detection area extractor that extracts the detection area to reduce noise and increase the detection effect. Color is converted to black and white, and the resolution of the detection area converted to black and white is changed to the second setting value to increase processing speed, reduce noise without reducing sensitivity, and effectively find edges for license plate recognition. It is possible to provide a high-speed distributed license plate recognition system utilizing cloud computing technology, which includes an image change unit that reduces the image size and a detection unit that detects the vehicle by determining the presence of a vehicle based on changes in pixel image values of the detection area. .

Here, the vehicle number recognition unit includes a recognition area extraction unit that extracts the recognition area by cutting the recognition area from the image, a license plate recognition unit that recognizes a license plate in the recognition area extracted using a deep learning algorithm, and the recognition area. It may include a license plate recognition unit that extracts the license plate area by cutting the area where the license plate is located, and recognizes the vehicle number in the license plate area.

The license plate recognition unit may recognize characters within the license plate area using a deep learning algorithm, and align the recognized characters from left to right based on set coordinates to complete the license plate number.

The detection unit calculates the amount of change in the pixel image value of the detection area and the amount of change in the average pixel image value of a set number of frames before receiving the image, and compares the amount of change in the pixel image value of the detection area with the amount of change in the average pixel image value. If the amount of change in the pixel image value of the detection area is greater than a set multiple of the amount of change in the average pixel image value, it is determined that there is a vehicle and the vehicle can be detected.

The number recognition unit extracts the license plate area by cutting it from the original image, recognizes characters within the license plate area in the original image using a deep learning algorithm, and sorts the recognized characters from left to right based on set coordinates to generate the vehicle number. can be completed.

The high-speed distributed license plate recognition system utilizing cloud computing technology according to the present invention can provide the following effects.

First, by centralizing vehicle detection and recognition functions on a cloud server, the server's performance can be utilized to the fullest and the field system can be made lighter, thereby reducing the cost of system construction and at the same time being free from frequent errors in the field system, thereby reducing maintenance costs. In addition, system integrated configuration management is possible.

Second, the high-speed distributed license plate recognition system using cloud computing technology according to the present invention can eliminate noise as much as possible and improve edge detection by resizing the extracted detection area in addition to resizing the initial image.

Third, unlike the conventional method of extracting a specific area from an image, the detection area is extracted by cutting the reduced image into thin strips at the pixel level and then calculating the vehicle based on the extracted detection area image to detect the vehicle, significantly increasing the throughput. It can be reduced to , and through this, the calculation processing speed can be improved.

Fourth, the detection rate of a vehicle can be improved by determining vehicle detection by comparing the amount of change in image value in the detection area and the amount of change in average image value of previous frames, and the range of detection targets such as people as well as vehicles can be expanded. there is.

Figure 1 is a diagram schematically showing the configuration of a high-speed distributed license plate recognition system using cloud computing technology according to an embodiment of the present invention.
Figure 2 is a block diagram showing the configuration of a vehicle detection unit in a high-speed distributed license plate recognition system using cloud computing technology according to an embodiment of the present invention.
Figure 3 is a block diagram showing the configuration of a license plate recognition unit in a high-speed distributed license plate recognition system using cloud computing technology according to an embodiment of the present invention.
Figure 4 is a procedure diagram showing the vehicle detection process using a high-speed distributed license plate recognition system utilizing cloud computing technology according to an embodiment of the present invention.
5 to 8 are diagrams illustrating a vehicle detection process in a vehicle license plate recognition method according to an embodiment of the present invention.
Figure 9 is a procedure diagram showing a license plate recognition process using a high-speed distributed license plate recognition system utilizing cloud computing technology according to an embodiment of the present invention.
Figures 10 to 13 are diagrams illustrating a license plate recognition process using a high-speed distributed license plate recognition system utilizing cloud computing technology according to an embodiment of the present invention.

Hereinafter, a preferred embodiment will be described with reference to the attached drawings so that the present invention can be understood in more detail.

Since the present invention can make various changes and have various embodiments, specific embodiments will be illustrated in the drawings and described in detail. However, this is not intended to limit the present invention to specific embodiments, and should be understood to include all changes, equivalents, and substitutes included in the spirit and technical scope of the present invention.

Terms such as first, second, etc. may be used to describe various components, but the components should not be limited by the terms. The above terms are used only for the purpose of distinguishing one component from another. For example, a first component may be named a second component, and similarly, the second component may also be named a first component without departing from the scope of the present invention. The term and/or includes any of a plurality of related stated items or a combination of a plurality of related stated items.

When a component is said to be "connected" or "connected" to another component, it is understood that it may be directly connected to or connected to the other component, but that other components may exist in between. It should be. On the other hand, when it is mentioned that a component is “directly connected” or “directly connected” to another component, it should be understood that there are no other components in between.

The terms used in this application are only used to describe specific embodiments and are not intended to limit the invention. Singular expressions include plural expressions unless the context clearly dictates otherwise. In this application, terms such as “comprise” or “have” are intended to designate the presence of features, numbers, steps, operations, components, parts, or combinations thereof described in the specification, but are not intended to indicate the presence of one or more other features. It should be understood that this does not exclude in advance the possibility of the existence or addition of elements, numbers, steps, operations, components, parts, or combinations thereof.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by a person of ordinary skill in the technical field to which the present invention pertains. Terms defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related technology, and should not be interpreted in an ideal or excessively formal sense unless explicitly defined in the present application. No.

Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the attached drawings. In order to facilitate overall understanding when describing the present invention, the same reference numerals are used for the same components in the drawings, and duplicate descriptions for the same components are omitted.

Referring to Figure 1, a high-speed distributed license plate recognition system utilizing cloud computing technology according to an embodiment of the present invention (hereinafter referred to as 'vehicle license plate recognition system') performs the vehicle detection and recognition functions of the field system on a server ( 200), the performance of the server 200 can be maximized, while system construction costs can be reduced by lightweighting the field system, and maintenance costs can be significantly reduced by being freed from frequent errors in the field system.

The license plate recognition system of the present invention may be configured to include an image capture unit 100, a cloud server 200, and a database 500.

First, the video capture unit 100 functions to capture images of driving vehicles and transmit the captured images.

The video capture unit 100 may be configured to include a camera unit including CCTV at each site and a communication unit that transmits the captured images.

The cloud server 200 serves to recognize and manage license plates by receiving and analyzing the images captured from each video capture unit 100 in the field through a wired or wireless network.

In detail, when the cloud server 200 detects a vehicle by comparing and analyzing the received images, it recognizes the vehicle license plate number, and a detailed description of this will be described later.

The database 500 serves to store various data such as server information of the cloud server 200, received image data, and recognized vehicle information.

Hereinafter, the cloud server 200 will be looked at in detail with reference to FIGS. 2 and 3.

The cloud server 200 includes a vehicle detection unit 300 that extracts a detection area from the received images and detects a vehicle in the detection area, and recognizes the vehicle from the image when the vehicle is detected by the vehicle detection unit 300. It may be configured to include a license plate recognition unit 400 that extracts an area and recognizes a license plate number in the recognition area.

Referring to FIG. 2, the vehicle detection unit 300 may be configured to include a resolution change unit 310, a detection area extraction unit 320, an image change unit 330, and a detection unit 340. .

First, the resolution change unit 310 changes the resolution of the image to a first setting value to reduce the image size, thereby increasing the detection speed and recognition speed.

The detection area extraction unit 320 serves to extract a detection area by cutting a random position from the image whose resolution has been changed and the size has been reduced by the resolution change unit 310.

The image change unit 330 serves to reduce noise by changing the image value of the detection area extracted by the detection area extraction unit 320.

Here, the image change unit 330 may be configured to convert the color of the detection area into black and white and change the resolution of the detection area whose color has been converted to black and white to a second setting value to reduce the image.

The detection unit 340 may detect the vehicle by determining that there is a vehicle based on a change in pixel image value of the detection area.

In detail, the detection unit 340 detects the vehicle by determining that there is a vehicle based on the change in the pixel image value of the detection area, the amount of change in the pixel image value of the detection area, and the set number of frames before receiving the image. The average pixel image value change is calculated, and the detection area pixel image value change amount is compared with the average pixel image value change amount. If the detection area pixel image value change amount is greater than a set multiple of the average pixel image value change amount, it is determined that a vehicle is present. The vehicle can be detected by making a decision.

The license plate recognition unit 400 serves to extract a recognition area from the image when a vehicle is detected by the vehicle detection unit 300 and recognize the license plate in the recognition area.

The vehicle number recognition unit 400 may be configured to include a recognition area extraction unit 410, a license plate recognition unit 420, and a license plate recognition unit 430.

The recognition area extractor 410 serves to extract a recognition area by cutting a set position from an image in which it is determined that a vehicle has been detected.

The license plate recognition unit 420 serves to recognize a license plate in the recognition area extracted using a deep learning algorithm.

The number recognition unit 430 serves to extract a license plate area by cutting the area where the license plate is located in the wounded machine recognition area and to recognize the vehicle number in the license plate area.

Here, the number recognition unit 430 may be configured to include a license plate recognition unit that aligns the letters from left to right to complete the letters when letters representing the vehicle number are recognized in the license plate area.

The number recognition unit 430 extracts the license plate area by cutting it from the original image, recognizes characters within the license plate area in the original image using a deep learning algorithm, and then moves the recognized characters from left to right based on set coordinates. You can complete the vehicle number by sorting.

Meanwhile, the details of each configuration of the license plate recognition system of the present invention described above will be examined in more detail in the license plate recognition method using the license plate recognition system of the present invention below.

First, in the license plate recognition method according to an embodiment of the present invention, the cloud server 200 extracts a detection area from the received image, detects a vehicle in the detection area, and when the vehicle is detected in this way, a recognition area is selected from the image. It can be configured by extracting and recognizing the vehicle number in the recognition area.

Referring to FIG. 4, when the cloud server 200 receives an image from the video capture unit 100 to detect a vehicle (S110), the cloud server 200 changes the received image to increase detection speed through the resolution change unit. The image can be reduced by changing the resolution to the first setting value (S120).

Thereafter, the detection area extraction unit 320 extracts the detection area from the image whose resolution has been changed and the size has been reduced (S130).

Here, the detection area extraction unit 320 can extract the detection area by cutting the reduced image into a strip shape with a full width and a set height (in pixel units) at an arbitrary position. This is to enable the detection unit 340 to more effectively determine the presence of a vehicle, considering that it detects the vehicle based on changes in pixel image values of the detection area, and further to improve processing speed by reducing the processing amount.

Meanwhile, the detection area extraction unit 320 can set the extraction position to an arbitrary position when extracting the detection area, and when the position is set in this way, the detection area can be extracted by cutting it into a strip shape at the same setting position.

When the detection area is extracted in this way, the image change unit 330 can reduce noise in the detection area by changing the image of the detection area (S140).

Changing the image of the detection area (S140) involves converting the color of the detection area to black and white (S141) and reducing the image by changing the resolution of the detection area whose color has been converted to black and white to a second setting value (S142). can do.

Here, the image change (S140) can achieve the effect of reducing noise by first converting the color of the detection area to black and white before changing the second resolution, and in the later step, when calculating the pixel image value, the brightness (brightness) is centered. Considering the calculation, color is not only unnecessary, but can also improve detection effectiveness.

Meanwhile, the second setting value can be set to the entire width (horizontal) and one pixel height (vertical).

Figure 5 is a diagram illustrating an image change process according to the above-described first resolution change step, detection area extraction step, and color conversion step.

Referring to the drawings, Figure 5(a) is a photo showing the received image, and the resolution at this time is 1920×1080. Figure 5(b) is a photo showing the state in which the resolution of the received image has been changed, which is the first resolution change step. At this time, the resolution is reduced to 480×270.

Figure 5(c) is a photograph in which a detection area is extracted by cutting out an area at a random location from an image reduced by changing resolution. Here, it is desirable to cut the detection area into a strip shape to have a resolution of 480×5. This is because it is a size that can prevent noise and sensitivity reduction and can attenuate noise without reducing sensitivity.

Figure 5(d) is a photograph showing the state in which the color of the detection area has been converted to black and white (RGB -> GRAY) in the color conversion step.

According to the above, in the present invention, the resolution is changed twice through a first resolution change (S120) that changes the resolution of the received image and a second resolution change (S1420) that changes the resolution of the extracted detection area. It is configured to do so. Accordingly, the present invention can increase processing speed by reducing processing amount, and can find edges more effectively by attenuating noise without reducing sensitivity.

After reducing the noise in the detection area according to the above process, the vehicle can be detected in the detection area, and the presence of a vehicle can be determined based on changes in pixel image values in the detection area.

Here, the pixel image value is a value including characteristic values of the corresponding pixel, and may include a brightness value converted to black and white and a color value (saturation value).

In detail, the vehicle detection calculates the amount of change in the pixel image value (pixel change amount, CPCA, current pixel change amount; detection area pixel change amount in the current frame) of the detection area (S151), and calculates the change amount of the pixel image value (pixel change amount, CPCA, detection area pixel change amount in the current frame) of the detection area (S151), The average pixel image value change is calculated (S152), and the vehicle can be detected by comparing the average pixel image value change with the pixel image value change in the detection area (S153).

Here, the detailed process of detecting a vehicle by comparing the amount of change in the pixel image value of the detection area and the average amount of change in the pixel image value is the average of the amount of change in the pixel image value of the detection area in the current frame and the amount of change in the pixel image value of the detection area in the set frames. Vehicles can be detected by comparing averages.

In addition, the average pixel image value change can be calculated from a table (PCAT; Pixel change amount table, circular cue) that collects the pixel image value changes of 30 frames before image reception (FIG. 7 reference).

Meanwhile, in the step of detecting the vehicle (S153), if the amount of change in the pixel image value of the detection area is greater than twice the amount of change in the average pixel image value, it is determined that there is a vehicle and the count is increased by 1 (S154), and this accumulated count is When it reaches 2 (S155), it is determined that the vehicle has been detected, and the detection count is initialized to 0 and then moves on to the next step, the license plate recognition step.

Here, the reason for setting the condition in which the amount of change in the pixel image value of the detection area is greater than 2 times the amount of change in the average pixel image value in increasing the count 1 is that the amount of edge change due to shaking is less than 2 times as an experimental value. This is because the amount of change in license plates with many edges occurs at more than twice the rate.

Meanwhile, unlike the above, if the amount of change in the pixel image value of the detection area is equal to or smaller than twice the amount of change in the average pixel image value, the count is initialized to 0 (Detect count 0) and the amount of change in the pixel image value of the detection area is entered into PCAT. Add.

Figure 6 is a diagram illustrating the process of changing the resolution of the detection area and calculating the amount of change in pixel image value of the detection area in the second resolution change step.

Referring to FIG. 6, the present invention can change the size before measuring the amount of change in the image (video) by converting and resizing the previously 480×5 detection area image to 30×1.

Here, the reason for resizing the detection area image to 30×1 is to increase the calculation speed without reducing the detection rate because the noise is greatly attenuated but the vertical edges for license plate recognition do not disappear much.

When the resolution of the detection area is changed and resized in this way, the amount of change in the pixel image value of the detection area is calculated for the calculation for edge detection. This process, as shown, changes the value of the corresponding index and the absolute value of the right index. You can calculate the overall average. At this time, if there is no change in the image (video), the average will be small, and the more edges are observed, the larger the value may be.

Figure 7 is a diagram showing the process of calculating the change in pixel image value and comparing the change in pixel image value in the detection area with the change in average pixel image value. Figure 7 is a diagram showing the process of calculating the change in pixel image value in the current detection area and comparing the average of the change in the previously calculated amount. It shows the process of calculating the average of the average image value change of 30 frames.

Next, Figure 8 is a diagram showing the process of updating the PCAT when the change amount obtained above is small. Referring to the figure, the index value of the PCAT is moved forward one space at a time, and at the end, the new change value currently obtained is added. (The index values shown in the drawing are arbitrary values for illustration purposes only.)

When a vehicle is detected through the above process, it goes through a process of recognizing the license plate number of the detected vehicle, which will be discussed with reference to FIG. 9.

Referring to FIG. 9, in the step of recognizing the license plate number, when a vehicle is detected in the above process, a recognition area is extracted by cutting a set position from the detected image (S210). This recognition area extraction process is shown in FIG. 10. As shown, the recognition area can be extracted by cutting the set position where the license plate is located from the detected image.

When the recognition area is extracted in this way, the license plate can be recognized in the recognition area extracted using a deep learning algorithm (S220), as shown in FIG. 11.

In this process, if it is determined that it is a license plate as a result of recognition (S221), the license plate area is extracted by cutting the area where the license plate is located in the recognition area as shown in FIG. 12, and the vehicle number is recognized in the license plate area (S230).

Here, in the extraction of the license plate area, the license plate area found in a low-resolution image (image) is cut from the original image, which allows more characters to be found in the original image than in the low-resolution license plate image.

Meanwhile, in the recognition of the license plate number (S230), the license plate area is cut and extracted from the original image (S231), and the license plate area in the original image is extracted using a deep learning algorithm as shown in (a) of FIG. 13. The characters can be recognized (S232), and as shown in (b) of FIG. 13, the recognized characters can be aligned from left to right based on the set coordinates to complete the vehicle number (S234).

The present invention has been described with reference to the embodiments shown in the drawings, but these are merely exemplary, and those skilled in the art will understand that various modifications and equivalent other embodiments are possible therefrom. Therefore, the true scope of technical protection of the present invention should be determined by the technical spirit of the attached patent claims.

100: video recording unit 200: cloud server
300: vehicle detection unit 310: resolution change unit
320: Detection area extraction unit 330: Image change unit
340: detection unit 400: vehicle number recognition unit
410: Recognition area extraction unit 420: License plate recognition unit
430: Number recognition unit 500: Database

Claims (5)

a video capture unit that captures images of driving vehicles and transmits the captured images;
A vehicle detection unit that receives the images captured from the on-site video capture unit through a wired or wireless network, extracts a detection area from the received images, and detects the vehicle in the detection area, and the vehicle is detected by the vehicle detection unit. a cloud server including a license plate recognition unit that extracts a recognition area from the image and recognizes a license plate number in the recognition area;
Including a database that stores server information and vehicle information of the cloud server,
The vehicle detection unit,
a resolution change unit that changes the resolution of the image to a first setting value to reduce the image in order to increase detection and recognition speed;
A detection area extraction unit that extracts a detection area by cutting the image whose resolution has been changed and its size reduced to reduce noise without reducing sensitivity into a strip of the entire width of the image and a set height at a random location;
In order to reduce noise and increase detection effectiveness, the color of the detection area extracted in the form of a band is converted to black and white, and the color is converted to black and white to increase processing speed, reduce noise without reducing sensitivity, and effectively find edges for license plate recognition. an image changer that reduces the image by changing the resolution of the detection area extracted in the form of a converted band to a second setting value;
It includes a detection unit that detects a vehicle by determining the presence of a vehicle based on changes in pixel image values of the detection area extracted in the form of a band,
The vehicle number recognition unit,
a recognition area extraction unit that cuts and extracts a recognition area from the image;
A license plate recognition unit that recognizes a license plate in the recognition area extracted using a deep learning algorithm,
A license plate recognition unit that extracts a license plate area by cutting an area where the license plate is located in the recognition area and recognizes a vehicle license plate number in the license plate area,
The detection unit,
Calculate the change in pixel image value of the detection area extracted in the form of a band and the change in average pixel image value of the detection area for a set number of frames before receiving the image, respectively, and calculate the change in pixel image value of the detection area and the change in pixel image value of the detection area for the set number of frames before receiving the image. Comparing the amount of change in the average pixel image value of the area, if the amount of change in the pixel image value of the detection area is greater than a set multiple of the amount of change in the average pixel image value of the detection area, the cloud is characterized in that it determines that a vehicle is present and detects the vehicle. High-speed distributed license plate recognition system using computing technology.
delete According to claim 1,
The vehicle number recognition unit,
High-speed distributed vehicle number using cloud computing technology, which recognizes characters in the license plate area using a deep learning algorithm and sorts the recognized characters from left to right based on set coordinates to complete the vehicle number. Recognition system.
delete According to claim 1,
The number recognition unit,
The license plate area is cut and extracted from the original image, characters within the license plate area are recognized in the original image using a deep learning algorithm, and the recognized characters are aligned from left to right based on set coordinates to complete the vehicle number. A high-speed distributed license plate recognition system using cloud computing technology.

Images