KR100413382B1 - System and Method for Integrated Remote Control of Watching Camera using Communication Network and Method for Processing Image Data Using the Same - Google Patents

Abstract

Integrated remote control system and method for surveillance cameras using a communication network that can remotely control and integrate various types of surveillance cameras using a communication network, and process image information of vehicles captured by the surveillance cameras in real time, and process image data using the same. Give a way.

The camera remote control system includes at least one or more cameras installed to monitor and capture the flow of the vehicle, mounted on the camera, and stored by the camera, and connected by a black box and a camera to control the camera, The camera-related information, the image information input from the black box, and the specific vehicle-related information are stored, and a central control unit for determining an optimal control condition for controlling the camera and processing the image information input from the black box is provided.

Such a system can centrally monitor the camera, efficiently control the camera remotely, and efficiently recognize the license plate of the vehicle with excellent image recognition rate.

Description

System and Method for Integrated Remote Control of Watching Camera using Communication Network and Method for Processing Image Data Using the Same}

The present invention relates to a surveillance camera control system and a data processing method using the same. More particularly, a surveillance camera for monitoring the flow of a vehicle is remotely controlled using a communication network, and the image of the license plate of the vehicle is recognized. The present invention relates to an integrated remote control system and method of a surveillance camera using a communication network for efficiently extracting data, and an image data processing method using the same.

In general, surveillance cameras used to recognize and manage a vehicle, such as monitoring the flow of a vehicle and capturing a specific scene, are installed on a roadside or on a road.

In order to control the surveillance cameras, the administrator directly accesses the camera or remotely controls only a part of the camera by using a specific communication means only for the camera. Therefore, in order to control a specific camera, since only a terminal having a corresponding communication means is available, heterogeneous equipment cannot be integrated and thus there is a problem in that the scope of remote control is very limited.

In addition, the remote control communication network does not have universality and immediately transmits a specific vehicle such as a wanted vehicle to the police network as soon as the camera recognizes it, or immediately delivers it to the owner by wire or wirelessly. There are many limitations in the use of the camera system because it cannot perform a function.

Meanwhile, film is used to store and output image data photographed using a surveillance camera. In this case, the film is used to store the image of the violating vehicle, so the film must be continuously consumed for continuous vehicle control. As a result, continuous operation costs are incurred, and the manager repeatedly accesses a remote camera and starts and stops the camera only when necessary. In order to facilitate the operation and stop of such a camera, some cameras are remotely controlled, but even in this case, the consumption of film cannot be solved.

The file system is a way to solve the problem of costly film. However, in the case of a camera using a file system, since different types of communication protocols are used according to the type of camera, the consumption of film can be suppressed, but there is a problem that it is difficult to integrate and control the camera from the center.

On the other hand, the license plate recognition device installed in the camera for the purpose of detecting the speed of the vehicle, the current situation is not high accuracy for the license plate recognition. In addition, in case of rain, the recognition rate of the license plate is drastically reduced, so the camera simply detects the speed violation, and photographs the vehicle in violation of the license plate, and the manager directly identifies the license plate and takes action accordingly.

As such, the current surveillance camera control system is difficult to control integrally, such as turning on and off various cameras, and takes a long time from detecting a speed violation by the camera to notifying the vehicle, and in case of a specific event The license plate automatic recognition by a camera is impossible, and the installed surveillance camera is not utilized efficiently.

Therefore, the following matters are currently required in the control of the surveillance camera. First, it is very important for government agencies to arrest the wanted vehicles faster for the purpose of public security. Therefore, it is necessary to improve the accuracy of the license plate recognition of the camera so that the wanted vehicle can be arrested immediately, and the license plate recognition information of the camera is transmitted to the police network in real time.

Second, the owner of the vehicle has a great interest in whether or not the driver who has rented the car has been caught by the speed detecting camera. Therefore, most vehicle drivers want to search these items in real time or check them using various communication means.

Third, parking lots, apartments, and companies that have registered specific vehicles require services that can encourage rapid communication in various congestion time zones by distinguishing external vehicles from internal vehicles. In this case, a system with a fast license plate recognition speed is needed.

Fourth, a new service using the registration information for the general vehicle may be possible. For example, there is a need for a system and a service capable of determining whether a vehicle passes by using only accurate license plate recognition without attaching a vehicle identification sticker at a highway toll gate.

Fifth, by analyzing the control information of the vehicle and recommending an optimal camera control method, an auxiliary function is required for a client to perform a more efficient control action by referring to the recommended function.

In addition, various services have recently been required for surveillance cameras, but the systems developed to date do not fulfill these recent demands, and therefore, there is an urgent need for systems and methods that can solve the above-described needs.

The present invention has been made to solve the above-mentioned problems, surveillance camera using a communication network that can be controlled remotely by converting all kinds of cameras into a single protocol without being restricted by time and space by remote control using a communication network The technical challenge is to provide an integrated remote control system and method.

Another technical problem of the present invention is to recognize a license plate of a specific designated vehicle such as a speeding vehicle or an arranging vehicle at a high speed, and to integrate a remote control system and method for a surveillance camera using a communication network capable of transmitting to all parties through a communication network; An image data processing method using the same is provided.

1 is a configuration diagram of a remote control system of a surveillance camera according to the present invention.

FIG. 2 is a detailed configuration diagram of the black box shown in FIG. 1.

3 is a detailed block diagram of the central control unit shown in FIG. 1.

4 is a flowchart illustrating a remote control method of the surveillance camera according to the present invention.

5 is a flowchart illustrating an image data processing method in a remote control system of a surveillance camera according to the present invention.

FIG. 6 is a flowchart for describing in detail an edge detection algorithm in the image data processing method illustrated in FIG. 5.

7 is a block diagram illustrating an edge detection algorithm of the present invention.

8 is a flowchart illustrating a vehicle number extraction process of the image data processing method illustrated in FIG. 5.

9 is a detailed flowchart of a vehicle number extraction process shown in FIG. 8.

10 is a diagram illustrating an example of classifying a character area by the character extraction module of the present invention.

11 is a view for explaining an example of applying the neural network algorithm to the license plate recognition process of the present invention.

12 is a block diagram illustrating a data fusion algorithm during a license plate recognition process of the present invention.

<Description of the symbols for the main parts of the drawings>

10-1 to 10-n: camera 20: central control unit

30-1 to 30-m: control terminal 40: communication network

50, 50-1 to 50-n: Black box 60: Buffer part

62: image buffer 64: extract image buffer

66: environmental data buffer 68: central processing unit

70: storage means 72: processing result storage means

74: program storage means 76: interface unit

78: web server 80: application program interface

82: Application 84: Database Management System

86: administrator authentication program 88: control information conversion program

90: control recommendation program 92: control program

93: image data extraction program 94: database

95: vehicle location program 96: manager DB

98: camera DB 100: image DB

102: vehicle DB 104: control DB

106: background image 108: input image

110: comparison unit 112: difference image

114: split image

The present invention for achieving the above-described technical problem is at least one camera installed to monitor and shoot the flow of the vehicle; It is mounted on the camera, monitors the control information for controlling the camera, stores the vehicle image information taken by the camera and vehicle license plate image information extracted from the vehicle image, black for controlling the camera box; And a location information and control information of the camera, image information input from the black box, and information on a specific vehicle requiring location, which are connected to the camera by a communication network and are stored in a rule for controlling the camera. And a central control unit for determining an optimum control condition and receiving and processing image information from the black box.

In addition, the present invention stores at least one camera, vehicle-wide image information mounted on the camera and taken by the camera and vehicle license plate image information extracted from the vehicle-wide image, the camera at the request of the camera administrator A black box for controlling, the central control unit connected to the camera and a communication network to store the camera information and the image information of the vehicle transmitted from the black box and determine an optimum control condition according to a rule for controlling the camera; A method for remotely controlling the camera in a remote control system of a surveillance camera, comprising: a first step of authenticating the manager as a manager who wants to control the camera is connected through a communication network; A second step of deciding whether to arbitrarily select a camera to be controlled from an administrator who has successfully authenticated or call and select a control recommendation processor provided from the central controller; A third step of providing the selected camera information to the manager when the manager arbitrarily selects a camera to be controlled; A fourth step of controlling the camera according to a control command input by the manager and storing the control command input by the manager; In the second step, the manager calls the control recommendation processor, and when control is requested from a specific camera, displays information of the camera requesting the control and an optimum control command determined by the control recommendation processor, and then proceeds to the fourth step. Proceeding; Determining a camera requiring control by the control recommendation processor when the manager calls a control recommendation processor in the second step and the administrator wants to search for a camera requiring control; And providing the manager with an optimal control command determined by the control recommendation processor for the determined camera, and proceeding to the fourth step.

In addition, the present invention is a method for processing the image data taken by the camera installed to monitor the flow of the vehicle to shoot, the pixel containing the brightness, saturation, color information of the entire image of the vehicle taken by the camera An edge detection process for separating a license plate image from the entire image of the vehicle by analyzing a pattern of input data received in a manner; And a vehicle number extraction process for extracting each character region from the license plate image, extracting possible characters by a plurality of neural network algorithms, and outputting the final character by combining the results of the neural network algorithms.

The present invention proposes an intelligent black box and a central control unit that process image data input through a camera and control main functions of the camera. In particular, in order to efficiently recognize and manage a large number of vehicles using a camera, image data is used to recognize a license plate of a vehicle and to quickly distinguish a license plate of a specific vehicle, and this information is used for a general communication network such as a wired or wireless Internet. To be effectively communicated to stakeholders.

In addition, we propose an algorithm (data fusion method) with a significantly better recognition rate of the license plate than the current method for processing image data, and the remote control method using a communication network for the purpose of reliability and versatility in camera remote control. do. In other words, the present invention provides a method for centrally monitoring fixed and mobile cameras using a general-purpose communication network such as the Internet, efficiently remotely controlling camera operation, and efficiently recognizing a license plate of a vehicle with excellent image recognition rate. To provide.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Surveillance camera integrated remote control system

1 is a structural diagram of a surveillance camera integrated remote control system according to the present invention.

As shown, the surveillance camera integrated remote control system includes at least one camera 10-1 to 10-n, a central control unit 20 and at least one control terminal 30-1 to 30-m, They have an environment in which they can communicate with each other via the communication network 40. Here, the communication network 40 means a general purpose network such as the Internet. In addition, the cameras 10-1 to 10-n include black boxes 50-1 to 50-n, respectively.

The manager controls the operation of each of the cameras 10-1 to 10-n by connecting to the central control unit 20 via the communication network 40 using the control terminals 30-1 to 30-m. The cameras 10-1 to 10-n are equipped with black boxes 50-1 to 50-n, respectively, and can be remotely controlled using the communication network 40. -1 to 10-n) can be controlled. In addition, it is possible to control the camera more efficiently using a recommendation processor that provides an optimal control method such as determining the driving time of the camera by analyzing the vehicle flow by the necessary conditions based on statistical data on the vehicle flow.

On the other hand, the black boxes 50-1 to 50-n have a buffer having a temporary storage function, and if necessary, transmit the entire image data of the vehicle including the license plate of the vehicle to the central control unit 20 using the communication network 40. Thus, the image information of the vehicle can be processed in real time. In addition, the cameras 10-1 to 10-n detect the speed of the vehicle while automatically monitoring the flow of the vehicle, thereby automatically recognizing the vehicle in violation of the speed, and immediately transmit the image data of the vehicle to the communication network 40. Through the central control unit 20 can be delivered.

In addition, when information about a specific registered vehicle such as an arrangement vehicle is input to the central control unit 20 through the control terminals 30-1 to 30-m, the central control unit 20 uses the black box 50-1 using the same. When analyzing the image data transmitted from ˜50-n), the position of the wanted vehicle is determined by applying the received vehicle data.

FIG. 2 is a detailed configuration diagram of the black box shown in FIG. 1.

The black box 50 is a remote control device installed in the cameras 10-1 to 10-n in order to monitor and control various control information, and through this, the camera 10-1 to 10-n information required by the administrator. Can monitor and control the cameras 10-1 to 10-n in a request form.

That is, the black box 50 is directly connected to each of the cameras 10-1 to 10-n to control the operation of the cameras 10-1 to 10-n, and the result of the work is centered through the communication network 40. It serves to transmit to the control unit 20, the data related to the image taken by the central processing unit 68, the cameras (10-1 to 10-n) for controlling the overall operation of the black box 50 temporarily A buffer unit 60 for storing data, a program for processing image data, a storage unit 70 for storing necessary data according to a request of the central control unit 20, and an interface unit for connection with the communication network 40 (76).

Here, the buffer unit 60 extracts only the license plate image data from the image buffer 62 and the overall image data of the vehicle which temporarily store the overall image data of the vehicle photographed by the cameras 10-1 to 10-n. And an extracted image buffer 64 for temporarily storing an image and an environment data buffer 66 for storing an environment value determined by weather factors such as camera related information, vehicle related information, and weather at the time of shooting. .

In addition, the storage means 70 extracts processing result storage means 72 and image data in which the overall image data of the vehicle and the image data related to the vehicle license plate are stored in order to provide necessary data according to the request of the central controller 20. And program storage means (72) in which the stored program is stored.

3 is a detailed block diagram of the central control unit shown in FIG. 1.

As shown, the central control unit includes a web server 78 and a database 94. The web server 78 includes an application program interface unit (API) 80, an application program 82, and a database management system (DBMS, 84). Although not shown, the web server 78 includes an interface for connecting to a communication network. The web server 80 operates as a server for monitoring and controlling each of the surveillance cameras 10-1 to 10-n, and accesses the application program 84 through the API 82 to access each control terminal 30-1 to Each camera 10-1 to 10-n is controlled by the manager's request through 30-m).

Here, the application program 82 is the administrator authentication program 86, the control information conversion program 88, the control recommendation program 90, the control program 92, the image data extraction program 93 and the vehicle positioning program ( 95). In addition, the database 94 includes an administrator DB 96, a camera DB 98, an image DB 100, a vehicle DB 102, and a control DB 104, each of which consists of one table. It may be present or may consist of multiple tables.

When the administrator who wants to control the specific cameras 10-1 to 10-n approaches the central control unit 20 through the control terminals 30-1 to 30-m, the manager authentication program 86 may include an administrator DB ( 96, it checks whether it is a legitimate administrator and permits access to specific cameras 10-1 to 10-n.

When a command is input through the control terminals 30-1 to 30-m, the control information converting program 88 converts the control information into corresponding camera-specific control information with reference to the camera DB 98.

The control recommendation program 90 analyzes the vehicle flow by the necessary conditions based on statistical data on the vehicle flow, determines the optimal control conditions such as the on / off time of the cameras 10-1 to 10-n, and presents them to the manager. It is a program.

The control program 92 directly controls the camera with reference to the camera DB 98 and the control DB 104 when a control command is input from an administrator or a control command is derived by the control recommendation program 90. The camera DB 98 includes model data indicating models of all the cameras 10-1 to 10-n that can be remotely controlled, control data indicating various control functions of the cameras 10-1 to 10-n, and a camera 10-. 1 to 10-n), operational data indicating operational information, and the like. In addition, the control DB 104 includes various control data necessary for operation of the cameras 10-1 to 10-n and image recognition.

The image data extraction program 93 stores and manages the vehicle image photographed by the camera and vehicle license plate image information extracted from the photographed vehicle in the image DB 100. In addition, the number of the actual vehicle is extracted by the neural network algorithm and the data fusion algorithm from the vehicle license plate image.

The vehicle location program 95 checks whether the vehicle recognized by the black box 50 is a specific vehicle recorded in the vehicle DB 102 by referring to the vehicle information stored in the vehicle DB 102, and determines whether the vehicle is a specific vehicle. Check the current location of the vehicle.

As described above, in order to achieve the purpose of efficient remote control and image recognition of the camera, the central control unit of the surveillance camera integrated remote control system of the present invention, the information about the various cameras to be controlled, the position information and control information of the camera A database 94 for storing data relating to the data and the like, an application 82 for making a decision on a control recommendation by defining rules for control using the database, and sending the determined instructions to the black box 50 of the camera. Interface means (not shown) for delivery.

In addition, the control terminals 30-1 to 30-m are provided with communication means for connecting to the central control unit 20 through the general-purpose communication network 40, and searches the database 94 of the central control unit 20, and the manager. A program, such as a web browser, which allows the command inputted by the user to be transmitted to the central control unit 20, is mounted.

Surveillance Camera Integrated Remote Control Method

4 is a flowchart illustrating a remote control method of a surveillance camera according to the present invention. Referring to FIGS. 1 to 3 again, the operation of the surveillance camera integrated remote control system according to the present invention will be described.

When the administrator connects to the central control unit 20 through the control terminals 30-1 to 30-m, the central control unit 20 permits or disallows control of the cameras 10-1 to 10-n by an administrator authentication process. It is determined whether or not (S101).

The administrator who receives the control permission accesses the interface unit 76 of the black box 50 through the central control unit 20 to check whether the corresponding cameras 10-1 to 10-n are operated and perform the necessary remote control. . To this end, the administrator first selects cameras 10-1 to 10-n to be monitored and controlled, or selects and controls the recommended processor (S102).

When the administrator arbitrarily selects the cameras 10-1 to 10-n to be controlled, the database of the camera is searched and displayed on the control terminals 30-1 to 30-m, and the administrator monitors them and then selects the appropriate ones. It waits until a control command is input (S103).

When the control command is input from the administrator, the control key is provided to perform the same functions of the remote cameras 10-1 to 10-n, and the control command is executed (S104). It is stored in the database (S105).

If the manager wants to control the cameras 10-1 to 10-n by the recommended processor in step 102, it is determined whether the call originated from the camera or from the manager (S106). If the determination result is a call by the camera, that is, when control is requested from specific cameras 10-1 to 10-n, a command for monitoring and controlling the camera is displayed (S107) and the process proceeds to step S103. Perform camera control by control command.

On the other hand, in the case of a call by the administrator, that is, when the administrator wants to search for a camera to be controlled by inputting a specific condition, the camera database is searched (S108). Next (S109). Thereafter, the result of the determined optimal control command is displayed on the web screen (S110), and the process proceeds to step S103 to perform the control command.

Image data processing method using surveillance camera integrated remote control system

A process of processing image data of the vehicle photographed by the surveillance camera is as follows.

When a moving vehicle is photographed by a camera and image data of the vehicle is input, the image data becomes image data of the entire vehicle including the license plate of the vehicle. This data is first transmitted to the image buffer 12 of the black box, and the central processing unit 68 performs edge detection operation to separate only the license plate image data therefrom. Transmission to the control unit 20.

The central control unit 20 extracts the license plate from the entire image of the vehicle using the image data extraction program 93 for the license plate recognition device, and accordingly the image of the entire vehicle stored in the image data buffer of the black box 50. Determine whether to receive or just delete and transmit to the black box (50).

The black box 50 stores the image data of the entire vehicle in the image buffer 62 until the command is received from the central control unit 20, and then transmits the image data of the entire vehicle from the central control unit 20. When requested or deleted, the command is executed accordingly. When the central control unit 20 requires an image of the entire vehicle, when the license plate image extraction result is a speed violation vehicle, when the penalty notice is issued, the central control unit 20 must provide the vehicle to the corresponding owner as evidence.

5 is a flowchart illustrating an image data processing method using a remote control system of a surveillance camera according to the present invention.

As the entire image of the vehicle is input by the camera (S201), the process of processing the image data is largely divided into two steps. The first is an edge detection process of separating the license plate from the entire image of the vehicle photographed by the camera (S202), and the second is a number extraction process of recognizing the actual number from the license plate image (S203).

In the present invention, the edge detection process is performed in the black box 50 of the camera, and the number extraction process is performed in the central control unit 20, because the image processing is divided into the central control unit 20 and the black box 50. In order to improve the work efficiency, and to significantly reduce the data processed by the black box 50 to transmit to the central control unit 20 to reduce the amount of data transmission. However, the present invention is not limited thereto, and the edge detection process and the number extraction process may be performed by the black box 50 or the central control unit 50 as necessary. Such modifications will be apparent to those skilled in the art.

More detailed description is as follows. The image data input by the camera is input to the image buffer 62 of the black box 50 and the edge detection operation is performed by the CPU 68. This edge detection operation is performed by analyzing a pattern of data coming in a pixel method as in the conventional pattern recognition method. In the present invention, in particular, all three types of patterns of input data are brightness, saturation, and color. To include.

When the image of the license plate recognized in this manner is transmitted to the central control unit 20, the central control unit 20 outputs the final vehicle number in a data fusion method. In addition, the result is compared with the vehicle DB 102 to distinguish whether or not the entire image of the vehicle is necessary and transmits the need to the black box 50. The black box 50 that receives the need for the entire vehicle image from the central controller 20 transmits the entire vehicle image stored in the image data buffer to the central controller 20 according to the request of the central controller 20 or by itself. Save or delete.

FIG. 6 is a flowchart for describing in detail an edge detection algorithm in the image data processing method illustrated in FIG. 5.

The edge detection algorithm is an algorithm that extracts only the license plate area from the entire image of the vehicle first recognized by the camera and transmits the license plate area to the central control unit 20. The first is a vehicle detection module (S20) for extracting a vehicle image from the image of the background when shooting, and the second is a license plate detection module for detecting the position of the license plate in the extracted vehicle image (S21), such a module is a black box It is mounted in the program storage means 74 of 50.

The vehicle detection module S20 uses an adaptive background method using a temporal median suitable for real time image processing. That is, the background image without the vehicle is compared with the instantaneous input image, and the image having the difference is separated and used as described with reference to FIG. 7. 7 is a block diagram illustrating a video data processing method of the present invention.

The vehicle detection module S20 first divides the input image 108 into units of pixels (S301), and compares each pixel with the background image 106 by a comparator (TM, 110) (S302). (Dt) is obtained (S303). Thereafter, the absolute value of the difference image Dt is binarized to obtain a divided image Mt (S304). That is, with respect to the input of the image signal, the background image Bt is updated by each frame through pixel-by-pixel comparison. In addition, the divided image Mt may be obtained by binarizing an absolute value of the difference value Dt between the input image It and the background image Bt.

When the image is detected in this manner, the license plate detection module S21 for detecting the position of the license plate of the vehicle is performed using the image.

The license plate detection module S21 is a module for detecting the position of the license plate in the schematic range of the vehicle obtained in the previous vehicle detection module S20. This module uses the property of writing on a single color plate, and observes the change on the X and Y axes along with the change in the property, and extracts the large change to the region of interest.

The specific execution process of the vehicle detection module is as follows. First, by using the environment information stored in the environment data buffer 66, for example, if it is a rainy or foggy day, all pixel unit data of brightness, saturation, and color are set to an environment value of 1; In step S305, pixel data of brightness is stored by setting an environment value to zero.

If the environment value is 1, the brightness accumulation value is listed on the Y-axis, and the area where the pattern change is severe is found. If the license plate image is not detected only by the brightness value, the pattern change of the brightness accumulation value is stored (S306), and the license plate image is found using the saturation accumulation value. If the license plate image is not detected even with the saturation value, the pattern change of the saturation accumulation value is stored (S307), and the vehicle image is detected using the color accumulation value. When the license plate image is not detected even when the color accumulation value is used, the pattern change of the color accumulation value is stored (S308), and the pattern change of each accumulation value is summed to find the most changeable area on the Y axis (S309). .

When the area on the Y axis is detected in this manner, the above steps are repeated in accordance with the pattern change of brightness, saturation, and color accumulation values in the X axis direction with respect to the detected area (S310, S311, S312). The X area is detected according to the change (S313). Thereafter, the final X and Y regions are found by the detected X and Y regions (S314). Here, the X and Y regions may be detected by performing the X region detection processes S310 to S313 first, and then performing the Y region detection processes S306 to S309.

The edge detection algorithm used in the present invention is characterized by considering all three types of input data: brightness, saturation, and color. In other words, by applying the environmental value of the weather in each pixel, the end point of the region having a special aspect was recognized as an edge by classifying each direction pattern on the X and Y axes.

On the other hand, when the environment value is 0, only the brightness accumulation value is applied to extracting the license plate image. That is, the pattern change of the brightness accumulation value on the Y axis is stored (S315), and the brightness accumulation value on the X axis is stored the pattern change (S316), thereby detecting the final X and Y regions (S317).

8 is a flowchart illustrating a vehicle number extraction process of the image data processing method illustrated in FIG. 5.

In the present invention, an individual neural network algorithm and a data fusion method using the results are used in the number extraction process using the image data of the license plate. That is, several neural network algorithms are processed in parallel with each image data as basic input data, and each result thereof is combined to finally calculate a number recognition result. In addition, these results can be stored in data so that similar license plates can be quickly recognized in the future.

Neural network algorithms are already known to have excellent license plate recognition. However, up to 99% of the recognition rate has not been reached yet, especially in rainy weather, there is a problem of low recognition rate of about 50%. Furthermore, the number plate recognition result is output differently depending on the number of hidden nodes as well as each input node of the neural network.

Therefore, the present invention uses the neural network algorithm as two models of MLP and LVQ, which are frequently used according to the structure of neural networks, and each of the four models (MLP1, MLP2, LVQ1, LVQ2) having various nodes of hidden layers. We divide neural network algorithms in parallel, and extract the number from the license plate image by introducing an algorithm that derives the final result based on each result. In other words, after the processing of the parallel neural network algorithm, the result is to apply a data fusion algorithm that synthesizes the results and draws the final conclusion. When data fusion algorithms are applied to license plate recognition, each neural network algorithm can be processed in parallel, so that excellent recognition rate can be expected without slowing down.

Vehicle number extraction process (S203) is an algorithm for extracting license plate information by processing the image of the license plate input to the central control unit, in detail, the step of extracting the character to find each character area from the license plate image (S401), each The neural network module execution step (S402) for recognizing data for each character area as a character, and the data fusion module execution step (S403) for deriving the resulting character by aggregating the results of each neural network module are divided into three stages. It is mounted in the video data extraction program 93 of the control unit 20. The detailed processing of each module is as follows.

9 is a detailed flowchart of a vehicle number extraction process shown in FIG. 8.

The character extracting module S401 is a module for finding a single character region in a license plate image recognized from an edge detection process. The character extracting module S401 divides the region up and down by a cumulative distribution of Y-axis brightness, and accumulates an X-axis cumulative distribution of characteristic values in the upper region. Classify the letter area by the X-axis cumulative distribution of the characteristic values in the and subareas. 10 is a diagram illustrating an example of classifying each letter area by the letter extracting module of the present invention. Referring to this, the letter extracting module S401 will be described in more detail as follows.

First, using only the cumulative distribution of the Y-axis brightness to find a point where the brightness is sharply reduced and separated into an upper region and a lower region (S501) (see A of FIG. 10). Then, in the upper region extracted on the Y-axis, using the cumulative distribution of the X-axis brightness (see B of FIG. 10), the point where the brightness rapidly decreases on the X-axis is separated into the upper region and the lower region (S502). Through this, the letter area of the upper area is distinguished.

Next, by using the X-axis brightness distribution in the lower region extracted on the Y-axis (see B of FIG. 10), the point where the brightness rapidly decreases on the X-axis is separated into the upper region and the lower region (S503). The character area of the lower area is distinguished.

The neural network module is an algorithm for recognizing data for each character area as a character. The neural network module divides each character area into a grid and sets an average value of characteristic values corresponding to each detailed area as an input value. 11 is a view for explaining an example of applying the neural network algorithm to the license plate recognition process of the present invention.

After each character region is extracted by the character extracting module S401, first, four neural network structures MLP1, MLP2, LVQ1, and LVQ2 are allocated to each processor in advance, whereby each processor is subsequently connected in parallel. Will perform the steps.

First, a learning step of repeatedly learning all letters and numbers that can be extracted from the license plate and obtaining weight values used in the neural network is performed (S504). Next, as a test step, the range and distribution of the output value for each letter are obtained by applying the image of the actual license plate (S505). After that, the data of the character area is put in the neural network, and the recognition result is stored. At this time, for the ambiguous result (when there are several similar output results similarly), the respective output values are stored (S506).

The data fusion module is a module that aggregates the results of each neural network module and makes a comprehensive judgment. The data fusion module is determined using a knowledge base technique. 12 is a block diagram illustrating a data fusion algorithm during a license plate recognition process of the present invention. In the data fusion module 124, the processing result of the neural network module is normalized so that the sum is 1, the recognition result values of the neural network module are sorted in descending order, and the recognition result corresponding to the result value is larger from the sorted result value. And stores the recognition result corresponding to the largest summation result by summing the result values among the stored recognition results.

A case where four neural network algorithms are executed in parallel in the previous neural network module is described as follows. In order to execute the data fusion model, data obtained by performing parallel processing on one character image with four neural network algorithms 116, 118, 120, and 122, that is, a pair of recognition results and neural network output values corresponding to the recognition results. Input to the fusion module 124 (S507).

Thereafter, the result value is normalized for each neural network algorithm result so that the sum is 1 (S508). This is to fairly compare the results of the four neural network algorithms. Next, the neural network is ranked by the recognition result, and the neural network algorithm is listed starting with the highest result value (S509). After listing the result values, the neural network compares the largest result value (S510). If three or more recognition results are the same, the recognition result of the corresponding result value is stored (S511).

If, as a result of comparing the largest result in each neural network, the two recognition results are the same, and if the result is 0.5 or more (S512), the recognition result for the corresponding result is stored (S511). Proceed to step S508 of normalizing the result value.

Subsequently, the result of all the neural networks having the same recognition result is added together (S513), and the recognition result corresponding to the largest value is stored (S514). Now, for all cases, one recognition value has been saved, so exit the data fusion module.

As described above, in the present invention, in order to integrate and remotely control the surveillance camera regardless of the type of camera, the camera is equipped with a black box and connected to the central control unit through the control terminal to control the required camera.

The black box includes a buffer unit for temporarily storing various image data and environmental data to reduce the load on the central control unit. The central control unit stores device information and control information of all cameras to be controlled. Through a control terminal equipped with a program such as a web browser, the camera manager can efficiently remotely control the camera as required or by receiving a recommendation result regarding the control from the central controller.

In addition, since the license plate of the vehicle can be recognized at a high speed, the vehicle data can be transmitted to all the devices and users who need the vehicle data using a communication network, and thus its versatility and usefulness can be said to be quite large.

As such, those skilled in the art will appreciate that the present invention can be implemented in other specific forms without changing the technical spirit or essential features thereof. Therefore, the above-described embodiments are to be understood as illustrative in all respects and not as restrictive. The scope of the present invention is shown by the following claims rather than the detailed description, and all changes or modifications derived from the meaning and scope of the claims and their equivalents should be construed as being included in the scope of the present invention. do.

As described above, the present invention remotely controls the surveillance camera using a communication network, so if a crime is found such as a speed violation of the vehicle, the violation can be automatically transmitted to the police and the vehicle operator by wireless or wired. . Accordingly, as soon as a vehicle that needs an urgent arrangement is detected, all the police stations nearby can be collectively notified, so it is easy to arrest immediately.

In addition, when the vehicle license plate recognition method of the present invention is applied to various parking lots and toll gates as well as surveillance cameras installed on the road, the recognition effect of the vehicle is maximized, so that effective control can be performed even when the number of vehicles is large.

In addition, it is possible to recognize the number of the vehicle desired by the user at high speed with the image recognition through the artificial intelligence of the camera, and to receive the image data of the recognized license plate through the communication network. Can be controlled remotely and the image data of the vehicle can be processed in real time.

Claims (16)

At least one camera installed to monitor and capture the flow of the vehicle; It is mounted on the camera, monitors the control information for controlling the camera, and stores the vehicle image information taken by the camera and vehicle license plate image information extracted from the vehicle image, black for controlling the camera box; And Connected to the camera by a communication network, the location information and control information of the camera, the image information input from the black box, the information about the specific vehicle that needs to know the location is stored, according to the rules for controlling the camera A central control unit for determining an optimum control condition and receiving and processing image information from the black box; Integrated remote control system of the surveillance camera using a communication network having a. According to claim 1, The black box may include a buffer for storing overall image data of the vehicle photographed by the camera, license plate image data extracted from the overall image data of the vehicle, camera information of the vehicle, and shooting conditions; Storage means for storing a program for processing the image data stored in the buffer unit and a processing result; And An interface unit for connecting with the central control unit; Integrated remote control system of the surveillance camera using a communication network having a. According to claim 1, The central control unit may include administrator authentication means for confirming whether an administrator to control the camera is a legitimate administrator and allowing access to a specific camera; Control information converting means for converting a control command input by the manager into control information for a specific camera; Control recommendation means for analyzing the flow of the monitoring target vehicle by the necessary conditions based on the statistical data on the vehicle flow to determine the driving time of the camera, and recommending the optimum control condition of the camera to the manager; Control means for transmitting a control command to the camera when a control command is input by the manager or when a command for controlling the camera is input from the manager according to an optimum control condition recommended to the manager; Image data extraction means for storing whole vehicle image information photographed by the camera and extracting, storing and managing vehicle license plate image information from the whole vehicle image information; And Vehicle location detecting means for extracting a current location of a specific vehicle requiring location by image information and vehicle license plate image information of the entire vehicle; Integrated remote control system of the surveillance camera using a communication network having a. According to claim 1, The central control unit includes a manager database for storing personal information of the manager controlling the camera; A camera database for storing model information, location information, and operation information of the camera; An image database for storing the entire vehicle image information and the vehicle license plate image information transmitted from the black box; A vehicle database for storing models and numbers of vehicles requiring urgent arrangements; And A control database for storing control commands necessary for operation of the camera and image recognition according to the model of the camera; Integrated remote control system of the surveillance camera using a communication network having a. The method according to any one of claims 1 to 4, The communication network is an integrated remote control system of a surveillance camera using a communication network, characterized in that the Internet. At least one camera, the vehicle is mounted on the camera to store the vehicle image information and vehicle license plate image information extracted from the vehicle image taken by the camera, a black box for controlling the camera at the request of the camera administrator And a central control unit connected to the camera by a communication network to store the camera information and the image information of the vehicle transmitted from the black box and determine an optimal control condition according to a rule for controlling the camera. A method for remotely controlling the camera in a remote control system, A first step of authenticating the manager as the manager who wants to control the camera is connected through a communication network; A second step of deciding whether to arbitrarily select a camera to be controlled from an administrator who has successfully authenticated or call and select a control recommendation processor provided from the central controller; A third step of providing the selected camera information to the manager when the manager arbitrarily selects a camera to be controlled; A fourth step of controlling the camera according to a control command input by the manager and storing the control command input by the manager; In the second step, the manager calls the control recommendation processor, and when control is requested from a specific camera, displays information of the camera requesting the control and an optimum control command determined by the control recommendation processor, and then proceeds to the fourth step. Proceeding; Determining a camera requiring control by the control recommendation processor when the manager calls a control recommendation processor in the second step and the administrator wants to search for a camera requiring control; And Providing the manager with an optimal control command determined by the control recommendation processor for the determined camera and proceeding to the fourth step; Integrated remote control method of the surveillance camera using a communication network comprising a. The method of claim 6, The control recommendation processor is a processor for deriving an optimal control method of the camera based on statistical data on the vehicle flow. As a method for processing image data captured by a camera installed to monitor and capture the flow of the vehicle, An edge detection process of separating the license plate image from the entire image of the vehicle by analyzing a pattern of the input data by receiving the entire image of the vehicle photographed by the camera in a pixel method including brightness, saturation, and color information; And A vehicle number extraction process for extracting each character region from the license plate image, extracting possible characters by a plurality of neural network algorithms, and then combining the results of the neural network algorithms to output a final character; Image data processing method using the integrated remote control system of the surveillance camera using a communication network comprising a. The method of claim 8, The edge detection process may include executing a vehicle detection module extracting a vehicle image from a background image of a location where the camera is installed; And Executing a license plate detection module for detecting the position of the license plate of the vehicle according to a weather condition at the time of shooting, a brightness, saturation, and color of the captured image from the extracted vehicle image; Image data processing method using the integrated remote control system of the surveillance camera using a communication network comprising a. The method of claim 9, The executing of the vehicle detection module may include: dividing an input image photographed by a camera in pixel units; Obtaining a difference image for each frame by comparing the background image of the location where the camera is installed and the input image; And Obtaining a segmented image by binarizing an absolute value of the difference image; Image data processing method using the integrated remote control system of the surveillance camera using a communication network comprising a. The method of claim 9, The executing of the license plate detection module may include: accumulating a brightness value input when the vehicle image is photographed, and detecting a license plate image of the vehicle by searching an area in which a pattern change is severe; If the license plate image of the vehicle is not detected by the brightness cumulative value, the pattern change of the brightness cumulative value is stored, the chroma cumulative value input when the vehicle image is photographed, and the area where the pattern change is severe is searched. Detecting a license plate image of the vehicle; When the license plate image of the vehicle is not detected by the saturation accumulation value, the pattern change of the saturation accumulation value is stored, the color accumulation value input when the vehicle image is taken, the area is changed, and the area where the pattern change is severe is searched. Detecting a license plate image of the vehicle; When the license plate image of the vehicle is not detected by the color accumulation value, the pattern change of the color accumulation value is stored, and the area with the greatest change on the Y axis by adding the pattern change by the brightness, saturation, and color accumulation values. Detecting a fourth step; A step (5) of detecting the region having the most change on the X axis according to the pattern change by performing the first step to the third step in the X axis direction with respect to the detected Y axis area; And A sixth step of detecting a final license plate image by the detected Y-axis region and X-axis region; Image data processing method using the integrated remote control system of the surveillance camera using a communication network comprising a. The method of claim 8, The vehicle number extracting process may include: executing a character extracting module for detecting each character region from the license plate image output in the edge detection process; Executing a neural network module for recognizing data extracted from the character area detected by the character extraction module as a character; And A data fusion module execution step of aggregating the results of the neural network module execution step to derive the resulting letters; Image data processing method using the integrated remote control system of the surveillance camera using a communication network comprising a. The method of claim 12, The executing of the character extracting module may include separating a point where the brightness sharply decreases into an upper region and a lower region by using only a cumulative distribution of brightness of Y-axis with respect to the license plate image recognized from the edge detection process; In the extracted upper region on the Y-axis, using the cumulative distribution of the brightness of the X-axis to separate the point where the brightness sharply decreases on the X-axis into the upper region and the lower region, thereby distinguishing the letter region of the upper region ; And In the extracted lower region on the Y-axis, using the X-axis cumulative distribution to separate the point where the brightness sharply decreases on the X-axis into the upper region and the lower region, thereby separating the character region of the lower region; Image data processing method using the integrated remote control system of the surveillance camera using a communication network comprising a. The method of claim 12, The neural network module execution step may include obtaining weight values used for the neural network by repetitive learning of all letters and numbers that can be extracted from the license plate of the vehicle; Obtaining a range and distribution of output values for each letter based on the image of the actual license plate extracted in the edge detection process; And Storing data recognized by inputting data of the text area extracted by the text extraction module into a neural network, and simultaneously storing ambiguous recognition results; Image data processing method using the integrated remote control system of the surveillance camera using a communication network comprising a. 13. The method of claim 12, wherein executing the data fusion module comprises: normalizing a processing result of the neural network module to add up to one; Sorting the recognition result values of the neural network module in descending order; Storing a recognition result corresponding to the result value from a larger value to the sorted result value; And Storing the recognition result corresponding to the largest summation result by adding up the result values among the stored recognition results; Image data processing method using the integrated remote control system of the surveillance camera using a communication network comprising a. The image data processing method using an integrated remote control system of a surveillance camera using a communication network according to any one of claims 6 to 15, wherein the communication network uses the Internet.