KR102107957B1 - Cctv monitoring system for detecting the invasion in the exterior wall of building and method thereof - Google Patents

Abstract

The present invention relates to a CCTV monitoring system for detecting a person who invades a building through an exterior wall, to effectively detect an object from a photographed image, and a method thereof. According to the present invention, the CCTV monitoring system comprises: a camera (100) capturing the outer wall of the building; a server (200) receiving a captured image from the camera; and a terminal (300) receiving a warning signal outputted from the server. The server (200) comprises: a region setting unit (210) setting a monitoring region in the image transmitted from the camera; an object detection unit (220) detecting a person from the set monitoring region; and a warning signal output unit (230) outputting the warning signal to the terminal when the person is detected. The object detection unit (220) comprises: a candidate window setting module (221) setting a region of a moving object; an object detection learning module (222) using an object learning module to detect the person from objects included in a candidate window; and an object acquisition module (223) removing a background from the current image and detecting the person. Detection performed in the object detection unit is performed by setting the candidate window on the region including the moving object, detecting the person included in the candidate window based on the object learning model, and detecting the person through background removal.

Description

CCTV monitoring system and method for intrusion detection of building exterior walls {CCTV MONITORING SYSTEM FOR DETECTING THE INVASION IN THE EXTERIOR WALL OF BUILDING AND METHOD THEREOF}

The present invention relates to a CCTV monitoring system for detecting an object invading a house through the outer wall of an apartment house.

Recently, the demand for CCTV monitoring systems for disasters, traffic, intruders, etc. continues to increase. The CCTV monitoring system prevents accidents in advance and responds quickly in the event of an accident to reduce damage.

CCTV cameras have a defined imaging range to monitor a surveillance area such as a doorway. The captured image is transmitted to the control server, and the control server detects abnormal behavior and intrusion detection of the object captured in the image.

Conventional intelligent image analysis technology is limited to one event and recognizes objects, making it difficult to monitor multiple channels simultaneously. In addition, when a plurality of event images are generated from the captured image, it is difficult to integrate and process them.

Patent Publication No. 10-2008-0072454, Frame interpolation device and frame interpolation method to estimate motion by separating background and moving object Registration Patent Publication No. 10-1130963, Unstructured object tracking device and method using form and feature information Patent registration No. 10-1180887, abnormal behavior detection device and method Patent registration No. 10-1690057, Long-distance transmission system of multi-channel high-definition (HD) video data for CCTV through high-definition image interpolation

An object of the present invention is to provide a CCTV monitoring system for effectively detecting an object in a captured image.

In addition, an object of the present invention is to provide a CCTV monitoring system capable of increasing a search speed when detecting an object from an image captured.

The problems to be solved of the present invention are not limited to the problems mentioned. Other technical problems not mentioned will be clearly understood by those skilled in the art from the following description.

The CCTV monitoring system according to the present invention is a CCTV monitoring system for detecting a person object invading a building exterior wall, wherein the CCTV monitoring system receives a camera 100 capturing an exterior wall of a building and an image captured from the camera. It includes a server 200 and a terminal 300 for receiving an alarm signal output from the server,

The server 200 includes an area setting unit 210 for setting a monitoring area from an image transmitted from a camera, an object detection unit 220 for detecting a human object in the set monitoring area, and when a person object is confirmed, to the terminal. It consists of a warning signal output unit 230 for outputting a warning signal,

The object detection unit 220 includes a candidate window setting module 221 for setting a moving object area, an object detection learning module 222 for detecting a human object among objects in the candidate window using the object learning module, and the current image. It consists of an object acquisition module (223) to remove the background and check the human object,

The detection of the object in the object detection unit is performed by setting a candidate window in an area including a moving object, detecting a human object included in the candidate window using the object learning model, and then checking the human object through background removal. .

As an embodiment, the human object may be detected within the set surveillance area.

A method for detecting a human object using a CCTV monitoring system according to the present invention is a method for detecting a human object that invades the exterior wall of a building, comprising: setting a surveillance area in the captured image (S100); And detecting an object (S200) and, if a human object is detected, outputting a warning signal (S300).

The detecting of the human object (S200) includes detecting an object from the current image and setting a candidate window on the object (S210), and detecting a human object using an object learning model in the candidate window (S220). , Consists of removing the background from the current image and identifying a human object (S230),

The detection of the human object detects the object from the current image, sets the candidate window, detects the human object within the set candidate window using the object learning module, and then removes the background from the image containing the detected human object. It is performed by checking the human object.

According to the present invention, it is possible to minimize the system resource and increase the search speed by setting only the surveillance area as the search target in the captured image.

Also, according to the present invention, a plurality of captured images are dividedly displayed on a monitor of a server, and an event marker is output on the edge of a divided image including a human object, so that an administrator can easily find the corresponding CCTV camera.

In addition, according to the present invention, by outputting an event marker on the edge of the area of a human object, the administrator can easily find the corresponding CCTV camera.

1 is a configuration diagram of a CCTV monitoring system according to the present invention.
2 is a configuration diagram of a server according to the present invention.
3 is a view showing an imaging range of a CCTV camera according to the present invention.
4 is a diagram showing detecting an object using a candidate window.
5 shows that an object image is obtained by differentiating a background image from a current image.
6 is a flow chart showing the operation of the CCTV monitoring system according to the present invention.
7 is a flowchart illustrating a method of detecting a human object in the current image.

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

In this specification, terms such as a part, a module, a server, a terminal, or a system may mean a combination of hardware and software driven by the hardware, and the term video means one video frame or a continuous set of video frames. can do.

The present invention relates to a CCTV monitoring system for detecting an object invading a house through the outer wall of an apartment house.

1 is a configuration diagram of a CCTV video system according to the present invention, and FIG. 2 is a configuration diagram of a server according to the present invention.

The CCTV video system according to the present invention includes a camera 100 for photographing the outer wall of a building, a server 200 for receiving an image captured from the camera, and a terminal 300 for receiving an alarm signal output from the server. Includes.

The camera 100 is installed to image the outer wall of the apartment house, such as a tenement house, an apartment. The camera is installed to have a predetermined angle of view, and images a preset area. More than one camera can be installed in multiple locations. For example, two cameras may be installed at the corners of the outer wall of the apartment house so as to image both sides of the corners in order to photograph different directions.

The image captured from the camera is transmitted to the server. The server 200 receives an image captured by one or more cameras 100 and detects an object from the captured image. Meanwhile, the object may be classified into a human object, an animal object, and an object object, and in the present invention, a human object that invades the home is a detection target.

The server 200 is an area setting unit 210 for setting a surveillance area from an image transmitted from a camera, an object detection unit 220 for detecting a human object in the set surveillance area, and when a human object is confirmed, alerts the terminal It includes a warning signal output unit 230 for outputting a signal.

Additionally, a server may be provided with a video recorder that stores the image transmitted from the camera and a monitor that displays the image. The video recorder may include a hard disk (HDD) for video storage. As an embodiment, the video recorder may include a network video recorder (NVR) and a digital video recorder (DVR).

The area setting unit 210 sets a predetermined monitoring area in the captured image. As an embodiment, the monitoring area is set as a monitoring area for the area including the outer wall of the building. Preferably, the lower part of the outer wall of the building can be set as a monitoring area.

The surveillance area includes only the building area to be monitored in the captured image, but does not include the area around the building, that is, the area of the sky, the ground, or the next building.

3 is a view showing an imaging range of a CCTV camera according to the present invention.

Referring to the drawing, the surveillance area of the CCTV camera shows that the lower part of the outer wall of the apartment is set as the surveillance area. It is assumed that the object to invade rises from the lower part of the building, and the lower part of the building is set as the surveillance area.

The object detection unit 220 detects a human object in a preset monitoring area. The captured image may include human objects such as men, women, the elderly, and children, animal objects such as dogs and cats, and object objects such as bicycles and cars. The object detection unit detects only human objects among moving objects in the surveillance area.

When a moving object is identified as a human object, the warning signal output unit 230 outputs a warning signal to the terminal. The terminal may be a terminal of a server administrator, and the warning signal includes an alarm sound and a detection position coordinate.

The administrator can confirm that an intruder has occurred in the corresponding location from the intrusion location and the alarm sound displayed on the map screen of the terminal.

Hereinafter, the object detection unit 220 of the server 200 according to the present invention will be described.

Detection of a human object according to the present invention is performed by setting a candidate window in an area containing a moving object, detecting a human object included in the candidate window using the object learning model, and then checking the human object through background removal do.

The object detection unit 220 includes a candidate window setting module 221 for setting a moving object area, an object detection learning module 222 for detecting a human object among objects in the candidate window using the object learning module, and a background from the current image. And an object acquiring module 223 for removing and identifying a human object.

The candidate window setting unit 221 sets a candidate window on the moving object when there is an object moving in the surveillance area in the captured image. For example, the candidate window can be set for each of human objects, animal objects, and object objects.

As an embodiment, the candidate window forms a grid grid of the same size in the input image, and uses a single-step method for predicting the number of candidate windows assigned to each grid grid.

4 is a diagram showing detecting an object.

When a person and a dog are moving together in the surveillance area, a candidate window is set for each person and dog.

The candidate window is set based on a grid, and the object is extracted from the candidate window. Referring to the drawings, in (a) and (b), the number of preset windows is predicted based on the grid, and only objects with high reliability are extracted from objects other than the background. (a) and (b) can be performed repeatedly.

As an embodiment, after dividing a video image into grids at regular intervals, pixels are sampled for each grid, and clustering is performed using the sampled pixels as a representative value of the grid. In another embodiment, a plurality of pixels are sampled, and clusters of K grids are clustered using the center of gravity of the plurality of pixels as a representative value of the grid. The center of gravity can be calculated as Center Of Gravity. In this case, the sampling time becomes longer, but the accuracy can be increased.

<Equation 1>

COG = ((X (11) + X (12) + X (13) + ... + X (1N)) / N, (X (21) + X (22) + X (23) + ... + X (2N) )) / N,…)

Here, X represents a pixel, and N represents a sampling number.

Subsequently, a candidate window is set based on the distance from the center of the grid cluster, and an object is extracted from the candidate window.

The object detection learning module 222 detects a human object among moving objects in the surveillance area. The object detection learning module may use an object learning model to which a convolutional neural network is applied. At this time, the extraction of the object can be detected using a pre-trained weight file.

The object detection learning model can learn images containing humans and animals. As an embodiment, the object learning model may be a model that learns images captured from various angles in a place such as a rock wall, a mountain, a park, a road, and an object.

Detecting human objects individually from images transmitted from multiple CCTV cameras is complex and requires a lot of computational processing. As an embodiment, when detecting a human object from 16 CCTV images, a method of synthesizing each channel into one image and detecting a human object from one synthesized image may be used to improve the object detection speed.

Meanwhile, the detection of an object using a convolutional neural network may incorrectly detect a background image as a human object. To compensate for this, the human object is accurately detected using the background difference method.

The object acquisition module 223 removes the background, verifies the object, and identifies the human object. First, the background image previously stored is differentiated from the current image. The object is separated from the background by subtracting the background image from the current image-the input image transmitted in real time from the camera. As an embodiment, the pre-stored background image may be updated at a predetermined cycle. The pre-determined background image may be set as an image in which an object is not detected from at least N or more images.

5 shows that an object image is obtained by differentiating a background image from a current image.

Background and object are captured in the current image. In addition, in the background image, only the background without the object is captured. The object acquisition module acquires the object image by differentiating the background image from the current image. 5 (c) shows an image in which the background is removed from the current image, that is, an object image.

Hereinafter, the operation of the CCTV monitoring system according to the present invention will be described.

6 is a flowchart illustrating the operation of the CCTV monitoring system according to the present invention, and FIG. 7 is a flowchart illustrating a method of detecting a human object in the current image.

The operation method of the system according to the present invention includes the steps of setting a surveillance area in the captured image (S100), detecting a moving human object in the set surveillance area (S200), and when a human object is detected, a warning signal It includes the step of outputting (S300),

The detecting of the human object (S200) includes detecting an object from the current image and setting a candidate window on the object (S210), and detecting a human object using an object learning model in the candidate window (S220). , Removing the background from the current image and identifying a human object (S230).

In the step (S100), the server receives the current image from a plurality of cameras, and sets the surveillance area from the current image. As an embodiment, the lower part of the outer wall of the building may be set as the monitoring area. By limiting the surveillance area to the lower part of the outer wall of the building, resources for object detection can be saved, and objects can be detected quickly.

In step S200, a moving object is detected in the set surveillance area. The moving object in the surveillance area may be a human object, an object object, or an animal object. In the present invention, only human objects are detected among these objects.

As an embodiment, an object is detected in the current image and a candidate window is set. Subsequently, the human object is detected in the candidate window set using the object learning module, and then the background is removed from the image containing the detected human object to check the human object.

As such, when a human object is detected in the current image, an event may be displayed on the edge of the image where the human object is detected among a plurality of divided images displayed on the monitor of the server. In addition, an event marker may be output on the edge of the human object area. The event marker may be a visual event effect such as flickering, and the administrator can easily find the corresponding video and CCTV camera.

In addition, when a human object is detected in the current image, the server outputs a warning signal to the terminal. The manager displays the current image capturing the intruded place on the map screen displayed by the terminal, and also outputs a warning sound.

The operation according to the human object detection method according to the embodiments may be implemented at least partially as a computer program and recorded in a computer-readable recording medium.

Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disks, and optical data storage devices. In addition, the computer-readable recording medium may be distributed and configured in a networked computer system, and the computer-readable code may be stored and executed in a distributed manner.

As described above, the present invention has been described in detail through specific embodiments, but the present invention is not limited to the above embodiments, and various modifications can be made by those skilled in the art within the scope of the technical spirit of the present invention.

100: camera
200: server
210: area setting unit
220: object detection unit
221: candidate window setting module
222: Object detection learning module
223: object acquisition module
230: warning signal output
300: terminal

Claims (3)

CCTV monitoring system for detecting human objects intruding into the outer walls of buildings,
The CCTV monitoring system,
It includes a camera (100) for photographing the outer wall of a building, a server (200) for receiving an image captured from the camera, and a terminal (300) for receiving an alarm signal output from the server,
The server 200 includes an area setting unit 210 for setting a monitoring area from an image transmitted from a camera, an object detection unit 220 for detecting a human object in the set monitoring area, and when a person object is confirmed, to the terminal. It consists of a warning signal output unit 230 for outputting a warning signal,
The object detection unit 220 includes a candidate window setting module 221 for setting a moving object area, an object detection learning module 222 for detecting a human object among objects in the candidate window using the object learning module, and the current image. It consists of an object acquisition module (223) to remove the background and check the human object,
The detection of the object in the object detection unit is performed by setting a candidate window in an area including a moving object, detecting a human object included in the candidate window using the object learning model, and then checking the human object through background removal,
The candidate window of the candidate window setting module 221 is performed by dividing a video image into grids at regular intervals, sampling pixels for each grid, and clustering K grid clusters using the center of gravity of the pixels as a representative value of the grid. However, the center of gravity is CCTV monitoring system, characterized in that calculated from <Equation 1> below.
<Equation 1>
COG = ((X (11) + X (12) + X (13) + ... + X (1N)) / N, (X (21) + X (22) + X (23) + ... + X (2N) )) / N,…)
Here, X represents a pixel, and N represents a sampling number.
The method according to claim 1,
CCTV monitoring system, characterized in that the human object is configured to detect within the set surveillance area.
As a method for detecting human objects that invade the outer wall of a building,
It includes the steps of setting a surveillance area in the captured image (S100), detecting a moving human object in the set monitoring area (S200), and outputting a warning signal when a human object is detected (S300). ,
The detecting of the human object (S200) includes detecting an object from the current image and setting a candidate window on the object (S210), and detecting a human object using an object learning model in the candidate window (S220). , Consists of removing the background from the current image and identifying a human object (S230),
The detection of the human object detects an object from the current image, sets a candidate window, detects the human object within the set candidate window using the object learning module, and then removes the background from the image containing the detected human object. Is performed by checking the human object,
The candidate window is performed by dividing the video image into grids at regular intervals, sampling pixels for each grid, and clustering K grid clusters using the pixel's center of gravity as a representative value of the grid. CCTV monitoring method characterized in that it is calculated from the equation (1).
<Equation 1>
COG = ((X (11) + X (12) + X (13) + ... + X (1N)) / N, (X (21) + X (22) + X (23) + ... + X (2N) )) / N,…)
Here, X represents a pixel, and N represents a sampling number.

Images