KR100994418B1 - System For Processing Imaging For Detecting the Invasion Of Building and Method Thereof - Google Patents

Abstract

Separates the background image from the captured image provided by the camera, generates a reference image to identify the moving object to be detected according to the difference between the captured image and the background image, and then generates snow, rain, birds and bugs from the reference image. After the removal of non-specific objects such as the like, the path of the object image remaining in the reference image is traced to determine whether or not the intrusion into the surveillance area. Therefore, the present invention can improve the reliability of the detection performance by reducing the malfunction due to the non-specific object, it is possible to increase the operation speed in the process of processing the photographed image.

Intrusion Detection, Object Extraction, Non-Specific Objects

Description

System For Processing Imaging For Detecting the Invasion Of Building and Method Thereof}

The present invention relates to a photographing image processing system and method for detecting an intrusion of a building, and more particularly, to a photographing image processing system and method for detecting an intrusion of a building to remove a nonspecific object from a photographing image provided from a camera. .

Intrusion or fire detection systems are typically installed inside or outside apartments or company buildings to protect the safety, property or confidentiality of individuals or companies. Intrusion detection systems typically include a camera, a method of continuously photographing a surveillance area through a camera, and a method of recording an image taken by the camera and generating an alarm when a moving object is detected using a sensor.

In the former method, the intruder is continuously detected by the camera and stored in the storage medium, and the intruder is detected by examining the recorded image stored in the medium when the intruder is stolen. However, in the case of the former method, it is difficult to prevent an intrusion in advance because a security guard or a manager must continuously monitor the captured image, and also requires a lot of storage space and occasional storage media management by continuous shooting. Shall be. Therefore, there is a problem in that it is difficult to find the photographed image after the intrusion occurs, and a lot of storage space and time and cost for managing the photographed image was consumed.

In the latter method, since only the moving object is detected by the sensor, only the moving object is photographed and stored, so that it does not require much storage space compared to the former method, and its management also has an easy advantage. However, when the operation of the sensor is not smooth or the sensor malfunctions or the sensor does not operate due to a defect, etc., there is a problem that may not operate properly when an incorrect intrusion and an intruder are generated from the outside.

In order to solve this problem, a method of detecting a change by motion detection or detecting a size after motion detection is introduced, or a background image is stored in advance, and the captured image is compared with the input image to change the image. A method of detecting an invasive object by detecting whether it is present has been introduced. However, this method is not intended to detect intrusions (persons such as intruders), but non-specific objects such as snow or rain, insects or birds or bugs that momentarily pass in front of the camera as intrusions to prevent malfunction. There is a problem that can be caused.

SUMMARY OF THE INVENTION An object of the present invention is to provide a photographing image processing system and method for detecting an intrusion of a building, which can improve reliability of detection performance by removing a non-specific object from a photographed image provided from a camera, thereby reducing malfunction.

Another object of the present invention is to provide a photographing image processing system and method for detecting an intrusion of a building, which can reduce the amount of computation in processing a photographed image, thereby increasing the computation speed.

According to an aspect of the present invention, there is provided a photographing image processing method for detecting an intrusion of a building, the method comprising: receiving a photographed image from a camera photographing the building; Separating a background image from the photographed image; Forming a reference image for identifying a moving object based on a difference between the photographed image and the separated background image; Extracting an object image from the reference image; Removing a non-specific object image which does not belong to a sensing object from the reference image; And determining whether an object image remaining in the reference image from which the non-specific object image has been removed invades a designated surveillance region, wherein the determining of the invasion comprises: specifying a center coordinate of the image of the remaining object in the designated image. It is characterized by determining whether the invasion according to the frequency of invading the surveillance area.

According to the present invention, the method may further include generating the background image by averaging the image frames extracted from the photographed image, and updating the background image by replacing the old image frame with a new image frame every predetermined period.

According to the present invention, the generating of the background image may include generating a background image using only the Y (brightness) signal of the extracted image frame, Cr (Y) of the extracted image frame, and Cr, One of the cases in which the background image is generated using all of Cb is applied.

According to the present invention, the generating of the reference image may include blurring the photographed image and converting the photographed image into black or white according to a difference between the Y signal of the blurred image and the Y signal of the background image. Generating a black and white image in which the difference between the image and the background image appears in black and white.

According to the present invention, the forming of the reference image may include removing the small sized image from the reference image to correct the image, and reducing the size of the reference image by performing a logical sum operation (OR) on the corrected image. do.

Further, according to the present invention, the step of removing the non-specific object image, if the size of the extracted object image is smaller than the predetermined minimum size or larger than the maximum size includes removing.

According to the present invention, the removing of the non-specific object may include removing the ratio of the horizontal size and the vertical size of the extracted object image larger than a preset ratio.

Also, according to the present invention, the removing of the non-specific object may include removing if the distance traveled by the same object image is greater than a predetermined distance in image frames in which the extracted object image is continuously displayed.

According to an aspect of the present invention, there is provided a photographing image processing system for detecting an intrusion of a building, the camera photographing the building; Separating the background image from the photographed image provided from the camera, forming a reference image for identifying a moving object based on a difference between the photographed image and the separated background image, extracting an object image from the reference image, An image processing unit for removing a non-specific object image that does not belong to the sensing object from the extracted object image and then determining whether the object image remaining in the reference image violates the designated surveillance region; And an event generating unit generating an event set to notify the occurrence of an invasion according to the determination of the image processing unit, wherein the image processing unit determines whether or not the intrusion is based on a frequency at which the center coordinates of the remaining object image invade the designated surveillance region. It is characterized by judging.

delete

According to the present invention as described above, by removing the non-specific object from the photographed image to accurately detect the object to detect the intrusion to reduce the malfunction can improve the reliability of the detection performance. In addition, the present invention can increase the operation speed by reducing the amount of computation in the process of processing the photographed image.

The above objects, features and other advantages of the present invention will become more apparent by describing the preferred embodiments of the present invention in detail with reference to the accompanying drawings. Hereinafter, a photographing image processing system and method for detecting an intrusion of a building according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

Referring to FIG. 1, a photographing image processing system to which the present invention is applied includes a camera 110, a signal converter 120, an image processor 130, an image storage unit 140, and a camera 110 for providing an image of a building. The event generator 150 is included.

The camera 110 photographs a designated area of a building that is a surveillance area, and a general CCTV camera, an infrared camera, or a day & night camera may be applied. Cameras are usually installed at a distance from a designated area of the building for effective shooting. The camera 110 may be one or more, and when a plurality of cameras are installed at the same time, it is also possible to select a signal photographed through any one camera by a selection switch (not shown).

The signal converter 120 converts the photographing signal input from the camera into a digital signal. And the size of the image is reduced by reducing the resolution by using only the Y signal component value of the converted signal. For example, after reducing the NTCT 720 * 480 (PAL 720 * 576) size of the converted digital signal to NTCT 352 * 240 (PAL 352 * 288), 7.5 frames, which is 1/4 of 30 frames per second (NTCT standard), are reduced. To the image processor 130. The image processor 130 performs signal processing using the received data.

The image processor 130 separates the object image and the background image through a stepwise process of the received data and then performs intrusion detection of the building. The image processor 130 includes an image filter 132, an object extractor 134, a non-specific object remover 135, and an intrusion detector 136. Specific operations of the image processor 130 will be described later.

The image storage unit 140 stores a photographed image, a background image, and the like. The captured image stored in the image storage unit 140 may not record all the captured images continuously, but may store only the captured image at the time of detecting an intrusion, thereby solving a privacy problem or a capacity problem.

The event generating unit 140 receives a signal from the image processing unit 130 and detects an intrusion or a fire in a specified area. For example, the event generator 140 may generate a buzzer, a warning sound using a speaker, and a relay operation. An event is transmitted by an external device operation, a method of transmitting an e-mail or a text using the Internet, and the like.

Hereinafter, a photographing image processing system and method for detecting an intrusion of a building according to the present invention will be described in detail with reference to FIG. 2.

First, the designated area of the building is photographed by the camera 110 (S210). As shown in FIG. 3, the camera 110 according to the present exemplary embodiment is installed outside the building BD so as to photograph a designated area of a building that requires monitoring. When the monitoring area is the interior of the building, the camera 110 may be installed inside the building. However, according to the present embodiment, the camera 110 may be installed to photograph the exterior of the building such as an apartment, a high-rise building, or a factory. It is desirable to be able to detect intrusions of buildings through fired gas lines or other structures or to detect fires. In addition, it is desirable to exclude windows, billboards, etc., whose brightness changes from time to time, such as apartments or buildings, from the surveillance area. The image captured by the camera is provided to the signal converter 120, and FIG. 6A shows an example of such a captured image.

Thereafter, the signal converter 120 converts the photographing signal input from the camera 110 into a digital signal (S220), and then reduces the resolution by using only the Y (brightness) signal component value among the converted signals to reduce the image size. Reduce (S230). That is, as an example, the signal converter 120 reduces the NTSC 720 * 480 (PAL 720 * 576) size to NTSC 352 * 240 (PAL 352 * 288) using only the Y component value of the converted digital signal, and then 30 per second. 7.5 frames, which are one-quarter of the frame (NTSC reference) data, are extracted and transmitted to the image processor 130. However, the present invention is not limited thereto, and (1/2) n to (1/4) n image frames of n input image frames per second may be appropriately used. By reducing the image size as described above, the processing speed of the image processor 130 may be reduced to increase the processing speed.

Thereafter, the image filtering unit 132 blurs the captured image by using an image filter that averages using the surrounding pixels (S240). The image shown in FIG. 6B is an example of the blurring of the captured image of FIG. 6A. The reason for blurring the captured image is to remove noise components generated during the vibration of the camera or the processing of the captured signal by the camera itself. This will be described in detail.

Previously, a method of calculating an average of pixel values for noise removal has been introduced. The conventional method divides and divides an image frame into 2 * 2, 3 * 3, and 5 * 5 sized regions as shown in FIG. 4A. Compute and average. In other words, the first table uses a 2 * 2 mask, which is obtained by dividing the area of the image frame into quarters and the sum of the areas divided by four. Similarly, the second table uses a 3 * 3 mask, which is obtained by averaging the area of the image frame into nine equal parts and the sum of the areas divided by nine. The third table uses a mask with 3 * 3 weights. When the pixel values are averaged in this way, the noise components cancel each other out to obtain a pixel value from which the noise components are removed. However, when the operation is performed as in the conventional method, there is a problem in that the operation speed is lowered due to the division operation.

4B is a table showing a method of averaging pixel values according to the present invention. First, as shown in the left table, the average pixel value s of the remaining areas except for the center area z is obtained. As shown in the table on the right, s is allocated to the remaining areas except the center, and the center area is calculated using the formula below the table. As shown, both the left and right tables represent the division by a multiple of two (8, 4).

In this way, the mask and the expression are set to operate with a multiple of 2, and the operation is performed by dividing into two so that the filter operation can be performed at high speed. Because dividing by a multiple of 2, the shift operation can be performed in only one clock.

Meanwhile, the image filtering unit 132 generates a background image from the blurred image, updates the background image at regular intervals, and stores the background image in the image storage unit 140 (S250). Referring to FIG. 5, an image frame is stored in each of the set N image buffers at each set time t (S251), and an image obtained by averaging the N images is stored as a background image (S253). For example, if t is 4 seconds and N is set to 16, then 1 frame per 4 seconds is sequentially stored in each of 16 image buffers after the noise is removed (so that 16 image frames fill the image buffer for 64 seconds). The average of the sixteen images is stored as a background image. In addition, after generating the first background image, the old image frames are sequentially replaced in the N image buffers, and the background image is updated at each set time t (S252) (S254). In the above example, instead of excluding the old image frame every four seconds, the average of 16 images including the new image frame is stored as a new background image. The background image is updated every four seconds.

Objects such as rain or snow are not taken at the same location, so you take several images and use them as a background image. The reason for this simple process is to reduce the amount of computation so that the image processing unit 130 processes data of multiple channels, and has the advantage of increasing accuracy in object detection by adapting to environmental changes through continuous changes of the background image. have. However, in case of snow or rain, the background image itself may be slightly blurred. The image shown in FIG. 6C is an example of a background image made from an image extracted from the images blurred in FIG. 6B.

In addition, the image filtering unit 132 generates a reference image for identifying the moving object according to the difference between the blurred image and the background image in addition to the function of blurring the captured image and generating and updating the background image (S260). Although the background image may not include the moving object image, the blurred image includes the moving object image. Therefore, in order to recognize the intrusion target by creating a reference image to easily identify the moving object in consideration of the characteristics of the image. It can be utilized.

First, the image filtering unit 132 generates a reference image by converting the blurred image and the background image into black or white depending on whether the difference in magnitude of the Y signal is greater than the set reference value (S261). For example, in the reference image, an area corresponding to the object image is white and an area other than the object image is displayed in black. An example of such a black and white reference image is shown in Fig. 6D.

The present embodiment introduces a method of making a reference image according to the magnitude of the Y signal, but is not limited thereto. For example, if the signal converter creates an image frame using both the Y signal and the Cr and Cb signals as color components, and selects the image frame to generate a color background image, there may be a moving object even though the Y signal is similar in size. In consideration of the present situation, the reference image may be generated in black or white based on the size of the Cr and Cb signals as color components, as well as the Y signal, to generate a reference image in which the difference between the photographed image and the background image is black and white.

In order to further reduce the amount of computation in the signal processing of the generated reference image, the image filtering 132 adjusts the image correction and the size of the reference image.

As shown in FIG. 6D, the first generated reference image includes an image of a small size that is not related to an object belonging to a sensing object, and an image correction is performed to remove it. As shown in FIG. 7, the entire image of the nine equal regions including the pixel is entirely white when the surrounding eight pixels j1 to j8 surrounding the center pixel k are all white. set to w). By applying this image correction method to correct the image of FIG. 6D, the corrected image of FIG. 6E is obtained.

Then adjust the size of the corrected reference image. The image resizing method uses a logical sum operation (OR) to reduce the resolution to reduce the size. As shown in FIG. 8, four blocks A1, A2, A3, and A4 composed of four pixels are reduced in correspondence with the four pixels Ba, Bb, Bc, and Bd of one block B1, respectively. For example, if any one of the four pixels of the arbitrary block A1 belonging to the original image is white, the pixel Ba of the block B1 belonging to the reduced image is determined to be white. The image of FIG. 6F is obtained from the image of FIG. 6E by applying the image resizing method, but does not operate on the remaining 3/4 region except for the 1/4 region of the upper left side of the entire image of FIG. 6F. As a result, the size of the reference image is reduced, which reduces the computation amount by about four times.

After the image is corrected in the image filtering unit 132 and the image size is adjusted, the object extracting unit 134 extracts the object image from the reference image (S270). This will be described in detail.

The object extracting unit 134 may label a region in which white pixels are clustered in the reference image, and four object images T1, T2, T3, and T4 may be specified as illustrated in FIG. 9A. .

Then, the object extraction unit 134 sets a rectangle including the object image for each object image. For example, as shown in FIG. 9B, the arbitrary object image T1 obtains the size of the rectangle, that is, the size of the object image, using the coordinates of the ends P1, P2, P3, and P4 in the four directions. As illustrated in FIG. 9C, the sizes of the various rectangles corresponding to the object images may be different.

The nonspecific object removing unit 135 removes nonspecific objects that do not belong to the sensing object from the object image extracted by the object extracting unit 134 in order to increase the reliability of intrusion detection (S280).

Here, the non-specific object refers to an object that is substantially unrelated to the object to be detected (for example, a person in case of intrusion, smoke, fire, etc. in case of fire) in detecting an intrusion or fire of a building. The removal process of the nonspecific object is performed step by step as follows.

First, when an image size of the extracted object image exceeds a preset minimum size MIN or maximum size MAX, the image is recognized and excluded as a nonspecific object. Because small animals, such as bees and rats, are displayed in a small size in the captured image, objects smaller than the predetermined minimum size are not included in the sensing object and are considered as non-specific objects. FIG. 10 illustrates a photographing image including a non-specific object. A small square overlapping the photographing image represents a minimum filter size for removing a non-specific object smaller than an object image to be detected. The large rectangle overlapping the captured image means the maximum filter size for removing non-specific objects larger than the object image to be detected. It corresponds to a non-specific object to be removed because it is smaller than the object image minimum filter size shown on the right side.

In addition, the nonspecific object removing unit 135 removes the non-specific object when the ratio of the horizontal size and the vertical size of the extracted object image is abnormal. As shown in FIG. 11, in the case of snow or rain falling in front of the camera, the camera is long taken by the camera. In other words, the vertical ratio is significantly larger than the horizontal ratio. Objects that are too long to one side are considered as nonspecific objects that do not belong to the detection object and are excluded.

In addition, the nonspecific object removing unit 135 considers and removes the non-specific object image when the size change of the object image is large in successive image frames. Referring again to FIG. 9C, the random object image D3 appears in the first image frame F1 and the second image frame F2 but disappears in the third image frame. In addition, the size of the object image D3 in the first image frame increases rapidly in the second image frame. This object image D3 corresponds to a nonspecific object. This is because moving objects such as snow, rain and birds appear and disappear instantaneously in the surveillance area of the camera. If the moving distance per unit time is larger than the preset distance, it can be regarded as a non-specific object and can be removed.

As described above, the captured image input in this embodiment uses 7.5 frames, which is 1/4 of 30 frames per second (NTCT basis), and the time for receiving data of the next frame from one frame is 0.1333 seconds (133 ms). . As shown in FIG. 12, the distance at which the object image to be measured moves for 0.133 seconds varies depending on the position where the camera is installed, but is limited in three consecutive image frames such as "A". Using this condition, moving a long distance such as "B" in two consecutive frames of the object's image under measurement is regarded as a non-specific object such as snow, rain, bugs, birds, etc., passing in front of the camera and removed. Done.

As described above, according to the present invention, if the non-specific object image is removed from the photographed image by using various methods, the accuracy of recognition of the object to be detected is increased and the amount of calculation is reduced, thereby increasing the computation speed.

After the nonspecific object removing unit 135 removes the nonspecific object image, the intrusion detecting unit 136 removes the nonspecific object image and tracks the path of the object image remaining in the reference image. In order to easily recognize the extracted object image, the outline of the object image may be displayed as a closed curve. The method of displaying the outline of the object image by the closed curve may be displayed by the method of displaying the outline of the real object by the closed curve and by the box shape surrounding the object image, and the present invention is not limited thereto. An example of the case where the captured image is displayed overlapped according to the latter method is shown in Fig. 6G.

The intrusion detecting unit 136 tracks the movement of the object image in successive image frames and determines whether the intrusion detection unit 136 has invaded the designated surveillance region (S290). In the method of determining an invasion, the coordinate corresponding to the center point of the extracted object image is determined with the incidence of invasion into the designated surveillance region, which is indicated by red in FIG. 6G. For example, if more than three) are detected, it is determined that the specified surveillance area is violated. Accordingly, the intrusion detection unit 136 provides an intrusion occurrence signal to the event generator 150.

If it is determined that the image of the object violates the surveillance area, the event generating unit 150 is set to notify the intrusion, for example, a buzzer built into a computer, a warning sound using a speaker, and an external device operation due to a relay operation. For example, an event may be transmitted by a method of transmitting a mail or a text using the Internet. In addition, the image storage unit 140 stores the captured image at the time of detecting the intrusion (S300).

On the other hand, the embodiment has been described mainly for the detection of intrusion of the building, in the case of fire can be detected due to the difference in the brightness of the outside of the building by the generation of smoke or flame. In this case, if the fire or smoke occurs slowly in the detection of fire, considering that the image change is not noticeable and can be treated as a background image, the image frame is selected at the interval of 4 seconds in the present embodiment. By selecting an image frame every time, a background image is generated so that the image change of a flame or smoke that is slow to change can be detected.

While preferred embodiments of the present invention have been described above, the present invention is not limited to the above specific embodiments. That is, a person having ordinary skill in the art to which the present invention pertains can make many changes and modifications to the present invention without departing from the spirit and scope of the appended claims, and all such appropriate changes and modifications are possible. Equivalents should be considered to be within the scope of the present invention.

1 is a block diagram of a photographing image processing system for detecting an intrusion of a building to which the present invention is applied.

2 is a flowchart illustrating a photographing image processing method for detecting an intrusion of a building according to an exemplary embodiment of the present invention.

3 is a diagram illustrating a state in which a camera according to the present embodiment is installed.

4A is a table for explaining a conventional method of removing noise.

4b is a table for explaining a method for removing noise according to the present invention;

5 is a view for explaining a method for generating a background image according to an embodiment of the present invention;

6A to 6G illustrate a series of processes for determining whether an object is invaded by separating an object image and a background image from a captured image according to the present invention.

7 is a view for explaining a method of correcting a reference image according to an embodiment of the present invention.

8 is a view for explaining a method of adjusting the size of a reference image according to an embodiment of the present invention.

9A to 9C are diagrams for describing a method for extracting an object image according to an exemplary embodiment of the present invention.

10 to 12 are diagrams for explaining a method of removing a non-specific object displayed by overlapping a captured image according to the present invention.

* Description of the symbols for the main parts of the drawings *

110: camera 120: signal conversion unit

130: image processing unit 140: image storage unit

150: event generator

Claims (10)

In the captured image processing method for intrusion detection of the building, Receiving a photographed image from a camera photographing the building; Separating a background image from the photographed image; Forming a reference image for identifying a moving object based on a difference between the photographed image and the separated background image; Extracting an object image from the reference image; Removing a non-specific object image which does not belong to a sensing object from the reference image; And Determining whether an object image remaining in the reference image from which the non-specific object image has been removed invades a designated surveillance region, The determining of the invasion may include determining whether an invasion occurs according to a frequency at which the center coordinates of the remaining object image invade the designated surveillance region. The method of claim 1, Generating the background image by averaging the image frames extracted from the photographed image, and updating the background image by replacing the old image frame with a new image frame every predetermined period. Way. The method of claim 2, Generating the background image Any one of generating a background image using only the Y (brightness) signal of the extracted image frame, and generating a background image using both the Y (brightness) signal of the extracted image frame and Cr, Cb as color components. Image processing method for intrusion detection of the building to apply the. The method of claim 1, The generating of the reference image may include blurring the photographed image and converting the photographed image into black or white according to a difference between the Y signal of the blurred image and the Y signal of the background image, thereby causing a difference between the photographed image and the background image. A photographing image processing method for detecting an intrusion of a building, the method comprising generating a black and white image appearing in black and white. The method of claim 1, The forming of the reference image may include correcting an image by removing a small sized image from the reference image, and reducing the size of the reference image by performing a logical sum operation on the corrected image. For shooting image processing method. The method of claim 1, And removing the non-specific object image, removing the extracted object image if the size of the extracted object image is smaller than the preset minimum size or larger than the maximum size. The method of claim 1, The removing of the non-specific object may include removing if the ratio of the horizontal size and the vertical size of the extracted object image is greater than a preset ratio. The method of claim 1, The removing of the non-specific object may include removing the image when the distance of the same object image is greater than a preset distance in image frames in which the extracted object image is continuously displayed. Way. delete In the shooting image processing system for detecting the intrusion of the building, A camera photographing the building; Separating the background image from the photographed image provided from the camera, forming a reference image for identifying a moving object based on a difference between the photographed image and the separated background image, extracting an object image from the reference image, An image processing unit for removing a non-specific object image that does not belong to the sensing object from the extracted object image and then determining whether the object image remaining in the reference image violates the designated surveillance region; And Including the event generator for generating a set event to notify the occurrence of intrusion in accordance with the determination of the image processor, And the image processor determines whether an invasion is made according to a frequency at which the center coordinates of the remaining object image invade the designated surveillance area.

Images