KR20060008268A - Smoke detecting method and system using ccd image - Google Patents

Abstract

The present invention relates to an automatic fire monitoring system, and more particularly, to a method and apparatus for automatically detecting smoke generation using visual and temporal characteristics of smoke generated in a fire.

In order to automatically monitor the fire from the CCD image taken by the CCD camera, the conventional technique through comparison with the background image is the solar lighting according to the weather and time such as the time, season and wind, when the background image is acquired in the outdoor environment. It is difficult to extract the background image under the influence of, and such a factor causes noise when calculating the difference image, and thus has a high possibility of false detection.

Accordingly, the present invention generates a stable background image from the CCD image to reduce the error situation due to noise, detect the amount of change in the color image, in consideration of the dynamic characteristics of the shape and size of the smoke, high accuracy and efficient automatic smoke detection method and To provide a device.

According to the present invention is excellent in the detection performance of the smoke even in the outdoors it is expected to contribute to the prevention of large fires such as forest fires.

CCD camera, fire monitor

Description

Smoke Detecting Method and System using CCD Image

1 is a block diagram of a smoke detection device in a CCD image according to the present invention.

2 is a flowchart of a smoke detection method in a CCD image according to the present invention;

3 is a flowchart of a background image generating method according to the present invention;

4 is a block average image configuration diagram according to the present invention.

<Description of the major reference numerals>

120: filter processing unit 130: background image generation unit

140: variation processing unit 150: candidate region and feature detection unit

160: smoke generation determination unit

The present invention relates to an automatic fire monitoring system, and more particularly, to a method and apparatus for automatically detecting smoke generation using visual and temporal characteristics of smoke generated in a fire.

CCD (charge coupled device) cameras have been widely used in fire monitoring systems recently because they have no afterimage and precision.

As a method of automatically monitoring a fire from a CCD image captured by a CCD camera, a method of detecting a change region using optical flow based on a luminance value and a previously obtained background image are used. A typical method is to set a reference image and calculate a difference in brightness values in units of frames, and to determine that a fire has occurred if the difference is large.

The method using the light flow tracking method of the brightness value requires a lot of time to calculate the light flow, and in case of a forest fire, most of the target areas have similar colors, brightness, and textures as forest areas, so there may be an error when tracking the light flow. .

The method of using the previously obtained background image as a reference image is a method of simply finding a changed place by calculating a difference between the background image and the current image frame and selecting a place having a large difference value. In this case, it is difficult to extract the background image due to the influence of the solar lighting according to the weather and time such as the time, season and wind from which the background image was acquired, and these factors generate noise when calculating the difference image, and thus have a high possibility of false detection. there was.

In the present invention, a stable background image is generated from a CCD image, which is commonly used in a fire (especially a forest fire) monitoring system, to reduce an error due to noise, to detect a change in color image, and to determine the dynamic characteristics of the shape and size of smoke. In view of this, a method and apparatus for automatically detecting smoke efficiently are proposed.

In order to solve the above technical problem, the present invention provides an automatic fire monitoring method using a CCD image, comprising: (a) a current image selection step of determining whether smoke is detected; (b) generating a background image using previous image frames where there is no smoke or motion; (c) a difference image calculation step of calculating an absolute value difference for each pixel between the background image and the current image; (d) a block average calculation step of calculating an average of a difference image in units of window blocks of a predetermined size; (e) obtaining a binary image by setting a value of 1 to a pixel having a difference image value above a threshold; (f) maintaining the previous background image if the number of blocks having a value of 1 in the binary image is greater than the threshold value, and updating the background image with the current image if it is smaller than the threshold value; (g) detecting a connection element of pixels having a value of 1 in the binary image and using the pixel coordinates of the connection element to obtain an image average color and standard deviation of the current image; (h) comparing the feature calculated from the connection element of the current image with the feature calculated from the connection element of the previous image to determine the presence of smoke; and provides a method for detecting smoke in a CCD image comprising a. .

In another aspect, the present invention provides an automatic fire monitoring apparatus using a CCD image, comprising: a background image generating unit generating a background image using an average image of an image frame in which no smoke or motion is present among the image frames; A variation processor configured to calculate a difference between the background image and the current image frame, calculate a block average, and calculate a binary image using a threshold; Applying a morphology operation to calculate a degree of change of the current image, to calculate a connection element of a predetermined size or more, and a candidate area and feature detection unit based on the size of the area, the geometric mean of the area, the image color average and the standard deviation in the area; A smoke generation determination unit determining a presence of smoke by comparing a feature calculated from a connection element of a current image frame with a feature calculated from a connection element of a previous frame; It provides a smoke detection device in a CCD image comprising a.

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

1 is a block diagram of a smoke detection apparatus in a CCD image according to the present invention, the CCD camera 100, the image acquisition unit 110, the filter processing unit 120, the background image generating unit 130, the variation processing unit 140, the candidate region and the feature detector 150, and the smoke generation determiner 160.

The CCD camera 100 is a solid-state image pickup device camera capable of obtaining color or gray scale images. The CCD camera 100 is fixed or configured to be rotated or vertically adjusted by a pan / tilt controller and a controller.

The image acquisition unit 110 is a frame grabber device capable of acquiring and storing image signals input from the CCD camera 100 as image frames at predetermined time intervals.

The filter processor 120 performs median filtering to remove spatial noise present in an image frame obtained from the image acquirer 110.

The background image generating unit 130 generates and updates a background image by averaging recent K image frames having no smoke or motion among the image frames. It is possible to reflect the change of the image over time by updating the background image by using a recent image with no smoke or motion. In addition, the noise can be reduced by using the average image through the average value rather than using a single image.

The variation processor 140 may calculate a variation of the current image frame from the background image by calculating a difference value between the background image and the current image frame. In order to exclude subtle changes due to noise, the average is calculated in units of blocks to detect a block having a change above a threshold and a reduced binary image is calculated. Using the binary image, the number of blocks having a change above the threshold is calculated and used as the degree of change of the current image frame. In the case of an image frame having a change degree lower than a threshold value, the background image generator 130 is used to update the background image.

When the degree of change is higher than the threshold, the candidate region and the feature detector 150 remove noise from the binary image through morphology calculation, and connect the connected components in consideration of 4-pixel or 8-pixel adjacentness of the binary image. After detecting, the connection elements having a predetermined size or less among the detected connection elements are judged as noise, and the characteristics of the size, geometric center, color average and standard deviation of the connection elements are detected in the connection elements having a predetermined size or more.

The smoke generation determiner 160 determines smoke generation by comparing the feature calculated in the connection element of the current image frame with the feature calculated in the connection element of the previous image frame.

FIG. 2 is a flowchart illustrating a method for detecting smoke in a CCD image according to the present invention, which will be described with reference to each component shown in FIG. 1 when necessary.

In the video image frame 200, an image having a gray or color scale obtained through the CCD camera 100 and the image acquisition unit 110 is obtained.

The spatial median filtering 210 is performed by the filter processor 120 to remove spatial noise included in an image frame input from a CCD camera in the video image frame obtaining step 200. It is preferable to apply a median filter having excellent nonlinear noise processing capability. The median filter is filtered using a 3 × 3 window to minimize time constraints, and in the case of color images, median filtering is performed for each channel.

The current image frame selection step 220 is an image that is a target to determine whether there is a current smoke among the image frames. This is selected at regular intervals, and the current image at time t can be expressed as I (x, y, t).

In the background image generation step 230, the background image generator 130 calculates the background image using an average of K image frames before the current image in which no smoke or motion exists.

3 shows a relationship between K image frames, a background image, and a difference image. Equations 1, 2, and 3 below present a method of generating a background image, respectively. Equation 1 is a method of generating a background image by directly calculating an average of past K images. Equation 2 shows a method of generating a background image B (x, y, t) using the previous background image B (x, y, t-1). Equations 1 and 2 should store K image frames in the past. Equation 3 uses the background image B (x, y, t-1) and the current image frame I (x, y, t) at time t when the past K images cannot be stored due to memory limitations. The background image B (x, y, t) is calculated.

(수식 1)

(Formula 1)

(수식 2)

(Formula 2)

(수식 3)

(Formula 3)

If there are n surveillance zones to be monitored by one camera, n background images are required. At this time, if there is enough memory, K image frames acquired in each set area are stored, and n background images are calculated using Equation 1, Equation 2 or Equation 3. If there is not enough memory, Pan / Tilt Whenever the camera is moved to a new area by the controller and controller, the background image can be newly constructed by initializing the K images stored in the buffer.

The difference image calculation 240 performs the absolute pixel-by-pixel between the background image B (x, y, t) generated using K image frames up to time t-1 and the current image I (x, y, t) at time t. Compute the change amount from the previous background image by calculating the difference difference. The gray image simply calculates the absolute value of the difference between the brightness values, and in the case of the RGB color image, it is calculated as shown in Equation 4. Other color models, such as HSI and YIQ, can be calculated and processed similarly to Equation 4.

(수식 4)

(Formula 4)

The block average calculation step 250 calculates an average image of the difference image in units of M × N window blocks. As shown in FIG. 4, when the size of the original image 400 is V × H, the block average image 410 is a reduced image having a size of V / M × H / N.

보다 큰 변화량을 갖는 블록은 수식 5에 의해 V/M × H/N 크기의 이진영상(binary image)을 계산한다.The binary image calculation 260 having an amount of change above the threshold has a threshold.

A block having a larger amount of change calculates a binary image of V / M × H / N size by Equation 5.

(수식 5)

(Formula 5)

보다 큰 변화량을 갖는 블록의 개수이다. nBlock은 현재 프레임의 배경영상으로부터 변화된 정도를 나타낸다. 따라서 nBlock >

이 면, 현재영상 I(x, y, t)를 배경영상 생성에 참여시키지 않는다. 즉 변화정도가 큰 현재영상은 시간 t+1에서 사용할 배경영상 생성에 참여시키지 않고, 시간 t+1에 사용할 배경영상 B(x, y, t+1)으로 이전의 배경영상 B(x, y, t)을 사용한다. 변화가 임계치 이하로 검출되는 경우에는 지속적으로 배경 영상을 갱신하고, 연기 등으로 인하여 임계치 이상으로 변화가 검출되는 경우에는 연기 등 이동이 없는 상태의 배경영상을 계속 유지하는 방법이다. NBlock in Equation 6

The number of blocks with a larger amount of change. nBlock represents the degree of change from the background image of the current frame. Thus nBlock >

In this case, the current image I (x, y, t) is not involved in generating the background image. That is, the current image with a large degree of change does not participate in the generation of the background image to be used at time t + 1, but the background image B (x, y, t + 1) to be used at time t + 1. , t). When the change is detected below the threshold, the background image is continuously updated. When the change is detected above the threshold due to smoke, the background image without a movement such as smoke is continuously maintained.

(수식 6)

(Formula 6)

The morphology operation application step 270 removes fine noise by using erosion and dilation using a 3 × 3 mask window in the binary image A (x, y, t).

는 시간 t의 현재영상 I(x, y, t) 에 변화량이 있는 블록들의 i번째 연결 요소이다. The connection element and feature detection step 280 is performed by the candidate region and the feature detection unit 150 to detect the connection element in consideration of the four-way or eight-way adjacentness of pixels having a binary image value of 1 in the binary image. . In equation 7

Is an i th connection element of blocks having a change amount in the current image I (x, y, t) at time t.

(수식 7)

(Formula 7)

의 화소의 수

가 연결 요소의 크기 임계치(

)보다 작은 연결 요소, 즉

<

는 삭제하여 크기가 작은 변화 영역은 삭제한다. 계산 속도를 증가시키기 위하여 각 영상 프레임에서 화소의 수

가 임계치(

) 보다 큰 연결 요소 중에서 가장 큰 2, 3개의 적은 수의 연결 요소만을 후보로 고려할 수 있다.Connection element

Number of pixels

Is the size threshold of the connection element (

Connection elements smaller than

<

Delete the small change area by deleting. Number of pixels in each image frame to increase computation speed

Is the threshold (

Only the smallest number of two or three connecting elements among the larger connecting elements can be considered as candidates.

에서 연기 여부를 판단하기 위한 특징으로 수식 8, 수식 9, 수식 10을 사용한다.If the size of the connection element is greater than the threshold

In Equation 8, Equation 9, and Equation 10 are used to determine whether to postpone.

(수식 8)Size of the i th connection element:

(Equation 8)

(수식 9)

(Formula 9)

과 표준편차

를 구한다. (수식 10)Average color of the image in the current image frame I (x, y, t) using the pixel coordinates of the i-th connection element

And standard deviation

Obtain (Formula 10)

In the calculation of Equation 10, since the pixel coordinates of the connection element are the pixel coordinates in the reduced block, they are multiplied by the window size (M × N) used in the block average image calculation 150.

와 시간 t에서의 j번째 연결 요소

사이에 대응되는 연결 요소는 수식 8의 연결 요소의 크기와 수식 9의 연결 요소의 기하학적 중심 위치를 이용하여 대응관계를 결정한다. 연기가 발생할 초기의 경우 수식 8의 크기 특징이 점점 커지는 효과가 있으며, 수식 9의 기하학적 중심 위치가 수평적으로 보단 연기의 특성상 수직적으로 변하는 특징이 있다. 또한 사전에 연기 색상에 대한 평균 색상, mSmokeColor와 stdSmokeColor를 연기를 포함하고 표본영상으로부터 계산하여 연기색상의 비교의 기준으로 사용한다.Delay determination 290 based on the feature is a step by the smoke generation determination unit 160, the i-th connection element at time t-1

Jth connection element at time t

The corresponding connection elements are determined using the size of the connection element of Equation 8 and the geometric center position of the connection element of Equation 9 to determine the correspondence relationship. In the early stages of smoke generation, the size characteristic of Equation 8 is gradually increased, and the geometric center position of Equation 9 is vertically changed due to the characteristics of the smoke rather than horizontally. In addition, the average color, mSmokeColor and stdSmokeColor , for smoke color is included in advance and calculated from sample images and used as a reference for comparison of smoke color.

(수식 11)

(Equation 11)

는 시간 t에서 구한 i번째 연결 요소

의 화소수이고,

는

에 대응하는 시간 t-1의 j번째 연결 요소

의 화소수이다. 수식 11은 연기 영역의 크기가 이전 프레임에서보다 점점 커지며, 연기 영역 내에서의 색상이 유사하고, 기존의 연기 색상과 표준편차 범위 내에 있으면 연 기로 판단한다. In Equation 11

Is the i th connection element at time t

Is the number of pixels

Is

J th connection element at time t-1 corresponding to

Is the number of pixels. Equation 11 determines the smoke as the smoke area becomes larger than in the previous frame, the color is similar in the smoke area, and is within the standard deviation range from the existing smoke color.

The invention can also be embodied as program code in a computer. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and the like, and may also be implemented in the form of a carrier wave (for example, transmission over the Internet). Include. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion. And functional programs, codes, and segments for implementing the present invention can be easily inferred by programmers in the art to which the present invention belongs.

Although specific terms have been used herein, they are used only for the purpose of describing the present invention and are not intended to limit the scope of the invention as defined in the claims or the claims. In addition, preferred embodiments of the present invention are disclosed for the purpose of illustration, those skilled in the art will be able to various modifications, changes, additions, etc. within the spirit and scope of the present invention, such modifications and modifications belong to the following claims Must see

According to the present invention made as described above, even in the presence of a large number of error determination elements in the image captured by the CCD camera can be excellent in the detection performance of smoke can lower the false detection rate of forest fire, the alarm signal automatically at the beginning of ignition It is possible to provide fire or forest fire detection system that can effectively monitor and operate wildfires by generating a fire, thereby minimizing human and physical damages of fire or wildfire that can cause enormous damage in case of occurrence.

Claims (10)

In the automatic fire monitoring method using a CCD image, (a) selecting a current image which is a determination target of smoke detection; (b) generating a background image using previous image frames where there is no smoke or motion; (c) a difference image calculation step of calculating an absolute value difference for each pixel between the background image and the current image; (d) a block average calculation step of calculating an average of a difference image based on window blocks having a predetermined size; (e) obtaining a binary image by setting a value of 1 to a pixel having a difference image value above a threshold; (f) maintaining the previous background image if the number of blocks having a value of 1 in the binary image is greater than the threshold value, and updating the background image with the current image if it is smaller than the threshold value; (g) detecting a connection element of pixels having a value of 1 in the binary image and using the pixel coordinates of the connection element to obtain an image average color and standard deviation of the current image; and (h) comparing the feature calculated from the connection element of the current image with the feature calculated from the connection element of the previous image to determine the presence of smoke. The method of claim 1, wherein the step (a) selects the current image at a predetermined time period and further includes a spatial media filtering step to provide a current image from which spatial noise is removed. . The method of claim 1, wherein the background image generation in step (b) is performed by any one of Equations 1, 2, and 3 below.

(Formula 1)

(Formula 2)

(Formula 3) Where I (x, y, t): current image at time t         B (x, y, t): Background image at time t         K: Number of previous images. The method of claim 1, wherein the step (g) further comprises a morphology calculation step of removing fine noise through erosion and dilation using a mask window on the connection element binary image. Smoke detection method in CCD image. [2] The method of claim 1, wherein the step (g) detects the connection element in consideration of four-direction or eight-direction adjacency centering on a pixel having a binary image value of 1. The method of claim 1, wherein the step (g) further includes a pixel number determination step of excluding a detected number of pixels of the connection element smaller than a threshold. The method of claim 1, wherein step (g) comprises:

,

) Is calculated by the following equation, and the average color and standard deviation of the current image are used by multiplying the pixel coordinates by the window block size used in step (d).

here,

: Number of pixels of connecting element. The method of claim 1, wherein in the step (h), in the comparison between the current image and the previous image, when the number of pixels of the connection element is increased, the color is similar, and is within the standard deviation range from the smoke color, the smoke is judged as smoke. Smoke detection method in a CCD image, characterized in that. An automatic fire monitoring apparatus using a CCD image, the apparatus comprising: a background image generation unit generating a background image using an average image of an image frame in which no smoke or motion exists among image frames; A variation processor configured to calculate a difference between the background image and the current image frame, calculate a block average, and calculate a binary image using a threshold; Applying a morphology operation to calculate a degree of change of the current image, to calculate a connection element of a predetermined size or more, and a candidate region and feature detection unit based on the size of the region, the geometric mean of the region, the image color average and the standard deviation within the region; A smoke generation determination unit determining a presence of smoke by comparing a feature calculated from a connection element of a current image frame with a feature calculated from a connection element of a previous frame; Smoke detection device in a CCD image comprising a. In a recording medium recording a program that can be read and executed by an automatic fire monitoring system using a CCD image, (a) selecting a current image which is a determination target of smoke detection; (b) generating a background image using previous image frames where there is no smoke or motion; (c) a difference image calculation step of calculating an absolute value difference for each pixel between the background image and the current image; (d) a block average calculation step of calculating an average of a difference image based on window blocks having a predetermined size; (e) obtaining a binary image by setting a value of 1 to a pixel having a difference image value above a threshold; (f) maintaining the previous background image if the number of blocks having a value of 1 in the binary image is greater than the threshold value, and updating the background image with the current image if it is smaller than the threshold value; (g) detecting a connection element of pixels having a value of 1 in the binary image and using the pixel coordinates of the connection element to obtain an image average color and standard deviation of the current image; and (h) comparing the feature calculated from the connection element of the current image with the feature calculated from the connection element of the previous image to determine the presence of smoke.

Images