JP2006268200A - Flame/gas smoke detecting system, and flame/gas smoke detecting method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a flame/gas smoke detecting system, and a flame/gas smoke detecting method capable of detecting leakage or the like of flame or gas smoke from a distant place by using a general-use photographing means. <P>SOLUTION: The photographing means 12 is provided for sequentially acquiring first, second, and third photographed images at certain time intervals in regard to the same scene. A difference image generating means 141 is provided for generating the first and second photographed images. An image binarization means 142 is provided for generating first and second binarized difference images. A composite image generating means 143 is provided for generating a logical product composite image, and a logical product composite image of the first and second binarized difference images. A filter means 144 is provided for removing high frequency components in a logical add composite image and in the logical product composite image, and generating a filter processed logical add composite image, and a filter processed logical product composite image. A scene change block specifying means 145 is provided for dividing the filter processed logical add composite image and the filter processed logical product composite image into plural areas, and specifying an area wherein the same scene changes. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a flame / gas smoke detection system and a flame / gas smoke detection method capable of accurately and easily detecting flames and gas smoke from a remote location using general-purpose imaging means.

  For example, in an oil refining plant or the like, even in a normal operation state, water vapor, flue gas, and the like are discharged from predetermined exhaust facilities, chimneys, and the like. In a chemical plant such as this type of oil refining plant, early detection is indispensable when an abnormality occurs in the operation state of various facilities or when a fire or the like occurs in various facilities or facilities.

  Although it is possible for a monitor to detect anomalies with a monitor by constantly photographing predetermined equipment and facilities with multiple cameras, flames and gas smoke are often difficult to see, and the monitor is not able to see from the monitor. I often release.

  For this reason, conventionally, for example, a technique for detecting the movement, shape, etc. of an object such as a leak of a flame or gas smoke from a remote place using a general-purpose camera is known.

  With this type of technology, an image captured by a normal monochrome camera or color camera can be subjected to image processing, and the captured image can be compared with a normal image.

However, in particular, when it is desired to detect flames or gas smoke, water vapor or people, automobiles, birds, or the like may be erroneously detected, and accurate processing (processing with high reliability) cannot be performed.
In general, the image comparison process is complicated and takes time, and there is a risk of causing a catastrophe once an accident such as a fire occurs. On the other hand, if the movement, shape, etc. of an object is detected too sensitively, erroneous detection will occur frequently. On the other hand, when an accident actually occurs, the problem is that the discovery is delayed.

  An object of the present invention is to provide a flame / gas smoke detection system and a flame / gas smoke detection method that can detect flames, gas smoke leaks, etc. from a remote location with high accuracy using general-purpose imaging means. It is in.

The flame / gas smoke detection system according to the first aspect of the present invention is summarized as (1) to (4).
(1) A flame / gas smoke detection system for detecting flame or gas smoke using a photographing means,
With respect to the same scenery, the photographing means for obtaining a reference photographed image and sequentially obtaining a plurality of detection photographed images at predetermined time intervals;
A reference captured image analysis means for dividing the reference captured image into a plurality of blocks according to a predetermined block division rule, and performing image analysis (texture analysis or the like) for each block;
Reference photographic image analysis result storage means for storing an image analysis result for the reference photographic image;
Scene change detection means for dividing the detection photographed image into a plurality of blocks according to the block division rule, and detecting a scene change of the detection photographed image based on an image change in each block;
When there is an image change for any of the blocks, a scene change block analysis unit that performs image analysis on the block that has the image change;
The image analysis result for the block related to the scene change by the scene change block analysis unit is compared with the image analysis result for the corresponding block stored in the reference photographed image analysis result storage unit. Flame / gas smoke determination means for specifying whether the change is a change due to a flame or a change due to gas smoke;
A flame / gas smoke detection system characterized by comprising: "

(2) “When the brightness change in each block is changed, the scene change detection means is configured such that when the number of pixels whose brightness has changed exceeds a predetermined ratio with respect to the total number of pixels constituting each block. The flame / gas smoke detection system according to (1), characterized in that it is determined that an image has changed. "

(3) To detect flame or gas smoke using three images of the first captured image, the second captured image, and the third captured image among the detection captured images acquired by the imaging unit (1) A flame / gas smoke detection system as described,
The scenery change detecting means is
A first difference image that is a difference between luminance values of the first and second captured images, and a second difference image that is a difference of luminance values between the second and third captured images are generated. Difference image generation means;
Image binarization means for binarizing the first difference image and the second difference image, respectively, and generating a first binarized difference image and a second binarized difference image;
Generating a logical sum composite image that is a logical sum of the first binarized difference image and the second binarized difference image, and / or the first binarized difference image and the second binarized difference A composite image generating means for generating a logical product composite image that is a logical product with the image;
Filter means for removing a high frequency component in the logical sum composite image to generate a filtered logical sum composite image and / or generating a filtered logical product composite image by removing a high frequency component in the logical product composite image;
In each block of the filtered logical sum composite image and / or the filtered logical product composite image, the number of pixels constituting the object appearing in each block is predetermined with respect to the total number of pixels constituting each block. A landscape change block identifying means for judging that there has been an image change when the ratio is exceeded,
A flame / gas smoke detection system characterized by comprising: "

(4) “The flame / gas smoke detection system according to any one of (1) to (3), wherein the image analysis includes a pattern recognition process.”

The flame / gas smoke detection system of the second aspect of the present invention is summarized in (5) or (6).
(5) A flame / gas smoke detection system for detecting flame or gas smoke using a photographing means,
With respect to the same scenery, the imaging means for acquiring captured images and sequentially acquiring a plurality of captured images at predetermined time intervals;
Scene change detection means for dividing the photographed image into a plurality of blocks according to the block division rule, and detecting a scene change of the photographed image based on an image change in each block;
Flame / gas smoke determination means for specifying whether the scene change is a change due to a flame or a change due to a gas smoke based on a spatial and / or temporal change of the group of blocks in which an image change occurs;
A flame / gas smoke detection system characterized by comprising: "

(6) “The flame or gas smoke is detected using three consecutive images of the first captured image, the second captured image, and the third captured image among the captured images acquired by the imaging unit” A flame / gas smoke detection system,
The scenery change detecting means is
A first difference image that is a difference between luminance values of the first and second captured images, and a second difference image that is a difference of luminance values between the second and third captured images are generated. Difference image generation means;
Image binarization means for binarizing the first difference image and the second difference image, respectively, and generating a first binarized difference image and a second binarized difference image;
Generating a logical sum composite image that is a logical sum of the first binarized difference image and the second binarized difference image, and / or the first binarized difference image and the second binarized difference A composite image generating means for generating a logical product composite image that is a logical product with the image;
Filter means for removing a high frequency component in the logical sum composite image to generate a filtered logical sum composite image and / or generating a filtered logical product composite image by removing a high frequency component in the logical product composite image;
In each block of the filtered logical sum composite image and / or the filtered logical product composite image, the number of pixels constituting the object appearing in each block is predetermined with respect to the total number of pixels constituting each block. A landscape change block identifying means for judging that there has been an image change when the ratio is exceeded,
A flame / gas smoke detection system characterized by comprising: "

The flame / gas smoke detection method of the first aspect of the present invention is summarized as (7) to (10).
(7) “A flame / gas smoke detection method for detecting flame or gas smoke using a photographing means,
A shooting step of acquiring a reference shot image and sequentially acquiring a plurality of detection shot images at predetermined time intervals for the same scene,
The reference captured image analysis step for dividing the reference captured image into a plurality of blocks according to a predetermined block division rule and performing image analysis for each block;
A reference captured image analysis result storage step for storing an image analysis result for the reference captured image;
A scene change detection step of dividing the detection photographed image into a plurality of blocks according to the block division rule, and detecting a scene change of the detection photographed image based on an image change in each block;
When there is an image change for any of the blocks, a scene change block analysis step for performing image analysis on the block having the image change;
The image analysis result for the block related to the scene change in the scene change block analysis step is compared with the image analysis result for the corresponding block stored in the reference captured image analysis result storage step, and the scenery A flame / gas smoke determination step for identifying whether the change is a change due to a flame or a change due to gas smoke;
A flame / gas smoke detection method characterized by comprising: "

(8) “In the scene change detection step, when the luminance in each block changes, the number of pixels whose luminance has changed exceeds a predetermined ratio with respect to the total number of pixels constituting each block. The flame / gas smoke detection method according to (7), characterized in that it is determined that an image has changed. "

(9) “Flame or gas smoke is detected using three consecutive images of the first, second, and third images among the detection images acquired in the imaging step (7). A flame / gas smoke detection method according to claim 1,
The scenery change detection step includes
A first difference image that is a difference between luminance values of the first and second captured images, and a second difference image that is a difference of luminance values between the second and third captured images are generated. A difference image generation step;
An image binarization step for binarizing the first difference image and the second difference image, respectively, and generating a first binarized difference image and a second binarized difference image;
Generating a logical sum composite image that is a logical sum of the first binarized difference image and the second binarized difference image, and / or the first binarized difference image and the second binarized difference A composite image generation step for generating a logical product composite image that is a logical product with the image;
A filter step of removing a high-frequency component in the logical sum composite image to generate a filtered logical sum composite image and / or generating a filtered logical product composite image by removing a high-frequency component in the logical product composite image;
In each block of the filtered logical sum composite image and / or the filtered logical product composite image, the number of pixels constituting the object appearing in each block is predetermined with respect to the total number of pixels constituting each block. When the ratio exceeds the ratio, it is determined that there has been an image change, and the block is a landscape change block specifying step,
A flame / gas smoke detection method characterized by comprising: "

(10) The flame / gas smoke detection method according to any one of (7) and (9), wherein the image analysis includes a pattern recognition process.

The flame / gas smoke detection method of the second aspect of the present invention is summarized in (11) or (12).
(11) “A flame / gas smoke detection method for detecting flame or gas smoke using a photographing means,
A shooting step for sequentially acquiring a plurality of captured images at predetermined time intervals for the same scenery,
A scene change detection step of dividing the photographed image into a plurality of blocks according to a predetermined block division rule, and detecting a scene change of the photographed image based on an image change in each block;
A flame / gas smoke determination step for determining whether the scene change is a change due to a flame or a change due to a gas smoke based on a spatial and / or temporal change of the group of blocks in which an image change occurs;
A flame / gas smoke detection method comprising: "

(12) As described in (11), flame or gas smoke is detected using three images of the first captured image, the second captured image, and the third captured image among the captured images acquired in the capturing step. A flame / gas smoke detection method,
The landscape change detection step includes:
A first difference image that is a difference between luminance values of the first and second captured images, and a second difference image that is a difference of luminance values between the second and third captured images are generated. A difference image generation step;
An image binarization step for binarizing the first difference image and the second difference image, respectively, and generating a first binarized difference image and a second binarized difference image;
Generating a logical sum composite image that is a logical sum of the first binarized difference image and the second binarized difference image, and / or the first binarized difference image and the second binarized difference A composite image generation step for generating a logical product composite image that is a logical product with the image;
A filter step of removing a high-frequency component in the logical sum composite image to generate a filtered logical sum composite image and / or generating a filtered logical product composite image by removing a high-frequency component in the logical product composite image;
In each block of the filtered logical sum composite image and / or the filtered logical product composite image, the number of pixels constituting the object appearing in each block is predetermined with respect to the total number of pixels constituting each block. A scene change block identifying step for determining that there has been an image change when the ratio is exceeded,
A flame / gas smoke detection method comprising: "

  According to the present invention, an image to be processed is divided into blocks, and detection is performed in units of the blocks. Therefore, detection can be performed at high speed, and real-time detection of fire and gas leakage is possible. Further, since the filtering process is also performed in units of blocks, the calculation is not burdened, and thus simple hardware can be used.

  In addition, since it is possible to perform detection in consideration of the movement of an object (object) appearing in the scenery, water vapor, or a person, a car, a bird, or the like is not erroneously detected as flame or gas smoke.

FIG. 1 is an explanatory view showing a first embodiment of the flame / gas smoke detection system of the present invention.
In FIG. 1, a flame / gas smoke detection system 1 includes a photographing means (camera) 11, a reference photographing image analysis means 12, a reference photographing image analysis result storage means 13, a scenery change detection means 14, and a scenery change block. Analyzing means 15 and flame / gas smoke determining means 16 are provided.

  The photographing unit 11 obtains the reference photographed image g0 for the same scene S, and sequentially obtains the photographed images for detection Gn with a time interval (for example, 0.2 second to 6 seconds). The reference captured image g0 is usually a normal image. The detection photographic images Gn are a first photographic image G11, a second photographic image G12, and a third photographic image G13 (examples of G11, G12, and G13 are shown in FIG. 2).

  As shown in FIG. 3, the reference photographed image analyzing means 12 divides the reference photographed image g0 into a plurality of blocks Bij (i = 1, 2,..., M, j = 1, 2, .., N) and image analysis is performed for each block. As will be described later, the texture processing of the reference captured image g0 can be performed by performing pattern recognition processing on the block specified by the scenery change block specifying unit 145.

  The reference captured image analysis result storage unit 13 stores the image analysis result for the reference captured image g0.

The scene change detection means 14 divides the captured image Gn for detection into a plurality of blocks Bij (i = 1, 2,..., M, j = 1, 2,..., N) according to a block division rule. A scene change of the detection photographed image Gn is detected based on the image change in each block. Specifically, the scenery change detection means 14 includes a difference image generation means 141, an image binarization means 142, a composite image generation means 143, a filter means 144, and a scenery change block identification means 145. .
The difference image generation unit 141 is a difference between the luminance values of the first difference image GD11 that is the difference between the luminance values of the first photographic image G11 and the second photographic image G12, and the luminance value of the second photographic image G12 and the third photographic image G13. A second difference image GD12 is generated (examples of GD11 and GD12 are shown in FIG. 2).

  The image binarization unit 142 binarizes the first difference image GD11 and the second difference image GD12, respectively, and generates a first binarized difference image GB11 and a second binarized difference image GB12 (GB11 and GB12). An example is shown in FIG.

  The composite image generation means 143 is a logical sum of the first binarized difference image GB11 and the first binarized difference image GB12, and the first binarized difference image GB11 and the first binarized difference image GB11. A logical product composite image Gand that is a logical product with the difference image GB12 is generated (an example of Gand is shown in FIG. 2). In the logical sum composite image Gor, for example, an object having no change in density or an object having a slow change in density, for example, a fast moving object, such as a pedestrian or a traveling vehicle, appears.

  On the other hand, in the logical product composite image Gand, for example, an object with a rapid change in density and a slow movement appears. No fast object appears.

  The filter unit 144 removes high-frequency components from the logical sum composite image Gor to create a filtered logical sum composite image FGor, and removes high-frequency components from the logical product composite image Gand to create a filtered logical product composite image FGand. To do. By removing the high-frequency component, it is possible to remove minute changes (such as noise) and to enlarge the characteristic part (emphasize the low-frequency component).

  In the present embodiment, spatial filter processing is performed on the emphasized portion (object change portion) of the logical sum composite image Gor and logical product composite image Gand. Specifically, after performing a reduction process a times, the noise is removed by performing a enlargement process, and the emphasized portion is enlarged by performing b reduction processes after performing b expansion processes. it can.

  The scene change block specifying unit 145 is configured such that, in each block of the filtered logical sum composite image and / or the filtered logical product composite image, the number of pixels constituting the object appearing in each block is the total number of pixels constituting each block. It can be determined that the image has changed when the ratio exceeds the predetermined ratio with respect to the number. In the present embodiment, the filtered logical sum composite image FGor can be divided into a plurality of blocks B11, B12,...

In the image division, for example, a captured image of 480 dots long × 720 dots wide can be divided into blocks B11, B12,..., Bmn of vertical p dots × horizontal q dots (for example, p = q = 30).
The number of vertical and horizontal dots (values of p and q) of the blocks B11, B12,..., Bmn varies in the range of, for example, 4 to 64, depending on the size of the detected flame or gas smoke figure. Can be made. Then, the block (at least one of B11, B12,..., Bmn) is specified.

  In this case, the number of blocks Bij that have undergone image change, the number of blocks Bij that have undergone image change, the number of previous detections, and the number of objects in the block Bij that has undergone image change in all or local regions. The change in the proportion of the object and the contour of the object due to the FFT can be used as information for recognizing the flame / gas smoke, so that the amount of information for recognizing the flame / gas smoke increases.

  That is, the count of the number of blocks Bij in which the image has changed in the entire region or the local region, the comparison of the number of blocks Bij in which the image has changed with the number in the previous detection, and the block Bij in which the image has changed The flame / gas smoke detection accuracy can be improved by appropriately combining the calculation of the ratio occupied by the object in FIG.

  In addition, when there are blocks that have undergone image changes independently (when there are no image changes in the surrounding blocks), this image change can be determined as noise. On the contrary, when the block Bij in which the image has changed continuously exists, it can be predicted from the continuous pattern of the block Bij whether this is due to flame or gas smoke or due to water vapor. Of course, in this case, accurate determination is performed by texture processing by the flame / gas smoke determination means 16, but it is effective for early detection.

  When there is an image change for any of the blocks Bij, the scene change block analysis unit 15 performs image analysis for the block Bij in which the image change has occurred. In the texture processing, the pattern recognition processing can be performed on the block Bij specified by the scenery change block specifying unit 145 among the blocks having an image change.

  The flame / gas smoke determination unit 16 stores the image analysis result of the block relating to the scene change by the scene change block analysis unit 15 for the corresponding block Bij stored in the reference captured image analysis result storage unit 13. Compared with the image analysis result, it is specified whether or not the scene change is a change caused by flame or a change caused by gas smoke.

  For example, in the case where no image appears in the filtered logical sum composite image FGor and pixels appear significantly in the filtered logical product composite image FGand, this specification is a change due to flame or a change due to gas smoke. Can be determined.

  As described above, in the present invention, image analysis is performed for each block Bij (that is, only the block Bij whose image is changed can be targeted), so that the memory capacity used can be reduced and dedicated hardware such as a DSP is used. In this case, high-speed processing can be performed.

The operation of the flame / gas smoke detection system 1 shown in FIG. 1 will be described below with reference to the flowchart of FIG.
(1) Shooting step S101: The shooting means 11 sequentially acquires a first shot image G11, a second shot image G12, and a third shot image G13.
(2) Reference photographed image analysis step S102: The reference photographed image analyzing means 12 divides the reference photographed image g0 into a plurality of blocks Bij and performs image analysis for each block.
(3) Reference photographed image analysis result storage step S103: The reference photographed image analysis result storage means 13 stores an image analysis result for the reference photographed image g0.

(4) Scenery Change Detection Step S104: The landscape change detection means 14 divides the detection photographed image into a plurality of blocks Bij, and detects a scene change of the detection photographed image based on the image change in each block.
The landscape change detection step S104 includes a difference image generation step S1041, an image binarization step S1042, a composite image generation step S1043, a filter step S1044, and a landscape change block specifying step S1045.
(4-1) In the difference image generation step S1041, the difference image generation means 141 generates the first difference image GD11 and the second difference image GD12.
(4-2) In the image binarization step S1042, the image binarization unit 142 generates the first binarized difference image GB11 and the second binarized difference image GB12.
(4-3) In composite image generation step S1043, the composite image generation unit 143 generates a logical sum composite image Gor and a logical product composite image Gand.
(4-4) In filter step S1044, the filter processing unit 144 creates a filtered logical sum composite image FGor and a filtered logical product composite image FGand.
(4-5) Scene change block identification step S1045 identifies a block having an image change by the landscape change block identification unit 145.

  Note that the number of vertical and horizontal dots (values of p and q) of the blocks B11, B12,..., Bmn generated by image division in the reference captured image analysis step S102 and the landscape change detection step S104 is the detected flame. Depending on the size of the figure that is gas smoke, it can be varied in the range of 4 to 64, for example. Then, the block (at least one of B11, B12,..., Bmn) is specified, and the process proceeds to the next process. In the landscape change detection step S104, the process of dividing into blocks is performed in step S1045 in this embodiment, but may be performed anywhere in the process of the landscape change detection step S104.

(5) Scene change block analysis step S105: The scene change block analysis means 15 performs image analysis of the block specified in the landscape change block specification step S1045.
Specifically, the following processing is performed for the block in which the image has changed.
(A) For example, when the original image is a monochrome image of 256 gradations (“NO” in S1051), gradation conversion is performed on 16 gradations (S1052), and image analysis (analysis of 14 types of shapes, patterns, etc.) Analysis of whether or not the object is a transparent object) (S1053), shape / pattern feature extraction is performed (S1054), and multi-level slice processing is performed (S1055).
In the processing using the multi-level slice in S1055, the density range of the captured image is divided into a plurality of sections, and a specific density level is assigned to each section. In addition to this, it is possible to detect motion by extracting an optical flow.

(B) When the original image is, for example, a color image (“YES” in S1051), image analysis (analysis of 14 types of shapes, patterns, etc., analysis of whether the object is a transparent object) is performed for each RGB (S1056). Shape / pattern feature extraction is performed (S1057), and processing by a multi-level slice method is performed (S1058). In addition to this, it is possible to detect motion by extracting an optical flow.
Conventionally, processing has been complicated because image analysis has been performed on the entire screen of a captured image, or image analysis has been performed on an area within the constantly changing shape of smoke. In the present invention, as can be seen from the above-described embodiment, the processing is made efficient by performing image analysis on a block (that is, a narrow region). As in the past, smoke recognition by image analysis is performed on a monochrome screen in the above (a) example, and RGB conversion in the above (b) example, in addition to shape recognition after binarization and optical flow extraction. Although it is performed on the subsequent three screens, it can also be performed on the screen after the HSL conversion.

(6) Flame / gas smoke determination step S106: It is determined whether the flame / gas smoke determination means 16 is a change due to flame or a change due to gas smoke.
Actually, flames and gas smoke have various shapes, colors, concentrations, patterns, and appearances. Therefore, a standard model of flame or gas smoke can be registered using a part or all of these information and compared, and detection can be performed with high accuracy.

  In addition, in the flame / gas smoke determination step S106, in addition to the above processing, a technique of a density histogram is also adopted, and flame and gas smoke can be detected with high accuracy by considering the density distribution. In this case, since it is sufficient to create a histogram for one or a plurality of blocks having image changes, high-speed processing can be performed.

  FIGS. 6A, 6B, and 6C show the above image processing results. As shown in FIG. 6A, when a change of an object always appears in the same block, it can be predicted that the change of the object is water vapor. Further, as shown in FIG. 6B, when an object changes in different blocks, there is a possibility of flame or gas smoke, and this detection result is appropriately displayed on the display. Note that FIG. 6C shows a case where the change of the object does not appear.

  In the flame / gas smoke detection system 1, a plurality of photographing means can be provided to photograph a plurality of places. In addition, as shown in FIG. 7, one photographing means 11 is used and rotated in the N direction to photograph three scenes S1, S2, and S3 at a constant time interval (for example, one minute interval). it can.

FIG. 8 is an explanatory view showing a second embodiment of the flame / gas smoke detection system of the present invention.
In FIG. 8, the flame / gas smoke detection system 2 detects flame or gas smoke using an imaging unit, and includes an imaging unit (camera) 21, a scene change detection unit 22, and a flame / gas smoke determination unit 23. And.

  The photographing means 21 can acquire a reference photographed image and sequentially obtain a plurality of detection photographed images at predetermined time intervals for the same scenery. In the present embodiment, the imaging means 21 detects flame or gas smoke using images of consecutive captured images G11, G12, G13,... As detection captured images.

The scene change detection means 22 divides the detection photographed images (captured images G11, G12, G13,...) Into a plurality of blocks Bij according to the block division rule, and changes the scene change of the detection photographed image to the image in each block. It can be detected based on the change.
The scene change detection unit 22 includes a difference image generation unit 221, an image binarization unit 222, a composite image generation unit 223, a filter unit 224, and a scene change block identification unit 225.

The difference image generation means 221 is a difference in luminance value between the first difference image GD11, which is a difference in luminance value between the first photographic image G11 and the second photographic image G12, and between the second photographic image G12 and the third photographic image G13. A certain second difference image GD12 is generated.
The image binarization means 222 binarizes the first difference image GD11 and the second difference image GD12, respectively, and generates a first binarization difference image GB11 and a second binarization difference image GB12.

The composite image generation unit 223 is a logical sum of the first binarized difference image GB11 and the second binarized difference image GB12, a logical sum composite image Gor, and the first binarized difference image GB11 and the second binarized image. A logical product composite image Gand that is a logical product with the difference image GB12 is generated.
The filter unit 224 removes the high frequency components in the logical sum composite image Gor to generate a filtered logical sum composite image FGor, and removes the high frequency components in the logical product composite image Gand to generate a filtered logical product composite image. .

  The scene change block specifying means 225 is configured such that, in each block Bij of the filtered logical sum composite image Gor and the filtered logical product composite image Gand, the number of pixels constituting the object appearing in each block constitutes each block Bij. It is determined that the image has changed when the ratio exceeds the predetermined ratio with respect to the total number of pixels.

  The flame / gas smoke determination means 23 determines whether the scene change is a change due to flame or a change due to gas smoke based on a spatial and / or temporal change of the group of blocks in which an image change occurs. To do.

  The flame / gas smoke determination means 23 can also detect flame / gas smoke based on the detection results of a plurality of times. For example, when image changes occur intermittently in multiple blocks in a specific area, it is possible to determine whether it is due to flame, gas smoke, or other factors depending on the frequency of the image change. Can be determined. When detecting an image change in units of pixels, the processing takes time. However, in the present invention, the processing target is limited to a specific block, so the processing does not take time.

In addition, when an image change occurs in almost all blocks (a% or more blocks) and a change occurs in a wide area (area of b% or more) in each block, the camera (photographing means 21) turns. In this case, the photographed image can be excluded from the detection target. Further, when the above state continues for a long time, a failure of the photographing means 21 is expected, so that an early detection of a device failure is possible.
Conversely, even if a long time (for example, a certain time T) has passed, c% or more of all blocks Bmn (i = 1, 2,..., M, j = 1, 2,..., N) If there is no change at all in the block, it can be predicted that the camera lens of the photographing means 21 has been attacked by spray or the like, which is suitable as a countermeasure against terrorism against industrial plants and the like.

  The operation of the flame / gas smoke detection system 2 shown in FIG. 8 will be described below with reference to the flowchart of FIG. In the flame / gas smoke detection method of the present invention, flame or gas smoke can be detected using an imaging means.

  (1) Shooting step S201: The shooting unit 21 sequentially acquires a plurality of shot images for the same scene at predetermined time intervals.

(2) Landscape change detection step S202: The scene change detection means 22 divides the photographed image into a plurality of blocks according to the block division rule, and detects a scene change of the photographed image based on the image change in each block. .
The landscape change detection step S202 includes a difference image generation step S2021, an image binarization step S2022, a composite image generation step S2023, a filter step S2024, and a landscape change block specifying step S2025.
(2-1) In the difference image generation step S2021, the difference image generation means 221 generates the first difference image GD11 and the second difference image GD12.
(2-2) In the image binarization step S2022, the image binarization unit 222 generates the first binarized difference image GB11 and the second binarized difference image GB12.
(2-3) In composite image generation step S2023, the composite image generation means 223 generates a logical sum composite image Gor and a logical product composite image Gand.
(2-4) In filter step S2024, the filter means 224 generates a filtered logical sum composite image FGor and a filtered logical product composite image FGand.
(2-5) Scene change block identification step S2025 identifies the block in which the image has changed by the scene change block identification means 225.

  (3) Flame / gas smoke determination step S203: The flame / gas smoke determination means 23 specifies whether the scene change is a change due to flame or a change due to gas smoke.

Hereinafter, a specific example of discrimination between smoke and water vapor by extracting shape features will be described.
FIG. 10 is an image showing a scene where smoke and water vapor are simultaneously generated on the screen. Although not clearly shown in FIG. 10, the water vapor disappears immediately and does not go far, but the smoke does not disappear and drifts far away. Based on this feature, smoke and water vapor are identified by focusing on the area of the shape feature.

  FIG. 11 shows the result of generating a logical sum composite image every 4 seconds for a scene where smoke is generated, and obtaining the smoke area. FIG. 12 shows a result of similarly generating a logical sum composite image for a scene where water vapor is generated and determining the area of the water vapor.

  FIG. 11 shows a feature such as “the smoke does not disappear and drifts far”, and FIG. 12 shows a feature that “the water vapor immediately disappears and does not go far”. That is, as can be seen by comparing FIG. 11 and FIG. 12, the difference in the area of smoke and water vapor also appears due to the difference in the above characteristics.

FIG. 13A is a graph showing the time change of the smoke area, and FIG. 13B is a graph showing the time change of the water vapor area. 13A and 13B, it can be seen that the amount of change in the area of smoke is large while the amount of change in the area of water vapor is small.
That is, it is effective to compare the dispersion of areas for the discrimination between smoke and water vapor, and the smoke and water vapor can be discriminated by variation in area, that is, dispersion.
The formula for determining the dispersion of the smoke and water vapor areas is shown below.

  FIG. 14A is a graph showing the calculation result of smoke dispersion, and FIG. 14B is a graph showing the calculation result of water vapor dispersion. Since the variation in area is greater for smoke than for water vapor, the overall variance in area is greater for smoke. As can be seen from these, comparison of the size and dispersion of the shape features is extremely effective as a method for distinguishing between smoke and water vapor. When performing detection based on an algorithm that focuses on the shape characteristics of smoke and water vapor, as described above, the captured image is divided into a plurality of blocks and processed in units of blocks, thereby enabling high-speed processing.

  As described above, the flame / gas smoke detection system 2 (see FIG. 8) should be used as an anti-terrorism measure against an industrial plant or the like by predicting that the camera lens of the photographing means 21 has been attacked by a spray or the like. Can do. In the present invention, it is possible to take more effective measures against acts such as forced movement and destruction of the photographing means 11 (FIG. 1) and the photographing means 21 (FIG. 6) by the following method.

FIG. 15 is a functional block diagram showing the device movement detection system 3 for forced movement / destruction. The device movement detection system 3 can be installed in a portable device such as a normal node book type computer. In this embodiment, the function of the detection device is narrowed down and applied to the monitoring system of the present invention. It shall be.
The photographing means (video camera 309) can be attached to the device movement detection system main body 31 via the external device connection interface I / F, or can be configured integrally with the device movement detection system main body 31.

  The device movement detection system 3 can include monitoring mode on means 301. In addition, the device movement detection system 3 operates when the monitoring mode is on, the device power switch operation invalidating means 302, the video camera driving means 303, the shooting area dividing means 304, and the scenery change detecting means 305. , A scene change block ratio calculation unit 306, a warning generation unit 307, a monitoring mode off unit 308, a video camera 309, a captured image storage unit 310, and a block image storage unit 311. In the present embodiment, among these means, the components other than the video camera 309 can be provided on the indoor management system side.

  The device power switch operation invalidating means 302 is not normally used in applications such as plant monitoring. However, when the device movement detection system 3 is mounted on a portable device such as a node book type computer, there is a deposit or the like. Works effectively in cases.

The monitoring mode on means 301 is used for shifting from the normal mode to the monitoring mode. Specifically, the monitoring mode on means 301 may function by operating a switch dedicated to the system main body 31. When the device movement detection system 3 includes a keyboard, special keys such as function keys may be used. And a general-purpose key such as an alphabet key. When the device movement detection system 3 includes a display, the device movement detection system 3 may function by a soft key or a soft button. Activated).
In the monitoring mode, as described above, the device power switch operation invalidating unit 302 can invalidate the operation of the device power switch. That is, in the monitoring mode, the device power cannot be turned off. Specifically, the device power switch operation invalidating means 302 can be configured to function by a bypass switch 322 provided in parallel with the device power switch 321. When the monitoring mode on means 301 is activated, the bypass switch 322 is turned on, and even if the device power switch 321 is turned off, the device power supply itself is not turned off.

  In the monitoring mode, the video camera driving unit 303 drives the video camera 309, and the video camera 309 captures the scenery at regular intervals and stores it in the captured image storage unit 310. The photographed image storage unit 310 stores at least two photographed images Gk and Gk + 1 that are temporally changed and processed by the photographing region dividing unit 304 and the landscape change detecting unit 305. In the present embodiment, the photographed image storage unit 310 stores two photographed images Gk and Gk + 1.

The shooting area dividing unit 304 includes a plurality of blocks as shown in FIG. 3 each of shot images of the video camera 309 (specifically, at least two shot images Gk and Gk + 1 stored in the shot image storage unit 310). Bij (i = 1, 2,..., M, j = 1, 2,..., N) is divided into block images GBk and GBk + 1 and stored in the block image storage means 311.
It is explanatory drawing of the block image GBk which divided | segmented the picked-up image Gk into the block Bij. When the block image GBk is stored in the block image storage means 311, the captured image Gk can be deleted.

  The image change detection unit 305 detects an image change for all of a plurality of blocks constituting the block image block image GBk. Specifically, the luminance value of the pixel Pqr (q, r = 1, 2,..., Z) of the block Bij (z × z rectangular blocks) constituting the block image GBk, and the block image GBk + 1. The difference from the luminance value of the pixel Pqr of the block Bij that constitutes. In the case of simplification (speeding up) of processing, when the luminance value is 8 bits, for example, the difference can be obtained by reducing the resolution to low bits (1 bit or 2 bits). When the resolution of the luminance value is high (for example, in the case of 8 bits), if the luminance value difference is “0” or a value close to “0”, the luminance value does not change and the luminance value difference is large. In some cases (for example, two or three digits), it can be determined that there is a luminance value. When the resolution of the luminance value is low (for example, in the case of 1 bit), the luminance value does not change when the luminance value difference is “0”, and the luminance value difference is “1”. It can be determined that there is a change in luminance value. When the number of pixels in which the brightness value of the block Bij has changed exceeds a predetermined number, it can be determined that there has been an image change in the block Bij.

The image change block number counting means 306 counts the number of blocks Bij in which an image change is detected.
When the count value (number of blocks Bij in which an image change has been detected) exceeds a threshold (for example, (M × N) ÷ 10) determined based on the total number of blocks (M × N), the TV camera 309 It can be judged that it is moving. Therefore, in this case, the warning generation means 307 can issue a warning. The warning can be made by voice and / or image. The warning generation means 307 may be configured integrally with the device 31 or may be installed at a remote location.

  For example, when a user authentication operation is performed by an administrator or the like, the monitoring mode off unit 308 can turn off the monitoring mode when the operation is appropriate. The user authentication operation may be a password input when the system main body 31 includes a keyboard (including a soft keyboard). Further, the user authentication operation may be a voice input when the system main body 31 includes a voice recognition device, or when the system main body 31 includes a biometric authentication device (fingerprint or iris recognition device). It may be a fingerprint or iris capture. Further, the user authentication operation may be insertion of an ID card into a card slot when the system main body 31 has a recognition function using an ID card.

  The video camera removal detection means 309 can detect when the video camera VC is disconnected from the external device connection interface I / F when the monitoring mode is on. In this case, the warning generation unit 307 can issue a warning.

It is explanatory drawing which shows 1st Embodiment of the flame and gas smoke detection system of this invention. The figure which shows the example which created the 1st, 2nd binarized difference image after creating the 1st, 2nd difference image by the 1st, 2nd, 3rd photographed image, and also created the logical product synthetic image It is. It is a figure which shows the example which divided | segmented the image from which the high frequency component was removed by the filter means into several area | regions. It is a flowchart which shows operation | movement of the flame and gas smoke detection system shown in FIG. (A) is a diagram illustrating three captured images obtained by capturing a scene in which smoke is generated at regular time intervals, and (B) is a diagram illustrating a binarized image corresponding to the first difference image and the second difference image. It is. (A) is an example in which an object change appears in the same block, (B) is an example in which an object change appears in a different block, and (C) is an example in which no object change appears. It is a figure which shows the example which rotates one imaging | photography means and performs said image processing for every fixed time. It is explanatory drawing which shows 2nd Embodiment of the flame and gas smoke detection system of this invention. It is a flowchart which shows operation | movement of the flame and gas smoke detection system shown in FIG. An image showing the scenery where smoke and water vapor are generated simultaneously on the screen. It is a figure which shows the result of having produced | generated the logical sum synthetic | combination image with respect to the scene where smoke is generate | occur | producing, and calculated | required the area of smoke. It is a figure which shows the result of having produced | generated the OR synthetic | combination image with respect to the scenery in which the water vapor | steam has generate | occur | produced, and calculated | required the area of water vapor | steam. (A) is a graph which shows the time change of the area of smoke, (B) is a graph which shows the time change of the area of water vapor | steam. (A) is a graph which shows the calculation result of smoke dispersion | distribution, (B) is a graph which shows the calculation result of dispersion | distribution of water vapor | steam. It is a functional block diagram which shows the apparatus movement detection system 3 with respect to forced movement, destruction, etc.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1, 2 Flame / gas smoke detection system 11, 21 Photographing means 12 Reference photographed image analysis means 13 Reference photographed image analysis result storage means 14, 22 Landscape change detection means 15 Landscape change block analysis means 16, 23 Flame / gas smoke Determination means 17 Flame / gas smoke determination means 141, 221 Difference image generation means 142, 222 Image binarization means 143, 223 Composite image generation means 144, 224 Filter means 145, 225 Landscape change block identification means g0 Captured image G1k kth Photographed image (k = 1, 2, 3)
GD1k kth difference image (k = 1, 2, 3)
GB1k k-th binarized difference image (k = 1, 2, 3)
Gor logical sum composite image Gand logical product composite image FGor filter processing logical sum composite image FGand filter processing logical product composite image B11, B12,..., Bmn block S, S1, S2, S3 Same view

Claims (12)

A flame / gas smoke detection system for detecting flame or gas smoke using an imaging means,
With respect to the same scenery, the photographing means for obtaining a reference photographed image and sequentially obtaining a plurality of detection photographed images at predetermined time intervals;
The reference captured image analysis means for dividing the reference captured image into a plurality of blocks according to a predetermined block division rule and performing image analysis for each block;
Reference photographic image analysis result storage means for storing an image analysis result for the reference photographic image;
Scene change detection means for dividing the detection photographed image into a plurality of blocks according to the block division rule, and detecting a scene change of the detection photographed image based on an image change in each block;
When there is an image change for any of the blocks, a scene change block analysis unit that performs image analysis on the block that has the image change;
The image analysis result for the block related to the scene change by the scene change block analysis unit is compared with the image analysis result for the corresponding block stored in the reference photographed image analysis result storage unit. Flame / gas smoke determination means for specifying whether the change is a change due to a flame or a change due to gas smoke;
A flame / gas smoke detection system characterized by comprising:   The scene change detection means, when the luminance in each block changes, the image change occurs when the number of pixels whose luminance has changed exceeds a predetermined ratio with respect to the total number of pixels constituting each block. The flame / gas smoke detection system according to claim 1, wherein the flame / gas smoke detection system is determined. 2. The flame or gas smoke according to claim 1, wherein flame or gas smoke is detected using three consecutive images of the first captured image, the second captured image, and the third captured image among the captured captured images acquired by the capturing unit. A gas smoke detection system,
The scenery change detecting means is
A first difference image that is a difference between luminance values of the first and second captured images, and a second difference image that is a difference of luminance values between the second and third captured images are generated. Difference image generation means;
Image binarization means for binarizing the first difference image and the second difference image, respectively, and generating a first binarized difference image and a second binarized difference image;
Generating a logical sum composite image that is a logical sum of the first binarized difference image and the second binarized difference image, and / or the first binarized difference image and the second binarized difference A composite image generating means for generating a logical product composite image that is a logical product with the image;
Filter means for removing a high frequency component in the logical sum composite image to generate a filtered logical sum composite image and / or generating a filtered logical product composite image by removing a high frequency component in the logical product composite image;
In each block of the filtered logical sum composite image and / or the filtered logical product composite image, the number of pixels constituting the object appearing in each block is predetermined with respect to the total number of pixels constituting each block. A landscape change block identifying means for judging that there has been an image change when the ratio is exceeded,
A flame / gas smoke detection system characterized by comprising:   The flame / gas smoke detection system according to claim 1, wherein the image analysis includes a pattern recognition process. A flame / gas smoke detection system for detecting flame or gas smoke using an imaging means,
With respect to the same scenery, the imaging means for acquiring captured images and sequentially acquiring a plurality of captured images at predetermined time intervals;
Scene change detection means for dividing the photographed image into a plurality of blocks according to the block division rule, and detecting a scene change of the photographed image based on an image change in each block;
Flame / gas smoke determination means for specifying whether the scene change is a change due to a flame or a change due to a gas smoke based on a spatial and / or temporal change of the group of blocks in which an image change occurs;
A flame / gas smoke detection system characterized by comprising: The flame / gas smoke according to claim 5, wherein flame or gas smoke is detected using three consecutive images of the first captured image, the second captured image, and the third captured image among the captured images acquired by the imaging unit. A detection system,
The scenery change detecting means is
A first difference image that is a difference between luminance values of the first and second captured images, and a second difference image that is a difference of luminance values between the second and third captured images are generated. Difference image generation means;
Image binarization means for binarizing the first difference image and the second difference image, respectively, and generating a first binarized difference image and a second binarized difference image;
Generating a logical sum composite image that is a logical sum of the first binarized difference image and the second binarized difference image, and / or the first binarized difference image and the second binarized difference A composite image generating means for generating a logical product composite image that is a logical product with the image;
Filter means for removing a high frequency component in the logical sum composite image to generate a filtered logical sum composite image and / or generating a filtered logical product composite image by removing a high frequency component in the logical product composite image;
In each block of the filtered logical sum composite image and / or the filtered logical product composite image, the number of pixels constituting the object appearing in each block is predetermined with respect to the total number of pixels constituting each block. A landscape change block identifying means for judging that there has been an image change when the ratio is exceeded,
A flame / gas smoke detection system characterized by comprising: A flame / gas smoke detection method for detecting flame or gas smoke using an imaging means,
A shooting step of acquiring a reference shot image and sequentially acquiring a plurality of detection shot images at predetermined time intervals for the same scene,
The reference captured image analysis step for dividing the reference captured image into a plurality of blocks according to a predetermined block division rule and performing image analysis for each block;
A reference captured image analysis result storage step for storing an image analysis result for the reference captured image;
A scene change detection step of dividing the detection photographed image into a plurality of blocks according to the block division rule, and detecting a scene change of the detection photographed image based on an image change in each block;
When there is an image change for any of the blocks, a scene change block analysis step for performing image analysis on the block having the image change;
The image analysis result for the block related to the scene change in the scene change block analysis step is compared with the image analysis result for the corresponding block stored in the reference captured image analysis result storage step, and the scenery A flame / gas smoke determination step for identifying whether the change is a change due to a flame or a change due to gas smoke;
A flame / gas smoke detection method characterized by comprising:   In the scene change detection step, when the luminance in each block changes, there is an image change when the number of pixels with the luminance change exceeds a predetermined ratio with respect to the total number of pixels constituting each block. The flame / gas smoke detection method according to claim 7, wherein the flame / gas smoke detection method is determined. The flame or gas smoke according to claim 7, wherein flame or gas smoke is detected using three images of the first captured image, the second captured image, and the third captured image among the detected captured images acquired in the capturing step. A gas smoke detection method comprising:
The scenery change detection step includes
A first difference image that is a difference between luminance values of the first and second captured images, and a second difference image that is a difference of luminance values between the second and third captured images are generated. A difference image generation step;
An image binarization step for binarizing the first difference image and the second difference image, respectively, and generating a first binarized difference image and a second binarized difference image;
Generating a logical sum composite image that is a logical sum of the first binarized difference image and the second binarized difference image, and / or the first binarized difference image and the second binarized difference A composite image generation step for generating a logical product composite image that is a logical product with the image;
A filter step of removing a high-frequency component in the logical sum composite image to generate a filtered logical sum composite image and / or generating a filtered logical product composite image by removing a high-frequency component in the logical product composite image;
In each block of the filtered logical sum composite image and / or the filtered logical product composite image, the number of pixels constituting the object appearing in each block is predetermined with respect to the total number of pixels constituting each block. When the ratio exceeds the ratio, it is determined that there has been an image change, and the block is a landscape change block specifying step,
A flame / gas smoke detection method characterized by comprising:   The flame / gas smoke detection method according to claim 7, wherein the image analysis includes a pattern recognition process. A flame / gas smoke detection method for detecting flame or gas smoke using an imaging means,
A shooting step for sequentially acquiring a plurality of captured images at predetermined time intervals for the same scenery,
A scene change detection step of dividing the photographed image into a plurality of blocks according to a predetermined block division rule, and detecting a scene change of the photographed image based on an image change in each block;
A flame / gas smoke determination step for determining whether the scene change is a change due to a flame or a change due to a gas smoke based on a spatial and / or temporal change of the group of blocks in which an image change occurs;
A flame / gas smoke detection method comprising: The flame / gas smoke according to claim 11, wherein flame or gas smoke is detected using three consecutive images of the first captured image, the second captured image, and the third captured image among the captured images acquired in the capturing step. A detection method,
The landscape change detection step includes:
A first difference image that is a difference between luminance values of the first and second captured images, and a second difference image that is a difference of luminance values between the second and third captured images are generated. A difference image generation step;
An image binarization step for binarizing the first difference image and the second difference image, respectively, and generating a first binarized difference image and a second binarized difference image;
Generating a logical sum composite image that is a logical sum of the first binarized difference image and the second binarized difference image, and / or the first binarized difference image and the second binarized difference A composite image generation step for generating a logical product composite image that is a logical product with the image;
A filter step of removing a high-frequency component in the logical sum composite image to generate a filtered logical sum composite image and / or generating a filtered logical product composite image by removing a high-frequency component in the logical product composite image;
In each block of the filtered logical sum composite image and / or the filtered logical product composite image, the number of pixels constituting the object appearing in each block is predetermined with respect to the total number of pixels constituting each block. A scene change block identifying step for determining that there has been an image change when the ratio is exceeded,
A flame / gas smoke detection method comprising: