JP4926603B2 - Smoke detector - Google Patents

Smoke detector Download PDF

Info

Publication number
JP4926603B2
JP4926603B2 JP2006222560A JP2006222560A JP4926603B2 JP 4926603 B2 JP4926603 B2 JP 4926603B2 JP 2006222560 A JP2006222560 A JP 2006222560A JP 2006222560 A JP2006222560 A JP 2006222560A JP 4926603 B2 JP4926603 B2 JP 4926603B2
Authority
JP
Japan
Prior art keywords
smoke
detection
image
luminance
area
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
JP2006222560A
Other languages
Japanese (ja)
Other versions
JP2008046917A (en
Inventor
主久 中野
賢治 寺田
貴俊 山岸
Original Assignee
国立大学法人徳島大学
能美防災株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 国立大学法人徳島大学, 能美防災株式会社 filed Critical 国立大学法人徳島大学
Priority to JP2006222560A priority Critical patent/JP4926603B2/en
Publication of JP2008046917A publication Critical patent/JP2008046917A/en
Application granted granted Critical
Publication of JP4926603B2 publication Critical patent/JP4926603B2/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02PCLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P20/00Technologies relating to chemical industry
    • Y02P20/50Improvements relating to the production of bulk chemicals
    • Y02P20/52Improvements relating to the production of bulk chemicals using catalysts, e.g. selective catalysts

Abstract

<P>PROBLEM TO BE SOLVED: To provide a smoke detection device for increasing the precision of smoke detection by providing two or more detection means (arithmetic means) in consideration of the characteristics of smoke. <P>SOLUTION: This smoke detection device for detecting the occurrence of smoke within a photographed range by carrying out image processing of an image photographed by a camera is provided with: a detection region setting means for setting a prescribed detection region within a photographed image; a moving detection image preparation means for detecting moving pixels within the detection region by carrying out frame difference; a first arithmetic means for calculating the first type of smoke detection determination elements relating to the detection region; and a second arithmetic means for calculating the second type of smoke detection determination elements relating to the detection region when prescribed conditions are satisfied by an arithmetic value in a certain detection region by the first arithmetic means. <P>COPYRIGHT: (C)2008,JPO&amp;INPIT

Description

  The present invention relates to a smoke detection apparatus that detects the generation of smoke by performing image processing on an image captured by a surveillance camera.

  Conventionally, there is a fixed low-brightness smoke source, there is a change in the area of the low-brightness above the smoke-source and the center of gravity is fluctuating, and the low-brightness area above the smoke-source is growing over time. As a condition, there is a device that detects smoke generation when there is a change in shape and area in a difference image of a visible image from a visible camera and there is an identity at the location of occurrence (for example, see Patent Document 1).

JP-A-7-245757

  A lot of smoke generated from a fire source as a fire type is generated in a very short time and moves rapidly (hereinafter, such smoke is called fluid smoke). On the other hand, smoke may gradually spread and spread at a place away from the fire source. In this case, the smoke (hereinafter referred to as slow smoke) moves very slowly and has little change in time units.

The smoke generated by the fire is black smoke in the case of an oil fire, whereas white smoke is white in the case of a normal fire in which wood or paper is the fire type. As described above, even if smoke is used, the color and movement differ depending on the generation factor and location, and there is a problem that it is difficult to detect smoke efficiently from one detection program.
An object of the present invention is to provide a smoke detection device that aims to improve the accuracy of smoke detection by providing two or more detection methods (calculation means) in consideration of smoke characteristics.

A smoke detection device according to the present invention is a smoke detection device that detects the generation of smoke within a range shot by performing image processing on an image shot by a camera. A predetermined detection region is included in the shot image. A detection area setting means for setting a frame, a motion detection image creating means for detecting a motion pixel in the detection area by performing a frame difference, and a first type of smoke detection determination element for the first type relating to the detection area And a second calculation means for calculating a second type smoke detection determination element in the detection area when a calculation value in a detection area by the first calculation means satisfies a predetermined condition; when by the first calculating means, calculation value at a certain detection area satisfies a predetermined condition, and a measurement line setting means for setting a measurement line to the detection region, originating smoke of the detection area To detect.

When detecting smoke, the smoke detection device according to the present invention performs a frame difference, extracts a region having a motion pixel in the detection region, and calculates a first type smoke detection determination element in the detection region. In addition, when the calculated value by the first calculating means satisfies a predetermined condition, the second type smoke detection determination element in the detection area is calculated. In this way, since it is detected whether there is smoke in the detection region using the two calculation means, the accuracy of smoke detection can be improved.
In addition, by setting a measurement line in the detection area, it is not necessary to calculate the pixels in the entire detection area, so that the processing time for image processing can be shortened.
Further, the second calculation means obtains a frequency spectrum by performing frequency analysis on the time series data of the average luminance of the pixels on the measurement line in the detection area, and obtains a predetermined low frequency band with respect to an integral value of the intensity of the frequency spectrum. When the ratio of the integral value to the total integral value is equal to or greater than a predetermined ratio, or the integral value of the intensity of the frequency spectrum is larger in the lower frequency band, it is determined that flowing smoke is generated in the detection region. For this reason, it is possible to distinguish an artificial light source whose integrated value of the intensity of the frequency spectrum becomes larger in the high frequency band and the middle frequency band from smoke, and mistakenly replace artificial light sources and moving objects that behave like smoke. Recognizing it as smoke can be prevented.

A: Regarding the Principle of Smoke Detection First, before explaining the smoke detection apparatus of the present embodiment, what principle is used to detect smoke from an image will be described with reference to FIGS. 1A and 1B show original images taken by a surveillance camera, FIG. 1A shows a state without smoke, and FIG. 1B shows smoke in a detection region. It shows the state of entering. 2 (a) and 2 (b) show the luminance distribution in the detection region where the horizontal axis indicates the luminance and the vertical axis indicates the number of pixels. FIG. 1 (a) and FIG. The drawing corresponds to b). Also, in FIG. 3, FIGS. 3A and 3B show the results of the differential processing of the detection region W1, respectively corresponding to FIGS. 1A and 1B. ing.

  Here, the detection region refers to a region W1 in which the periphery of the window glass is surrounded by a rectangle in the room which is the monitoring region shown in FIGS. 1A and 1B, and smoke is generated in the room. This is an area (area) to be monitored. In FIG. 2A, it can be seen that the pixels are distributed over a wide range from a high value to a low value in a state where there is no smoke in the room. When the luminance value of each pixel is added and the average of the luminance values (average luminance value), which is the value divided by the total number of pixels, is obtained, naturally, the variance that is a deviation from the average value shows a large value. Become.

  On the other hand, when smoke enters the detection area of FIG. 1B, the area becomes blurred. When this is seen in the luminance distribution diagram of FIG. 2B, the range of luminance that can be taken is narrower than in the state without smoke. Similarly, when the average value of luminance is obtained and the variance is calculated, it can be seen that the deviation (dispersion) from the average value becomes small. With the inflow of such smoke, the fact that the luminance distribution range narrows can be said in both cases of black smoke and white smoke, but in the case of black smoke, the brightness value shifts in the direction of decreasing, In the case of white smoke, the luminance value shifts in the increasing direction. In addition, if the detection area is completely filled with smoke, the luminance distribution range is further narrowed, and is considered to converge to a specific luminance value.

In addition, the detection area W1 including the window glass has a large luminance difference between the outside bright part and the indoor dark part, and when the differential process (edge process) is performed, it corresponds to the outline of the window glass part. An edge is formed (see FIG. 3A). However, even if the differentiation process is performed in a smoke-containing state, the difference in luminance is not large and the edge is not so much compared to the normal state without smoke (see FIG. 3B). That is, it is considered that the edge amount decreases when smoke is generated.
Further, the smoke fluctuation was small, and it was confirmed that the low frequency band was larger than the high frequency band when frequency analysis was performed.

Based on such a viewpoint, the inventor of the present application, when smoke is generated,
(1) The field of view is blurred and the transmittance or contrast is lowered.
(2) the luminance value converges to a certain value;
(3) The luminance distribution range is narrowed and the luminance dispersion is reduced,
(4) The average value of luminance changes from a normal smoke-free state,
(5) In the detection area, the total amount of edges decreases,
(6) The intensity of the low frequency band is greater than the intensity of the high frequency band,
I started. Judging from these comprehensively, smoke can be detected.

B: Basic configuration of the present invention Embodiment 1
FIG. 4 is a block diagram of the fire detection apparatus according to Embodiment 1 of the present invention. As shown in FIG. 4, the fire detection apparatus according to Embodiment 1 of the present invention captures a fire occurrence monitoring range, for example, at a rate of 30 frames per second and outputs image data for each frame. A smoke detection device 3 is provided for processing the data to detect the generation of smoke and issue an alarm based on the detection. The camera 2 is composed of, for example, a CCD camera or a CMOS camera.

The smoke detection device 3 includes a central processing unit (CPU) 4, a ROM 5, a RAM 6, a built-in timer 7, and an input / output interface (I / O) 8. From the computer in which the frame grabber 9 and the external storage device 10 are built-in. It is configured.
The frame grabber 9 acquires image data from the NTSC video signal output from the camera 2. The image data is composed of, for example, one line of 640 pixels and one frame of 480 lines, and the pixels are represented by luminance of 256 gradations. The ROM 5 stores a procedure for processing operations performed by the CPU 4 as a program. The CPU 4 reads the program and advances the procedure for processing operations based on the program.

FIG. 5 shows various storage units secured in the smoke detection device 3 according to Embodiment 1 of the present invention.
As shown in FIG. 5, the smoke detection device 3 includes a flow smoke image storage unit 11 that stores luminance data related to a flowing smoke image (hereinafter referred to as “flowing smoke image data”), and luminance data related to a slow smoke image (hereinafter referred to as “slow”). Slow smoke image storage unit 12 that stores smoke image data), reference image storage unit 13 that stores luminance data related to the reference image (hereinafter referred to as reference image data), and luminance data related to the difference image (hereinafter referred to as difference image data) A reference image storage unit 14 for storing a reference maximum luminance, a reference minimum luminance, a reference average luminance and a reference variance (hereinafter referred to as reference data) for each detection area of the reference image, A detection area storage unit 16 for storing information about the detection area, a histogram storage unit 17 for storing the average and variance of the appearance frequency of the histogram, and the detection area in time series Average luminance distribution storage unit 18 that stores the mean and variance of the luminance of the detected number storage unit 19 that stores the detection count each time a fire detection procedure is set as a storage area.

  As shown in FIG. 6, the rough configuration of the smoke detection device 3 includes a detection preparation means 21 that updates reference image data and sets a detection area, and image data output from the camera 2 is sluggish with the flowing smoke image storage unit 11. It comprises image storage means 22 for storing in the smoke image storage section 12, smoke detection means 23 for detecting slow smoke or flowing smoke, and fire determination means 24 for determining fire when slow smoke or flowing smoke is detected.

Here, before describing the image processing method of the present embodiment in detail, a pre-stage of performing image processing will be described. The processing performed by the detection preparation means 21 will be described with respect to the processing performed in the previous stage. The detection preparation unit 21 includes a reference image update unit 26 and a detection area setting unit 27.
In the smoke detection device 3, the means for extracting the smoke generation region that becomes the abnormality generation region from the detection region is based on difference processing between images. In particular, in the detection of flowing smoke that occurs near a fire source and moves quickly and has a large area change, difference processing between frames is performed.

For example, as shown in FIG. 1A, the reference image is an image that is usually taken in a state where no abnormality has occurred, and is stored in the reference image storage unit 13. This reference image does not always store the image, but is updated sequentially using means of a known technique to eliminate fine effects such as sunshine.
The reference image update means 26 updates the reference image periodically based on the time of the timer 7, in short description. When the reference image update time comes after a predetermined time has elapsed, the latest image data is read from the flowing smoke image storage unit 11 and the reference image data is updated. At this time, the reference image to be updated with the original reference image is updated. The reference image is updated only when there is almost no change.

  Next, the detection area setting means 27 will be described. The detection area setting means 27 is executed before starting the detection process, and determines which area of the image is set as the detection area in the reference image or the latest photographed image. When trying to monitor the entire image, the amount of calculation for image processing becomes enormous, which is difficult. Therefore, in this embodiment, a plurality of small areas, for example, 64 * 64 horizontal pixel areas are collectively set as a detection area in the image, and this area is arbitrarily set in the image. In this way, by treating a part of the region as a detection region instead of the entire image, the amount of calculation can be reduced and the influence of noise can be reduced. The size of the detection region is not limited, but a size of 2 to the Nth power is convenient when performing frequency analysis or the like. Further, two or more detection areas may be connected and handled as one detection area.

  In FIG. 1A, detection areas are indicated by W1 and W2. Here, W1 is a rectangular region surrounding the window glass, and W2 is a region surrounding the corner on the floor side of the room with a rectangle. Here, the detection area W can be set in any part of the image by changing its size and shape. Preferably, the detection area W is a place where smoke to be detected is generated, and smoke is detected. It should be set in a place that is easy to do.

  It is desirable to set the detection area in consideration of detection sensitivity, and an area having a large change in luminance is set as the detection area. As described with reference to the smoke detection principle of FIGS. 1 and 2, as the smoke flows in, the area has a narrow luminance distribution range. That is, in a region where the luminance distribution is narrow from the normal state where there is nothing, even if smoke enters, the change is scarce and it is difficult to detect smoke. On the other hand, the window glass region W1 contains external light and has a wider luminance distribution than the W2 region at the corner of the room. Therefore, when smoke enters, changes in the luminance distribution are easily detected. Specifically, the detection area setting means 27 performs spatial differential processing on the first reference image data and sets an area having a large spatial differential value as a detection area.

  The detection area setting means 27 excludes the predetermined reference luminance or higher from the initial reference image data in order to prevent the influence of luminance saturation. For example, a spatial differential value is obtained by performing spatial differential processing excluding pixels with luminance of 250 or more, and a region having the largest sum of spatial differential values is set as one detection region. Specifically, the reference image is scanned over the entire area, for example, an area composed of 64 × 64 pixels, for a total of 4096 pixels, to obtain the sum of the spatial differential values of each area. Then, the region having the largest sum of the spatial differential values is set as the first detection region, and then, the region having the largest sum of the spatial differential values among the regions not overlapping the first detection region is set as the second detection region. And set. By repeating this operation, a predetermined number of detection areas are set, and information on the set detection areas is stored in the detection area storage unit 16.

  The detection preparation unit 21 includes a reference data calculation unit 28. The reference data calculation means 28 calculates a reference maximum brightness, a reference minimum brightness, a reference average brightness, and a reference variance for each detection area. The reference maximum brightness, the reference minimum brightness, the reference average brightness, The reference variance is obtained and stored in the reference data storage unit 15.

Next, the configuration of the image storage unit 22 will be described. The image storage unit 22 includes a luminance correction unit 31, a flowing smoke image storage unit 32, and a slow smoke image storage unit 33.
The brightness correction unit corrects the brightness by performing inverse gamma correction on the image data input from the camera 2. The image data input from the camera 2 is subjected to inverse gamma correction because the image data captured and output by the camera 2 is subjected to gamma correction so as to match the human visibility when the image data is displayed on the monitor. Because. When the image data subjected to the gamma correction is subjected to image processing as it is, distortion due to the gamma correction is included in the processing result. Therefore, the inverse gamma correction is performed by the luminance correction unit 31 to restore the actual luminance.

  The fluid smoke image storage means 32 stores the brightness-corrected image data in the fluid smoke image storage unit 11 each time. The fluid smoke image storage means 32 stores, for example, 300 frames of image data whose brightness has been corrected by the brightness correction means 31 in the fluid smoke image storage unit 11, and when deleting a new image, the old image data is deleted first. The smoke image data is updated by the first-in first-out procedure.

The slow smoke image storage means 33 stores, in the slow smoke image storage unit 12, image data whose luminance has been corrected for each predetermined slow smoke capture period. Since the slow smoke hardly moves, the slow smoke image storage means 33 extracts the image data whose brightness has been corrected by the brightness correction means 31 every 3 seconds, for example, stores it for 90 frames, and further stores a new image. Sometimes, the slow smoke image data is updated by the first-in-first-out procedure in which old image data is deleted first.
Slow smoke is smoke that spreads into the room very slowly compared to flowing smoke. For this reason, for example, even if the difference process is performed every 30 msec, there is almost no change area, so that the image is extracted and stored every 3 seconds from the photographed image. On the other hand, since the moving smoke moves very quickly, the image is memorized every time shooting is performed in 30 msec.

Next, the configuration of the smoke detection means 23 will be described. The smoke detection means 23 includes an obstacle detection means 35, a flowing smoke detection means 36, and a slow smoke detection means 37. The obstacle detection means 35 detects the presence / absence of an obstacle in the flowing smoke image for each detection area by performing differential processing with the reference image when the flowing smoke image data is newly stored.
The obstacle detection means 35 has a difference means, reads the reference image data and the latest image data of the flowing smoke image storage unit 11, and for each detection region, the difference in luminance between the two image data or between the image data The cross-correlation coefficient is calculated, and when the difference is equal to or larger than a predetermined threshold value or when the correlation value is equal to or smaller than the threshold value, the pixel is determined as a pixel in which motion is detected. Note that, for the detection region including the detection pixel that has changed, the differential image data from which noise has been removed by performing the degenerate expansion process is created and stored in the differential image storage unit 14.

  For example, when a person passes through the monitoring area, naturally, if a difference process with the reference image is performed, the person is detected as an obstacle in the detection area. Simply by performing the difference processing, it is not known whether the area (pixel) whose luminance value has changed is a person or smoke. For this reason, various processes to be described later are performed, but it is clear that if there is any change area at the stage where the captured input image and the reference image are differentially processed, it is at least not slow smoke. This is because the slow smoke moves slowly, and a change area does not appear unless a long time elapses between frames in which the difference is performed. Therefore, when an obstacle is detected by the obstacle detection means 35, smoke detection by the following slow smoke detection means 37 is not performed.

  The flowing smoke detection means 36 detects the presence or absence of generation of flowing smoke by image processing. The slow smoke detecting means 37 detects the presence or absence of slow smoke by image processing when no obstacle is detected by the obstacle detecting means 35 and new slow smoke image data is stored.

C: Configuration of fluid smoke detection means (primary judgment)
As shown in FIG. 7, the flowing smoke detection means 36 is composed of 10 means. First, each of the five units serving as the primary determination unit, the motion detection image creation unit 42, the average luminance / dispersion calculation unit 43, the area occupation ratio calculation unit 44, the differential value sum ratio calculation unit 45, and the primary determination unit 46 will be described. To do. The primary determination unit 46 compares the values calculated by the calculation units 43 to 45 with predetermined values to determine the occurrence of smoke in the detection region.

  The motion detection image creation means 42 performs a difference process between frames in the captured image stored in the fluid smoke image storage unit 11 to extract motion detection pixels in the detection region. Then, motion detection image data relating only to the motion pixel is created and stored in the difference image storage unit 14. In addition, the difference process between frames performs a difference process between several frames on the time axis, for example, about 5 to 8 frames (150 to 240 msec), and creates, for example, 64 motion detection images.

The average luminance / dispersion calculating means 43 calculates the average and variance of the luminance in the time direction of the moving pixels from 64 pieces of motion detection image data, and calculates the moving average of the average luminance and the variance. These moving averages are compared with the average luminance of the reference image and the variance of the reference image, and are determined by the primary determination means 46. The primary determination means 46 determines whether the moving average of the average luminance of the image is within a predetermined range for each detection area. If smoke flows into the detection area, the luminance value fluctuates, so that the average luminance naturally fluctuates. By treating this as a moving average, the variation in average luminance can be found in this way, and it can be used as a standard for determining whether the object that has entered the detection area is smoke.
In addition, the primary determination unit 46 determines, for each detection region, whether or not the moving average of the image luminance variance is smaller than the variance of the reference image. Similarly to the average luminance, if the smoke flows into the detection area, the value fluctuates (decreases). By treating this as a moving average, the variation in dispersion can be understood and used as a guide for determining whether the object entering the detection area is smoke.

  Moreover, the area occupation ratio calculation means 44 calculates | requires the occupation ratio which a motion detection pixel occupies with respect to the pixel of the whole detection area. That is, in the rectangular pixel area of 64 * 64 (vertical) that constitutes the detection area, how many pixels moved by the difference processing, that is, pixels whose luminance has changed, occupy the rectangular area in terms of number. To calculate the occupation ratio. The occupation ratio is determined by the primary determination means 46. The primary determination unit 46 determines, for each detection area, whether the occupation ratio occupied by the motion detection pixels in the detection area is equal to or greater than a predetermined occupation ratio, for example, 10%. Here, if it is determined that the occupation ratio is at least 10% or more, it can be determined that some object has occurred in the detection region.

  The differential value summation ratio calculation means 45 obtains the sum of the spatial differential values of the region in the reference image, that is, the region where motion is detected, and the spatial differential value of the region of the recent image, and calculates the sum of the region of the reference image. A ratio between the sum of the spatial differential values and the sum of the spatial differential values of the area of the recent image is calculated. This ratio is determined by the primary determination means 46. The primary determination means 46 determines whether or not the ratio of the sum of the spatial differential values varies by a predetermined ratio or more, for example, 5% or more for each detection region. If smoke flows into the detection area, the amount of the spatial differential value (edge) will decrease from the reference time, so by looking at the ratio of the spatial differential value, whether the object that entered the detection area is smoke It becomes a standard of judgment.

  In the primary determination means 46, the occupation ratio occupied by the motion detection pixels is greater than or equal to a predetermined occupation ratio, the ratio of the sum of the spatial differential values is greater than or equal to the predetermined ratio, and the moving average of the average luminance varies more than the predetermined ratio, When the moving average of the variance is reduced by a predetermined ratio or more, it is determined that there is a possibility that flowing smoke is generated. Since the flowing smoke has a feature that the flow can be seen, it is possible to detect moving objects by determining the occupation ratio of the motion detection pixels in the detection region. However, the detected objects include all moving objects.

The ratio between the sum of the spatial differential values of the reference image and the sum of the spatial differential values of the image containing the flowing smoke is reduced when the flowing smoke enters the detection area, the spatial differential value decreases, and the sum also decreases. Flowing smoke can be detected by detecting the decrease in the amount of smoke.
Whether or not there is smoke in the detection area may be determined by the determination of only the primary determination unit 46, but the primary determination unit 46 basically determines only by information based on the luminance change. There is a possibility that a luminance change that turns on and off the illumination in the monitoring area is erroneously detected as smoke. Therefore, by performing more advanced image processing such as frequency analysis, the accuracy of smoke detection is improved as a secondary determination.

D: Configuration of fluid smoke detection means (secondary determination)
The flowing smoke detection means 36 includes a measurement line setting means 51, a frequency analysis means 52, and a frequency determination means 53 as secondary determination means. The measurement line setting unit 51 is a unit that performs a difference process between frames and sets a measurement line for a detection region where a moving pixel exists. Here, an example of how to set the measurement line will be described with reference to FIG. FIG. 8A shows original image data when motion pixels are detected in a plurality of detection regions arranged in a matrix. FIG. 8B shows difference image data between the original image and the reference image shown in FIG. FIG. 8C illustrates a measurement line set from the difference image data in FIG.
In the center of the original image, two rectangular areas constituting the detection area are provided in the vertical direction and three in the horizontal direction. Although the detection areas are arranged in contact with each other in FIG. 8A, the detection areas may be arranged with a gap in each rectangular area. In FIG. 8A, it can be observed that smoke is generated from the fire type below the lower left detection region and flows in the upper right direction. Here, in FIG. 8B, the area indicated by black smoke is a pixel with motion extracted by the inter-frame difference processing. The simplest measurement line is to draw a horizontal line or a vertical line passing through the center of the rectangle in the detection area including these moving pixels, and use that line as the measurement line. However, this method does not necessarily include many moving pixels on the measurement line. The measurement line is preferably set to draw a line so that a number of motion pixels are included on the line as follows, and this measurement line is a line indicating the movement of the smoke flow to be detected. It becomes.

  A simple method of drawing a measurement line will be described with reference to this drawing. Now, with respect to a detection area including moving pixels, the detection area is divided into four rectangles, two vertically and horizontally. At this time, in four rectangles, the number of moving pixels is counted, and two rectangles containing more moving pixels are extracted. Then, a straight line is drawn so as to connect the centers of two rectangles (small regions) with many moving pixels to obtain a measurement line. For example, in the detection area C of FIG. 8B, there are four rectangles with four moving pixels, that is, the upper right corner and the lower left corner. The measurement line will be drawn. As a result, a measurement line is finally set in the detection area as shown in FIG. The meaning of drawing this measurement line is that if frequency analysis is performed on all moving pixels, the amount of calculation increases.Therefore, by selecting only effective pixels on the measurement line in the detection area, the number of pixels is reduced. The purpose is to reduce the amount of calculation.

  There are many other ways to draw measurement lines, and there are the following methods. The measurement line setting means 51 reads the latest image data and the image data of several frames before stored in the flow smoke image storage procedure several times before from the fluid smoke image storage unit 11 for each detection region including the motion detection pixels, Two-dimensional FFT processing is performed on each of them, a complex complex number obtained by two-dimensional FFT processing of image data several frames before is obtained, and the complex number obtained by two-dimensional FFT processing of the latest image data and several frames before A product of conjugate complex numbers related to image data is obtained. Further, the measurement line setting means 51 obtains an image showing the cross-correlation coefficient distribution by performing inverse FFT processing on this product, and obtains a line connecting the pixel located at the center in the image to the pixel having the highest luminance. And stored as a movement vector. Furthermore, the measurement line setting means 51 reads the movement vector stored at the time of the most recent measurement line setting for the detection area from the movement vector, and sets the average vector of these movement vectors as the measurement line. It may be. At this time, the measurement line may be a normal line of the movement vector.

Next, the frequency analysis means 52 will be described. The frequency analysis means 52 extracts luminance time-series data using the latest image data and a plurality of image data retroactive to the pixels on the measurement line for each detection region, performs frequency analysis, and performs pixel analysis. A frequency spectrum is calculated, and an average frequency spectrum thereof is calculated. The frequency determination means 53 determines whether the intensity of the frequency spectrum is equal to or lower than a predetermined intensity or whether the integrated value of the predetermined low frequency band intensity is greater than the integrated value of the high frequency band intensity.
That is, frequency analysis is performed for each pixel of the luminance of the pixel on the measurement line to obtain a frequency spectrum, and further, an average spectrum thereof is obtained, and the peak of the average luminance when flowing smoke enters, for example, as shown in FIG. And appearing in a frequency band exceeding 0 Hz and 2 Hz or less. By performing such frequency analysis as a secondary determination, periodic fluctuations in the amount of light of the artificial light source can be excluded from detection targets.

The flowing smoke detection means 36 further includes a histogram creation means 54, an average / dispersion calculation means 55, and a secondary determination means 56 as secondary determination means. The histogram creation means 54 creates, as a time-series histogram, the number of pixels in which motion has been detected by analysis between the latest image data and a plurality of image data retroactively with respect to the pixels on the measurement line for each detection region. The average / dispersion calculating means 55 obtains the average and variance of the number of motion detection pixels from the time series histogram.
Then, the secondary determination unit 56 determines whether or not the variance (variance / average value) with respect to the average value of the number of pixels is equal to or less than a predetermined threshold value. The secondary determination means 56 determines that flowing smoke has been generated when the intensity of the frequency spectrum is greater than a predetermined ratio and the average luminance and dispersion of the pixels are within a predetermined range. To do.
A time-series variation of the appearance frequency of the motion detection pixel on the measurement line when the flowing smoke enters the detection region will be described with reference to FIG. For example, in the detection area F of FIG. 8C, the size of the detection area itself is 64 * 64 in the vertical direction, and here, the length of the measurement line is 64, which is the maximum included on the measurement line. The number of pixels is 64. Here, when the corresponding detection area F of the difference image in FIG. 8B is viewed, it can be seen that not all the moving pixels are present on the measurement line.
Histogram creating means 54 counts the number of motion pixels in this measurement line, for example, as shown in FIG. 10 (a), at the time t 1, which plots the number in the vertical axis direction. In this way, the numbers at other times t 2 , t 3 ... In the horizontal axis direction are also plotted to complete the histogram. When such a histogram is completed, when the detection in the detection region is smoke, the time-series variation of the appearance frequency of the motion pixel does not take a large value. On the other hand, when the behavior with a large time-series fluctuation of the appearance frequency as shown in FIG. 10B shows a detected behavior, it is considered that the detected light is a rapidly changing rotating lamp. The periodic light quantity fluctuation of the artificial light source can be separated from the detection target.

E: Configuration of Slow Smoke Detection Unit The slow smoke detection unit 37 includes a luminance average variance calculation unit 61 and a determination unit 62 as shown in FIG. The characteristic of slow smoke is that the smoke concentration is low and it is difficult to capture the movement of the smoke.
The luminance average variance calculating means 61 calculates and calculates the average luminance and variance of the entire detection region for each detection region at every cycle (3 seconds) at which a slow smoke image is captured. The calculated average luminance and variance are determined by the determining means 62. That is, the determination unit 62 determines whether the ratio between the average luminance and the reference average luminance is within a predetermined range for each detection region. Further, for each detection region, it is determined whether or not the variance is within a predetermined range with respect to the variance of the reference image.

The slow smoke detecting means 37 includes an average luminance frequency analyzing means 64. The average luminance frequency analysis means 64 stores the average luminance of the latest image data in the detection area and the predetermined plural number in advance in the flowing smoke image storage unit 11 for each detection area for each period of capturing the slow smoke image. For example, the average luminance of 63 pieces of image data is calculated, and the average luminance frequency spectrum is generated by frequency analysis of the 64 average luminances. And the determination means 62 determines whether the intensity | strength of an average luminance frequency spectrum is below predetermined intensity | strength. The determination unit 62 divides the average luminance frequency spectrum into a plurality of frequency bands, and determines whether or not the integrated value of the intensity in the low frequency band is larger.
As shown in FIG. 12 (a), the frequency spectrum related to the average time series data of the luminance of the detection region has an intensity of a predetermined threshold value or less as shown in FIG. 12 (a), or two frequencies (2 Hz, 8Hz), the integrated value of the intensity becomes larger as the frequency is lower, as shown in FIG. 12B. On the other hand, the artificial light source has a large middle or high frequency band, and periodic fluctuations of the artificial light source and slow smoke can be distinguished by detecting the characteristics of this frequency.

The slow smoke detecting means 37 includes a transmittance calculating means 67 and a convergent luminance value calculating means 68. The transmittance calculating means 67 calculates the transmittance according to the equation (1) for each detection region for each period of capturing the slow smoke image. In addition, the convergence luminance value calculation unit 68 calculates a convergence luminance value according to the equation (2) for each detection region for each period in which the slow smoke image is captured. Then, the determination unit 62 determines whether or not the calculated transmittance is equal to or less than a predetermined value. The determination unit 62 determines whether or not the calculated convergence luminance value is equal to or greater than a predetermined threshold value.
Transmittance = (maximum luminance value of latest image−minimum luminance value of latest image) / (maximum luminance value of reference image−minimum luminance value of reference image) (1)
Convergence luminance value = {maximum luminance value of latest image− (maximum luminance value of reference image × transmittance) / (maximum luminance value of reference image−minimum luminance value of reference image) (2)

The transmittance expressed by the formula (1) becomes small when the slow smoke enters the detection region. This is because the difference between the maximum brightness and the minimum brightness when slow smoke enters is smaller than the difference between the maximum brightness and the minimum brightness before slow smoke enters. However, the decrease in transmittance may also occur when the illumination is dark.
The convergence brightness value represented by the expression (2) increases or decreases with respect to the convergence brightness before the slow smoke enters when slow smoke other than the black smoke that is extremely black enters the detection region, but decreases below a predetermined threshold. do not do. On the other hand, when the illumination becomes dark, it decreases below a predetermined threshold. By utilizing this characteristic, it is possible to distinguish between the slow smoke near white and the fact that the lighting has become dark.

In the determination means 62 of the slow smoke detection means 37, the ratio of the average luminance to the average luminance of the reference image is within a predetermined range, and the variance is within a predetermined range with respect to the variance of the reference image, and the average When the intensity of the luminance frequency spectrum is less than the predetermined intensity or the integrated value of the intensity of the low frequency band is large, it is determined that the possibility of slow smoke is high.
Or, when the intensity of the average luminance frequency spectrum is less than the predetermined intensity or the integrated value of the intensity of the low frequency band, and the transmittance is less than the predetermined threshold and the convergence luminance value is more than the predetermined threshold, slow smoke is generated. It is determined that there is a high possibility.
Here, the warning determination means 72 is such that the intensity of the average luminance frequency spectrum is less than the predetermined intensity or the integrated value of the intensity of the low frequency band, the transmittance is within the predetermined range, and the convergence luminance value is less than the predetermined threshold value. If a warning is issued. The warning is issued because the convergence luminance value is less than the predetermined value, and therefore there is a possibility of black smoke as smoke.

As described above, in the present embodiment, the flowing smoke with a fast smoke flow and the slow smoke with a slow smoke flow are discriminated separately, and each has a plurality of calculation means for detecting smoke. In particular, the smoke detection element for slow smoke consists of transmittance, convergence luminance value, frequency spectrum, average value of average brightness, and dispersion of average brightness, and the smoke detection detection element for flowing smoke is a shift of frequency spectrum and average brightness. It consists of the average value, the moving variance of the average luminance, the occupation ratio of the motion detection pixels, the ratio of the sum of the spatial differential values, and the variance / average value of the histogram.
It is possible to detect smoke most accurately by using all of these calculation means and determining that smoke has been generated in the detection area when all the calculation means satisfy the conditions of the determination means. It becomes. However, these computing means can be used in combination of two or more as appropriate, and can sufficiently improve the accuracy of smoke detection as compared with a single computing process. Can be prevented from being mistakenly recognized as smoke.

  Referring back to FIG. 6 again, the fire determination means 24 determines the number of detection areas determined to be flowing smoke by the flowing smoke detection means 36 and the detection areas determined to be slow smoke generation by the slow smoke detection means 37. The number is stored in the detected number storage unit 19. The fire discriminating means 24 reads the number of past detection areas from the detection number storage unit 19 as time-series data, and the number of detections for a certain period continues for a predetermined number of times exceeding a predetermined upper limit value. When the density of the detection area determined to be present exceeds a predetermined value, a fire is determined.

Next, a fire detection procedure performed by the fire detection device will be described with reference to FIGS. When the fire detection procedure starts, as shown in FIG. 13, the reference image update procedure (S101) is executed, and then the image storage procedure (S102), the smoke detection procedure (S103), and the fire discrimination procedure (S104) are executed in sequence. Thus, one fire detection procedure is completed, and the fire detection procedure is repeatedly executed.
First, the reference image update procedure will be described with reference to the flowchart shown in FIG.
When the reference image update procedure is started, it is determined in step S201 whether or not the reference image update time has arrived. When the reference image update procedure has not arrived, the reference image update procedure is terminated, and when it has arrived, the process proceeds to step S202.
In step S202, the latest image data is read from the fluid smoke image storage unit 11, and after calculating the image correlation and confirming that there is no intruder, the latest image data is stored in the reference image storage unit 13 as reference image data. Rewrite the reference image data.
First, only when setting the detection region, after this step S202, the reference image data is read from the reference image storage unit 12, and the reference image data is subjected to spatial differentiation processing to obtain a spatial differential value for each pixel. Ask for.
In step S203, the reference maximum brightness, the reference minimum brightness, the reference average brightness, and the reference variance of the set detection area are obtained, stored in the reference data storage unit 15, and the reference image update procedure is terminated.

Next, the image storage procedure will be described with reference to the flowchart of FIG.
When the image storage procedure is started, it is determined in step S301 whether or not image data has been input from the camera 2. If not, step S301 is repeated, and if it has been input, the process proceeds to step S302.
In step S302, the input image data is set as the latest image data, and the latest image data is subjected to luminance correction by inverse gamma correction.
In step S303, it is determined whether or not the slow smoke image capture timing has arrived. If it has arrived, the process proceeds to step S304, and if not, the process proceeds to step S306.
In step S304, the latest brightness corrected image data is stored in the slow smoke image storage unit 12 by FIFO (this procedure is referred to as a slow smoke image storage procedure).
In step S305, 1 is set to the slow smoke flag.
In step S306, the latest image data whose luminance has been corrected is stored in the flowing smoke image storage unit 11 by FIFO, and the image storing procedure is terminated (this procedure is referred to as a flowing smoke image storing procedure).

Next, the smoke detection procedure will be described with reference to the flowchart of FIG.
When the smoke detection procedure is started, the setting of various flags that leave the determination result is set to zero.
In step S401, based on the detection area information stored in the detection area storage unit 16, one detection area that has not yet undergone smoke detection processing is designated.
In step S402, the reference image data is read from the reference image storage unit 13, the latest image data from the flowing smoke image storage unit 11, the reference image data for each pixel, and the latest image data are subjected to a difference process. Find pixels that are equal to or greater than the threshold value.
In step S403, a degenerate / expanding process is performed on pixels having a difference equal to or greater than a predetermined threshold value, noise is removed to determine a motion detection pixel, and the difference image data is stored in the difference image storage unit 14 (this procedure is referred to as a difference). This is referred to as an image storage procedure). The noise processing performed here excludes a small pixel block generated by the difference processing, and leaves only a region where the pixels are connected to a difference image to some extent.

In step S404, the flow smoke detection procedure is executed, and the process proceeds to step S405.
In step S405, it is determined whether or not 1 is set in the slow smoke flag. When the slow smoke flag is set to 1, the process proceeds to step S406, and when the slow smoke flag is set to 0, the process proceeds to step S408.
In step S406, obstacle detection is performed by the obstacle detection means, and the process proceeds to step S407 only when no obstacle is detected in the image. Otherwise, the process proceeds to step S408.
In step S407, the slow smoke detection procedure is executed, and the process proceeds to step S408.
In step S408, it is determined whether or not all detection areas have been specified. If not, the process returns to step S401, and if specified, the process proceeds to step S409. In other words, since a plurality of detection areas where smoke should be detected is set in the image, smoke detection is sequentially performed on these detection areas one by one. When the smoke detection procedure has been performed for all of the plurality of detection areas, in step S409, the slow smoke flag is set to 0 and the smoke detection procedure is terminated.

Next, the flow smoke detection procedure will be described with reference to the flowcharts of FIGS.
When the flow smoke detection procedure is started, in step S501, the latest image data of the detection area and the image data stored for each of the latest multiple flow smoke image storage procedures are read from the flow smoke image storage unit 11, and motion detection pixels are detected. The average and variance of the brightness of the latest image data corresponding to the image data and the image data stored in the most recent plurality of flowing smoke image storage procedures are calculated.
In step S502, a moving average of average luminance and a moving average of variance are calculated from the average and variance of luminance.
In step S503, the difference image data of the detection area is read from the difference image storage unit 14, and the occupation ratio of the motion detection pixels to the pixels in the entire detection area is calculated.
In step S504, the reference image data of the detection area is read from the reference image storage unit 13 and the latest image data is read from the flowing smoke image storage unit 11, and each is subjected to spatial differential processing to obtain a spatial differential value. Are obtained, and the ratio of the sum is calculated.

In step S505, the occupation flag is set to 1 when the occupation ratio occupied by the motion detection pixels is equal to or greater than a predetermined ratio.
In step S506, 1 is set in the differential value flag when the ratio of the sum of the spatial differential values fluctuates beyond a predetermined threshold.
In step S507, the average flag is set to 1 when the moving average of the average brightness fluctuates within a predetermined ratio with respect to the average brightness of the reference image.
In step S508, the variance flag is set to 1 when the moving average of variance decreases within a predetermined ratio with respect to the variance of the reference image.
In step S509, it is determined whether 1 is set for all of the occupation flag, the differential value flag, the average flag, and the variance flag. If all are set to 1, the process proceeds to step S510, and at least one is set to 0. When is set, the smoke detection procedure is terminated.

In step S510, a measurement area measurement line is set.
In step S511, the latest image data of the detection area and the image data stored for each of a plurality of flow smoke image storage procedures closest to the detection area are read from the flow smoke image storage unit 11, and the luminance is determined for each pixel on the measurement line. The frequency spectrum is calculated by frequency analysis, and the average of the frequency spectrum for all pixels is calculated.
In step S512, the frequency spectrum is corrected by excluding unnecessary portions of the spectrum from the frequency spectrum.
In step S513, it is determined whether or not the integrated value of the intensity of the frequency band below the predetermined frequency exceeds the integrated value of the intensity of the frequency band exceeding that frequency, and the integrated value of the intensity of the frequency band below the predetermined frequency is When the integrated value of the intensity of the frequency band exceeding the frequency is exceeded, the process proceeds to step S514, and otherwise, the flow smoke detection procedure is terminated. Here, rather than simply looking at the ratio between the low frequency and the high frequency, the frequency band may be divided into three and the integrated value may be seen in each band.

In step S514, the difference image data of the latest image in the detection area and the difference image data stored for each of a plurality of difference image storage procedures closest to the detection area are read from the difference image storage unit 14, and measurement of each difference image data is performed. A histogram is created with the number of motion detection pixels on the line as the frequency.
In step S515, the histogram frequency average value and frequency variance are calculated, and the ratio (variance / average value) is calculated and stored in the histogram storage unit 17.
In step S516, it is determined whether or not the ratio between the frequency average and the frequency variance is within a predetermined range. When the ratio is within the predetermined range, the process proceeds to step S517. Otherwise, the flow smoke detection procedure is performed. Exit.
In step S517, it is determined that flowing smoke is generated in the detection area, the flowing smoke detection register is incremented, and the flowing smoke detection procedure is terminated.

Next, the slow smoke detection procedure will be described with reference to the flowcharts of FIGS. In step S 601, the latest image data is read from the slow smoke image storage unit 12, the luminance average and variance are calculated, and stored in the luminance average variance storage unit 18.
In step S602, the luminance average and variance are read from the luminance average variance storage unit 18 and include the latest data and stored for each of the latest slow smoke detection procedures, and the average luminance moving average and variance moving average are obtained. calculate.
In step S603, when the moving average of the average luminance varies within a predetermined range with respect to the average luminance of the reference image, 1 is set to the average flag.
In step S604, when the moving average of the variance decreases within a predetermined range with respect to the variance of the reference image, 1 is set in the variance flag.

In step S605, the image data stored in each of a plurality of recent flow smoke storage procedures including the latest image data of the detection region is read from the flow smoke image storage unit 11, the average of each luminance is calculated, and the average luminance A frequency spectrum is generated by performing frequency analysis of the time-series data regarding.
In step S606, the frequency spectrum is corrected by excluding unnecessary portions of the spectrum from the frequency spectrum.
In step S607, the frequency spectrum is divided into three frequency band spectra by two frequencies of 2 Hz and 8 Hz, and when the value obtained by integrating the intensity of each frequency band is larger in the lower frequency band, 1 is set in the frequency flag.

In step S608, the maximum luminance value and the minimum luminance value are obtained from the latest image data in the detection area, the maximum luminance value of the reference image and the minimum luminance value of the reference image are read from the reference data storage unit 15, and the transmittance is expressed by the formula ( Calculate according to 1). Further, the convergence luminance value is calculated according to the equation (2).
In step S609, the transmittance flag is set to 1 when the transmittance is equal to or less than a predetermined value.
In step S610, the convergence luminance flag is set to 1 when the convergence luminance value is greater than or equal to a predetermined threshold value.
In step S611, it is determined whether 1 is set in the average flag, the dispersion flag, and the frequency flag, or 1 is set in the frequency flag, the transmittance flag, and the convergence luminance flag, and the average flag, the dispersion flag, and the frequency flag are set. If 1 is set, or if 1 is set in the frequency flag, the transmittance flag, and the convergence luminance flag, the process proceeds to step S612. Otherwise, the process proceeds to step S613.
In step S612, the slow smoke detection register is incremented because there is a possibility that slow smoke is generated, and the slow smoke detection procedure is terminated.
In step S613, it is determined whether or not 1 is set in the frequency flag and the transmittance flag. If 1 is set in the frequency flag and the transmittance flag, the process proceeds to step S614. End the detection procedure.
In step S614, a warning signal is issued to warn that black smoke may be generated, and the slow smoke detection procedure is terminated.

Next, the fire determination procedure will be described with reference to the flowchart of FIG.
When the fire discrimination procedure is started, in step S701, the number set in the flowing smoke detection register (the number of detection areas determined to have flowing smoke) and the number set in the slow smoke detection register (with slow smoke) The number of detection areas determined to be) is read out and stored in the detection number storage unit 19 as the number of detected flowing smoke and the number of detected slow smoke.
In step S702, it is determined whether or not the latest detected number of flowing smoke or the latest number of detected slow smoke is equal to or greater than a predetermined threshold (number). If at least one is equal to or greater than the predetermined threshold, the process proceeds to step S703. When it is less than the predetermined threshold, the fire discrimination procedure is terminated.
In step S <b> 703, the number of detected flowing smoke and the number of detected slow smoke stored in the detection number storage unit 19 for each of the most recent fire determination procedures are read out.
In step S704, whether or not the value stored in the latest predetermined number of fire detection procedures for the latest number of detected flowing smoke or the latest number of slow smoke detected above a predetermined threshold is continuously above a predetermined threshold. If it is continuously equal to or greater than the predetermined threshold value, the process proceeds to step S705, and otherwise, the fire determination procedure is terminated.
In this way, it is possible to determine whether there is a fire at a certain point in time when the number of detection areas determined as having smoke is greater than or equal to a predetermined value and spatially determining that there is smoke. The number of detection areas determined to be equal to or greater than a predetermined number of times in a continuous time, and when it can be determined that smoke is present even in time, that is, the number of detection areas with smoke within a predetermined period is high. Sometimes it is determined that a fire has occurred.
In step S705, it is determined that a fire has occurred, an alarm is issued, and the fire determination procedure is terminated.

In the flowchart of FIG. 21 described above, when the number of detection areas in which smoke is detected is continuously equal to or greater than the threshold value, and when such a state continues for a predetermined number of times, smoke is displayed in the image. Although it was determined that there was a fire, the method for determining the fire may be as follows. This fire discrimination method does not particularly require continuity, and uses density in space and time as a discrimination condition.
In FIG. 22, it is assumed that a plurality of, for example, a plurality of nine detection areas A to I are set in the image. In this figure, at time t 0 ~t 5, it shows how going smoke filled the room to be photographed by the camera.
The fire discrimination threshold in the smoke detection device is set to 6/9, for example. “9” indicates the number of detection areas, and “6” indicates the number of detection areas in which smoke is detected by the smoke detection unit and is determined to have smoke.

Now, at time t 2, only the detection region E is determined that there is smoke. Here, when determining fire, the total value of the number of detection areas determined to have smoke for three frames is used. That is, it is “2” because it is the total value of the detection areas with smoke for three frames from t 0 to t 2 . Since “2” is smaller than the threshold value 6, it is not determined as a fire.
If the same calculation is made sequentially, “5” is obtained at t 3 , “5” is obtained at t 4 , and “6” is obtained at t 5 (= 3 + 1 + 2). That is, in the case of such smoke generation would t 5 Oite, fire is determined.
This fire discrimination method considers the number of detection areas with smoke in a certain time and the number of detection areas with smoke in a predetermined time, in other words, detection of presence of smoke in an image. When the density of the region (temporal and spatial density) exceeds a predetermined value, the generation of smoke is detected. If smoke always because its size (area) is changed, for example, it is determined that there is smoke in the detection region H at time t 1, at time t 2, the smoke is moving in the detection area E Thus, the detection region H may disappear. In addition, smoke always spreads spatially over time, and the number of detected areas tends to increase. Therefore, such a discrimination method is an effective discrimination method for detecting smoke. Conceivable.

  In addition, from the viewpoint of this temporal and spatial density, the discrimination method is used not only for the final fire discrimination but also for each step used in the smoke detection means, for example, judgment of transmittance, average value and threshold value. You may make it do. For example, instead of simply setting a flag when the transmittance falls below a predetermined value, a plurality of, for example, four detection data, when three or more times are detected below a predetermined value, that is, there is a continuous decrease. When it is seen, the flag may be set for the first time.

Embodiment 2. FIG.
The fire detection apparatus according to the second embodiment of the present invention is different from the fire detection apparatus according to the first embodiment in the detection area setting means, and the other parts are the same. Omitted. The detection area setting means according to the second embodiment scans the updated reference image data over the entire area, calculates the contrast of each area, and sets the area with the highest contrast as the first detection area. Further, an area having a large contrast without overlapping the first detection area is set as the second detection area, and this is repeated until a predetermined number of detection areas are set.

  The image data includes a low contrast region with a small change in luminance and a high contrast region with a large change in luminance. However, when smoke is generated, it tends to change so that the contrast becomes small. Therefore, it is possible to set a detection area as a detection area that has a large change in brightness rather than a low contrast area that has a small change in brightness. Is easy to detect and effective.

Embodiment 3 FIG.
The fire detection apparatus according to the third embodiment of the present invention is different from the fire detection apparatus according to the first embodiment in the differential value sum ratio calculation means, and is otherwise the same. Description is omitted. The differential value sum ratio calculating means according to the third embodiment obtains a ratio between the spatial differential value of the detection area of the reference image and the spatial differential value of the detection area of the nearest image for each pixel, and calculates the ratio for the entire detection area. And the ratio of the sum of the spatial differential values with respect to the detection area is obtained.

Embodiment 4 FIG.
The fire detection device according to the fourth embodiment of the present invention is different from the fire detection device according to the first embodiment in the measurement line setting means, and is otherwise the same. Omitted. The measurement line setting means according to the fourth embodiment obtains the center of gravity of a region with a weight of 1 as a motion detection pixel and a weight of 0 as a pixel where no motion is detected for each detection region. A horizontal, vertical, or diagonal line segment that passes through the center of the area closest to the straight line connecting is set as a measurement line.

Embodiment 5 FIG.
The fire detection device according to the fifth embodiment of the present invention is different from the fire detection device according to the first embodiment except for the fire discrimination means, and is otherwise the same. To do.
The fire discriminating means according to the fifth embodiment detects the number of detection areas determined to be flowing smoke by the flowing smoke detection means 36 and the number of detection areas determined to be slow smoke by the slow smoke detecting means 37. Stored in the number storage unit 19, reads the number of detections related to the past detection area from the detection number storage unit 19 as time series data, obtains a regression line regarding the detection number, and fires when the slope of the regression line is larger than a predetermined inclination Is determined.

  In the above-described embodiment, the smoke detection device has been described as detecting smoke generated at the time of a fire. However, smoke detects smoke generated from, for example, a chimney, piping, plant equipment, or electronic equipment. You may do it.

It is a figure explaining the principle of smoke detection in this invention. It is a figure explaining the principle of smoke detection in this invention. It is a figure explaining the principle of smoke detection in this invention. It is a block diagram of the fire detection apparatus concerning Embodiment 1 of this invention. It is various memory | storage parts ensured by the smoke detection apparatus concerning Embodiment 1. FIG. 1 is a functional block diagram of a smoke detection device according to Embodiment 1. FIG. FIG. 3 is a functional block diagram of fluid smoke detection means according to the first embodiment. 3 is an explanatory diagram of a measurement line according to Embodiment 1. FIG. It is a graph which shows the frequency spectrum regarding the brightness | luminance of the pixel on the measurement line of the detection area | region containing a motion detection pixel. It is a graph which shows a time-dependent change of the frequency where a motion detection pixel appears on the measurement line of the detection area | region containing a motion detection pixel. 2 is a functional block diagram of slow smoke detection means according to Embodiment 1. FIG. It is a graph of the frequency spectrum obtained by carrying out frequency analysis of the brightness | luminance average time series data of the detection area | region where slow smoke entered. 3 is a flowchart illustrating a fire detection procedure according to the first embodiment. 3 is a flowchart illustrating a reference image update procedure according to the first embodiment. 3 is a flowchart illustrating an image storage procedure according to the first embodiment. 3 is a flowchart illustrating a smoke detection procedure according to the first embodiment. 3 is a flowchart showing the first half of a flow smoke detection procedure according to the first embodiment. 4 is a flowchart showing the second half of the flow smoke detection procedure according to the first embodiment. 3 is a flowchart showing the first half of a slow smoke detection procedure according to the first embodiment. 4 is a flowchart showing the second half of the slow smoke detection procedure according to the first embodiment. 3 is a flowchart illustrating a fire determination procedure according to the first embodiment. 3 is a diagram for explaining a fire discrimination method according to the first embodiment.

Explanation of symbols

  2 Camera, 3 Smoke detection device, 4 Central processing unit (CPU), 5 ROM, 6 RAM, 7 Built-in timer, 8 Input / output interface (I / O), 9 Frame grabber, 10 External storage device, 11 Fluid smoke image storage Unit, 12 slow smoke image storage unit, 13 reference image storage unit, 14 difference image storage unit, 15 reference data storage unit, 16 detection area storage unit, 17 histogram storage unit, 18 luminance average dispersion storage unit, 19 detection number storage unit , 21 Detection preparation means, 22 Image storage means, 23 Smoke detection means, 24 Fire discrimination means, 26 Reference image update means, 27 Detection area setting means, 28 Reference data calculation means, 31 Luminance correction means, 32 Fluid smoke image storage means 33 Slow smoke image storage means 35 Motion detection means 36 Flowing smoke detection means 37 Slow smoke detection means 42 Motion detection Image creation means, 43 Average luminance / dispersion calculation means, 44 Area occupation ratio calculation means, 45 Differential sum total ratio calculation means, 46 Primary determination means, 51 Measurement line setting means, 52 Frequency analysis means, 53 Frequency determination means, 54 Histogram Creation means, 55 average / dispersion calculation means, 56 pixel variation determination means, 61 luminance average dispersion calculation means, 62 determination means, 64 average luminance frequency analysis means, 67 transmittance calculation means, 68 convergence luminance value calculation means, 72 warning determination means.

Claims (3)

  1. In the smoke detection device that detects the generation of smoke within the range captured by image processing the image captured by the camera,
    Detection area setting means for setting a predetermined detection area in the captured image;
    A motion detection image creating means for performing a frame difference and detecting a motion pixel in the detection region;
    First computing means for computing a first type smoke detection determination element in the detection area;
    A second computing means for computing a second type of smoke detection determination element in the detection area when a computed value in a detection area by the first computing means satisfies a predetermined condition;
    A measurement line setting unit configured to set a measurement line in the detection region when a calculated value in a detection region by the first calculation unit satisfies a predetermined condition ;
    A smoke detection device for detecting the generation of smoke in the detection area.
  2. The second type smoke detection determination element comprises at least a frequency spectrum, and the second calculation means obtains a frequency spectrum by frequency analyzing time series data of pixels on the measurement line in the detection region. The smoke detection device according to claim 1 , comprising:
  3. The first type of smoke detector determining element, the moving average of the average luminance, the moving average of the dispersion, according to claim 1 or 2, characterized in that it is constituted from either one or more occupancy ratio or the differential sum ratio Smoke detection device.
JP2006222560A 2006-08-17 2006-08-17 Smoke detector Active JP4926603B2 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2006222560A JP4926603B2 (en) 2006-08-17 2006-08-17 Smoke detector

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP2006222560A JP4926603B2 (en) 2006-08-17 2006-08-17 Smoke detector

Publications (2)

Publication Number Publication Date
JP2008046917A JP2008046917A (en) 2008-02-28
JP4926603B2 true JP4926603B2 (en) 2012-05-09

Family

ID=39180599

Family Applications (1)

Application Number Title Priority Date Filing Date
JP2006222560A Active JP4926603B2 (en) 2006-08-17 2006-08-17 Smoke detector

Country Status (1)

Country Link
JP (1) JP4926603B2 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017084094A1 (en) * 2015-11-20 2017-05-26 富士通株式会社 Apparatus, method, and image processing device for smoke detection

Families Citing this family (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5015838B2 (en) * 2008-03-31 2012-08-29 国立大学法人徳島大学 Smoke detector
JP5215707B2 (en) * 2008-03-31 2013-06-19 能美防災株式会社 Smoke detector
JP5015839B2 (en) * 2008-03-31 2012-08-29 国立大学法人徳島大学 Smoke detector
JP2010049395A (en) * 2008-08-20 2010-03-04 Hochiki Corp Fire detection device
CN101726357B (en) 2008-10-14 2014-01-08 能美防灾株式会社 Smoke detecting apparatus
JP4729610B2 (en) * 2008-10-14 2011-07-20 国立大学法人徳島大学 Smoke detector
JP4653207B2 (en) * 2008-10-16 2011-03-16 国立大学法人徳島大学 Smoke detector
TWI385602B (en) * 2008-10-17 2013-02-11 Ind Tech Res Inst Smoke detecting method and system
KR100982347B1 (en) 2009-02-17 2010-09-15 (주)위니텍 smoke sensing method and system
JP5286113B2 (en) * 2009-03-13 2013-09-11 能美防災株式会社 Smoke detector
JP5356094B2 (en) * 2009-03-31 2013-12-04 能美防災株式会社 Fire detection system
JP5167186B2 (en) * 2009-03-31 2013-03-21 能美防災株式会社 Smoke detector
JP5329466B2 (en) * 2010-03-31 2013-10-30 能美防災株式会社 Smoke detector
JP5286342B2 (en) * 2010-09-29 2013-09-11 能美防災株式会社 Smoke detection frequency component identification method and smoke detection device
DE102011000090B4 (en) * 2011-01-11 2016-03-24 Leica Microsystems Cms Gmbh Method and device for scanning microscopic imaging of an object
JP2012243049A (en) * 2011-05-19 2012-12-10 Fuji Heavy Ind Ltd Environment recognition device and environment recognition method
JP2014232026A (en) * 2013-05-28 2014-12-11 株式会社リコー Attachment detection device, moving body apparatus control system, and moving body
JP5699187B2 (en) * 2013-07-24 2015-04-08 能美防災株式会社 Smoke detector
CN110084160A (en) * 2019-04-16 2019-08-02 东南大学 A kind of video forest rocket detection method based on movement and brightness significant characteristics

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS6486687A (en) * 1987-09-28 1989-03-31 Nec Corp Environment monitoring method
JP2999392B2 (en) * 1995-05-18 2000-01-17 日本無線株式会社 How to detect dust and smoke
JP3909633B2 (en) * 1999-05-17 2007-04-25 小糸工業株式会社 Fire detection device
JP3436519B2 (en) * 2000-09-21 2003-08-11 国際航業株式会社 Observation system for plumes
JP2004030225A (en) * 2002-06-26 2004-01-29 Mitsubishi Heavy Ind Ltd White smoke detection method and apparatus
JP4422002B2 (en) * 2004-11-24 2010-02-24 富士フイルム株式会社 Smoke detection device, smoke detection method, and program

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017084094A1 (en) * 2015-11-20 2017-05-26 富士通株式会社 Apparatus, method, and image processing device for smoke detection
CN108140291A (en) * 2015-11-20 2018-06-08 富士通株式会社 Mist detecting device, method and image processing equipment

Also Published As

Publication number Publication date
JP2008046917A (en) 2008-02-28

Similar Documents

Publication Publication Date Title
Ohmi et al. Particle-tracking velocimetry with new algorithms
US8477232B2 (en) System and method to capture depth data of an image
Ferryman et al. Pets2009: Dataset and challenge
JP4966012B2 (en) System and method for searching for changes in surveillance video
US9679202B2 (en) Information processing apparatus with display control unit configured to display on a display apparatus a frame image, and corresponding information processing method, and medium
CN104123544B (en) Anomaly detection method and system based on video analysis
Celik et al. Fire detection in video sequences using a generic color model
US9390333B2 (en) System and method for robust motion detection
EP3021575B1 (en) Image processing device and image processing method
US20200157929A1 (en) System and Method for Measuring Characteristics of Cuttings From Drilling Operations with Computer Vision
JP4373840B2 (en) Moving object tracking method, moving object tracking program and recording medium thereof, and moving object tracking apparatus
Li et al. Foreground object detection in changing background based on color co-occurrence statistics
EP1131802B1 (en) Smoke detection
JP2009519510A (en) Abnormal crowd behavior detection
US20030141980A1 (en) Smoke and flame detection
JP2009266052A (en) Abnormal behavior detector
CN102163361B (en) Image-type fire detection method based on cumulative prospect image
ES2666499T3 (en) Method for identification of contamination in a lens of a stereoscopic camera
JP2006163452A (en) Tracking device and tracking method
CN107258077B (en) System and method for Continuous Auto Focus (CAF)
KR100948128B1 (en) Smoke detecting method and device
KR100936108B1 (en) Object detection device, and object detection device for elevator
JP4811653B2 (en) Object detection device
CN102348128A (en) Surveillance camera system having camera malfunction detection function
TWI419082B (en) Moving object detection method and image processing system for moving object detection

Legal Events

Date Code Title Description
A621 Written request for application examination

Free format text: JAPANESE INTERMEDIATE CODE: A621

Effective date: 20090618

A977 Report on retrieval

Free format text: JAPANESE INTERMEDIATE CODE: A971007

Effective date: 20111005

A131 Notification of reasons for refusal

Free format text: JAPANESE INTERMEDIATE CODE: A131

Effective date: 20111122

A521 Written amendment

Free format text: JAPANESE INTERMEDIATE CODE: A523

Effective date: 20120119

TRDD Decision of grant or rejection written
A01 Written decision to grant a patent or to grant a registration (utility model)

Free format text: JAPANESE INTERMEDIATE CODE: A01

Effective date: 20120207

A01 Written decision to grant a patent or to grant a registration (utility model)

Free format text: JAPANESE INTERMEDIATE CODE: A01

A61 First payment of annual fees (during grant procedure)

Free format text: JAPANESE INTERMEDIATE CODE: A61

Effective date: 20120208

R150 Certificate of patent or registration of utility model

Ref document number: 4926603

Country of ref document: JP

Free format text: JAPANESE INTERMEDIATE CODE: R150

Free format text: JAPANESE INTERMEDIATE CODE: R150

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20150217

Year of fee payment: 3

R250 Receipt of annual fees

Free format text: JAPANESE INTERMEDIATE CODE: R250

R250 Receipt of annual fees

Free format text: JAPANESE INTERMEDIATE CODE: R250

R250 Receipt of annual fees

Free format text: JAPANESE INTERMEDIATE CODE: R250

R250 Receipt of annual fees

Free format text: JAPANESE INTERMEDIATE CODE: R250

R250 Receipt of annual fees

Free format text: JAPANESE INTERMEDIATE CODE: R250

R250 Receipt of annual fees

Free format text: JAPANESE INTERMEDIATE CODE: R250