JP2001076157A - Device for detecting object as sort of smoke - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately detect an object as a sort of smoke and a person without misdetection. SOLUTION: The same monitored place is imaged by a television camera C while shifting time to obtain images A and B, and the luminance difference of each pixel is operated to obtain an image a1. The image a1 is binarized to obtain a binary image a2. The number (n) of divided areas 38 and 39 is counted by using the image a2, and when a counted value (n) is equal to or more than the preliminarily defined value n0, an object as the sort of smoke is decided as detected. The value n0 may be, for instance, one value between 2 to 5. When the value (n) is less than the value n0 and also is not zero, a person is decided as detected.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to the emission of white smoke and water vapor generated when a cryogenic fluid such as smoke from a fire or liquefied natural gas leaks into the atmosphere. The present invention relates to a device for detecting generated white smoke and the like and a device for detecting a person.

[0002]

2. Description of the Related Art The difference method for calculating a difference for each pixel between images is the simplest method for comparing a reference image and an input image, and is positioned as a basic process of moving object detection.
Since this method extracts all moving objects, when there is a moving object (disturbance) other than the detection target on the image, erroneous detection increases.

[0003] As a detection device of a smoke-like substance by image processing using typical image processing, there is an apparatus for detecting smoke emission at the occurrence of a fire, leak of steam, and the like. By incorporating calculations such as moving object removal and fixed object contour extraction into these devices,
Although disturbance is removed, it is mainly used for indoor and semi-outdoor conditions.

[0004] In the case where the method is carried out outdoors, the reliability is enhanced by extracting a high-temperature portion with an infrared camera. Further, in order to prevent erroneous detection, a method of detecting an abnormality when a detection target becomes larger than a disturbance element is also known. However, since detection is performed after the abnormality becomes relatively large, it takes time to discover.

[0005]

SUMMARY OF THE INVENTION An object of the present invention is to provide an apparatus for detecting a smoke-like object capable of accurately detecting a smoke-like object without erroneous detection even if it is relatively small. That is.

Another object of the present invention is to provide a person detecting device capable of detecting a person accurately without erroneous detection.

[0007]

According to the present invention, there is provided a brightness difference calculating means for calculating a brightness difference of each pixel of each of the images A and B shifted by a predetermined time Δt to obtain an image a1 represented by the brightness difference. A binarization operation means for binarizing an image a1 represented by the luminance difference to obtain a binary image a2 in response to an output of the luminance difference operation means;
A counting means for counting the number of separated areas of the value image a2;
A determination means for determining that a smoke-like object has been detected when the value is 0 or more.

According to the present invention, the brightness difference calculating means calculates the brightness difference of each pixel of a plurality of images A and B shifted by a predetermined time Δt, ie, calculates the brightness difference between two images A and B. An image a1 composed of pixels having a luminance difference is obtained by subtracting the difference between the respective luminances of the corresponding pixels. The images A and B may be obtained by, for example, an image signal captured by a visible light imaging unit or an infrared imaging unit, or may be obtained by reading out image data stored in a memory.
The image a1 represented by the luminance difference thus obtained is
The image is converted into a binary image a2 by the binarization calculating means, and the number n of the separated areas of the binary image a2 is counted by the counting means. According to the experiment of the present inventor, when the count value n is equal to or greater than the predetermined value n0, it can be determined that a smoke-like object has been detected, and if the count value n is less than the predetermined value n0. For example, it can be determined that a person has been detected. Therefore, even if the shape of the image of the smoke-like object is small,
The smoke-like object can be accurately detected without erroneous detection. The brightness difference may be an absolute value.

Further, according to the present invention, the binarization calculation means includes a histogram creation means for creating a histogram of the brightness difference of the image a1 represented by the brightness difference and the number of pixels, and a brightness response in response to the output of the histogram creation means. The value of the luminance difference near the minimum value of the number of pixels in the range where the difference is small is defined as a first threshold L
A first threshold value setting means for setting as 1 and a value of a luminance difference larger than the first threshold value L1 by a predetermined value α in response to an output of the first threshold value setting means. L2
And an image a1 represented by a luminance difference in response to the output of the first and second threshold setting means, and the first and second thresholds L1, L2
And a level discriminating unit that sets a pixel having a luminance difference within the extraction range of the luminance difference between the pixels as one logical value and a pixel having a luminance difference outside the extraction range of the luminance difference as the other logical value. And

According to the present invention, as will be described later with reference to FIGS. 6 and 7, in an image a1 represented by a luminance difference obtained by subtraction of a pixel having a signal representing luminance having a plurality of gradations. A histogram of the luminance difference and the number of pixels is created, and a value near the minimum value of the number of pixels in a range where the luminance difference is small is set as a first threshold L1, and a value determined in advance from the first threshold L1 A value larger by α is set as the second threshold L2. According to the experiment of the present inventor, it has been confirmed that the extraction range of the luminance difference between the first and second thresholds L1 and L2 corresponds to a smoke-like object. Smoke-like objects have shades, and therefore light smoke is transmitted through the smoke, and the scene behind the smoke-like object may be imaged by the imaging means. , The extraction range is considered to correspond to a smoke-like object. On the other hand, scenes in which no smoke-like object is generated, and moving objects such as people or automobiles do not transmit light, and such characteristics are different from those of the smoke-like object. The present invention more reliably prevents erroneous detection of a smoke-like object based on such characteristics of the smoke-like object. The one logical value and the other logical value are logic “1” and logic “0”.

Further, according to the present invention, the binarizing operation means includes a histogram creating means for creating a histogram of the luminance difference of the image a1 represented by the luminance difference and the number of pixels; Threshold setting means for setting a range between two minimum values of the number as an extraction range of the luminance difference;
In response to the output of the threshold value setting means, an image a1 represented by the luminance difference is converted to a pixel having a luminance difference within the extraction range of the luminance difference between the first and second thresholds L1 and L2. Level discriminating means for setting a pixel having a luminance difference outside the luminance difference extraction range as the other logical value as a logical value.

According to the present invention, based on the characteristics of the smoke-like object described above, the threshold value for binarizing the image a1 represented by the luminance difference is set between the two minimum values of the number of pixels in the histogram. The extraction area, which distinguishes the smoke-like object from scenery and other objects,
False detection of a smoke-like object can be prevented.

Further, the present invention is characterized in that the predetermined value n0 of the determination means is one of 2 to 5.

According to the present invention, a binary image a2 of a smoke-like object is provided.
The number n of the separated regions is 2 to 5 or more, preferably 2 or 3 or more according to the experiment of the present inventor. When such a plurality of separated regions are detected, It is determined that the object has been detected. In this way the detection of smoke-like objects is even more reliable.

Further, in the present invention, the counting means includes a center-of-gravity calculating means for calculating the center of gravity of each of the separated areas, and a center-of-gravity number counting means for counting the number of the centers of gravity in response to the output of the center-of-gravity calculating means. It is characterized by the following.

According to the present invention, when counting the number of separated regions, the center of gravity of the region, that is, the centroid of an image in a two-dimensional figure is calculated and calculated, and the number of the centers of gravity is counted. The number of separated areas can be easily obtained.

Further, according to the present invention, the judging means further includes a detecting means for detecting a smoke-like object based on the shape of the luminance distribution of the luminance of the images A and B and the number of pixels given to the luminance difference calculating means. When the numerical value is equal to or more than a predetermined value n0 and the detection of the smoke-like object is performed by the detection means, it is determined that the object is a smoke-like object.

According to the present invention, as described later with reference to FIG. 12, the smoke distribution is determined by the shape of the luminance distribution in the histogram of the luminance and the number of pixels of the original images A and B for which the luminance difference is to be calculated. It has been found by experiments of the present inventor that it is possible to distinguish and detect a state object and a person. By performing such a determination method in combination with a method of detecting a smoke-like object and a person based on the magnitude of the count value n, it is possible to more accurately detect the smoke-like object and the person. it can.

Also, the present invention provides a luminance difference calculating means for calculating a luminance difference for each pixel of each of the images A and B shifted by a predetermined time Δt to obtain an image a1 represented by the luminance difference, A binarization operation means for binarizing the image a1 represented by the luminance difference to obtain a binary image a2 in response to the output of the means, and a separation of the binary image a2 in response to the output of the binarization operation means A counting means for counting the number of regions obtained, and a person detecting device which responds to the output of the counting means and determines that a person has been detected when the counted value is less than a predetermined value n0.

According to the present invention, when the above-mentioned count value n is less than the predetermined value n0, it is possible to accurately detect a person without erroneous detection. The predetermined value n0 is
Instead of zero, it may be, for example, two or more values. That is, when the binary image a2 is detected as a single area, it is determined that the image is a person.

Further, according to the present invention, the binarizing operation means includes a histogram creating means for creating a histogram of the luminance difference of the image a1 represented by the luminance difference and the number of pixels, and a luminance responding to the output of the histogram creating means. Threshold value setting means for setting, as the threshold value L2, a value of the luminance difference larger by a predetermined value α from the value L1 of the luminance difference near the minimum value of the number of pixels in the range where the difference is small; In response to the output, the image a1 represented by the luminance difference
A pixel having a luminance difference exceeding 2 is set to one logical value,
Level discriminating means for setting a pixel having a luminance difference equal to or smaller than the threshold value L2 to the other logical value.

According to the present invention, a binarization operation is performed by setting a value of a luminance difference larger by a predetermined value α from a value L1 of the luminance difference near the minimum value of the number of pixels in a range where the luminance difference in the histogram is small as a threshold value L2. Therefore, according to the experiment of the present inventor, it becomes possible to accurately discriminate the smoke-like object from the person and detect only the person.

Further, according to the present invention, the binarization calculating means includes a histogram creating means for creating a histogram of the luminance difference of the image a1 represented by the luminance difference and the number of pixels, and a luminance responding to the output of the histogram producing means. Threshold value setting means for setting a value near the minimum value of the number of pixels in the range where the difference is large to a threshold value L2; Threshold L2
And a level discriminator that sets a pixel having a luminance difference equal to or less than the threshold value L2 to the other logical value.

According to the present invention, binarization is performed by setting a value near the minimum value of the number of pixels in a range where the luminance difference of the histogram is large as the threshold value L2. The range in which the luminance difference is large is a range in which the minimum value of the number of pixels is at least the second or more from the smaller luminance difference.
Is the local minimum. As a result, the person can be distinguished from the smoke-like object and other objects, and can be accurately detected without erroneously detecting the person.

[0025]

FIG. 1 is a block diagram showing the overall configuration of an embodiment of the present invention. In a vast factory such as an LNG (liquefied natural gas) base, a plurality j of industrial television cameras C1 to Cj as visible light imaging means are provided, and these television cameras C1 to Cj are switched by a switching means D.
It is displaced and driven by R1 to DRj. Reference C1
ＣCj, DR1ＤＲDRj, etc. may be collectively indicated by omitting the suffixes 1１〜j. Switching means DR1 to DRj
Are individually controlled by the television camera control device 2 in the control device 1. In the central control room 3, an image signal from each television camera C is supplied to the signal processing circuit 4, and each of the captured images is subjected to arithmetic processing in the processing means 5 to detect an abnormality. The result of the abnormality detection is displayed on the visual display means 6 realized by a liquid crystal or a cathode ray tube. The display means 6 also includes a television camera C
Is displayed. Furthermore, when an abnormality is detected, a notification of the abnormality detection is visually displayed. The notification signal of the abnormality detection is also given to a monitor 7 which is a display means realized by a liquid crystal or a cathode ray tube connected to the television camera control device 2 in the control device 1 to be notified. On the monitor 7, the work monitoring state in the factory is displayed with priority, and the notification at the time of abnormality detection is visually displayed as described above. Particularly, in this embodiment, of the plurality of j television cameras C, an image signal from one or a plurality of (for example, one) television cameras C1 is switched by a display means via a switching means DR1 for display / abnormality detection switching. Provided to a certain monitor 7, the operator can visually observe the image of the monitored location captured by the television camera C1. At this time, the remaining image signals from one or more (for example, plural) television cameras C2 to Cj are selectively processed by the switching means DR2 to DRj from the signal processing circuit 4 of the central control room 3 for display / abnormality detection. Means 5 are provided. Switching means DR2 to DRj are:
One image signal of the television cameras C2 to Cj is selected and transmitted from the television camera control device 2 to the signal processing circuit 4 as described above. Signal processing circuit 4 and processing means 5
Are provided in a number equal to the number of the remaining television cameras C2 to Cj, the image signals from the television cameras C2 to Cj can be individually supplied to the signal processing circuit 4 and the processing means 5. In the above-described embodiment, the signal processing circuit 4 and the processing unit 5 are provided in common for the remaining television cameras C2 to Cj, so that the configuration can be efficiently realized.

Next, when the operator selects the television camera C2 for displaying an image, the image signal of the television camera C2 is given to the monitor 7 via the television camera control device 2 via the switching means DR2, and is thus provided. The operator can visually observe the image of the monitored place captured by the television camera C2. At this time, the television cameras C1
Of the Cj, the television camera C on which the above-described image display is performed
TV cameras C1, C3, C4,...
j from the switching means DR1, DR3, D
The signals are selectively switched by R4,..., DRj, supplied to the signal processing circuit 4 from the television camera control device 2, and the processing unit 5 automatically detects an abnormality in the monitored place. In the drawings, the references C3, C4, DR3, DR4 are:
Not shown to simplify the drawing.

In this way, a plurality j of the television cameras C1
To Cj, one or a plurality of television cameras, for example, an image signal from C1 is displayed on the monitor 7, and the remaining one or more television cameras C2 to Cj are displayed.
Are used by the processing means 5 for automatic abnormality detection, and for the automatic abnormality detection, the remaining plurality of television cameras C2 to Cj are sequentially switched one by one. To sequentially perform abnormality detection. Further, image display by such a monitor 7 and processing means 5
Cameras C1 to Cj with automatic abnormality detection by
Are selectively switched, for example, sequentially. In this way, the plurality of television cameras C1 to Cj are shared for displaying images on the monitor 7 and for automatically detecting abnormalities by the processing means 5, whereby it is possible to efficiently configure the present apparatus. it can.

When an abnormality is detected in the image processed by the processing means 5, a signal indicating the detection of the abnormality is given to the television camera control device 2 via a line 15 and the abnormality is detected. The switching operation by the switching means DR is prohibited so that the television camera C capturing the monitored location R keeps capturing the same monitored location R.

The console 8 is connected to the television camera control device 2. By the operation of the console 8 by the observer, the displacement and selection of the television camera C can be manually performed by the switching means DR1 to DRj. Such a console 8 functions as input means.

FIG. 2 is a simplified diagram showing a part of the configuration of the embodiment of the present invention shown in FIG. The television camera C1 is attached to an upper portion of a structure 9 such as a building or a device in a factory such as an LNG base, and is angularly displaced around, for example, a vertical axis 11 to switch the monitored locations R1 and R2 to be imaged. Images can be taken. Further, a television camera C2 is provided on a support 12 erected in a factory, for example, by being angularly driven by a switching means DR2 around a vertical axis 13. Therefore, it is possible to automatically switch the monitored places R3 and R4 to take an image.
In the prior art, the monitored locations of these television cameras C were fixed, but in one embodiment of the present invention, the monitored locations R of each camera C were switched by the switching means DR.
Can be sequentially switched to take an image, so that monitoring can be performed over as many places as possible using as few cameras C as possible.

FIG. 3 is a diagram for explaining the overall operation of the embodiment of the present invention shown in FIGS. 1 and 2. Moving from step i1 to step i2, the television camera control device 2 checks the location R to be monitored by each television camera C.
Is automatically switched and set by the switching means DR. An image signal from one of the cameras C is sent from the television camera control device 2 via the signal processing circuit 4 to the processing means 5.
, The image is subjected to arithmetic processing in step i3, and the occurrence of an abnormal state is determined. Step i4
In step i5, when the processing unit 5 detects an abnormality, a signal notifying the occurrence of the abnormality is supplied to the display unit 6, the television camera control device 2, and the monitor 7 in step i5. In this way, the display unit 6 and the monitor 7 notify the occurrence of the abnormality.
The occurrence of this abnormality is also displayed together with information on the monitoring place R by the television camera C. During the period in which the abnormality is detected at the monitoring location R, steps i3, i4,
i5 is executed, and the abnormal state is constantly monitored. The display means 6 and the monitor 7 also display a visible light captured image of the monitored location R captured by the television camera C.

If the processing means 5 does not detect any abnormality in step i4, the process proceeds to step i6. The time W1 required for the arithmetic processing for detecting the abnormality of the processing means 5 in step i3 described above is, for example, about 5 seconds. In step i6, a predetermined time W2 is set from the time when one monitored place R is imaged by the television camera C1.
Until (W2> W1) has elapsed, the switching of the monitored location R by the switching means DR is not performed. After the elapse of the time W2, the process returns from step i6 to step i2, and the imaging operation of the next monitored place R by the television camera C is switched. For example, when the monitored places R1 and R2 of a certain television camera C1 are switched by the switching means DR1 and no abnormality is detected, such switching operation is sequentially performed every time W2. Monitored location R1, R by TV camera C1
After the switching operation of Step 2 is completed, the switching means DR2 switches the monitored places R3 and R4 by the next television camera C2.
When no abnormality is detected, the switching is sequentially performed every time W2. In this way, the switching means D for each television camera C
The monitored places R are sequentially switched by R, and it is determined whether the common processing unit 5 has detected an abnormality.

The operation of detecting an abnormality in the monitored location R by the processing means 5 will be described with reference to FIGS. FIG.
Table 1 shows the imaging time of each of the images A, B, and a1 and the calculation for each pixel.

[0034]

[Table 1]

FIG. 4 is a diagram showing a state in which images a1 and a2 are created from images A and B. The images A and B are the same image by one television camera, for example, C1, with a time lag by a predetermined time Δt. This is an image obtained by imaging a monitoring place, for example, R1. During the time when these images A and B are captured, the switching means DR1 stops and the television camera C1 is stopped.
Do not displace and keep stopped. The pixels constituting each of the images A and B derive a signal representing a plurality of, for example, 256 luminances, that is, grayscales of density.

The image A includes an image 21 and has a time Δt1
The image B which is shifted only by a distance includes an image 22 identical to the image 21. The image B includes an image 23 that is not included in the image A. Therefore, by calculating the luminance difference of each pixel of these images A and B as described above, the image a
1 is obtained. In this image a1, one of the two images A and B, for example, the image 23 included only in B exists, and the images 21 and 22 commonly included in both images A and B are Removed. By binarizing each pixel of the image a1 using the thresholds L1 and L2, a binary image a2 can be obtained. Image 23 in image a1
Are binarized and obtained as an image 24 in the binary image a2. Pixels are indicated by reference numerals 26 and 27 in the binary image a2. The black pixel 26 has one logical value, for example, a logical “1”, and the white pixel 27 has one of the other logical values, for example, a logical “0”.

At this time, the area of logic "1" is the pixel where the change has occurred, and its minimum circumscribed rectangle is the change-occurring area 3
Recognized as 3. Note that a more reliable change occurrence region can be obtained by using a plurality of continuous images and performing an OR operation on the difference image (see FIG. 8).

FIG. 5 is a flow chart for explaining the operation of the processing means 5. Step j1 to Step j
The process proceeds to step S2, where the data of the image A is stored and stored in the memory 17. In step j3, the time series difference of the image a1 is calculated using the data stored in the memory 17 of the images A and B, that is, the brightness difference of each pixel is calculated. Thus, for the image a1, data representing the luminance difference between the pixels of the images A and B on which the image a1 is based is stored. The image a1 is shown in FIG. 4 (3). The data of each pixel of each image a1 may be the absolute value of the luminance difference, but may be a positive or negative value.

At step j4 in FIG. 5, each pixel of the image a1 represented by the luminance difference is binarized according to the threshold values L1 and L2. Thus, the image a1 is binarized to obtain 2
The value image is represented by reference numeral a2.

FIG. 6 is a diagram for explaining the operation of the binarization operation of the processing means 5. This binarization operation is described in FIG.
At step j4. The processing unit 5 first proceeds from step k1 to k2 in FIG. 6, and creates a histogram of the luminance difference and the number of pixels of the image a1 represented by the luminance difference.

FIG. 7 is a diagram showing a histogram of the image a1. The horizontal axis in FIG. 7 is the luminance difference, and the luminance difference is 0 to 2
55 are set at equal intervals. The vertical axis in FIG. 7 represents the number of pixels at equal intervals. Image a1 in which a smoky object is imaged in images A and B and is therefore represented by a luminance difference
9, the number of pixels is divided into three groups GP1, GP2, and GP3. In a state where the group GP1 is obtained, in the images A and B, the object to be imaged is obtained by a non-moving object, for example, an image of a landscape, a road, a house, or the like. Group GP2 indicates a state when a smoke-like object is imaged. The group GP3 indicates a state when a moving person is imaged.

When the smoke-like object is not imaged, the group GP2 is not obtained in the histogram of FIG.
Is obtained. When a moving person is not imaged, the group GP3 is not obtained and the characteristic 19 is obtained. The present inventor has discovered that it is important to detect the groups GP2 and GP3 in the histogram shown in FIG. 7 in detecting a smoke-like object and a person in this manner. With such a histogram, the image a1 represented by the luminance difference is level-discriminated by the threshold values L1 and L2, and a binary image a2 using only the pixels of the group GP2 is obtained.

In order to obtain such a binary image a2, a first threshold value L1 is first calculated and obtained in step k3 of FIG. The first threshold value L1 is selected as a value of the luminance difference near the minimum value 29 of the number of pixels in the range where the luminance difference is small in FIG. The range in which the luminance difference is small is a range including the first minimum value 29 when the luminance difference is traced from the smaller one to the larger one. In addition, erroneous detection due to noise can be prevented. Is a range of values for identifying the group GP according to.

After the first threshold value L1 is set in this way, a value of a luminance difference larger by a predetermined value α from the first threshold value L1 is set as a second threshold value L2. The second threshold value L2 is obtained by calculation in step k4 in FIG. L2 = L1 + α (1)

In the extraction range 40 between the first and second threshold values L1 and L2 thus obtained, the image a1 represented by the luminance difference is level-discriminated to only the image of the group GP2 caused by the smoke-like object. A binary image a2 is obtained as described above.

The threshold value L1 is changed to the group GP1 or G1 in FIG.
A method of calculating from P2 will be described. Group GP1, G
The number of P2 is k, the luminance of the h-th pixel in the groups GP1 and GP2 is xkh, the average luminance of the group GPk (k = 1, 2 in this case) is x1k, and the number of pixels of the group GPk is Is nk, the variance σ ² a is as follows.

When the average luminance of all pixels is x1, the variation between the groups GP1 and GP2 is the inter-class variance σ
It can be evaluated ² b as an index.

The dispersion ratio F0, which is the ratio between σ ² a and σ ² b, is defined by the following equation. F0 = σ ² b / σ ² a (4)

The threshold value L1 is selected so that the dispersion ratio F0 becomes maximum. In another embodiment described later, a similar method can be used to determine the threshold value L2 from the groups GP2 and GP3.

FIG. 9 is a diagram for explaining the principle for discriminating between the smoke-like object 36 and the person 37. FIG. 9 (1)
Assume that a smoke-like object 36 and a person 37 are imaged as shown in FIG. The image of the smoke-like object 36 has a distribution of brightness density, whereas the image of the person 37 has a relatively uniform brightness density. Both the smoke-like object 36 and the person 37 move as time passes.

FIG. 9 (2) shows the smoke-like object 36 and the person 37.
Is performed as described above, and an image a2 obtained thereby is shown. The area 33a is a change occurrence area corresponding to the smoke-like object 36, and the area 33b is a change occurrence area corresponding to the person 37. Change generation area 33 of smoke-like object 36
a has a relatively closely separated area 38 corresponding to the image of this smoke-like object 36. In this embodiment, there are a total of six separate areas 38 corresponding to the smoke-like objects 36. On the other hand, the change occurrence area 33b can be obtained based on the image 39 of the person 37. This change occurrence area 3
The image 39 corresponding to the person 37 in 3b is not separated and is a single image. Therefore, the number n of the separated regions 38 for each of the change occurrence regions 33a and 33b is counted,
It is understood that a determination as to whether the object is a smoke-like object 36 or a person 37 can be made based on the count value n.

FIG. 10 shows each of the change occurrence areas 33a and 33b.
5 is a flowchart for explaining the operation of counting the number n of the separated areas 38 and 39 in FIG. Step r
The process moves from step 1 to step r2, where the change occurrence areas 33a, 33
Determine b. In the next step r3, the areas 38 and 39 in the change occurrence areas 33a and 33b are determined.
Therefore, in step r4, the center of gravity of each separated area 38 is calculated. Also, a single area 39 in the change occurrence area 33b
Is also calculated. In step r5, the change occurrence areas 33a, 33b of the areas 38, 39 thus obtained are
The number n of the centers of gravity within is counted. Step r
At 6, the series of operations is terminated.

FIG. 11 is a flowchart for explaining the operation of the processing means 5 for discriminating between a smoke-like object and a person. In step r5 of FIG. 10 described above, the number n of the counted centroids is equal to or larger than a predetermined value n0 (n
≧ n0) is determined in step u2. If the count value, which is the number n of the centers of gravity, is equal to or greater than a predetermined value n0,
In step u3, it is determined that a smoke-like object has been detected. If n <n0, the process moves from step u2 to step u4, and it is determined that a person has been detected. Thus, in step u5, a series of operations ends. n0 is 2
-5, preferably 2 or 3.

In another embodiment of the present invention, instead of calculating the center of gravity and counting it, the number of separated areas may be directly counted.

In still another embodiment of the present invention, FIG.
The first and second thresholds L1,
L2 represents a range between the two minimum values 29 and 41 of the number of pixels in the created histogram, and a luminance difference extraction range 40.
It may be. In this way, level discrimination is performed using pixels having a luminance difference within the extraction range 40 as one logical value and pixels having a luminance difference outside the extraction range 40 as the other logical value.

In another embodiment of the present invention, in steps u3 and u4 of FIG. 11, the determination of a smoke-like object and a person can be performed more accurately by using another method described below. Good.

FIG. 12 is a diagram for explaining a method for discriminating between a smoke-like object and a person according to another embodiment of the present invention. The histogram of the number of pixels corresponding to the luminance shown in FIG.
5 is a histogram showing the luminance and the number of pixels in the image. 2 above
The change occurrence area 33 is determined by a value calculation or the like. The luminance of each pixel having a plurality of gradations of the image B included in the change occurrence area 33 and the number of pixels corresponding to the luminance are obtained as shown in FIG. When an image of a smoke-like object is detected, the characteristic of FIG. 12A is obtained, and when a person is detected, the characteristic of FIG. 12B is obtained. The inclinations (Y1 / X1) and (Y2) of the number of pixels from the tops q1 and q2 on the side of low luminance to the side of high luminance on each characteristic.
/ X2). By giving each of the gradients (Y1 / X1) and (Y2 / X2) thus obtained to predetermined values and performing level discrimination, it is possible to determine whether the object is a smoke-like object or a person. In this way, when a smoke-like object is further detected in step u3 in the determination result in step u2, the accuracy of detection of the smoke-like object is improved. In step u4, a person is determined in step u2.
As shown in (2), the further the person is determined, the more accurately the person is detected.

[0058]

According to the first aspect of the present invention, a luminance difference of each pixel of the images A and B shifted in time is calculated to obtain an image a1 composed of pixels having a luminance difference, and a binarization operation is performed. Do 2
A value image a2 is obtained, and when the number n of the separated areas in the binary image a2 is equal to or greater than a predetermined value n0, it is determined that a smoke-like object has been detected. Is relatively small, accurate detection can be performed without erroneous detection.

According to the second aspect of the present invention, a histogram of the brightness difference of the image a1 represented by the brightness difference and the number of pixels is created, and the value near the minimum value of the number of pixels in the range where the brightness difference is small is calculated. By defining the first threshold value L1 and further defining the second threshold value L2 which is larger by a predetermined value α, the extraction range L1 to L2 in which the smoke-like object is present is determined, thereby erroneously detecting the smoke-like object. Therefore, it is possible to more reliably and accurately detect.

According to the third aspect of the present invention, the number of pixels in the histogram of the image a1 represented by the luminance difference is 2
An extraction range is set between the two minimum values. For example, the two minimum values may be two minimum values in order from the one with the smallest luminance difference, whereby erroneous detection of a smoke-like object can be reliably prevented. Become like

According to the fourth aspect of the present invention, the smoke-like object 2
The number of separated areas of the value image a2 is 2 to 5 or more, and when the number n of such separated areas is a predetermined value n0 of 2 to 5 or more, a smoke-like object is detected. Judgment is made, and false detection is prevented.

According to the fifth aspect of the present invention, when counting the number of separated areas, the center of gravity of the area is calculated,
Thus, the number of separated areas can be easily obtained.

According to the sixth aspect of the present invention, a smoke-like object and a person can be distinguished from each other based on the shape of the luminance distribution in the histogram. Detection can be performed even more accurately.

According to the present invention, when the count value n is less than the predetermined value n0, it is determined that the person is a person, and it is known that the person has entered a factory, for example. And the above detection can be performed.

According to the eighth aspect of the present invention, the threshold value L2 for detecting a person is set to a predetermined value α from the luminance difference value L1 near the minimum value of the number of pixels in the range where the luminance difference is small in the histogram. Is set to a threshold value L2, and when the number of pixels having a luminance difference exceeding the threshold value L2 is equal to or greater than a predetermined value, it can be determined that a person exists.

According to the ninth aspect of the present invention, the threshold value L2 is set in a range where the luminance difference of the histogram is large.
Thus, a person can be accurately detected without erroneous detection.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an overall configuration of an embodiment of the present invention.

FIG. 2 is a simplified diagram showing a part of the configuration of the embodiment of the present invention shown in FIG. 1;

FIG. 3 is a diagram for explaining the overall operation of the embodiment of the present invention shown in FIGS. 1 and 2;

FIG. 4 is a diagram showing a state in which images a1 and a2 are created from images A and B.

FIG. 5 is a flowchart for explaining the operation of the processing means 5;

FIG. 6 is a diagram for explaining the operation of the binarization operation of the processing means 5;

FIG. 7 is a diagram showing a histogram of an image a1.

FIG. 8 is an image diagram for obtaining a change area from a plurality of images by an OR operation.

FIG. 9 is a diagram for explaining the principle for distinguishing a smoke-like object from a person;

FIG. 10 is a flowchart for explaining an operation of counting the number n of separated areas 38 and 39 in each of the change occurrence areas 33a and 33b.

FIG. 11 is a flowchart for explaining an operation performed by the processing means 5 to discriminate between a smoke-like object and a person.

FIG. 12 is a diagram for explaining a method for distinguishing between a smoke-like object and a person according to another embodiment of the present invention.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 control device 2 television camera control device 3 main control room 4 signal processing circuit 5 processing means 6 visual display means 7 monitor 36 smoke-like object 37 people 38, 39 area A to F Time-shifted gradation image a1 Image a2 Binary image C1 to Cj Visible light television camera DR1 to DRj Switching means R1 to R4 Monitored location

Continuation of the front page (72) Inventor Kiyoshi Yada 4-1-2, Hirano-cho, Chuo-ku, Osaka-shi, Osaka Inside Osaka Gas Co., Ltd. (72) Inventor Naoya Fujimaru 4-1-2, Hirano-cho, Chuo-ku, Osaka-shi, Osaka No. Osaka Gas Co., Ltd. (72) Kuniyoshi Okamoto, Inventor 4-1-2, Hirano-cho, Chuo-ku, Osaka City, Osaka Prefecture Osaka Gas Co., Ltd. Mitsubishi Heavy Industries, Ltd.Nagasaki Shipyard (72) Inventor Takashi Mori 1-1, Akunouramachi, Nagasaki, Nagasaki Prefecture Mitsubishi Heavy Industries, Ltd.Nagasaki Shipyard F-term (reference) 5L096 BA02 CA05 DA03 EA43 FA14 FA37 FA53 FA60 GA06 GA08 GA51 HA03 HA07

Claims (9)

[Claims] 1. A luminance difference calculating means for calculating a luminance difference for each pixel of each of the images A and B shifted by a predetermined time Δt to obtain an image a1 represented by the luminance difference, and an output of the luminance difference calculating means. And a binarization operation means for binarizing the image a1 represented by the luminance difference to obtain a binary image a2, and a separated area of the binary image a2 in response to the output of the binarization operation means And counting means for responding to the output of the counting means and determining that a smoke-like object has been detected when the count value is equal to or greater than a predetermined value n0. Smoke-like object detection device. 2. A binarizing operation means, comprising: a histogram creating means for creating a histogram of the brightness difference of the image a1 represented by the brightness difference and the number of pixels; and a response to the output of the histogram creating means, wherein the brightness difference is small. The value of the luminance difference near the minimum value of the number of pixels in the range is defined as the first
A first threshold value setting means for setting as the threshold value L1, and a first threshold value L in response to an output of the first threshold value setting means.
A second threshold value setting means for setting a value of a luminance difference larger by a predetermined value α from 1 as a second threshold value L2; A pixel having a luminance difference within the luminance difference extraction range between the first and second thresholds L1 and L2 is defined as one logical value, and the luminance difference outside the luminance difference extraction range is represented by 2. A smoke-like object detecting apparatus according to claim 1, further comprising level discriminating means for setting the pixel having the other value as the other logical value. 3. A binarizing operation means, comprising: a histogram creating means for creating a histogram of the luminance difference between the image a1 represented by the luminance difference and the number of pixels; Threshold value setting means for setting a range between the two minimum values as a luminance difference extraction range; and an image a1 represented by the luminance difference in response to an output of the threshold value setting means. Level discriminating means for setting a pixel having a luminance difference within the extraction range of the luminance difference between the values L1 and L2 as one logical value, and a pixel having a luminance difference outside the extraction range of the luminance difference as the other logical value. The smoke-like object detection device according to claim 1, further comprising: 4. The predetermined value n0 of the determination means is 2
2. The method according to claim 1, wherein the value is one of the values from 1 to 5.
The detection device for a smoke-like object according to any one of claims 1 to 3. 5. The counting means includes: a center-of-gravity calculating means for calculating a center of gravity of each of the separated areas; and a center-of-gravity number counting means for counting the number of centers of gravity in response to an output of the center-of-gravity calculating means. The detection device for a smoke-like object according to any one of claims 1 to 4. 6. The judging means further comprises a detecting means for detecting a smoke-like object based on the shape of the luminance distribution between the luminance of the images A and B and the number of pixels given to the luminance difference calculating means, and wherein the count value is determined in advance. The smoke-like object according to one of claims 1 to 5, wherein the smoke-like object is determined when the smoke-like object is detected by the detection means when the smoke-like object is equal to or more than a predetermined value n0. Detection device. 7. A brightness difference calculating means for calculating a brightness difference of each pixel of each of the images A and B shifted by a predetermined time Δt to obtain an image a1 represented by the brightness difference, and an output of the brightness difference calculating means. And a binarization operation means for binarizing the image a1 represented by the luminance difference to obtain a binary image a2, and a separated area of the binary image a2 in response to the output of the binarization operation means A counting means for counting the number of persons, and a person detecting device which responds to the output of the counting means and determines that a person has been detected when the counted value is less than a predetermined value n0. 8. A binarizing operation means, comprising: a histogram creating means for creating a histogram of the brightness difference of the image a1 represented by the brightness difference and the number of pixels; and a response to the output of the histogram creating means, wherein the brightness difference is small. A luminance difference value larger by a predetermined value α from a luminance difference value L1 near the minimum value of the number of pixels in the range is set to a threshold L2
In response to the output of the threshold setting means, the image a1 represented by the luminance difference is set to a pixel having a luminance difference exceeding the threshold L2 as one logical value, 8. The apparatus according to claim 7, further comprising level discriminating means for setting a pixel having a luminance difference equal to or smaller than the threshold value L2 to the other logical value. 9. A binarization operation means, a histogram creation means for creating a histogram of the brightness difference of the image a1 represented by the brightness difference and the number of pixels, and a response to the output of the histogram creation means, wherein the brightness difference is large. A value near the minimum value of the number of pixels in the range is defined as a threshold value L2
And a pixel having a luminance difference exceeding the threshold L2 in the image a1 represented by the luminance difference in response to an output of the threshold setting means, 8. The apparatus according to claim 7, further comprising level discriminating means for setting a pixel having a luminance difference equal to or smaller than the threshold value L2 to the other logical value.