JP2012073891A - Frequency component specification method for smoke detection and smoke detector - Google Patents

Abstract

PROBLEM TO BE SOLVED: To simply acquire a frequency component of fluctuation spectrum of smoke without using Fourier transform processing to detect smoke flowing with airflow.SOLUTION: A smoke detector comprises: difference image creation means 21 for creating frame difference images with different sampling intervals in an attention region based on time series data formed of multiple images imaged in a time series by a monitoring camera; difference quantity calculation means 22 for calculating a difference quantity of the attention region for the respective frame difference images created with the different sampling intervals; and target frequency component specification means 23 for specifying an inverse number of a sampling interval corresponding to the largest difference quantity in the calculated respective difference quantity as a frequency component indicating a peak of time spectrum of the attention region.

Description

  The present invention relates to a smoke detection device that detects the presence or absence of smoke during a fire by performing image processing on an image captured by a surveillance camera, and to specify a frequency component for smoke detection used in the smoke detection device In particular, the present invention relates to a smoke detection frequency component identification method and a smoke detection device that focus on the flow rate of smoke.

  From the viewpoint of initial fire extinguishment in the event of a fire or prevention of escape delay in a fire accident, early detection of fire (flame) or smoke is very important. Thus, in the field of smoke detection devices, research has been conducted on the early detection of flames or smoke by performing image processing on images captured by a surveillance camera.

  As an example, based on time-series data of an image obtained by photographing a monitoring area at a predetermined sampling period (for example, a period of 1/30 seconds), interframe difference processing, subband filter processing, or Fourier There is a conventional technique for detecting flame and smoke by performing frequency analysis processing such as conversion processing (see, for example, Patent Documents 1 and 2).

  In Patent Document 1, a flame is detected by performing inter-frame difference processing and cumulative addition processing of the difference results by paying attention to the characteristic that the temporal change of the flame image is small in the flame main body and large in the flame outline. is doing.

  Further, in Patent Document 2, the flame has a high signal intensity level at a low frequency, and the signal intensity level decreases as the frequency becomes higher, and the flame is detected by dividing the image into a plurality of frequency bands. is doing. And in patent document 2, the technique regarding the subband filter process which divides | segments into a several frequency band is disclosed by combining a low-pass filter and a high-pass filter. By using such a subband filter process, a simple frequency analysis that does not require Fourier transform is possible.

  Furthermore, Patent Document 2 also discloses a technique that uses Fourier transform processing in determining whether or not the extracted abnormal region is a smoke region.

  Also, if a predetermined low frequency component is extracted from the power spectrum calculated by Fourier transform of the time series data of the average luminance and the intensity is below a predetermined value, there is a high possibility that smoke has occurred. There is also a prior art that judges that.

JP 11-224389 A JP 2006-293646 A

However, the prior art has the following problems.
Considering the case where smoke flowing in an air current is a detection target, it is important to evaluate the continuous flow of smoke, and the above-described frame difference processing or frequency analysis processing is also used as a method for detecting such movement. Often used.

  The frame difference is a method of performing difference processing on images having different imaging times and extracting only a portion having a difference. For example, Patent Document 1 described above is used for flame extraction or the like. And in this patent document 1, based on the time series data of the image by a fixed sampling period, the flame | frame outline part is extracted by performing the difference process between frames and the accumulation addition process of a difference result.

  However, when the smoke flowing in the air current is to be detected, the moving speed of the smoke becomes a problem, unlike the flame-specific characteristic that the temporal change in the image is small in the flame body and large in the flame contour. . That is, if the flame is a detection target, the contour portion can be emphasized and detected by performing the cumulative addition process of the difference results. However, when the smoke flowing in the airflow is the detection target, the position of the smoke moves in accordance with the speed of the airflow, so that the appropriate addition cannot be performed in the difference result cumulative addition process.

  In addition, when the sampling interval between images to be subjected to difference processing is constant, there is a characteristic peculiar to smoke that the amount of difference (difference amount) varies depending on the speed of smoke flowing in the airflow. More specifically, the difference processing based on a short sampling interval is suitable for detecting smoke that moves fast due to the updraft caused by the flame. It is difficult to extract the smoke that is the detection target. On the other hand, in the difference processing based on a long sampling interval, when smoke as a detection target is accompanied by a fast movement, the movement cannot be reliably captured.

  Furthermore, when the smoke flowing in the airflow is to be detected, as described in Patent Document 2, it is effective to specify the frequency component that maximizes the smoke fluctuation spectrum by Fourier transform processing. Become. However, in order to perform the Fourier transform process, a large amount of calculation and a lot of calculation time associated therewith are required.

  Further, as a simple frequency analysis method with a reduced amount of computation instead of the Fourier transform process, a subband filter process as described in Patent Document 2 can be considered. However, this subband filter processing needs to prepare in advance a combination of a low-pass filter and a high-pass filter suitable for division into a plurality of frequency bands.

  The present invention has been made in order to solve the above-described problems. The frequency component of the smoke fluctuation spectrum can be easily obtained without using the Fourier transform process or the subband filter process, and can be flown into the airflow. It is an object of the present invention to obtain a smoke detection frequency component identification method and a smoke detection device that can detect smoke.

  The smoke detection frequency component specifying method according to the present invention specifies a frequency component for determining whether smoke is generated in a region of interest in an image by performing image processing on an image captured by a surveillance camera. A method of identifying frequency components for smoke detection, wherein a difference image generation step of generating frame difference images at different sampling intervals in a region of interest based on time series data consisting of a plurality of images taken in time series by a surveillance camera For each frame difference image generated at different sampling intervals by the difference image generation step, a difference amount calculation step for calculating the difference amount of the attention area, and a difference amount calculated in the difference amount calculation step The reciprocal of the sampling interval corresponding to the maximum difference amount is the frequency indicating the peak of the time spectrum of the region of interest. Those having a target frequency component specifying step of specifying a component.

  The smoke detection device according to the present invention is a smoke detection device that detects the presence or absence of smoke generation in a region of interest in an image by performing image processing on an image captured by the monitoring camera. An image memory for storing a plurality of images captured in series as time series data, and a difference image generating means for generating frame difference images at different sampling intervals in a region of interest based on the time series data stored in the image memory For each frame difference image generated at different sampling intervals by the difference image generating means, and a difference amount calculating means for calculating the difference amount of the attention area, and a difference amount calculated by the difference amount calculating means. Then, the reciprocal of the sampling interval corresponding to the maximum difference amount is used as a frequency component indicating the peak of the time spectrum of the region of interest. The target frequency component specifying means to be determined and the frequency component specified by the target frequency component specifying means are within a predetermined range, and the difference amount calculated for the frequency component by the difference amount calculating means is greater than or equal to a predetermined threshold value. In this case, the apparatus includes a smoke generation detection unit that determines that smoke has been generated in the region of interest.

  According to the smoke detection frequency component identification method and the smoke detection device according to the present invention, frame difference amounts at different sampling intervals are calculated based on time-series data of an image in a region of interest in the image, and the difference amount By calculating the reciprocal of the sampling interval that maximizes the frequency component that indicates the peak of the time spectrum in that region, the frequency component of the smoke fluctuation spectrum can be simplified without using Fourier transform processing or subband filter processing. Thus, it is possible to obtain a smoke detecting frequency component specifying method and a smoke detecting device capable of detecting smoke flowing in an air stream.

It is a block diagram of the smoke detection apparatus in Embodiment 1 of this invention. It is explanatory drawing which showed typically the technical feature of the frequency characteristic detection part of the smoke detection apparatus in Embodiment 1 of this invention. It is a figure which shows the comparison of the difference amount computed by the difference amount calculation means of the frequency characteristic detection part of the smoke detection apparatus in Embodiment 1 of this invention. In the smoke detection apparatus in Embodiment 1 of this invention, it is explanatory drawing which showed typically the method of obtaining the frame difference image in a different sampling interval using the time series data of an image. In the smoke detection apparatus in Embodiment 1 of this invention, it is explanatory drawing at the time of applying a down sampler as a method of obtaining the frame difference image in a different sampling interval using the time series data of an image.

  Hereinafter, a preferred embodiment of a smoke detection device of the present invention will be described with reference to the drawings.

Embodiment 1 FIG.
FIG. 1 is a configuration diagram of a smoke detection device according to Embodiment 1 of the present invention. The smoke detection apparatus according to the first embodiment includes an image memory 10, a frequency characteristic detection unit 20, and a smoke generation detection unit 30. The image memory 10 is configured as an image memory for a plurality of frames so that images captured by the camera 1 in a time series in a predetermined sampling cycle can be stored as time series data for a certain past period.

  The frequency characteristic detection unit 20 includes a difference image generation unit 21, a difference amount calculation unit 22, and a target frequency component identification unit 23. Specifically, first, the difference image generation means 21 in the frequency characteristic detection unit 20 is stored in the image memory 10 for a past fixed period of time taken by the camera 1 at a predetermined sampling period. Based on the time-series data of the image, frame difference images at different sampling intervals are generated in the region of interest in the image. This frame difference image can be obtained as a difference in luminance value of each pixel in the attention area. Note that this luminance value may be binary or multivalued.

Next, the difference amount calculation unit 22 calculates the difference amount of the attention area for each frame difference image generated at different sampling intervals by the difference image generation unit 21. This difference amount can be obtained as the sum of the luminance values of the difference images in the region of interest.
Or you may obtain | require as a sum total of the number of pixels which changed in the attention area of a difference image.

  Finally, the target frequency component specifying unit 23 calculates the reciprocal of the sampling interval corresponding to the maximum difference amount among the difference amounts obtained for the frame difference images at different sampling intervals, as a time spectrum of the attention area. It is specified as a frequency component indicating a peak.

  In this way, the frequency characteristic detection unit 20 calculates “the frame difference amount at different sampling intervals in the region of interest in the image, and calculates the reciprocal of the sampling interval at which the difference amount is the maximum in the peak of the time spectrum of the region. It is obtained as a frequency component indicating “.

  The smoke generation detection unit 30 subsequent to the frequency characteristic detection unit 20 calculates the value of the frequency component specified by the target frequency component specifying unit 23 within the predetermined range and the specified frequency component. If the difference amount is equal to or greater than a predetermined threshold value, it is determined that smoke has occurred.

  On the other hand, if the value of the frequency component specified by the target frequency component specifying unit 23 is outside the predetermined range, the smoke generation detection unit 30 determines that there is no cause of smoke within this attention area. Further, even if the smoke generation detection unit 30 has a frequency component value specified by the target frequency component specifying unit 23 within a predetermined range, the magnitude of the difference amount in the specified frequency component is a predetermined threshold value. If it is less than that, it is determined that there is no smoke factor within this attention area.

  By providing such a configuration, the smoke detection apparatus of the first embodiment can easily obtain the frequency component of the smoke fluctuation spectrum without using Fourier transform processing or subband filter processing, and whether smoke is present or not. Can be judged. In particular, it can be said that the smoke detection apparatus according to the first embodiment has an image processing function suitable for detecting smoke flowing in the airflow.

  Next, technical features of the frequency characteristic detection unit 20 described above will be described in more detail. FIG. 2 is an explanatory view schematically showing technical features of the frequency characteristic detection unit 20 of the smoke detection device according to Embodiment 1 of the present invention. When a frame difference is performed in a region of interest between consecutive images, if there is a motion, a residual occurs in the difference image, and the motion can be extracted.

  When the frame difference is performed, the residual is generated in both the region existing in the attention region only at the previous time and the region existing in the attention region only at the next time. Considering the relationship between the frame interval and the motion at this time, four patterns A to D as shown in FIG. 2 can be considered. In FIG. 2, in order to simplify the description, it is assumed that the detection object to be moved is shown as a circle, and the background other than the detection object has the same luminance. Further, the detection target indicated by a circle moves to the right side as time passes.

  Further, (a1) to (d1) shown on the left side of FIG. 2 indicate detection objects at the previous time (solid line display) and detection objects at the next time (dotted line display) at the frame intervals of the patterns A to D. Each position is shown. Also, (a2) to (d2) shown on the right side of FIG. 2 are detection objects at the previous time (solid line display) and detection objects at the next time (dotted line display) at the frame intervals of the patterns A to D. (The part shown in black corresponds to the difference, and the sum total corresponds to the difference amount).

[Pattern A] The case where the frame interval is sufficiently faster than the motion of the detection target (see FIGS. 2A1 and 2A2), that is, the sampling interval is very short.
In this case, there is almost no movement of the detection target, and the detection target at the previous time and the detection target at the next time are not significantly shifted, which is almost equal to a stationary state. For this reason, the difference amount in the difference image is also a small value.

[Pattern B] When the frame interval is faster than the motion of the detection target but later than the frame interval of the pattern A (see FIGS. 2 (b1) and (b2)), that is, the sampling interval is slightly longer than the pattern A. Represents the case.
In this case, the amount of motion of the detection target is larger than that of the pattern A, but the detection target at the previous time and the detection target at the next time are still partially wrapped. For this reason, the difference amount in the difference image becomes a larger value than the previous pattern A.

[Pattern C] This represents a case where the frame interval is further slower than the frame interval of the pattern B (see FIGS. 2C1 and 2C2), that is, a case where the sampling interval is longer than the pattern B.
In this case, the amount of movement of the detection target is larger than in the case of the pattern B, and the detection target at the previous time and the detection target at the next time are not wrapped. For this reason, the difference amount in the difference image becomes a larger value than the previous pattern B, and the difference amount for two detection targets is obtained.

[Pattern D] This represents a case where the frame interval is slower than the motion of the detection target and further slower than the frame interval of the pattern C (see FIGS. 2 (d1) and (d2)), that is, the sampling interval is very long. ing.
In this case, the amount of movement of the detection target is larger than in the case of the pattern C, and the detection target at the next time has finally moved out of the attention area. For this reason, the difference amount in the difference image becomes a smaller value than the previous pattern C, and only the difference amount for one detection target is obtained.
Thus, it can be seen that the amount of difference varies depending on the difference in the length of the frame interval depending on the speed at which the detection target moves. Therefore, when detecting smoke, it is necessary to take a difference image at a frame interval that increases the difference amount.

  FIG. 3 is a diagram showing a comparison of the difference amounts calculated by the difference amount calculation means 22 of the frequency characteristic detection unit 20 of the smoke detection device according to Embodiment 1 of the present invention. Specifically, the difference amounts in the four patterns A to D in FIG. 2 are compared. When the difference image is obtained based on the binary image, the difference amount corresponds to, for example, the size of the difference area indicated by black in the difference image in (2a) to (2d) of FIG. To do.

  Of these four patterns A to D, the largest difference amount can be detected when there is a relationship of pattern C. In other states, the difference amount is smaller than pattern C. Yes. From this result, the frequency component obtained as the reciprocal of the frame interval in the pattern C can be easily obtained as the frequency component indicating the peak of the time spectrum in that region.

  In FIG. 2, the case where the frame intervals of the pattern A are the smallest and the frame interval of the pattern D is the largest is described. In order to collect difference images based on such different frame intervals, it is conceivable to obtain difference images between frames at different sampling intervals from the time-series data of the images.

  FIG. 4 is an explanatory view schematically showing a method of obtaining frame difference images at different sampling intervals using time-series data of images in the smoke detection apparatus according to Embodiment 1 of the present invention. The case where nine images I1 to I9 acquired at the sampling interval ΔT are stored in the image memory 10 as time-series data of images is illustrated.

Next, in the example shown in FIG. 4, the difference image generation means 21 in the frequency characteristic detection unit 20 generates the difference images ΔI1, ΔI2, ΔI3,... Based on the time series data of the images as follows. To generate.
ΔI1 = I1-I2 (corresponding to frame interval = ΔT)
ΔI2 = I1-I3 (corresponding to frame interval = 2ΔT)
ΔI3 = I1-I4 (corresponding to frame interval = 3ΔT)

  By performing such difference processing, it is possible to easily generate frame difference images at different sampling intervals using time-series data of images. In the example of FIG. 4, the frame interval is increased by ΔT as ΔT, 2ΔT, 3ΔT,..., For example, ΔT, 3ΔT, 5ΔT,. Similarly, the frequency component indicating the peak of the time spectrum in the region can be easily obtained by increasing the number as ΔT, 2ΔT, 4ΔT, 8ΔT.

  When the frame interval is increased twice, the difference image generating means 21 in the frequency characteristic detection unit 20 can be configured to use a down sampler. FIG. 5 is an explanatory diagram in a case where a down sampler is applied as a method for obtaining frame difference images at different sampling intervals using time-series data of images in the smoke detection apparatus according to Embodiment 1 of the present invention. . The down sampler is to extract images in time series from the group of images taken in time series by the camera 1 at a predetermined sampling cycle at intervals longer than those at the time of imaging. Thereby, even after an image having a predetermined sampling period is acquired by one camera, time-series data having different frame intervals can be generated. That is, time series data is thinned out, and a difference image is generated between the thinned images. By repeating this process, a plurality of frame difference images having different sampling intervals can be calculated.

  As shown in FIG. 5, the difference image generating means 21 is configured to use a down sampler, so that time-series data obtained by doubling the frame interval can be easily obtained based on the time-series data of the sampling interval ΔT. Further, frame difference images with different sampling intervals can be easily generated based on these time series data.

  As described above, according to the first embodiment, based on the time-series data of an image, a plurality of frame differences at different sampling intervals are acquired for a region of interest, and the sampling frequency that maximizes the amount of difference in the region is It can be easily obtained only by the addition / subtraction operation. The frequency component that is the reciprocal of the sampling period thus obtained corresponds to the frequency component that indicates the peak of the time spectrum in that region. Therefore, by paying attention to the peak spectrum and performing such calculation, it is possible to easily specify the approximate value of the frequency component used for smoke detection without performing complicated frequency conversion.

  In the first embodiment described above, the difference image generation unit 21 obtains the difference between images taken at the respective sampling times captured by the camera based on the time-series data of the image. It is not limited to such a difference image. The difference image generation means 21 can perform spatial differentiation on the image captured by the camera instead of using the image itself captured by the camera, and can obtain the difference between the images after spatial differentiation. The effect of can be obtained.

  Moreover, in Embodiment 1 mentioned above, in order to simplify description, although the smoke detection in the attention area was demonstrated about the case where one attention area was provided in the image, this invention is limited to this. Is not to be done. It is also possible to divide the image into a plurality of attention areas consisting of small areas and perform smoke detection for each of the attention areas. In this case, it is possible to detect smoke based on an appropriate determination criterion that differs from region to region by appropriately setting a predetermined range of frequency components and a predetermined threshold value of the difference amount for each region of interest. .

  Further, the plurality of attention areas can be set so that parts thereof overlap each other.

  Moreover, in Embodiment 1 mentioned above, generation | occurrence | production of smoke was judged by calculating the frequency component which is the reciprocal number of a sampling period, For example, the difference amount which becomes the maximum among the obtained difference amounts If the sampling interval corresponding to is within a predetermined range and the detected difference amount exceeds a predetermined value, it may be determined that there is smoke.

  DESCRIPTION OF SYMBOLS 1 Camera, 10 Image memory, 20 Frequency characteristic detection part, 21 Difference image generation means, 22 Difference amount calculation means, 23 Target frequency component specification means, 30 Smoke generation | occurrence | production detection part

Claims (6)

A frequency component identification method for smoke detection that identifies a frequency component for determining the presence or absence of smoke generation in a region of interest in an image by performing image processing on an image captured by a surveillance camera,
A difference image generation step for generating frame difference images at different sampling intervals in the region of interest based on time series data composed of a plurality of images taken in time series by the monitoring camera;
A difference amount calculating step of calculating a difference amount of the region of interest for each frame difference image generated at the different sampling intervals by the difference image generating step;
Target frequency component specification that specifies the reciprocal of the sampling interval corresponding to the maximum difference amount among the difference amounts calculated in the difference amount calculation step as the frequency component indicating the peak of the time spectrum of the region of interest A method for identifying frequency components for smoke detection, comprising the steps of: In a smoke detection device that detects the presence or absence of smoke generation in a region of interest in an image by performing image processing on the image captured by the monitoring camera,
An image memory for storing a plurality of images captured in time series by the monitoring camera as time series data;
Difference image generation means for generating frame difference images at different sampling intervals in the region of interest based on the time-series data stored in the image memory;
For each frame difference image generated at the different sampling intervals by the difference image generation means, a difference amount calculation means for calculating a difference amount of the attention area;
Among the difference amounts calculated by the difference amount calculation means, when the sampling interval corresponding to the maximum difference amount is within a predetermined range, and the calculated difference amount is equal to or greater than a predetermined threshold value Comprises a smoke generation detector that determines that smoke has been generated in the region of interest. In a smoke detection device that detects the presence or absence of smoke generation in a region of interest in an image by performing image processing on the image captured by the monitoring camera,
An image memory for storing a plurality of images captured in time series by the monitoring camera as time series data;
Difference image generation means for generating frame difference images at different sampling intervals in the region of interest based on the time-series data stored in the image memory;
For each frame difference image generated at the different sampling intervals by the difference image generation means, a difference amount calculation means for calculating a difference amount of the attention area;
Target frequency component specification that specifies the reciprocal of the sampling interval corresponding to the maximum difference amount among the difference amounts calculated by the difference amount calculation means as the frequency component indicating the peak of the time spectrum of the region of interest Means,
When the frequency component specified by the target frequency component specifying unit is within a predetermined range and the difference amount calculated for the frequency component by the difference amount calculating unit is a predetermined threshold value or more, A smoke detection device comprising: a smoke generation detection unit that determines that smoke has been generated in a region of interest. The smoke detection device according to claim 3,
The difference image generation means performs spatial differentiation on the image of the time series data stored in the image memory, and based on the time series data after spatial differentiation, frame difference images at different sampling intervals are obtained in the attention area. A smoke detector characterized by generating. The smoke detection device according to claim 3 or 4,
The difference image generating means has a down sampler that generates time series data obtained by doubling the frame interval based on the time series data of the sampling interval ΔT, and when the frame intervals generated by the down sampler are different. A smoke detection device that generates frame difference images at different sampling intervals in the region of interest based on sequence data. The smoke detection device according to any one of claims 3 to 5,
The difference image generation means divides the image into a plurality of attention areas consisting of small areas, generates a frame difference image with a different sampling interval for each of the plurality of attention areas,
The difference amount calculating means calculates a difference amount for each of the plurality of attention areas,
For each of the plurality of regions of interest, the target frequency component specifying unit calculates the reciprocal of the sampling interval corresponding to the maximum difference amount among the difference amounts calculated by the difference amount calculation unit. As a frequency component indicating the peak of the time spectrum of the region of interest,
The smoke generation detection unit is configured such that the frequency component specified for each of the plurality of regions of interest by the target frequency component specifying unit is within a predetermined range set in advance for each of the plurality of regions of interest, and the difference amount When the difference amount calculated for each of the plurality of regions of interest with respect to the frequency component by the calculation unit is equal to or greater than a predetermined threshold value set in advance for each of the plurality of regions of interest, the smoke for each of the plurality of regions of interest A smoke detection device characterized by the fact that it has occurred.

Images