JP2011215806A - Smoke detection device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a smoke detection device for improving the detecting precision of smoke generation based on the extraction of a featured value than a conventional device under the consideration of the behavior of smoke.SOLUTION: The smoke detection device includes: a smoke featured value calculation means (30); a storage part (20) for storing a prescribed determination reference value; a smoke detection means (40) for, on the basis of the comparison result of the featured value extracted by the smoke featured value calculation means with the reference determination value, specifying a first candidate area in which smoke is highly likely to have been generated; and an adjacency degree of similarity processing means (50) for calculating a texture featured value quantitatively showing that features relating to luminance is similar with respect to each of a plurality of areas, and for specifying the area where difference quantity between the texture featured value and a texture featured value calculated about the adjacent areas is equal to or less than allowable difference quantity as a second candidate area whose degree of similarity is high. The smoke detection means determines that the area among the first candidate areas and the second candidate areas configured of at least the prescribed number of areas as the area where smoke is highly likely to have been generated.

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 and extracting a feature amount caused by smoke, and in particular, depending on an observation situation and an observation time. The present invention relates to a smoke detection device capable of further stabilizing fluctuating detection accuracy.

  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 or smoke is very important. Thus, in the field of smoke detection devices, research has been conducted on early detection of smoke by performing image processing on an image captured by a surveillance camera.

  As an example, there is a conventional smoke detection device that detects smoke by installing a camera in a tunnel or the like and performing image processing on an image captured by the camera. In image processing for detecting smoke, generally, a reference image (reference image) is stored in advance, a difference image between the latest captured image and the reference image is calculated, and a region where a change has occurred is calculated. Extraction detects smoke (for example, see Patent Document 1).

  In addition, the reference image is regularly updated in order to cope with the temporal change of the reference image due to the influence of sunlight.

Thus, by performing image processing on the image captured by the camera and performing smoke detection, the following two merits can be obtained.
1) By visually confirming the image of the surveillance camera, it is possible to grasp the smoke detection status in a remote place.
2) It is possible to divert already installed surveillance cameras and construct efficient equipment.

Japanese Patent No. 3909665

However, the prior art has the following problems.
The captured smoke has different visibility depending on the background. In addition, the smoke that is the detection target is greatly affected by the airflow, and its behavior is probabilistic that varies depending on the observation situation and the observation time. For this reason, the feature amount extracted for detecting the generation of smoke varies depending on the observation state and the observation time. Therefore, in order to stabilize the detection accuracy of smoke generation, it is important to fully consider the behavior of smoke.

  In the prior art, as one measure for considering the behavior of smoke, by attaching the same label to a set of regions connected to each other in the region where the feature amount due to smoke is extracted, Some perform a “labeling process” in which a plurality of areas are grouped and extracted. This process is equivalent to a labeling process for a binary image.

  Then, by performing grouping by concatenation processing using such labeling processing, and extracting the part where the region after grouping is a predetermined size or more, the region from which the feature amount is extracted Even so, the area less than the predetermined size can be excluded from areas where smoke is highly likely to be generated. Accordingly, it is possible to perform appropriate detection in consideration of smoke-specific behavior that smoke spreads beyond a predetermined region.

  However, even when smoke detection is performed using such a labeling process, the detection accuracy of the feature quantity itself used for the labeling process may vary depending on the observation situation and the observation time. Therefore, as a result, there is a possibility that the smoke detection accuracy may be deteriorated without completely excluding the influence of the feature amount that varies depending on the observation state and the observation time.

  The present invention has been made to solve the above-described problems, and in consideration of smoke behavior, it is intended to improve the detection accuracy of smoke generation based on feature amount extraction, compared to conventional devices. The purpose is to obtain a smoke detector capable of performing the above.

  The smoke detection device according to the present invention includes a smoke feature amount calculation unit that extracts a feature amount related to smoke for each of a plurality of regions set in an image captured by a monitoring camera, and a plurality of regions. A storage unit in which a predetermined reference determination value for determining the presence or absence of smoke generation is stored in advance for each area, and a predetermined reference determination value stored in the storage unit is extracted from each of the plurality of areas. , Comparing with the feature quantity extracted by the smoke feature quantity calculation means, identifying a first candidate area that is highly likely to generate smoke for each of the plurality of areas based on the comparison result, and further specifying the identified first candidate In a smoke detection device comprising smoke determining means for determining that a region composed of a predetermined number of regions or more among the regions is a region where smoke is highly likely to be generated, the image was captured by a surveillance camera For each of a plurality of regions set in the image, a texture feature amount that quantitatively indicates that the features related to luminance are similar is calculated, and a difference amount from the texture feature amount calculated for the adjacent region is calculated. , Further comprising an adjacent similarity processing unit that identifies an area in which the calculated difference amount is within the allowable difference amount as a second candidate region having a high similarity in characteristics regarding luminance with the adjacent region, and the smoke determination unit includes the first candidate Among the areas that are areas and the second candidate areas, an area that is composed of a predetermined number of areas or more is determined to be an area that is highly likely to generate smoke.

  According to the smoke detection device of the present invention, the first smoke-specific first means that the smoke spreads beyond the predetermined area by including the adjacent similarity processing means that can extract the area having the high similarity of the characteristic with respect to the adjacent area. And the second characteristic unique to smoke that the characteristics related to brightness are similar in the adjacent small area due to the spread of smoke. It is possible to obtain a smoke detection device capable of improving the detection accuracy of smoke generation based on the extraction of feature amounts than the device.

It is a block diagram of the conventional smoke detection apparatus. It is explanatory drawing which showed the area | region division in 1 screen in Embodiment 1 of this invention, and the pixel structure of each area | region. It is explanatory drawing regarding the conventional labeling process by a smoke determination means. It is explanatory drawing about the smoke detection process which considered the smoke specific behavior in the conventional smoke detection apparatus. It is a block diagram of the smoke detection apparatus in Embodiment 1 of this invention. It is explanatory drawing about the smoke detection process which considered the smoke specific behavior in the smoke detection apparatus of Embodiment 1 of this invention.

Hereinafter, a preferred embodiment of a smoke detection device of the present invention will be described with reference to the drawings.
The present invention relates to a smoke detection apparatus including a detection method capable of improving the detection accuracy of smoke generation in consideration of the stochastic behavior of smoke affected by the observation situation and the observation time. This detection method is an effective technique for detecting smoke flowing in the airflow.

Embodiment 1 FIG.
When smoke is observed from a distance, such as a surveillance camera, the smoke that is the detection target is not limited to the small area in the captured screen, but is present over a plurality of small areas, and the plurality of small areas Then, it shows a behavior peculiar to smoke that feature quantities related to luminance are similar. Therefore, in the present invention, paying attention to such smoke-specific behavior, by introducing a feature quantity that is different from the feature quantity used for smoke detection, as a feature quantity used to perform the similarity processing of the region, The technical feature is that the detection accuracy of smoke generation can be improved.

  Therefore, before describing the above-described technical features of the present invention, feature amounts used for smoke detection will be described first.

  FIG. 1 is a configuration diagram of a conventional smoke detector. The smoke detection apparatus of FIG. 1 includes an image memory 10, a storage unit 20, a smoke feature amount calculation unit 30, and a smoke determination unit 40. The image memory 10 is configured as an image memory for a plurality of frames so that images captured by the camera 1 can be stored as time-series data for a certain period in the past.

  The image to be monitored is taken into the image memory 10 as a one-screen image divided in pixel units. FIG. 2 is an explanatory diagram showing area division within one screen and a pixel configuration of each area in Embodiment 1 of the present invention. FIG. 2A shows a plurality of divided areas set in advance in one screen, and the screen is divided into a plurality of rectangular areas in a vertical and horizontal matrix. On the other hand, FIG. 2B shows that each area is divided into pixel units.

  The smoke feature quantity calculation means 30 extracts a feature quantity related to smoke for each region of the captured image stored in the image memory 10. Then, the smoke determination unit 40 performs a comparison process between a predetermined reference determination value stored in advance in the storage unit 20 and the feature amount extracted by the smoke feature amount calculation unit 30, and the possibility that smoke has occurred. Identify areas with high Typical examples of the feature quantity extraction method by the smoke feature quantity calculation means 30 include the following four.

[Extraction method 1: Smoke detection based on luminance distribution of pixels]
The smoke feature quantity calculation means 30 calculates the luminance variance of the pixels in each area for each area. As the image for calculating the luminance dispersion, the smoke feature amount calculation means 30 basically uses the latest captured image. However, by using time-series data of images stored in a plurality of image memories 10, a plurality of captured images can be used retroactively. In this way, the smoke feature quantity calculation means 30 outputs the calculated luminance dispersion or the standard deviation obtained from the brightness dispersion as a feature quantity relating to smoke.

  The smoke determination unit 40 described later determines whether or not the luminance variance calculated by the smoke feature amount calculation unit 30 or the standard deviation obtained from the luminance variance is within a predetermined range. When it is inside, it can be determined that there is a high possibility that smoke has been generated.

  Although not shown, the smoke determination unit 40 is provided with a final smoke determination unit. The final smoke determination unit, when there is a predetermined number or more of areas where the smoke determination unit 40 determines that the possibility that smoke has been generated is high continuously for a predetermined time, finally, smoke is finally generated in that area. This is the part that is judged to be.

[Extraction method 2: Smoke detection based on temporal dispersion of average luminance of pixels]
The smoke feature quantity calculation means 30 calculates the average luminance of the pixels in each area for each area. Next, the smoke feature quantity calculation means 30 uses the time-series data of the images stored in the plurality of image memories 10 to go back to the past and average the plurality of captured images over a certain period in the same region. The brightness is calculated, and time series data of average brightness is generated for each target area. Then, the smoke feature amount calculating unit 30 calculates the luminance variance of the generated time series data of the average luminance.

  In this way, the smoke feature quantity calculating means 30 outputs the brightness variance calculated based on the time series data of the average brightness or the standard deviation obtained from the brightness variance as the feature quantity related to smoke.

  The smoke determination means 40 described later has a luminance variance calculated based on the time series data of the average luminance calculated by the smoke feature amount calculation means 30 or a standard deviation obtained from the luminance variance within a predetermined range. It is possible to determine whether or not there is a high possibility that smoke has been generated.

[Extraction method 3: Smoke detection based on low frequency intensity of average luminance of pixels]
The smoke feature amount calculation means 30 generates time series data of average luminance for each target area in the same manner as the extraction method 2 described above. Then, the smoke feature quantity calculation means 30 performs Fourier transform on the generated time series data of the average luminance to calculate a power spectrum.

  Next, the smoke feature quantity calculating means 30 extracts a predetermined low frequency component from the power spectrum calculated based on the time series data of the average luminance, and calculates the intensity for the mode. In this way, the smoke feature quantity calculation means 30 outputs the intensity of the low frequency component calculated based on the time series data of the average luminance as a feature quantity related to smoke.

  And the smoke determination means 40 mentioned later has high possibility that smoke generate | occur | produced when the intensity | strength calculated based on the time series data of the average brightness | luminance calculated by the smoke feature-value calculation means 30 is below a predetermined value. It can be judged.

[Extraction method 4: Smoke detection based on difference average with reference image]
For each region, the smoke feature amount calculation means 30 obtains a luminance difference value between each pixel in the region and the corresponding pixel of the reference image stored in the image memory 10 in advance. Further, the smoke feature amount calculation means 30 calculates an average value of the luminance difference values for each region. In this way, the smoke feature quantity calculation means 30 outputs the average value of the luminance difference values as the feature quantity related to smoke.

  And the smoke determination means 40 mentioned later can judge that the possibility that smoke generate | occur | produced is high, when the average value calculated by the smoke feature-value calculation means 30 is larger than a predetermined value.

  Next, conventional labeling processing will be described with reference to the drawings. FIG. 3 is an explanatory diagram relating to a conventional labeling process performed by the smoke determination means 40. Based on the extraction result of the smoke feature quantity calculation unit 30, the smoke determination unit 40 groups small areas that are determined to have a high possibility that smoke has been generated.

  Specifically, the smoke determination means 40 attaches the same label to a set of small areas that are connected to each other among the small areas that are determined to have a high possibility of smoke. Group multiple small areas. FIG. 3 shows an example of grouping into eight groups. For example, the second grouped region (the region where “2” is described in FIG. 2) has three small regions. It will be combined.

  FIG. 4 is an explanatory diagram of a smoke detection process that takes into account the behavior unique to smoke in a conventional smoke detection apparatus. FIG. 4A is a mapping of a small area determined by the smoke determination unit 40 based on the detection result of the smoke feature amount calculation unit 30 according to the extraction methods 1 to 4 described above, which is highly likely to have occurred. Results are shown.

  Next, FIG.4 (b) has shown the result by which the smoke determination means 40 performed the labeling process with respect to the mapping result of Fig.4 (a). Here, a state of grouping into four regions L1 to L4 is shown. In the following description, it is assumed that the L3 region corresponds to a portion where smoke is generated, and the other L1, L2, and L4 portions are regions extracted due to a false detection factor.

  Next, FIG. 4 (c) shows that smoke determining means 40 excludes regions grouped in a size of 2 regions or less based on the labeling result of FIG. 4 (b), so that L1 and L3 Only the two areas show a state where it is finally determined that the possibility that smoke has occurred is high.

  As described above, the smoke determination means 40 performs the labeling process as shown in FIG. 4B, so that even if it is a small region temporarily determined that the possibility that smoke has been generated is high. Regions that are less than the size (corresponding to L2 and L4 in FIG. 4B) can be excluded. Thereby, in the conventional detection process, it is possible to consider the smoke-specific behavior that the smoke spreads beyond a predetermined area by performing the labeling process and connecting the small areas.

  However, the basis for performing the concatenation of the small regions is the feature amount extraction result as shown in FIG. Therefore, if there is a possibility that the extraction accuracy of the feature quantity itself may fluctuate depending on the observation situation and observation time, even if a labeling process based on the feature quantity is performed, it is not always possible to appropriately detect the smoke-specific behavior. It cannot be said that it is processing.

  In FIG. 4C, the L2 and L4 regions can be excluded, but the L1 region, which is the region extracted by the same false detection factor, is the same size as the L3 region that should be detected originally. Therefore, the labeling process based on the detection result of FIG. 4A cannot be completely removed.

  On the other hand, FIG. 4 (d) shows a case where the smoke determination means 40 has performed a process of excluding areas grouped with a size of 3 areas or less based on the labeling result of FIG. 4 (b). Yes. When such a process is performed, the L3 region that should be detected together with L1 that is the region extracted due to the erroneous detection factor is excluded from the detection target, and the L1 region is not erroneously detected. The L3 area to be detected is not detected.

  Therefore, next, an appropriate detection process in consideration of the smoke-specific behavior proposed in the present invention will be specifically described. FIG. 5 is a configuration diagram of the 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 storage unit 20, a smoke feature amount calculation unit 30, a smoke determination unit 40, and an adjacent similarity processing unit 50. Compared to the previous FIG. 1 showing the configuration of the conventional smoke detection apparatus, the smoke detection apparatus in the first embodiment is different in that it further includes an adjacent similarity processing means 50. Therefore, the function of the adjacent similarity processing unit 50 will be mainly described below.

  Adjacent similarity processing means 50 separates the smoke-specific first feature that the smoke spreads beyond a predetermined region, and the smoke-specific first feature that the characteristic related to the brightness is similar in the adjacent small region due to the spread of the smoke. By considering the second feature, the similarity processing of the small area is performed. For example, the adjacent similarity processing unit 50 can use the following texture features when performing the similarity processing based on the second feature.

  In the above equations (1) to (5) showing various texture characteristics, L means the number of gradations, l means the luminance, and P (l) means the frequency of the luminance l in the region. The adjacent similarity processing means 50 calculates a texture feature amount for each region by any one of the above formulas (1) to (5). In this calculation, not only the latest image taken but also the time series data consisting of a plurality of images taken in time series by the surveillance camera may be used for averaging. it can.

  In addition, the average value of the time series data of texture features is used, but in addition to this, it is also possible to use the amount of variation such as using the variance or gradient of the amount of change in the time series data. Good.

  The texture feature amount calculated in this way is a value depending on the luminance of each pixel in the region. Therefore, by obtaining the similarity of the texture feature amount to the adjacent region based on the texture feature amount of each region, the smoke-specific feature that “the feature related to brightness is similar in the adjacent small region due to the spread of smoke” The second feature can be considered.

  Specifically, the adjacent similarity processing unit 50 can obtain the similarity between the attention area and the adjacent area as follows. Here, as an example, a case will be described in which the similarity is calculated for four adjacent areas on the top, bottom, left, and right with respect to the attention area. In this case, the adjacent similarity processing unit 50 obtains the difference amounts ΔT1 to ΔT4 between the texture feature amount in the attention area and the texture feature amounts in the four adjacent areas.

  The adjacent similarity processing means 50 has a high similarity as the second feature when the difference amounts ΔT1 to ΔT4 obtained for the four adjacent areas are all within the predetermined difference amount. And the similarity of the attention area is mapped as “1”. On the other hand, if at least one of the difference amounts ΔT1 to ΔT4 obtained for the four adjacent areas exceeds the predetermined difference amount, the adjacent similarity processing unit 50 is similar as the second feature. The degree of similarity is determined to be low, and the similarity of the attention area is mapped as “0”. An example of this mapping result is as shown in FIG.

  Although the difference amounts ΔT1 to ΔT4 are all within the predetermined difference amount, another predetermined difference amount smaller than the predetermined difference amount is set. For example, three of the difference amounts are different from the predetermined difference amount. It may be determined that the degree of similarity is high when the amount is within the amount.

  Next, smoke detection processing in consideration of the first and second characteristics unique to smoke in the first embodiment of the present invention will be described with reference to the drawings. FIG. 6 is an explanatory diagram of smoke detection processing that takes into account the behavior unique to smoke in the smoke detection device according to Embodiment 1 of the present invention. FIG. 6A is a mapping of a small area that the smoke determination unit 40 determines that the possibility that smoke has been generated is high based on the extraction result of the smoke feature amount calculation unit 30 according to the extraction methods 1 to 4 described above. The result is shown and is the same as the previous FIG.

  Next, FIG. 6B shows a result of the labeling process performed by the smoke determination unit 40 on the mapping result of FIG. 6A, which is the same as FIG. 4B.

  Next, FIG. 6C shows a mapping result of an area determined by the adjacent similarity processing unit 50 as having a high similarity as the second feature. Here, the two regions T1 and T2 indicate a state in which it is determined that the similarity as the second feature is high.

  FIG. 6B shows a region where it is determined by the smoke determination unit 40 that there is a high possibility that there is smoke, but the region mapped in FIG. 6C is determined that there is a high possibility that there is smoke. It includes areas that were not done. FIG. 6C is a diagram in which only a region having a high degree of similarity is simply mapped on the basis of an extraction result or the like by the smoke feature amount calculation unit 30 regardless of the presence or absence of smoke.

  Next, in FIG. 6 (d), the smoke determination means 40 has the mapping result of FIG. 6 (b) (corresponding to the first candidate area) and the mapping result of FIG. 6 (c) (corresponding to the second candidate area). An area extracted as a result of performing the AND process is shown.

  Further, FIG. 6 (e) shows that only the L3 region is obtained by the smoke determining means 40 excluding regions grouped in a size of 2 regions or less based on the AND processing result of FIG. 6 (d). Is the state where it is finally determined that the possibility that smoke has occurred is high.

  In this way, the smoke determination unit 40 reflects the mapping result (corresponding to the first candidate region) in FIG. 6B reflecting the “first characteristic unique to smoke that the smoke spreads beyond a predetermined region”, and “ The mapping result (second candidate) of FIG. 6C by the adjacent similarity processing means 50 reflecting the second characteristic unique to smoke that the characteristics related to luminance are similar in adjacent small regions due to the spread of smoke. In consideration of both, it is possible to finally identify an area where smoke is highly likely to have occurred.

  As a result, the L3 region is reliably detected by the function of the mapping result of FIG. 6C reflecting the second feature, and the L1 region that has been erroneously detected in FIG. It can exclude from the object of the area | region where possibility that smoke generate | occur | produced is high, and can improve a detection accuracy.

  As described above, according to the first embodiment, the smoke-dependent texture feature is considered in consideration of the second feature unique to smoke that the feature related to brightness is similar in the adjacent small region due to the spread of smoke. An adjacent similarity processing unit is provided that can calculate the amount for each region and extract a region having a high similarity with the adjacent region based on the comparison result with the texture feature amount calculated in the adjacent region. As a result, both the first characteristic peculiar to smoke that the smoke spreads over a predetermined area and the second characteristic peculiar to smoke that the characteristic related to the brightness is similar in the adjacent small area due to the spread of the smoke. Therefore, it is possible to improve the detection accuracy of smoke generation based on the feature amount extraction.

  In the first embodiment described above, the case where the texture feature amount is calculated using any one of the above formulas (1) to (5) has been described, but the present invention is not limited to this. For example, the similarity can be determined using a plurality of texture feature quantities of the above formulas (1) to (5). Further, the texture feature amount is not limited to those of the above formulas (1) to (5), and the same effect can be expected as long as it is an evaluation scale indicating that the feature amount related to luminance is similar. .

  Further, the predetermined difference amount used when the adjacent similarity processing unit 50 determines the similarity need not be uniform in all regions. By storing an individual predetermined difference amount set in advance for each region in the storage unit, it is possible to perform an appropriate smoke detection process for each region according to the imaging target. More specifically, according to the imaging target, setting an individual predetermined difference amount suitable for each region from the texture feature amount of each region when it is assumed that smoke has actually occurred. Thus, an appropriate smoke detection process that suppresses false detection and non-detection can be realized.

  In order to determine the similarity of adjacent areas, the difference amount of the texture feature amount is used. However, the difference is not limited to the difference amount. For example, if the ratio of the texture feature amount is within a certain range, the texture feature amount may be similar. Well, the feature amount is not limited to the difference amount.

  DESCRIPTION OF SYMBOLS 1 Camera, 10 Image memory, 20 Storage part, 30 Smoke feature-value calculation means, 40 Smoke determination means, 50 Adjacent similarity processing means.

Claims (5)

Smoke feature quantity calculating means for extracting a feature quantity related to smoke for each of a plurality of regions set in an image captured by the monitoring camera;
In each of the plurality of regions, a storage unit in which a predetermined reference determination value for determining the presence or absence of smoke generation is stored in advance for each region;
In each of the plurality of regions, a predetermined reference determination value stored in the storage unit is taken out, compared with the feature amount extracted by the smoke feature amount calculation unit, and the plurality of regions based on the comparison result The first candidate area that is highly likely to generate smoke is identified for each of the first candidate areas, and the smoke may have been generated in the identified first candidate areas that are configured with a predetermined number of areas or more. In a smoke detection device comprising smoke determination means for determining a high area,
The image processing apparatus further includes an adjacent similarity processing unit that identifies each of a plurality of areas set in an image captured by the monitoring camera as a second candidate area having a high similarity in characteristics related to the adjacent area and brightness,
The smoke determination means determines that a region composed of a predetermined number of regions or more among the first candidate region and the second candidate region is a region that is highly likely to generate smoke. A smoke detection device characterized by that. Smoke feature quantity calculating means for extracting a feature quantity related to smoke for each of a plurality of regions set in an image captured by the monitoring camera;
In each of the plurality of regions, a storage unit in which a predetermined reference determination value for determining the presence or absence of smoke generation is stored in advance for each region;
In each of the plurality of regions, a predetermined reference determination value stored in the storage unit is taken out, compared with the feature amount extracted by the smoke feature amount calculation unit, and the plurality of regions based on the comparison result The first candidate area that is highly likely to generate smoke is identified for each of the first candidate areas, and the smoke may have been generated in the identified first candidate areas that are configured with a predetermined number of areas or more. In a smoke detection device comprising smoke determination means for determining a high area,
A texture feature amount that quantitatively indicates that a feature relating to luminance is similar to each of a plurality of regions set in an image captured by the monitoring camera is calculated, and a texture feature amount calculated for an adjacent region is calculated. Further comprising an adjacent similarity processing means for specifying a region where the calculated difference amount is within an allowable difference amount as a second candidate region having a high similarity in features related to luminance with the adjacent region,
The smoke determination means determines that a region composed of a predetermined number of regions or more among the first candidate region and the second candidate region is a region that is highly likely to generate smoke. A smoke detection device characterized by that. The smoke detection device according to claim 1 or 2,
The smoke detection apparatus according to claim 1, wherein similarity evaluation criteria are stored in advance in the storage unit for each of a plurality of regions. The smoke detection device according to claim 2,
In the storage unit, an individual allowable difference amount is stored in advance for each of the plurality of regions,
The adjacent similarity processing means identifies the second candidate region using the individual allowable difference amount. The smoke detection device. The smoke detection device according to claim 3 or 4,
The adjacent similarity processing means uses any one of an average value, contrast value, variance, energy, or entropy calculated based on the luminance of pixels included in each region as the texture feature amount. Smoke detection device.

Images