JP5167163B2 - Smoke detector - Google Patents

Description

  The present invention relates to a smoke detection device that detects the generation of smoke by performing image processing on an image captured by a surveillance camera, and in particular, generates smoke based on a comparison between a reference image and a current image. The present invention relates to a smoke detection device for detection.

  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 luminance difference image between the latest captured image and the reference image is calculated, and a region where a change has occurred The smoke is detected by extracting. (For example, refer to 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.
It is preferable that the image of the surveillance camera is always taken in an environment where the illumination conditions and the visual field are stable. However, depending on the monitoring location, there are not only such favorable conditions, but within the monitoring range, there are places where moving objects such as people come and go, or places where the sunshine conditions change with time.

  In the prior art, even for a luminance change partially generated in an area that is not suitable for monitoring, it is erroneously detected that smoke has been generated from the luminance difference result even though smoke has not been generated. There is a possibility. In addition, it is conceivable that an area that is not suitable for monitoring may be masked so that smoke detection is not performed, but this results in a limited monitoring range.

  The present invention has been made in order to solve the above-described problems, and identifies whether the change area in the monitoring area is caused by humans or smoke. It aims at obtaining the smoke detection apparatus which can implement | achieve a detection.

  The smoke detection apparatus according to the present invention provides a feature amount related to smoke for each of a plurality of areas set in common in a current image captured by a surveillance camera and a reference image stored in advance in a storage unit. Based on the smoke feature quantity calculating means to be extracted, the feature quantity in the current image and the feature quantity in the reference image extracted by the smoke feature quantity calculation means in each of the plurality of areas, and a predetermined reference judgment value A smoke detection device that includes a region-specific smoke determination unit that detects a region where smoke is highly likely to be generated, and each time a current image is captured, the region-specific smoke determination unit By mapping the areas that are determined to be areas where the possibility of smoke generation is high, the maximum separation distance between the mapped areas is n times or more (n is 1 or more). If lasted integer) or more, and further comprising a mapping determining section for determining that smoke has occurred in the current image captured.

  According to the smoke detection device of the present invention, the presence or absence of a region having a predetermined spread is obtained from the mapping result of a region where smoke is likely to be generated using the characteristics of smoke having a spread over a plurality of regions. By determining the above, it is possible to obtain a smoke detection device that can realize stable smoke detection by identifying human and smoke with high accuracy.

It is a block diagram of the smoke detection apparatus in Embodiment 1 of this invention. It is explanatory drawing of the area division | segmentation in Embodiment 1 of this invention. It is explanatory drawing of the distance of the dispersion | distribution between classes calculated by the smoke feature-value calculation means in Embodiment 1 of this invention. It is explanatory drawing of the determination method by the mapping determination means in 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.

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 storage unit 20, a smoke feature amount calculation unit 30, a region-specific smoke determination unit 40, and a mapping determination unit 50. Furthermore, a monitoring camera 1 that captures an image to be monitored is connected to the smoke detection device.

  The image memory 10 is a storage unit that stores a current image picked up by the monitoring camera 1 and stores an image in a normal state where no smoke or the like is generated as a reference image in advance. The smoke feature quantity calculation means 30 is a feature quantity related to smoke for each of a plurality of areas set in common in the current image captured by the monitoring camera 1 and in the reference image stored in advance in the image memory 10. To extract.

  FIG. 2 is an explanatory diagram of area division according to the first embodiment of the present invention. As shown in FIG. 2, the entire area of one screen imaged by the monitoring camera is divided into a plurality of areas determined in advance according to the monitoring target, the visual field size, and the like. Then, the smoke feature amount calculation means 30 extracts a feature amount related to smoke for each of the divided regions for each of the reference image and the current image.

  In the present invention, the case of calculating the distance between classes as an example of the feature quantity related to smoke will be described. FIG. 3 is an explanatory diagram of the distance between class variances calculated by the smoke feature quantity calculation unit 30 according to Embodiment 1 of the present invention. FIG. 3A shows the distance of interclass variance for a certain region of the reference image. FIG. 3B shows the distance between classes for one region of the current image when a moving object such as a person comes and goes. Furthermore, FIG. 3C shows the distance of interclass dispersion for one region of the current image when smoke is generated.

  First, a case where the luminance distribution of each pixel in a certain region is obtained based on a reference image without a person or smoke in FIG. 3A will be described. The luminance distribution of the entire region includes the first predetermined luminance distribution 61a configured in the range from the highest luminance pixel to the first predetermined number of pixels and the second predetermined number of pixels centered on the luminance having the largest number of pixels. Is included. Therefore, the smoke feature amount calculating means 30 obtains the respective centroids of the first luminance distribution 61a and the second luminance distribution 62a, and subsequently obtains a distance 63a between the centroids from these centroids, and FIG. ) Is defined as the distance between classes in the image.

  Next, in FIG. 3B, a case will be described in which the luminance distribution of each pixel in the same area as the reference image is obtained based on the current image when a moving object such as a person comes and goes. Here, the luminance distribution of the entire region includes the first luminance distribution 61b configured from the highest luminance pixel to the first predetermined number of pixels and the second luminance centering on the luminance having the largest number of pixels. A second luminance distribution 62b composed of a predetermined number of pixels is included. Therefore, the smoke feature amount calculation means 30 obtains the distance 63b between the centers of gravity of the first luminance distribution 61b and the second luminance distribution 62b, and defines it as the distance between the classes in the image of FIG.

  Similarly, in FIG. 3C, when the luminance distribution of each pixel in the area of the same portion as the reference image is obtained based on the current image when smoke is generated, the luminance distribution of the entire area has the highest luminance. A first luminance distribution 61c configured from a high pixel to a first predetermined number of pixels and a second luminance distribution 62c configured by a second predetermined number of pixels centering on the luminance having the largest number of pixels. It is included. Therefore, the smoke feature amount calculating means 30 obtains the distance 63c between the centers of gravity of the first luminance distribution 61c and the second luminance distribution 62c, and defines the distance between the classes in the image of FIG.

  Here, paying attention to the change in luminance distribution, as shown in FIG. 3B, when a moving object such as a person comes and goes within the area, the size of one area is larger than the size of the person. Therefore, the luminance change factor is generated only in a part of the area with respect to the size of the entire area shown in FIG. 2, and the change in the luminance distribution as the entire area is small. As a result, the distance 63b between the centers of gravity is considered to be a value close to the distance 63a between the centers of gravity in the reference image.

  On the other hand, as shown in FIG. 3 (c), when smoke is generated at a place where the camera takes an image, the smoke spreads over the entire screen, so that the luminance change factor in the entire area shown in FIG. As a result, the luminance distribution of the entire region changes. Further, due to the influence of smoke, the average luminance of the entire region is lowered, there is almost no difference between the high luminance portion and the low luminance portion, and the intercentroid distance 63c is shorter than the intercentroid distance 63a in the reference image. .

  Therefore, the area-specific smoke determination unit 40 determines that the distance between the centroids (corresponding to 63b and 63c) in the current image extracted by the smoke feature amount calculation unit 30 in each of the plurality of areas is the distance between the centroids in the reference image (63a). If it is shorter than the predetermined ratio, it can be determined that there is a high possibility that smoke has been generated.

  In this way, by using the characteristics of smoke that spreads over multiple areas and the characteristics of moving objects such as people that cause local brightness changes, from the change in the distance between the centers of gravity based on the brightness distribution, It is possible to identify an area where smoke is likely to be generated.

  Next, the mapping determination unit 50 will be described. As described above, the smoke that is the detection target has a characteristic that spreads over a plurality of regions. That is, when the monitoring image is divided into a plurality of images as shown in FIG. 2, the plurality of areas are determined to be areas where the possibility of smoke is high by the area-specific smoke determination means 40. Therefore, the mapping determination means 50 maps the area identified as having a high possibility of the occurrence of smoke by the area-specific smoke determination means 40 on the screen, and finally the smoke is determined from the extent of the mapping result. The presence or absence of occurrence is judged.

  FIG. 4 is an explanatory diagram of a determination method performed by the mapping determination unit 50 according to Embodiment 1 of the present invention. The mapping determination means 50 maps the areas specified by the area-specific smoke determination means 40 as having a high possibility that smoke has been generated for each of the plurality of divided areas as shown in FIG.

  FIG. 4A is a mapping result when a moving object such as a person enters the monitoring target screen, and FIG. 4B is a mapping result when smoke is generated in the monitoring target screen. is there. 4 (a) and 4 (b), the blacked out area corresponds to the area specified by the area-specific smoke determining means 40 as having a high possibility that smoke has been generated. Normally, the area-specific smoke determination means 40 does not detect a person, but assumes a case where it is detected erroneously depending on the position of the person, the number of persons, or how to move.

  Even if a moving object such as a person invades, even if it is mapped that smoke is likely to be generated, the mapped area in this case is likely to remain within a certain local range. In FIG. 4A, the maximum separation distance between the mapped areas, that is, the distance between the areas farthest in the vertical direction and the horizontal direction in the mapped area is two areas in the horizontal direction. A case where there are four areas in the direction is illustrated.

  On the other hand, in the case where the smoke is actually generated and the mapping is performed on the assumption that the smoke is likely to be generated, the mapped area is likely to have a wide range due to the characteristics of the smoke. FIG. 4B illustrates a case where the maximum separation distance between the mapped areas is 7 areas in the horizontal direction and 4 areas in the vertical direction.

  Therefore, the mapping determination means 50 can finally determine whether or not smoke is generated by determining whether or not the maximum separation distance between the mapped areas has become a predetermined value or more. For example, it is possible to determine that smoke is generated when the predetermined values in the horizontal direction and the vertical direction are set separately and the conditions are satisfied in both directions. In the example of FIG. 4, smoke is generated in the case of FIG. 4A in which a moving object such as a person has entered by setting the predetermined value to be 5 areas or more in the horizontal direction and 3 areas or more in the vertical direction. It can be determined that smoke has been generated only in the case of FIG.

  Alternatively, it is possible to determine that smoke is generated when one predetermined value is set and the condition is satisfied in either the horizontal direction or the vertical direction. In the example of FIG. 4, smoke is generated in the case of FIG. 4A in which a moving object such as a person has entered by setting a predetermined value common to the horizontal direction and the vertical direction to 5 areas or more. Only in the case of FIG. 4B in which smoke is actually generated, it can be determined that smoke has been generated. Alternatively, in the values in the vertical direction and the horizontal direction, a value that has a large value may be compared with a predetermined value, and if the value is larger than the predetermined value, it may be determined that there is smoke.

Furthermore, the following improvement measures can be taken as a method for improving the detection accuracy of smoke generation.
First, as a first improvement measure, the mapping determination unit 50 determines that the maximum separation distance between the mapped regions is equal to or greater than a predetermined value twice based on the mapping result for the current image obtained in time series. If it continues, it can be finally determined that smoke has been generated. In this way, by adding time-series information, it is possible to reduce the possibility of erroneously detecting a transient disturbance as smoke.

  As a second improvement measure, the mapping determination unit 50 excludes an isolated region in which the adjacent region is not mapped from the mapping target region when mapping. Here, the isolated region is a region in which the surrounding eight regions are not determined to have any smoke in the target region, such as the region of the human leg portion in FIG. I mean. When the state in which the maximum separation distance between the regions in the mapping result after excluding the isolated region has become a predetermined value or more continues once or more, it can be finally determined that smoke has been generated. In this way, by removing the isolated region, it is possible to make an appropriate determination using the characteristics of smoke that spreads over a plurality of regions, and it is possible to improve smoke detection accuracy.

  Further, as a third improvement measure, the smoke feature amount calculation means 30 sequentially stores the distance between the centers of gravity calculated each time the current image is captured in the storage unit 20 as time series data. Then, the area-specific smoke determination means 40 calculates a moving average of the distance between the center of gravity based on the time series data of the distance between the center of gravity calculated for the current image stored in the storage unit 20, and calculates the distance between the center of gravity When the moving average of is shorter than a predetermined ratio compared to the distance between the centers of gravity calculated with respect to the reference image, it can be determined that there is a high possibility that smoke has occurred. Thus, by adopting the moving average, it is possible to reduce the possibility of erroneously detecting a transient disturbance as smoke.

  As described above, according to the first embodiment, a region having a predetermined spread is obtained from a mapping result of a region where smoke is likely to be generated using the characteristics of smoke having a spread over a plurality of regions. By determining whether or not there is a smoke detection device, it is possible to obtain a smoke detection device that reduces false detection and improves detection accuracy.

  In addition, in order to identify a region where smoke is highly likely to be generated, the distance between the centers of gravity based on the luminance distribution is employed as the feature amount regarding smoke. As a result, it is possible to reliably extract only the overall luminance change factor generated over a plurality of areas due to the generation of smoke without erroneously detecting a local luminance change factor that causes the movement of a moving object such as a person. be able to.

  In the above description, the distance between the center of gravity has been described as the feature quantity related to smoke. However, as the feature quantity in the present invention, the contrast ratio between both luminance distributions is calculated as the brightness gradient instead of the distance between the center of gravity. Is also possible. That is, the ratio of both the contrast ratio of the reference image and the contrast ratio of the current image is calculated as a luminance gradient, and when this luminance gradient becomes a gradient within a predetermined range, the area-specific smoke determination means 40 It can be determined that the region is likely to have occurred.

  Even when such a luminance gradient is adopted as a feature quantity related to smoke, the above-described first to third improvement measures can be applied.

In the present embodiment, the distance between the centroids from the luminance distribution is calculated as the smoke feature amount using the reference image, but it is not particularly necessary to use the reference image, and the average luminance of the latest image is not required. The value may be calculated as a feature amount of smoke.
Further, regarding the calculation of the maximum separation distance of the region, it may be calculated as the Euclidean distance.

  DESCRIPTION OF SYMBOLS 1 Surveillance camera, 10 image memory, 20 memory | storage part, 30 Smoke feature-value calculation means, 40 Area | region smoke determination means, 50 Mapping determination means, 61a, 61b, 61c 1st luminance distribution, 62a, 62b, 62c 2nd Luminance distribution, 63a, 63b, 63c Distance between centroids.

Claims (5)

Smoke feature quantity calculating means for extracting feature quantities related to smoke for each of a plurality of areas set in the current image captured by the monitoring camera;
A region where smoke is highly likely to occur in each of the plurality of regions based on the feature amount in the current image extracted by the smoke feature amount calculation unit and a predetermined reference determination value in each of the plurality of regions. A smoke detection device comprising a region-specific smoke determination means for detecting
Each time the current image is captured, the area determined as the area where smoke is likely to be generated by the area smoke determination unit is mapped for all of the plurality of areas, and the maximum separation between the mapped areas is mapped. Mapping determination means for determining that smoke has occurred in the captured current image when the state where the distance is equal to or greater than the predetermined value continues n times (where n is an integer equal to or greater than 1). A smoke detector. The smoke detection device according to claim 1,
The smoke feature amount calculation means extracts a feature amount related to smoke for each of a plurality of regions set in the reference image that is set in common with the region in the current image and stored in advance in the storage unit. ,
The region-specific smoke determination means is based on the feature amount in the current image and the feature amount in the reference image extracted by the smoke feature amount calculation means in each of the plurality of regions, and a predetermined reference determination value. A smoke detection device that detects an area where smoke is likely to be generated in each of a plurality of areas. The smoke detection device according to claim 2,
The mapping determination unit excludes an isolated region in which an adjacent region is not mapped from a mapping target region when mapping a region that is determined to be a region where smoke is likely to be generated by the region-specific smoke determination unit, When the state where the maximum separation distance between the mapped areas after excluding the isolated area has become a predetermined value or more continues n times (n is an integer of 1 or more), in the captured current image A smoke detection device that judges that smoke has been generated. The smoke detection device according to claim 2 or 3,
The smoke feature amount calculating unit obtains a luminance distribution in each of the plurality of regions in the current image and the reference image, and performs a first predetermined determination from a pixel having the highest luminance in the luminance distribution. A first luminance distribution composed of a range of the number of pixels and a second luminance distribution composed of a second predetermined number of pixels centering on the luminance having the largest number of pixels are extracted, and between the two luminance distributions Calculate the distance between the center of gravity,
The area-specific smoke determination means is configured such that, in each of the plurality of areas, the center-of-gravity distance calculated for the current image is shorter than the center-of-gravity distance calculated for the reference image by a predetermined ratio or more. When this happens, it is determined that the area is likely to generate smoke. The smoke detection device according to claim 2 or 3,
The smoke feature amount calculating means calculates a contrast ratio between both luminance distributions as a luminance gradient instead of the distance between the center of gravity,
The area-specific smoke determination means determines that smoke is highly likely to be generated when the calculated brightness gradient is within a predetermined range in each of the plurality of areas. Smoke detection device.

Images