Smoke detection device, method for detecting smoke from a fire, and computer program

Abstract

Smoke detection device (1) for detecting smoke (9) from a fire in a monitored area (U) with a camera interface (5) for receiving a sequence of images (3) with temporally successive individual images (4) from a surveillance camera (2), wherein the individual images (4) show the surveillance area (U), with a first evaluation module (10a), wherein the first evaluation module (10a) performs a first motion estimation to determine a movement of a smoke section (7) in the monitoring area (U) on the basis of two individual images (4) of the image sequence (3), wherein the two individual images (4) have a first time interval (dt1) between each other, characterized by a second evaluation module (10b), wherein the second evaluation module (10b) performs a second motion estimation to determine a movement of a smoke section (7) in the monitoring area (U) based on two individual images (4) of the image sequence (3), wherein the two individual images (4) have a second time interval (dt2) between each other.

Description

State of the art

[0001] The invention relates to a smoke detection device for detecting smoke from a fire in a monitored area, comprising a camera interface for receiving a sequence of temporally successive individual images from a surveillance camera, wherein the individual images depict the monitored area, and a first evaluation module, wherein the first evaluation module performs a first motion estimation to determine the movement of a smoke compartment in the monitored area based on two individual images of the image sequence, wherein the two individual images have a first time interval between them. The invention further relates to a method for detecting a fire, preferably with the smoke detection device, and a computer program.

[0002] Multiple measurement principles have been implemented for fire detection. For example, it is possible to automatically detect a fire via its temperature using a temperature sensor, via smoke particles using a scattered light sensor, or even video-based using digital image processing.

[0003] Document WO 2008 / 037 293 A1, which likely represents the closest state of the art, proposes a method and a device for detecting smoke using a video camera. In this method, at least one video image is recorded by a video camera monitoring an area. Subsequently, at least one moving area within this video image is examined for the probable presence of smoke by determining its direction and size. If the test is positive, at least a portion of this moving area is evaluated based on at least one piece of information characteristic of smoke.

[0004] A smoke detection method is disclosed in US patent 2008 / 0 137 906 A1. EP 2 372 642 A1 discloses a method and a system for detecting moving objects. DE 10 2007 041 893 A1 discloses a method for detecting and / or tracking moving objects in a surveillance scene with intruders. DE 10 2013 004 073 A1 discloses a method for detecting activities captured in video streams. Disclosure of the invention

[0005] The invention proposes a smoke detection device with the features of claim 1, a method with the features of claim 10, and a computer program with the features of claim 11. Preferred or advantageous embodiments of the invention are described in the dependent claims, the following description, and the accompanying figures.

[0006] The invention proposes a smoke detection device suitable and / or designed for detecting smoke from a fire within a monitored area. The smoke is generated by smoke emissions, particularly combustion products of the fire. The monitored area can be an enclosed space, such as a hall, or alternatively, an open area. Optionally, the smoke detection device is designed to trigger a fire alarm based on the smoke detection. In particular, the smoke detection device is designed as an automated fire alarm.

[0007] Fire detection is thus achieved indirectly through the detection of smoke or smoke fragments, in particular through the movement of the smoke or smoke fragment(s) within the monitored area. Overall, the smoke detection device is based on the premise that a fire leads to smoke development, with the smoke developing and, in particular, moving within the monitored area. Since smoke typically has a higher temperature than its surroundings, it rises vertically and / or at right angles. It has been observed that smoke moves at a typical speed of approximately 0.5 meters per second, and this movement of the smoke, especially its speed and direction, is characteristic of the smoke and thus constitutes a feature for its detection.

[0008] The smoke detection device is specifically designed as a digital data processing device, such as a computer, a microcontroller, etc. The smoke detection device has a camera interface configured to receive a sequence of images from a surveillance camera. The camera interface can be internal or external. The image sequence can be transmitted wirelessly or via cable, particularly over a network connection. The individual images depict the monitored area. Preferably, each individual image shows the same section of the monitored area and is thus captured by a stationary surveillance camera with fixed lens settings. The individual images can be black and white or color images.In particular, the sequence of images with the individual pictures comes from a single surveillance camera.

[0009] The smoke detection device comprises a first evaluation module, which is preferably implemented as a software module. Alternatively, the function of the first evaluation module can also be implemented in hardware.

[0010] The first evaluation module is designed, both programmatically and / or circuit-wise, to perform an initial motion estimation to determine the movement of a smoke fragment within the monitored area based on two individual frames of the image sequence. Specifically, the first evaluation module is designed to detect the movement of the smoke fragment within the monitored area based on the initial time interval between two individual frames of the image sequence. The two individual frames have an initial time interval between them. In particular, the first evaluation module determines the movement of a smoke fragment within the monitored area between the two individual frames. The smoke fragment can be a single pixel segment of the smoke in the individual frames or a spatially extended sub-area of ​​the smoke.In continuous operation of the smoke detection device, it is preferably provided that the determination of the movement of the smoke section in the monitoring area is continued via further individual images, whereby, however, the movement between the individual images is determined with the first time interval between them.

[0011] The initial time interval is chosen to allow for reliable detection of the smoke's movement and / or ascent, based on the smoke's inherent motion and / or upward movement described above. Specifically, the initial time interval is chosen so that the smoke's movement and / or the smoke's upward movement within the monitored area results in a sufficient change in its position within the individual frames—particularly in the image coordinates. For example, it is advantageous for determining the movement if, within the initial time interval, the smoke or smoke's upward movement changes by at least one pixel in the image coordinates at the aforementioned rate of ascent. Conversely, determination becomes more difficult if the initial time interval is chosen so that the smoke or smoke's upward movement...If the smoke segment between the individual images has traveled more than 10 or 20 pixels within the first time interval, then finding and assigning (corresponding to) the smoke segments between the individual images becomes difficult.

[0012] Within the scope of the invention, it is proposed that the smoke detection device comprises a second evaluation module, which can be configured in the same variations or embodiments as the first evaluation module. The second evaluation module is configured, programmatically and / or circuit-wise, to perform a second motion estimation to determine the movement of a potentially further smoke section within the monitored area, based on two individual images from the image sequence, wherein the two individual images have a second time interval between them. The first and second time intervals have different durations.

[0013] One aspect of the invention is that the previously described number of pixels in the individual frames, which the smoke segment travels between two frames, depends strongly on the distance of the smoke from the surveillance camera. With regard to the mapping of the surveillance area onto the individual frames, smoke rising in the immediate vicinity leads to a significant and evaluable change in the image position of the smoke segment within a shorter time interval, whereas smoke at a greater distance requires considerably more time to significantly change the image position in the individual frames.

[0014] Based on this consideration, it is therefore advantageous to use at least or exactly two evaluation modules that analyze the image sequence in parallel, using different time intervals between the individual images in order to be able to analyze different distance ranges in the monitoring area in an adapted manner.

[0015] In one possible embodiment of the invention, the first time interval is thus shorter than the second time interval, with the first evaluation module being designed for fire detection in the near range and the second evaluation module for detection in the far range. The advantage of this embodiment is that the smoke detection device is adapted for two different distance ranges and can therefore operate much more accurately than if only a single time interval were used. For example, the second time interval is at least two or three times longer than the first time interval.

[0016] In one possible embodiment of the invention, the smoke detection device comprises the surveillance camera, wherein the first time interval is configured such that a change in height of a smoke segment in the near range at a rate of 0.5 meters per second corresponds to a change in one pixel, and / or the second time interval is configured such that a change in height of a smoke segment in the far range at a rate of 0.5 meters per second corresponds to a change in one pixel. For example, the near range is defined as between 5 and 25 meters and the far range as between 20 and 45 meters. The proposed embodiment is to be understood as exemplary and can be adapted to real measurement situations.

[0017] It is particularly preferred that the first and second evaluation modules operate in parallel and / or are configured as two parallel processing lines, especially two parallel pipelines. The smoke detection device is optionally limited to precisely these two processing lines and / or pipelines or may include further processing lines and / or pipelines.

[0018] In one possible embodiment of the invention, motion estimation is implemented as a block-matching method. A particularly preferred method for motion estimation is the determination of a first and second optical flow field. The optical flow field is understood to be a vector field of the velocity vectors of visible points, in particular smoke segments, of the monitored area projected into the image plane of the individual images in the reference frame of the imaging optics of the surveillance camera. With reference to a pixel segment of the smoke as a smoke segment in the individual images, or to a spatially extended sub-region of the smoke as a smoke segment, the optical flow field comprises a vector that indicates the velocity and direction of the pixel segment or spatially extended sub-region of the smoke that this area has traveled between the two individual images.The calculation of optical flow in individual images is known to those skilled in the art and is widely used in digital image processing, so that the smoke detection device can be implemented very reliably by means of the calculation of optical flow.

[0019] In one possible embodiment of the invention, the evaluation module(s) for motion estimation are designed based on individual images, wherein the complete object content is represented in the individual images. Thus, stationary, quasi-stationary, and moving objects are present in the individual images when they are processed by the evaluation module(s).

[0020] In a possible further development or alternative of the invention, the smoke detection device comprises a difference module for eliminating a scene background, particularly one that is static or quasi-static and contains static and / or quasi-static objects. For example, the difference module can model the scene background by averaging previous frames and subtracting the modeled scene background from the current frames. In this way, background-cleaned intermediate frames, in particular difference frames, can be generated, with the evaluation module(s) for motion estimation being based on these background-cleaned intermediate frames. Alternatively, difference frames between two temporally successive frames can also be generated as the background-cleaned intermediate frames.The advantage of this advanced training lies in the fact that, by using background-cleaned intermediate images, thin and / or transparent smoke, in particular, can be detected more effectively, thus allowing for better determination of any movement. It is possible for at least two evaluation modules to both work with background-cleaned individual images, or for only a subset of the majority of evaluation modules to work with background-cleaned individual images, while another subset works with complete individual images.

[0021] In a preferred embodiment of the invention, the smoke detection device comprises a detection module, wherein the detection module is configured to detect the smoke and thus the fire based on motion estimation. If the motion estimation includes, for example, the determination of an optical flow field, the smoke can be detected as a feature by a characteristic collection of flow vectors, which are vertically or substantially vertically oriented and have a predetermined length, since the smoke is usually arranged in a continuous area, rises vertically upwards, and has a uniform rate of ascent.

[0022] The detection module is specifically designed to detect smoke, and thus a fire, in the immediate vicinity based on the motion estimation of the first evaluation module, and smoke, and thus a fire, in the distant vicinity based on the motion estimation of the second evaluation module. In particular, the detections occur simultaneously or in parallel, but independently of each other.

[0023] In particular, the detection module is designed to determine a smoke probability based on motion estimation and, for example, detect smoke when a predefined probability is exceeded. Optionally, it can also be provided that further parameters, such as the size of the area in the individual images with a sufficient smoke probability and / or the duration of the detected smoke probability, are included in the calculation.

[0024] A further aspect of the invention relates to a method for detecting smoke in a monitored area, which is preferably implemented by the smoke detection device as previously described or according to one of the preceding claims. In this method, a sequence of images with temporally successive individual images is evaluated, the individual images depicting the monitored area. A first motion estimation is performed, whereby movement in the monitored area is determined based on two individual images of the image sequence, wherein the two individual images have a first temporal interval between them. Preferably in parallel and / or overlapping with this, a second motion estimation is performed, wherein movement in the monitored area is determined based on two individual images of the image sequence, wherein the two individual images have a second, different temporal interval between them.Optionally, in a subsequent step, smoke and thus fire detection is carried out based on the first and second motion estimates.

[0025] Another object of the invention is a computer program with program code means for carrying out the previously described method, preferably on a smoke detection device or on a data processing system.

[0026] Further features, advantages, and effects of the invention will become apparent from the following description of a preferred embodiment of the invention and the accompanying figures. These figures show: Fig. 1 a schematic block representation of a smoke detection device as an embodiment of the invention; Fig. 2 a single image of the smoke detection device in the Fig. 1. To explain the basic operating principle of the smoke detection device in the Fig. 1; Fig. 3. An illustration to explain the procedure for operating the smoke detection device in the Fig. 1.

[0027] In the Fig. Figure 1 shows a smoke detection device 1 as a first embodiment of the invention in a highly schematic representation. The smoke detection device 1 can optionally include a surveillance camera 2, but this can also be designed as a separate component and connected to the smoke detection device 1.

[0028] The surveillance camera 2 provides a sequence of images 3 with temporally successive individual frames 4 and transmits this to the smoke detection device 1. The smoke detection device 1 has a camera interface 5 for receiving the image sequence 3. The camera interface 5 can be configured as a wired, internal, external, or wireless communication interface. The image sequence 3, for example, contains individual frames 4 with a frequency of 30 Hertz or 60 Hertz.

[0029] Surveillance camera 2 is directed at a surveillance area U, where in this example a near area N and a far area F are represented within surveillance area U, with the far area F being further away from surveillance camera 2 than the near area N. In the near area N, a person 6 and a smoke flare 9 are schematically depicted as smoke with a smoke section 7 to visualize the scale. Similarly, in the far area F, a person 6 and a smoke flare 9 are also shown as smoke with a smoke section 7. The near area is located, for example, at a distance of 5 to 25 meters from surveillance camera 2. The far area F, on the other hand, is located at a distance of, for example, 20 to 45 meters from surveillance camera 2.

[0030] The Fig. Figure 2 shows a highly schematic representation of a single image 4 of the monitored area U. It can be seen that the person 6 and the smoke compartment 7 in the near field N are depicted larger in single image 4 than the person 6 and the smoke compartment 7 in the far field F. This is due to the imaging, in particular the imaging optics of the surveillance camera 2.

[0031] The smoke detection device 1 is designed to detect smoke, such as the smoke flare 9 of a fire, by determining the movement of smoke compartments 7 within the monitoring area U into or from the smoke compartment 7. The procedure is based on a movement estimation to determine the movement within the monitoring area U. In the smoke detection device 1 in the Fig. 1. The optical flow, in particular one or more optical flow fields, is calculated as a motion estimation. When calculating the optical flow, displacements of image segments or pixels from one frame 4 to the next frame 4 are calculated and visualized, for example, by vectors 8. In the Fig. 2 are possible vectors 8 for both the near range N and the far range F for the movement of the smoke section 7 and other smoke sections.

[0032] It is assumed that both smoke sections 7 move vertically at the same rate of ascent of approximately 0.5 meters per second. The diagram shows that, for a uniform evaluation of the optical flow in the single image 4, the flow vectors 8 of the near field N and the far field F differ significantly in length. Specifically, the vectors 8 in the near field N are considerably longer than the vectors 8 in the far field F. The optical flow method is limited when evaluating single images 4 because changes in the position of pixels from one single image 4 to the next should be at least on the order of one pixel.On the other hand, the changes should not be too large, otherwise the pixels in successive individual images 4 can no longer be assigned to each other, so that no vectors 8 can be derived.

[0033] Therefore, the time interval between two individual images 4 for determining the optical flow must be adjusted to the distance of the smoke section 7 from the surveillance camera 2. In particular, the time interval between two individual images 4 for determining the optical flow or for determining the vectors 8 must be such that the length of the vectors 8 in the individual image 4 is greater than 1 pixel and less than, for example, 10 or 20 pixels. As a result, this consideration leads to the conclusion that when creating an optical flow field with a uniform time interval between the individual images 4, a compromise is always necessary between the best possible processing of smoke sections 7 in the near field N and smoke sections 7 in the far field F.

[0034] In the Fig. Figure 1 shows the smoke detection device 1 as an embodiment of the invention, wherein it has two evaluation modules 10a, b which process the image sequence 3 or a selection of individual images 4 of the image sequence 3 in parallel to each other, wherein the first evaluation module 10a is tuned for the near range N and the second evaluation module 10b for the far range F. The distance-dependent tuning is based on the Fig. 3 explained: In the Fig. Figure 3 shows a highly schematic representation of the temporal progression t of the evaluation of the image sequence 3 of the individual images 4. The first evaluation module 10a implements a first processing line 11a, whereby the vectors 8 are each determined from two individual images 4, which have an initial time interval dt1 for the subsequent area N. It is intended that the vectors 8 are calculated at least three times per second in order to obtain a reliable result. Furthermore, the Fig. 3. A second processing line 11b, which is implemented by the second evaluation module 10b. In the second processing line 11b, the vectors 8 are also calculated, but a second, larger time interval dt2 is provided between two individual images 4. The calculation in the second processing line 11b is also preferably performed at least three times per second.

[0035] By choosing different time intervals dt1, dt2 between the individual images 4 in the processing lines 11a, b, the expected length of the vectors 8 is adjusted and / or optimally matched to the calculation of the optical flow.

[0036] Thus, the shorter initial time intervals dt1 allow the lengths of vectors 8 to be adjusted or even set to the same level in the far range F compared to the longer second time intervals dt2 of vectors 8. This allows the lengths of vectors 8 for both the near range N and the far range F to be advantageously matched for calculating the optical flux. As can be seen from the Fig. As can be seen in Figure 3, the two processing lines 11a and 11b operate in parallel or at least overlapping in time. The results from the first and second evaluation modules 10a and 11b are transferred to a detection module 12, which detects the smoke compartment 7 and thus a fire based on the optical flow field. In principle, it is possible to use further evaluation modules 10a and 11b to advantageously coordinate further distance ranges with further time intervals.

[0037] An exemplary configuration of the smoke detection device 1, with a total opening angle alpha of 70° for the surveillance camera 2, specifies the use of a first time interval dt1 of 80 milliseconds for a near range N of 5 meters to 25 meters and a second time interval dt2 of 300 milliseconds for a far range F of 20 meters to 45 meters. Thus, significantly larger distance ranges can be covered by combining the parallel optical flow algorithms. Further processing of the optical flow results only needs to be performed once in the detection module 12.

[0038] It is generally possible to apply the evaluation modules 10a and 10b to the unprocessed or largely unprocessed individual images 4. A possible enhancement involves prefacing evaluation modules 10a and 10b, or just one of them, with a difference module (not shown). This difference module removes a static scene background from the individual images 4, allowing the evaluation modules 10a or 10b to operate on background-cleaned images. This enhancement has the advantage of improving the clarity and recognizability of smoke sections 7, particularly in cases of transparent smoke.

Claims

[1] Smoke detection device (1) for detecting smoke (9) from a fire in a monitored area (U) with a camera interface (5) for receiving a sequence of images (3) with temporally successive individual images (4) from a surveillance camera (2), wherein the individual images (4) show the surveillance area (U), with a first evaluation module (10a), wherein the first evaluation module (10a) performs a first motion estimation to determine a movement of a smoke section (7) in the monitoring area (U) on the basis of two individual images (4) of the image sequence (3), wherein the two individual images (4) have a first time interval (dt1) between each other, characterized by a second evaluation module (10b), wherein the second evaluation module (10b) performs a second motion estimation to determine a movement of a smoke section (7) in the monitoring area (U) based on two individual images (4) of the image sequence (3), wherein the two individual images (4) have a second time interval (dt2) between each other. [2] Smoke detection device (1) according to claim 1, characterized by , that the first time interval (dt1) is smaller than the second time interval (dt2), wherein the first evaluation module (10a) is designed for the detection of the smoke (9) in a near range (N) and the second evaluation module (10b) is designed for the detection of the smoke (9) in a far range (F). [3] Smoke detection device (1) according to any one of the preceding claims, characterized bythe surveillance camera (2), wherein the first time interval (dt1) is configured such that a change in height of the smoke section (7) in the near range (N) with a rate of ascent of 0.5 m / s corresponds to the change of a pixel in the single image (4) and / or that the second time interval (dt2) is configured such that a change in height of the smoke section (7) in the far range (F) with a rate of ascent of 0.5 m / s corresponds to the change of a pixel in the single image (4). [4] Smoke detection device (1) according to any one of the preceding claims, characterized by that the first and second evaluation modules (10a,b) operate in parallel and / or are configured as two parallel processing lines (11a,b). [5] Smoke detection device (1) according to any one of the preceding claims, characterized by, that the first and / or the second evaluation module (10a,b) is or are designed to determine a first or second optical flow field as a motion estimator to determine the motion of the smoke section (7). [6] Smoke detection device (1) according to any one of the preceding claims, characterized by , that the first and / or the second evaluation module (10a,b) are designed for motion estimation based on the complete individual images (4). [7] Smoke detection device (1) according to any one of the preceding claims, characterized by a difference module for eliminating a scene background of the monitored area (U) in the individual images (4) and for creating background-cleaned individual images, wherein the first and / or the second evaluation module (10a,b) is designed for motion estimation based on the background-cleaned individual images. [8] Smoke detection device (1) according to one of the preceding claims, characterized by a detection module (12), wherein the detection module (12) is configured to detect the smoke (9) based on motion estimation. [9] Smoke detection device (1) according to claim 8, characterized by , that the detection module (12) is designed to detect smoke (9) in the near range (N) based on the motion estimation of the first evaluation module (10a) and smoke (9) in the far range (F) based on the motion estimation of the second evaluation module (10b). [10] Method for detecting smoke (9) from a fire in a monitoring area (U), preferably with a smoke detection device (1) according to one of the preceding claims, wherein an image sequence (3) with temporally successive individual images (4) is evaluated, wherein the individual images (4) show the monitoring area (U), wherein a first motion estimation is carried out, wherein a movement in the monitoring area (U) is determined on the basis of two individual images (4) of the image sequence (3), wherein the two individual images (4) have a first time interval (dt1) from each other, and wherein a second motion estimation is carried out, wherein a movement in the monitoring area (U) is determined on the basis of two individual images (4) of the image sequence (3), wherein the two individual images (4) have a second, different time interval (dt2) from each other. [11] Computer program with program code means to perform all steps of a method according to claim 10 when the program is executed on a computer and / or on a smoke detection device (1) of any one of claims 1 to 9.

Images