Method for verifying fire detection in a scene and evaluation device

Abstract

The invention relates to a method for verifying fire detection in a monitoring area (12), wherein an image of the monitoring area (12) is recorded at at least two points in time, wherein a suspected area (23) for smoke is determined in the image and a color difference of the suspected area (23) between the at least two points in time is determined, wherein a change in smoke density is determined taking into account the determined color difference and wherein a positive verification of a fire only takes place if the change in smoke density exceeds a prescribed minimum.

Description

State of the art

[0001] It is already known to detect smoke using optical sensors, where smoke penetrates the sensor and influences an optical measuring path. In cases where the smoke does not reach the concentration at the sensor required to trigger the sensor, particularly in the early stages of a fire, it is known, for example from DE 10 2016 207 705 A1, to use a camera to record a sequence of images of a monitored area and to detect smoke based on the specific movement patterns caused by the smoke in the recorded sequence of images by means of appropriate evaluation in successive images. This makes it possible to detect a fire early, even outdoors or in large buildings, when the smoke concentration required for detection has not yet been reached at an optical smoke sensor. Disclosure of the inventionAdvantages of the invention

[0002] The inventive method for verifying fire detection in a monitored area has the advantage that a change in smoke density can be easily determined, thus ensuring a high degree of certainty in the actual detection of a fire. This avoids false alarms. This higher degree of reliability is achieved by evaluating the change in smoke density. If the focus is solely on detecting smoke, the presence of vapors or dust can lead to them being falsely identified as smoke. This could, for example, be harmless moisture evaporation that manifests as vapor, or harmless dust that disperses throughout a room.Such vapors or dusts can result in background coverage comparable to smoke, similar color effects, and / or fundamentally similar movement patterns to those of smoke occurring in a fire. However, comparative studies of smoke and vapors / dusts have shown that, due to the lower thermal driving force, the change in smoke density in vapors or dusts generally occurs more slowly than in smoke originating from a fire source. Thus, visually comparable effects caused by vapors or dusts can be distinguished from actual smoke originating from a fire source by analyzing the smoke density.Such a distinction between smoke from a fire and vapors / dust can be determined by ensuring that a specified threshold for a change in smoke density is exceeded for a positive fire detection. If this threshold is not exceeded, no positive verification of fire detection occurs, and an alarm is not triggered.

[0003] By limiting such an analysis to a subsection of the monitoring area, namely a highly restricted area where smoke development is suspected, a change in color in the suspected area can easily be used to determine the change in smoke density, thus enabling a simple technical determination of smoke density. This can significantly improve the accuracy of fire detection with minimal effort, particularly in terms of computational effort.

[0004] Further advantages arise from the dependent claims. For example, in addition to evaluating the change in smoke density, it is advantageous to provide a predefined threshold value for the smoke density in order to positively detect a fire. This means that not only the change in smoke density must exceed a specified threshold value, but also the absolute value of the smoke density. This ensures that sudden color changes, for example, due to changes in lighting, wind, or similar factors, do not lead to a false, positive verification of a fire detection. A warning is only issued when sufficient smoke is present.

[0005] Furthermore, it is advantageous to consider an expected or specific color, in particular color values ​​of a color scale, of a smoke and / or a color of a background against which the smoke or potential smoke is examined when determining the smoke density or the change in smoke density. This can improve the validity of the determination of the smoke density. The color of the smoke or the color of the background is preferably selected depending on the respective material of the background, for example, a stored material. Furthermore, it is advantageous to also consider a spectrum of the ambient light, for example whether daylight or sodium vapor lighting is used.

[0006] Furthermore, it is advantageous to consider the suspected area as a group of neighboring pixels in a captured image, or possibly even a selected individual pixel. The small data volume thus available allows for a fast and reliable calculation, even for multiple images, to be easily performed automatically.

[0007] It is particularly advantageous to determine the suspected area within the surveillance area by evaluating the optical flow in consecutively acquired images of the surveillance area. Preferably, an area is selected in which a predetermined minimum optical flow, i.e., an optical change, is exceeded in consecutive images.

[0008] It is also advantageous, after a positive verification of fire detection for tracking the fire progression, to also consider the smoke density for an assessment of the further fire development, e.g., to determine whether a fire is spreading quickly or whether the fire may have extinguished again.

[0009] Corresponding advantages arise for an evaluation device for connection to a camera for implementing the method according to the invention and for using the method in a video-based fire detection method. drawing

[0010] Embodiments of the invention are illustrated in the drawings and explained in more detail in the following description. Fig. 1 an evaluation device for carrying out the method connected to a camera directed at a surveillance area, Fig. 2 a procedure, Fig. 3a and Fig. 3b shows an example of a temporal change of an object in the monitoring area to which the method according to the invention is applied. Embodiments of the invention

[0011] In the Fig. 1 shows a device for fire detection, in which a camera 10 directs its field of view 11 onto an object 12, whereby the Fig. 1, the surface of the object 12 forms a surveillance area. The camera 10 records images of the surveillance area and thus of the object 12 at predetermined time intervals, for example at intervals between 0.5 and 10 seconds, in particular one second. The recorded image data are forwarded to a monitoring device 13, which has a computing unit 14 and a storage unit 15, wherein a program is stored in the storage unit 15, according to which the computing unit 14 evaluates the images. This is done, for example, by searching for areas within the recorded images that have a critical color, i.e., a color of fire or smoke. Furthermore, the recorded image data are examined over time for optical changes, in particular the optical flow, in order to detect movements.Thus, for example, an area 16 can be determined in which movement is present and in which a smoke color is detected. Corresponding information is forwarded from the computing unit 14 to an alarm device 17. However, information is also sent via an interface 18 to an evaluation device 20 for verification of fire detection. The evaluation device 20 also has a computing unit 21 and a memory 22. The interface 18 informs the evaluation device which area within the area 16 may represent a suspected fire area, for example the one in the . Fig. 1. The evaluation device 20 also receives image information of the monitored area via the data connection 19 and can therefore now specifically evaluate the suspected area 23.

[0012] In one embodiment, an analysis device 24 is provided, which also has a computing device 25 and a memory 26 for storing a corresponding program and corresponding comparison data. The analysis device 24 analyzes the material of the object 12, for example, by evaluating the image information or has corresponding stored information. It determines a color of the material of the object 12 and also a color of the smoke that arises when the material of the object 12 catches fire, by the computing device accessing relevant information stored in the memory 26. For example, the material can be wood, coal, plastic parts, waste materials, or the like, each of which can have a characteristic color, but also each produce a characteristic smoke color in the event of a fire.This information is transmitted to the evaluation device 20 via an interface 27. If a color value of the background and the smoke is not transmitted via the interface 27 in the embodiment, sample values ​​can also be stored in the memory 22.

[0013] The observed, supposed smoke in the suspected area 23 can be present in different densities. Since smoke is not necessarily optically 100% dense, it lies as a semi-transparent object above the background, namely the object 12. The color of a pixel in the suspected area 23 is approximately determined according to the following formula: x=(1−α)*s+α*b, where α is the transparency value of the smoke, s is the color of the smoke, and b is the color of the background of the object 12. The color value is preferably a color value from a color scale of the camera 10, for example, with 256 colors, a color value on a scale from 0 to 255. Depending on the color setting of the camera, a larger color scale can also be used. The color value is a value that results from the mathematical assignment of a color table of the camera 10 for the corresponding pixel. If necessary, this color value can also be described as an RGB color vector with color values ​​for red, green, and blue.

[0014] In the following, it is assumed that the color of the smoke and the color of the background do not change during the measurement itself. What can change, however, is the transparency, namely when the smoke density changes at the position of the reference area. For a pixel within the suspected area 23, a change in transparency between two times i and j can be written as xi−xj=(αi−αj)*(s−b).

[0015] An approximate change in smoke density can be determined as (αi−αj)=(s−b) / (xi−xj).

[0016] In principle, this information can be determined for each pixel of an image, but in particular for each pixel of the suspected area 23.

[0017] If a predetermined limit value for the change in smoke density is now determined for the comparison area to be exceeded, i.e., a change in the change in smoke density derived from the color difference between smoke and background exceeds a predetermined level, the evaluation device 20 outputs a confirmation of the fire information already transmitted by the monitoring device 13 to the alarm device 17. The alarm device 17 then informs an alarm unit 31 via an interface 30, for example, by issuing a visual alarm via a warning light 32 and / or an acoustic alarm. Furthermore, an image representation of the monitored area and thus of the object 12 detected by the camera 10 can also be shown on a display 33.

[0018] The limit value for smoke density changes can also be adjusted to the time of day, temperature, or other variable circumstances. For example, when storing fresh wood chips or hay in a warehouse, increased vapor generation is to be expected. The limit value can therefore be increased compared to a standard value for these materials if the probability of a fire occurring is lower than the probability of the stored material emitting vapor, particularly at low temperatures. If, on the other hand, dry material is stored, the occurrence of dust or vapor is possible, but not entirely unlikely. The limit value can therefore be set lower.

[0019] In a further embodiment, it is also possible to integrate the various components 13, 20, 24 in a computing unit and to represent each of them as software modules.

[0020] In the Fig. 2 shows an exemplary embodiment of a method sequence according to the invention. Starting from an initialization step 50, video-based fire detection is first carried out in a detection step 51. Video-based fire detection preferably focuses on smoke-like events. For this purpose, for example, a neural network is used in which learned images of fire situations are evaluated. Furthermore, it is also possible to evaluate an optical flow. In a subsequent test step 52, it is checked whether a fire has been detected. If this is not the case, the system branches back to the detection step 51. If a fire is suspected, a first image is initially recorded in a first recording step 53 and a second image at a later time in a second recording step 54. If necessary, further images can also be recorded subsequently.In a subsequent first test step 55, the two images, or possibly several additional images, are compared with regard to a suspected area in order to obtain an approximate determination of the smoke density in the manner described above. If a change in the smoke density beyond a predetermined level is detected, the process branches to an alarm step 56; otherwise, the process branches back to the detection step 51 and no alarm is initially triggered. In a further embodiment, the first test step 55 can also initially branch to a smoke density determination step 57. If it is determined in a subsequent second test step 58 that the smoke density has exceeded a predetermined absolute level, the process branches to the alarm step 56; otherwise, the process branches back to the first recording step 53.

[0021] An example of an item 40 is shown in the Fig. 3a and Fig. 3b at different times. In the Fig. 3a shows a cloud of smoke or vapor 41 on a top side 42 of the object 40. The cloud 41 has regions 43 with a high smoke density and regions 44 with a low smoke density. At the time of recording the Fig. 3b, this picture has changed. Cloud 41 now also has high-density areas 43' and low-density areas 44', but in different positions. If this smoke density change occurs very slowly, it is probably uncritical. If the smoke density change occurs more quickly, a fire is probably present. The detection of cloud 41 alone is not sufficient for reliable, positive fire detection. In addition, there must be a sufficiently rapid change in the smoke density within cloud 41. According to the invention, confirmation occurs when the smoke density change is above a specified limit. QUOTES CONTAINED IN THE DESCRIPTION

[0000] This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions. Cited patent literature

[0000] DE 10 2016 207 705 A1

[0001]

Claims

[1] Method for verifying fire detection in a monitoring area (12), wherein an image of the monitoring area (12) is recorded at at least two points in time, wherein a suspected area (23) for smoke is determined in the image and a color difference of the suspected area (23) between the at least two points in time is determined, wherein a change in smoke density is determined taking into account the determined color difference and wherein a positive verification of a fire only takes place if the change in smoke density exceeds a prescribed minimum. [2] Method according to claim 1, characterized by that as an additional criterion for a positive verification of a fire, a smoke density determined in the suspected area (23) must exceed an absolute limit value. [3] Method according to one of the preceding claims, characterized bythat for determining the smoke density an expected or determined color of a smoke (16) and a color of a background (12) of the smoke are taken into account. [4] Method according to claim 3, characterized by that the color of the smoke and / or the color of the background (12) is determined depending on the material of the background (12) and / or depending on a spectrum of the ambient light. [5] Method according to one of the preceding claims, characterized by that one or more adjacent pixels of the captured image in an area in which possible smoke development is determined are defined as a suspicion area (23). [6] Method according to claim 5, characterized by that the suspected area (23) is determined by an evaluation of the optical flow in successive acquired images of the surveillance area (12). [7] Method according to claim 6, characterized bythat a candidate for the suspected area (23) is an area in which a predetermined minimum of the optical flow is exceeded. [8] Method according to one of the preceding claims, characterized by that after a positive verification of a fire detection, an evaluation of the smoke density is taken into account for an assessment of the further development of the fire. [9] Evaluation device for connection to a camera (10) and with an interface (30) for outputting a positive verification of a fire development for carrying out the method according to one of the preceding claims. [10] Use of the method according to any one of claims 1-8 in a video-based fire detection method.

Images