Smoke monitoring alarm device and method

Abstract

The embodiment of the present invention provides a smoke monitoring alarm device and method, which relates to the field of environmental monitoring and emergency response technology, and can promptly detect fires with less smoke and accurately detect the fire point, thereby performing refined fire extinguishing operations. The method of the embodiment of the present invention includes: receiving raw smoke concentration data collected by a smoke concentration sensor and a current image collected by an image acquisition unit; inputting the connectivity features of the differential image of the current image into a pixel resolution model to obtain risk area data at the current moment; when it is determined that the raw smoke concentration data exceeds a concentration threshold or the risk area data meets the image alarm condition, sending a fire alarm message to the user terminal; converting the raw smoke concentration data into smoke concentration data, performing weighted clustering on the monitoring center point coordinates of each smoke concentration sensor using the smoke concentration as a weight, obtaining the cluster center point at the current moment, and performing fire extinguishing actions according to the positional relationship between the cluster center point and the risk area.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental monitoring and emergency treatment, and in particular to a smoke monitoring and alarm device and method. Background Art

[0002] Current indoor fire prevention and control methods typically use a combination of smoke sensors and sprinklers. This approach has two drawbacks. First, the smoke concentration generated by a fire is dependent on the nature of the burning material and the indoor oxygen concentration. Fires involving certain substances or with higher oxygen concentrations produce less smoke, which doesn't trigger the smoke sensor and prevents the fire from being extinguished in a timely manner. Second, due to insufficient accuracy in detecting the fire point, the existing technology requires sprinklers to spray water across the entire area after the sensor is triggered. This often causes damage in situations where water protection is required. These losses could be avoided if more accurate fire point detection and refined fire extinguishing could be achieved. Summary of the Invention

[0003] In view of this, an embodiment of the present invention provides a smoke monitoring alarm device and method, which can promptly detect fires with less smoke and accurately detect the fire point and then perform refined fire extinguishing operations.

[0004] To achieve the above objectives, according to one aspect of the present invention, a smoke monitoring alarm device is provided.

[0005] The smoke monitoring alarm device of an embodiment of the present invention includes: a plurality of smoke concentration sensors installed at different positions of the target space, a plurality of image monitoring discriminators installed at different positions of the target space, a data fusion device electrically connected to the plurality of image monitoring discriminators, a central controller electrically connected to the smoke concentration sensors and the data fusion device, and a fire extinguishing actuator controlled by the central controller; wherein, each image monitoring discriminator includes an image acquisition unit and a fire discrimination unit, and the fire discrimination unit includes an image processing layer and a pixel resolution model based on a machine learning algorithm; after the image processing layer obtains the current image captured by the image acquisition unit, it subtracts the current image from a normal image pre-stored locally and pointing to the same area to obtain a differential image; the image processing layer traverses the pixel points of the differential image with a convolution kernel of a fixed size, flattens the pixel values ​​within the range circled by the convolution kernel to form a connectivity feature of each pixel point and inputs the connectivity feature into the pixel resolution model; the pixel resolution model calculates the connectivity feature and outputs each pixel The recognition result of the point and each suspicious pixel point determined based on the recognition result, and all the suspicious pixel points are merged into the suspicious area data at the current moment and transmitted to the data fusion device; the data fusion device fuses the suspicious area data output by each image monitoring discriminator into the risk area data at the current moment according to the installation position and shooting angle of each image monitoring discriminator, and sends an alarm signal to the central controller when it is judged that the risk area data meets the preset image alarm condition; when the central controller receives the alarm signal sent by the data fusion device or the alarm signal based on the smoke concentration sensor, it sends a fire alarm message to the user terminal; and the central controller obtains the risk area data sent by the data fusion device and the smoke concentration data formed by the smoke concentration sensor, and performs weighted clustering on the monitoring center point coordinates of each smoke concentration sensor with the corresponding smoke concentration as the weight to obtain the cluster center point at the current moment, and controls the fire extinguishing actuator to perform the corresponding fire extinguishing action according to the positional relationship between the cluster center point at the current moment and the risk area.

[0006] Preferably, the device further includes: a plurality of front-end controllers corresponding to and connected to the plurality of smoke concentration sensors one by one; each front-end controller is electrically connected to the central controller, receives raw smoke concentration data sent by the connected smoke concentration sensor, converts the raw smoke concentration data into the smoke concentration data, and transmits it to the central controller; each front-end controller sends an alarm signal based on the smoke concentration sensor to the central controller when it determines that the received raw smoke concentration data exceeds a preset concentration threshold.

[0007] Preferably, if the central controller determines that the cluster center point at the current moment is within the risk area, the risk area is expanded into a target area according to the first coefficient, and the fire extinguishing actuator is controlled to spray the fire extinguishing medium at the target area; if the central controller determines that the cluster center point at the current moment is outside the risk area and the minimum distance between the cluster center point and the risk area is less than the distance threshold, a spherical area with the cluster center point as the center is generated according to a radius of a specified length, the spherical area is connected to the risk area at the current moment as an extended area, the extended area is expanded into a fire extinguishing area according to the second coefficient, and the fire extinguishing actuator is controlled to spray the fire extinguishing medium at the fire extinguishing area; wherein the second coefficient is greater than the first coefficient; if the central controller determines that the cluster center point at the current moment is outside the risk area and the minimum distance between the cluster center point and the risk area is not less than the distance threshold, the fire extinguishing actuator is controlled to spray the fire extinguishing medium to fully cover the target space.

[0008] Preferably, the pixel resolution model is a lightweight model trained by knowledge distillation based on a pre-trained heavyweight model based on a machine learning algorithm, and the number of parameters of the pixel resolution model is less than one percent of the number of parameters of the heavyweight model.

[0009] Preferably, the image warning condition includes: the spatial volume of the risk area exceeds a preset volume threshold, or the spatial shape of the risk area is a preset target shape.

[0010] To achieve the above object, according to another aspect of the present invention, a smoke monitoring and alarm method is provided.

[0011] The smoke monitoring and alarm method of the embodiment of the present invention includes: receiving raw smoke concentration data collected by multiple smoke concentration sensors installed at different positions in the target space, and current images collected by multiple image acquisition units installed at different positions in the target space; subtracting the current image from a normal image pre-stored locally and pointing to the same area to obtain a differential image; traversing the pixel points of the differential image with a convolution kernel of a fixed size, flattening the pixel values ​​within the range of the convolution kernel to form a connectivity feature of each pixel point and inputting the connectivity feature into the pixel resolution model; obtaining the recognition result of each pixel point output by the pixel resolution model for calculating the connectivity feature and the determination based on the recognition result. For each suspicious pixel point, all suspicious pixel points are merged into the suspicious area data at the current moment, and the suspicious area data are fused into the risk area data at the current moment according to the installation position and shooting angle of each image acquisition unit; when it is judged that the original smoke concentration data exceeds the preset concentration threshold or the risk area data meets the preset image alarm condition, a fire alarm message is sent to the user terminal; and the original smoke concentration data is converted into smoke concentration data, and the monitoring center point coordinates of each smoke concentration sensor are weighted clustered with the corresponding smoke concentration as the weight to obtain the cluster center point at the current moment, and the corresponding fire extinguishing action is performed according to the positional relationship between the cluster center point at the current moment and the risk area.

[0012] Preferably, the corresponding fire extinguishing action is performed according to the positional relationship between the cluster center point and the risk area at the current moment, including: if it is judged that the cluster center point at the current moment is within the risk area, the risk area is expanded into a target area according to a first coefficient, and a preset fire extinguishing actuator is controlled to spray a fire extinguishing medium at the target area; if it is judged that the cluster center point at the current moment is outside the risk area, and the minimum distance between the cluster center point and the risk area is less than a distance threshold, a spherical area with the cluster center point as the center is generated according to a radius of a specified length, the spherical area is connected to the risk area at the current moment as an extended area, the extended area is expanded into a fire extinguishing area according to a second coefficient, and the fire extinguishing actuator is controlled to spray a fire extinguishing medium at the fire extinguishing area; wherein the second coefficient is greater than the first coefficient; if it is judged that the cluster center point at the current moment is outside the risk area, and the minimum distance between the cluster center point and the risk area is not less than the distance threshold, the fire extinguishing actuator is controlled to spray a fire extinguishing medium to fully cover the target space.

[0013] Preferably, the pixel resolution model is a lightweight model trained by knowledge distillation based on a pre-trained heavyweight model based on a machine learning algorithm, and the number of parameters of the pixel resolution model is less than one percent of the number of parameters of the heavyweight model; the image alarm conditions include: the spatial volume of the risk area exceeds a preset volume threshold, or the spatial shape of the risk area is a preset target shape.

[0014] To achieve the above objective, according to another aspect of the present invention, an electronic device is provided.

[0015] An electronic device of the present invention includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the smoke monitoring alarm method provided by the present invention.

[0016] To achieve the above objective, according to yet another aspect of the present invention, a non-transitory computer-readable storage medium is provided.

[0017] A non-transitory computer-readable storage medium of the present invention stores computer instructions, and the computer instructions are used to enable the computer to execute the smoke monitoring and alarm method provided by the present invention.

[0018] According to the technical solution of the present invention, one embodiment of the above invention has the following advantages or beneficial effects:

[0019] On the basis of the traditional smoke concentration sensor, multiple image monitoring discriminators at different positions are added. The image monitoring discriminator consists of an image acquisition unit and a fire discrimination unit. The former is used to acquire images, and the latter is used to extract features from the image and perform pixel-level precision differential judgment on the normal image of the same area without fire and the current image. Finally, a suspicious area composed of suspicious pixels and a three-dimensional risk area formed by fusing the suspicious areas of each monitoring position and monitoring angle are obtained. When the spatial volume of the risk area is greater than the threshold or the smoke concentration sensor detects a high smoke concentration, the central controller immediately alarms the user terminal. The image monitoring discriminator thus makes up for the shortcomings of the smoke concentration sensor. When there is less smoke, the image monitoring discriminator can detect flames or objects deformed by fire earlier based on pixel-level image differential comparison, thereby solving the problem of fire detection when there is less smoke. Furthermore, the present invention achieves accurate detection of the fire point and refined fire extinguishing operations in the following ways. After a fire is detected by an image monitoring discriminator or a smoke density sensor, the coordinates of the monitoring center points of each smoke density sensor are weightedly clustered with the smoke density as the weight to obtain the cluster center point at the current moment. The cluster center point is the most likely fire point detected by the smoke density sensor, and the risk area obtained from the image monitoring discriminator is the most likely fire point distribution area detected by the image monitoring discriminator. The confidence of the above risk areas obtained based on the high-performance machine learning model and pixel-level accuracy is high. At this time, the central controller determines the positional relationship between the cluster center point and the risk area: when the cluster If the cluster center is within the risk area, it means the fire is not dispersed and is likely in its infancy. In this case, precise firefighting can be carried out within a certain range outside the risk area. If the cluster center is outside the risk area but not far away, it means the fire is spreading but can still be controlled. In this case, a spherical area is first generated with the cluster center as the center. The spherical area is then connected to the risk area to form an extended area. The extended area is then expanded to a large extent before firefighting is carried out. If the cluster center is outside the risk area and is far away, it means the fire has developed to a certain extent and there is an uncontrollable risk. In this case, firefighting can be carried out directly in the entire target space. This way, the spray area of ​​the fire extinguishing medium can be minimized while ensuring timely and effective firefighting, thus avoiding property losses caused by the large-scale and full-coverage spraying of the fire extinguishing medium.

[0020] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings are provided for a better understanding of the present invention and are not intended to limit the present invention.

[0022] Figure 1 2 is a schematic diagram of the composition of a smoke monitoring alarm device according to an embodiment of the present invention;

[0023] Figure 2 2 is a schematic diagram of the steps of a smoke monitoring and alarm method according to an embodiment of the present invention;

[0024] Figure 3 Schematic diagram of the hardware structure of an electronic device used to implement the smoke monitoring and alarm method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The following description of exemplary embodiments of the present invention is made in conjunction with the accompanying drawings, in which various details of the embodiments of the present invention are included to facilitate understanding. These details should be considered as merely exemplary. Therefore, it should be appreciated by those skilled in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0026] Figure 1 is a schematic diagram of the composition of a smoke monitoring alarm device according to an embodiment of the present invention, see Figure 1 .

[0027] The smoke monitoring alarm device of an embodiment of the present invention includes: multiple smoke concentration sensors installed at different positions in the target space, multiple image monitoring discriminators installed at different positions in the target space, a data fusion device electrically connected to the multiple image monitoring discriminators, a central controller electrically connected to the smoke concentration sensors and the data fusion device, and a fire extinguishing actuator controlled by the central controller.

[0028] Specifically, the smoke concentration sensors are used to collect smoke in a corresponding area and generate raw smoke concentration data. Each smoke concentration sensor can be connected to an independent front-end controller, which is electrically connected to the central controller. The front-end controller receives the raw smoke concentration data from the connected smoke concentration sensor, converts the raw smoke concentration data into smoke concentration data, and transmits the smoke concentration data to the central controller.

[0029] In particular, each front-end controller sends an alarm signal based on the smoke concentration sensor to the central controller when it determines that the received raw smoke concentration data exceeds a preset concentration threshold, so as to warn the user in time.

[0030] In an embodiment of the present invention, each image monitoring discriminator includes an image acquisition unit and a fire discrimination unit. The image acquisition unit can be a known camera or other device. The fire discrimination unit includes an image processing layer and a pixel resolution model based on a machine learning algorithm. In order to achieve high-precision pixel-level image discrimination, it is necessary to make the pixel resolution model of the distributed actual scene have the computing power of a heavyweight machine learning model (heavyweight can be defined by model parameters, for example, a model with a parameter value greater than 1 million. Similarly, a lightweight model can also be defined by model parameters, for example, a model with a parameter value less than 10,000). To achieve this goal, the existing technology can only have two solutions: remotely accessing the interface of the cloud server or deploying the heavyweight model on site. However, the former will greatly affect the real-time performance of fire handling, and the latter requires the use of ultra-high-performance CPU, GPU and other hardware on site, which is very unrealistic.

[0031] To address the above issues, the present invention first trains a heavyweight model in the background. The model structure can adopt BERT, CNN, etc., and then constructs a lightweight network based on low-layer DNN, RNN, etc. as a pixel resolution model on site. The pixel resolution model can be formed by jointly training the heavyweight model and the training set in a knowledge distillation manner. In this way, the pixel resolution model is more suitable for deployment on site based on the performance of heavyweight models such as BERT, with low cost, easy maintenance, and no need for high-performance hardware support. In practical applications, the number of parameters of the pixel resolution model can be less than one percent of the number of parameters of the heavyweight model.

[0032] After acquiring the current image captured by the image acquisition unit, the image processing layer subtracts the current image from a pre-stored normal image pointing to the same area to obtain a differential image. Subsequently, the image processing layer traverses the pixels of the differential image with a convolution kernel of a fixed size (e.g., 3*3 or 5*5), flattens the pixel values ​​within the convolution kernel's range in a preset order to form a connectivity feature for each pixel, and inputs the connectivity feature into the pixel resolution model. The pixel resolution model calculates the input connectivity feature and outputs the recognition result for each pixel (including whether it is the same pixel or a different pixel) as well as each suspicious pixel determined based on the above recognition result. All suspicious pixels are then merged into the current suspicious area data, which is transmitted to the data fusion unit.

[0033] The data fusion unit, which can be built using FPGA hardware or conventional software, fuses the suspicious area data output by each image surveillance discriminator into current risk area data (this data is three-dimensional spatial coordinate data) based on the installation location and shooting angle of each image surveillance discriminator. It then sends an alarm signal to the central controller when the risk area data meets preset image alarm conditions. These image alarm conditions can include: the spatial volume of the risk area exceeds a preset volume threshold, or the spatial shape of the risk area matches a preset target shape (such as a flame shape or the initial shape of an object in the corresponding surveillance area).

[0034] Upon receiving an alarm signal from the data fusion unit or an alarm signal based on the smoke concentration sensor, the central controller immediately sends a fire alarm message to the user terminal. Through the above solution, multiple image monitoring discriminators at different locations are added to the traditional smoke concentration sensor. The image acquisition unit in the image monitoring discriminator is used to acquire images, and the fire discrimination unit in the image monitoring discriminator is used to extract features from the image and perform pixel-level precision differential judgment between a normal image of the same area without a fire and the current image. Ultimately, a suspicious area composed of suspicious pixels and a three-dimensional risk area are obtained by fusing the suspicious areas of each monitoring position and monitoring angle. When the spatial volume of the risk area exceeds a threshold or the smoke concentration sensor detects a high smoke concentration, the central controller immediately sends an alarm to the user terminal. This compensates for the shortcomings of the smoke concentration sensor through the image monitoring discriminator. When there is less smoke, the image monitoring discriminator can detect flames or objects deformed by fire earlier based on pixel-level image differential comparison, thereby solving the problem of fire detection when there is less smoke.

[0035] On the other hand, the central controller obtains the risk area data sent by the data fusion unit and the smoke concentration data sent by the front-end controller, and performs weighted clustering on the coordinates of the monitoring center points of each smoke concentration sensor using the corresponding smoke concentration as the weight to obtain the cluster center point at the current moment. The above clustering method can adopt the DBSCAN clustering algorithm. It can be understood that the cluster center point is the most likely fire point detected by the smoke concentration sensor, and the risk area obtained from the image monitoring discriminator is the most likely fire point distribution area detected by the image monitoring discriminator. The confidence level of the above risk areas is high based on the high-performance machine learning model and pixel-level precision judgment. At this point, the central controller can control the fire extinguishing actuator to perform the corresponding fire extinguishing action based on the positional relationship between the cluster center point and the risk area at the current moment.

[0036] In actual scenarios, if the central controller determines that the cluster center at the current moment is within the risk area, it means that the fire is not dispersed and is most likely in the embryonic stage. At this time, it is sufficient to precisely execute fire extinguishing within a certain range outside the risk area. Therefore, the central controller expands the risk area into a target area according to the first coefficient, and controls the fire extinguishing actuator to spray the fire extinguishing medium at the target area. Among them, the first coefficient is the expansion ratio, which is a number greater than 1. The above expansion can take the cluster center as the starting point, and expand outward with the line connecting the cluster center and the boundary point in any direction of the risk area. After the expansion, the ratio of the length of the boundary point in this direction from the cluster center to the original connection length is the first value. In addition, the fire extinguishing actuator can be a sprinkler head, a water pump, etc.

[0037] If the central controller determines that the cluster center at the current moment is outside the risk area and the minimum distance between the cluster center and the risk area is less than the distance threshold, it means that the fire has begun to spread but is still controllable. At this time, the central controller first generates a spherical area with the cluster center as the center according to the radius of the specified length. Then, the spherical area is connected to the risk area at the current moment to form an extended area, and the extended area is expanded into a fire extinguishing area according to the second coefficient. Finally, the fire extinguishing actuator is controlled to spray the fire extinguishing medium at the fire extinguishing area. Among them, the above connection refers to opening up the space between the spherical area and the risk area. The connection method can be executed according to the preset strategy. The above second coefficient is also the expansion ratio, which is greater than the first coefficient. The expansion method of the above extended area is the same as above.

[0038] If the central controller determines that the current cluster center is outside the risk area and the minimum distance between the cluster center and the risk area is at least a threshold, indicating that the fire has progressed to a certain extent and poses an uncontrollable risk, the central controller directly controls the fire extinguishing actuator to spray the fire extinguishing medium across the entire target area. This process minimizes the spray area while ensuring timely and effective fire extinguishing, thus preventing property damage caused by widespread, full-coverage spraying of the fire extinguishing medium.

[0039] Figure 2 This is a schematic diagram of the steps of the smoke monitoring and alarm method according to an embodiment of the present invention. The execution entities of each step are the front-end controller, the image monitoring discriminator, the data fusion device and the central controller. Figure 2 .

[0040] Step S201: Receive raw smoke concentration data collected by multiple smoke concentration sensors installed at different locations in the target space, and current images collected by multiple image acquisition units installed at different locations in the target space. Step S202: Subtract the current image from a normal image pre-stored locally and pointing to the same area to obtain a differential image; traverse the pixels of the differential image with a fixed-size convolution kernel, flatten the pixel values ​​within the range defined by the convolution kernel to form a connectivity feature for each pixel, and input the connectivity feature into the pixel resolution model. Step S203: Obtain the recognition result of each pixel point calculated by the pixel resolution model on the connectivity feature and output, as well as each suspicious pixel point determined based on the recognition result, merge all suspicious pixels into the suspicious area data at the current moment, and fuse the suspicious area data into the risk area data at the current moment based on the installation position and shooting angle of each image acquisition unit. Step S204: When it is determined that the raw smoke concentration data exceeds a preset concentration threshold or the risk area data meets the preset image alarm condition, a fire alarm message is sent to the user terminal; and the raw smoke concentration data is converted into smoke concentration data, and the monitoring center point coordinates of each smoke concentration sensor are weighted clustered with the corresponding smoke concentration as the weight to obtain the cluster center point at the current moment, and the corresponding fire extinguishing action is performed according to the positional relationship between the cluster center point at the current moment and the risk area.

[0041] In an embodiment of the present invention, the corresponding fire extinguishing action is performed according to the positional relationship between the cluster center point and the risk area at the current moment, including: if it is determined that the cluster center point at the current moment is within the risk area, the risk area is expanded into a target area according to a first coefficient, and a preset fire extinguishing actuator is controlled to spray a fire extinguishing medium at the target area; if it is determined that the cluster center point at the current moment is outside the risk area and the minimum distance between the cluster center point and the risk area is less than a distance threshold, a spherical area with the cluster center point as the center is generated according to a radius of a specified length, the spherical area is connected to the risk area at the current moment as an extended area, the extended area is expanded into a fire extinguishing area according to a second coefficient, and the fire extinguishing actuator is controlled to spray a fire extinguishing medium at the fire extinguishing area; wherein the second coefficient is greater than the first coefficient; if it is determined that the cluster center point at the current moment is outside the risk area and the minimum distance between the cluster center point and the risk area is not less than the distance threshold, the fire extinguishing actuator is controlled to spray a fire extinguishing medium to fully cover the target space.

[0042] In a specific application, the pixel resolution model is a lightweight model trained by knowledge distillation based on a pre-trained heavyweight model based on a machine learning algorithm. The number of parameters of the pixel resolution model is less than one percent of the number of parameters of the heavyweight model; the image alarm conditions include: the spatial volume of the risk area exceeds a preset volume threshold, or the spatial shape of the risk area is a preset target shape.

[0043] In the technical solution of the present invention, multiple image monitoring discriminators at different positions are added on the basis of the traditional smoke concentration sensor. The image monitoring discriminator consists of an image acquisition unit and a fire discrimination unit. The former is used to acquire images, and the latter is used to extract features from the image and perform pixel-level precision differential judgment on the normal image of the same area without fire and the current image. Finally, a suspicious area composed of suspicious pixels and a three-dimensional risk area formed by fusing the suspicious areas of each monitoring position and monitoring angle are obtained. When the spatial volume of the risk area is greater than the threshold or the smoke concentration sensor detects a high smoke concentration, the central controller immediately alarms the user terminal. The image monitoring discriminator thus makes up for the shortcomings of the smoke concentration sensor. When there is less smoke, the image monitoring discriminator can detect flames or objects deformed by fire earlier based on pixel-level image differential comparison, thereby solving the problem of fire detection when there is less smoke. Furthermore, the present invention achieves accurate detection of the fire point and refined fire extinguishing operations in the following ways. After a fire is detected by an image monitoring discriminator or a smoke density sensor, the coordinates of the monitoring center points of each smoke density sensor are weightedly clustered with the smoke density as the weight to obtain the cluster center point at the current moment. The cluster center point is the most likely fire point detected by the smoke density sensor, and the risk area obtained from the image monitoring discriminator is the most likely fire point distribution area detected by the image monitoring discriminator. The confidence of the above risk areas obtained based on the high-performance machine learning model and pixel-level accuracy is high. At this time, the central controller determines the positional relationship between the cluster center point and the risk area: when the cluster If the cluster center is within the risk area, it means the fire is not dispersed and is likely in its infancy. In this case, precise firefighting can be carried out within a certain range outside the risk area. If the cluster center is outside the risk area but not far away, it means the fire is spreading but can still be controlled. In this case, a spherical area is first generated with the cluster center as the center. The spherical area is then connected to the risk area to form an extended area. The extended area is then expanded to a large extent before firefighting is carried out. If the cluster center is outside the risk area and is far away, it means the fire has developed to a certain extent and there is an uncontrollable risk. In this case, firefighting can be carried out directly in the entire target space. This way, the spray area of ​​the fire extinguishing medium can be minimized while ensuring timely and effective firefighting, thus avoiding property losses caused by the large-scale and full-coverage spraying of the fire extinguishing medium.

[0044] According to an embodiment of the present invention, the present invention further provides an electronic device and a readable storage medium.

[0045] The electronic device of the present invention includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the smoke monitoring alarm method provided by the present invention.

[0046] The non-transitory computer-readable storage medium of the present invention stores computer instructions, and the computer instructions are used to enable the computer to execute the smoke monitoring and alarm method provided by the present invention.

[0047] like Figure 3 FIG. 1 is a schematic diagram of the hardware structure of an electronic device for implementing the method for business rule management according to an embodiment of the present invention. Figure 3 The electronic device includes: one or more processors 31 and a memory 32, Figure 3 A processor 31 is taken as an example. The memory 32 is a non-transitory computer-readable storage medium provided by the present invention.

[0048] The electronic device of the present invention may further include an input device 33 and an output device 34 .

[0049] The processor 31, memory 32, input device 33 and output device 34 may be connected via a bus or other means. Figure 3 The bus connection is taken as an example.

[0050] Memory 32, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer executable programs, and modules, such as the program instructions / modules corresponding to the business rule management method in the embodiments of the present invention. Processor 31 executes the non-transitory software programs, instructions, and modules stored in memory 32 to execute various server functional applications and data processing, thereby implementing the business rule management method in the above-mentioned method embodiments.

[0051] The memory 32 may include a program storage area and a data storage area, wherein:

[0052] The program storage area can store the operating system and the application program required for at least one function; the data storage area can store data created by the use of the device managed according to business rules.

[0053] In addition, the memory 32 may include a high-speed random access memory, and may also include a non-transitory memory, such as at least one disk storage device,

[0054] Flash memory devices, or other non-volatile solid-state storage devices. In some embodiments, the memory 32 optionally includes a memory remotely located relative to the processor 31, and these remote memories can be connected to the device for business rule management via a network.

[0055] Examples of the network include but are not limited to the Internet, corporate intranet, local area network,

[0056] Mobile communication networks and their combinations.

[0057] The input device 33 can receive input digital or character information and generate

[0058] Keys related to user settings and function control of devices managed by business rules

[0059] Signal input. The output device 34 may include a display device such as a display screen.

[0060] The above one or more modules are stored in the memory 32 , and when executed by the one or more processors 31 , the smoke monitoring and alarm method in any of the above method embodiments is executed.

[0061] The above product can execute the smoke monitoring alarm method provided by the embodiment of the present invention, and has the corresponding functional modules and beneficial effects of executing the smoke monitoring alarm method.

[0062] The technical details not described in detail in this embodiment can be found in the embodiment of the present invention.

[0063] The smoke monitoring and alarm method provided in the embodiment.

[0064] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A smoke monitoring alarm device, characterized in that: include: A plurality of smoke concentration sensors installed at different positions in the target space, a plurality of image monitoring discriminators installed at different positions in the target space, a data fusion device electrically connected to the plurality of image monitoring discriminators, a central controller electrically connected to the smoke concentration sensors and the data fusion device, and a fire extinguishing actuator controlled by the central controller; wherein, Each image monitoring discriminator includes an image acquisition unit and a fire discrimination unit, wherein the fire discrimination unit includes an image processing layer and a pixel resolution model based on a machine learning algorithm; After the image processing layer obtains the current image captured by the image acquisition unit, it subtracts the current image from a normal image pre-stored locally and pointing to the same area to obtain a difference image; the image processing layer traverses the pixels of the difference image with a convolution kernel of a fixed size, flattens the pixel values ​​within the range of the convolution kernel to form a connectivity feature of each pixel, and inputs the connectivity feature into the pixel resolution model; The pixel resolution model calculates the connectivity features and outputs the recognition result of each pixel and each suspicious pixel determined based on the recognition result, and merges all the suspicious pixels into the suspicious area data at the current moment and transmits it to the data fusion device; The data fusion unit fuses the suspicious area data output by each image monitoring discriminator into risk area data at the current moment according to the installation position and shooting angle of each image monitoring discriminator, and sends an alarm signal to the central controller when it is determined that the risk area data meets the preset image alarm conditions; The central controller sends a fire alarm message to the user terminal when receiving the alarm signal sent by the data fusion device or the alarm signal based on the smoke concentration sensor; and the central controller obtains the risk area data sent by the data fusion device and the smoke concentration data formed by the smoke concentration sensor, performs weighted clustering on the monitoring center point coordinates of each smoke concentration sensor using the corresponding smoke concentration as a weight, obtains the cluster center point at the current moment, and controls the fire extinguishing actuator to perform the corresponding fire extinguishing action according to the positional relationship between the cluster center point at the current moment and the risk area; If the central controller determines that the cluster center point at the current moment is within the risk area, the risk area is expanded into a target area according to the first coefficient, and the fire extinguishing actuator is controlled to spray the fire extinguishing medium at the target area; if the central controller determines that the cluster center point at the current moment is outside the risk area and the minimum distance between the cluster center point and the risk area is less than the distance threshold, a spherical area with the cluster center point as the center is generated according to a radius of a specified length, the spherical area is connected to the risk area at the current moment as an extended area, the extended area is expanded into a fire extinguishing area according to the second coefficient, and the fire extinguishing actuator is controlled to spray the fire extinguishing medium at the fire extinguishing area; wherein the second coefficient is greater than the first coefficient; if the central controller determines that the cluster center point at the current moment is outside the risk area and the minimum distance between the cluster center point and the risk area is not less than the distance threshold, the fire extinguishing actuator is controlled to spray the fire extinguishing medium to fully cover the target space.

2. The device according to claim 1, characterized in that The device further comprises: a plurality of front-end controllers corresponding to and connected to the plurality of smoke concentration sensors in a one-to-one manner; Each front-end controller is electrically connected to the central controller, receives the raw smoke concentration data sent by the connected smoke concentration sensor, converts the raw smoke concentration data into the smoke concentration data, and transmits the data to the central controller; Each front-end controller sends an alarm signal based on the smoke density sensor to the central controller when it determines that the received smoke density raw data exceeds a preset density threshold.

3. The device according to claim 2, characterized in that The pixel resolution model is a lightweight model trained by knowledge distillation based on a pre-trained heavyweight model based on a machine learning algorithm. The number of parameters of the pixel resolution model is less than one percent of the number of parameters of the heavyweight model.

4. The device according to claim 1, characterized in that The image warning conditions include: the spatial volume of the risk area exceeds a preset volume threshold, or the spatial shape of the risk area is a preset target shape.

5. A smoke monitoring and alarm method, characterized in that: include: Receiving raw smoke concentration data collected by a plurality of smoke concentration sensors installed at different positions in the target space, and current images collected by a plurality of image acquisition units installed at different positions in the target space; Subtract the current image from a normal image pre-stored locally and pointing to the same area to obtain a differential image; Traversing the pixels of the differential image with a convolution kernel of a fixed size, flattening the pixel values ​​within the range defined by the convolution kernel to form a connectivity feature of each pixel, and inputting the connectivity feature into the pixel resolution model; Obtaining an identification result of each pixel point output by the pixel resolution model through calculation of the connectivity feature and each suspicious pixel point determined based on the identification result, merging all the suspicious pixels into suspicious area data at the current moment, and fusing the suspicious area data into risk area data at the current moment based on the installation position and shooting angle of each image acquisition unit; When it is determined that the raw smoke concentration data exceeds a preset concentration threshold or the risk area data meets a preset image alarm condition, a fire alarm message is sent to the user terminal; in addition, the raw smoke concentration data is converted into smoke concentration data, and the coordinates of the monitoring center points of each smoke concentration sensor are weighted clustered using the corresponding smoke concentration as a weight to obtain the cluster center point at the current moment, and the corresponding fire extinguishing action is executed according to the positional relationship between the cluster center point at the current moment and the risk area; The corresponding fire extinguishing action is performed according to the positional relationship between the cluster center point and the risk area at the current moment, including: if it is judged that the cluster center point at the current moment is within the risk area, the risk area is expanded into a target area according to a first coefficient, and a preset fire extinguishing actuator is controlled to spray a fire extinguishing medium at the target area; if it is judged that the cluster center point at the current moment is outside the risk area and the minimum distance between the cluster center point and the risk area is less than a distance threshold, a spherical area with the cluster center point as the center is generated according to a radius of a specified length, the spherical area is connected to the risk area at the current moment as an extended area, the extended area is expanded into a fire extinguishing area according to a second coefficient, and the fire extinguishing actuator is controlled to spray a fire extinguishing medium at the fire extinguishing area; wherein the second coefficient is greater than the first coefficient; if it is judged that the cluster center point at the current moment is outside the risk area and the minimum distance between the cluster center point and the risk area is not less than the distance threshold, the fire extinguishing actuator is controlled to spray a fire extinguishing medium to fully cover the target space.

6. The method according to claim 5, characterized in that The pixel resolution model is a lightweight model trained by knowledge distillation based on a pre-trained heavyweight model based on a machine learning algorithm, and the number of parameters of the pixel resolution model is less than one percent of the number of parameters of the heavyweight model; The image warning conditions include: the spatial volume of the risk area exceeds a preset volume threshold, or the spatial shape of the risk area is a preset target shape.

7. An electronic device, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of claim 5 or 6.

8. A non-transitory computer-readable storage medium, characterized in that The non-transitory computer-readable storage medium stores computer instructions for causing the computer to execute the method of claim 5 or 6.

Images