Smart park smoke and fire detection system and method based on visual large model

Abstract

The invention relates to the technical field of smart park safety, in particular to a smart park smoke and fire detection system and method based on a visual large model, and the system comprises a detection assembly, a switch, a master control server, and a warning module, and the output end of the detection assembly is in communication connection with the input end of the switch through a network. The output end of the switch is in communication connection with the input end of the master control server through a network; the designed danger grading module is divided into a fire smoke and fire range grading unit and a fire smoke and fire type unit, the fire smoke and fire range grading unit can judge the grade of the smoke and fire, and the fire smoke and fire type unit can judge the type of fire which the fire smoke and fire is generated. Therefore, fire smoke and fire can be detected, the range and type of the smoke and fire can be obtained, workers can use a proper fire extinguishing mode to rapidly extinguish fire, the fire extinguishing response speed is increased, the fire extinguishing speed is increased, and losses in a park are reduced.

Description

Technical Field

[0001] The present invention relates to the field of smart park safety technology, and in particular to a smart park smoke and fire detection system and method based on a visual large model. Background Art

[0002] ‌‌A smart park refers to an advanced park development concept that utilizes new-generation information and communication technologies, with the capabilities of rapid information collection, high-speed information transmission, highly centralized computing, intelligent transaction processing, and ubiquitous service provision, to achieve timely, interactive, and integrated information perception, transmission, and processing within the park. In order to ensure the safety within the park, a series of safety measures and equipment such as fire monitoring systems will be set up in the park.

[0003] For example, a fireworks detection system based on a combined neural network and multi-sensor fusion, whose authorization announcement number is CN202210643060.X and whose authorization announcement date is 20220812, includes an installation fixed component, a focus anti-reflective precise temperature monitoring and acquisition module, a fusion multi-sensor auxiliary monitoring device, an emergency negative pressure air suction chemical particle adsorption fire extinguishing device and a cloud service processing unit. The present invention belongs to the field of fireworks detection technology, and specifically provides a method of gathering the key monitoring areas into one point on the infrared sensor by total reflection, greatly improving the accuracy of fireworks detection, avoiding the occurrence of misjudgment, and reducing the burden on staff. Through the fusion of neural networks and multi-sensor fusion, the accuracy of detection is further improved. For small-scale fireworks, negative pressure can be used to avoid combustion-supporting methods to effectively and non-destructively extinguish the fire. At the same time, harmful particles are adsorbed to avoid harm to the human body, ensuring the timeliness and effectiveness of fire extinguishing. A fireworks detection system based on a combined neural network and multi-sensor fusion.

[0004] For example, a power plant fire and smoke detection system based on front-end processing, with an authorization announcement number of CN202021386796.6 and an authorization announcement date of 20210528, includes at least one normally open camera device, at least one normally closed camera device, and a main controller; the normally open camera device includes a first processor and a first camera; the normally closed camera device includes a second processor, a second camera, and a relay; the normally open camera device and the normally closed camera device are both integrated devices; the normally open camera device is used to detect suspicious points of fire and smoke in the power plant in real time and their location information, and send the suspicious point location information to the main controller; the main controller is used to control the normally closed camera device at the corresponding position to open according to the suspicious point location information, and receive the alarm information sent by the normally closed camera device; the normally closed camera device is used to detect the fire and smoke situation at the suspicious point, and send an alarm information to the main controller if there is fire and smoke. Based on front-end processing, it has fast speed and high quality; the architecture of the normally open camera device and the normally closed camera device is adopted, so that no suspicious points of fire and smoke will be missed, and the accuracy of detection is guaranteed.

[0005] Although the above-mentioned fireworks detection system and the fireworks detection system in the prior art can detect fireworks, it is impossible to judge the type of fire fireworks during the detection process, resulting in workers being unable to choose appropriate fire extinguishing methods according to the type of fire and unable to extinguish the fire in time, which will cause great losses. At the same time, the detection of fireworks is mostly too single and can only determine whether the fireworks are open flames. It is unable to detect fireworks that are not open flames, resulting in fireworks becoming open flames before they can be detected, wasting fire extinguishing time. Therefore, it is urgent to design a smart park fireworks detection system and method based on a visual big model to solve the above problems. Summary of the invention

[0006] The purpose of the present invention is to provide a smart park smoke and fire detection system and method based on a visual large model to solve the above-mentioned deficiencies in the prior art.

[0007] In order to achieve the above object, the present invention provides the following technical solutions: An embodiment of the present invention provides a smart park smoke and fire detection system based on a visual large model, comprising a detection component, a switch, a main control server, and an alarm module, wherein the output end of the detection component is communicatively connected to the input end of the switch through a network, the output end of the switch is communicatively connected to the input end of the main control server through a network, and the output end of the main control server is communicatively connected to the input end of the alarm module through a network; The detection component includes a mounting tube, one end of the mounting tube is threadedly connected to a glass cover, a camera is installed inside the mounting tube by bolts, and the camera model is preferably a computing module installed inside the mounting tube by bolts, and one end of the mounting tube is threadedly connected to a threaded cover, and the computer module can be repaired by unscrewing the threaded cover. The detection component also includes a connecting frame, an electric box is installed on the outer wall of one side of the connecting frame by bolts, and a connecting rod is arranged at the bottom of the electric box, and an electric push rod is installed on the outer wall of one side of the connecting rod by a hinge, and the electric push rod model is preferably THK, when the electric push rod is started, the output end of the electric push rod will pull up or push one end of the mounting tube to move, so that the mounting tube swings up and down around the connecting rod with the hinge seat, so as to expand the long and short distance of camera monitoring, the bottom end of the connecting rod is installed with a hinge seat by a rotating shaft, and the bottom end of the hinge seat is installed with the mounting tube by bolts, and the output end of the electric push rod is rotatably connected with the mounting tube by a hinge, A high beam is installed on the outer wall of one side of the bottom of the installation tube by bolts. Under the control of the computing module, the high beam will light up and emit light to illuminate the fire ground, so that workers can find the fire location in time at night to extinguish the fire. A drive motor is installed on the bottom side of the inner wall of the power box by bolts. When the drive motor is started, the connecting rod will carry the electric push rod and the installation tube to make circular motions, so that the camera shooting position changes, so as to better monitor the park. The output end of the drive motor is fixedly connected to the connecting rod through a coupling. A signal module is installed on the top side of the inner wall of the power box by bolts. The signal module uses a local area network signal transmitter, such as a router. A battery is provided on the bottom side of the inner wall of the power box. The battery is a lithium battery, which can provide power for the operation of the detection component. The battery is externally connected to a photovoltaic device, and the battery is electrically connected to the signal module, the drive motor, the high beam, and the electric push rod through wires. The computing module includes an image processing module, a fireworks judgment module, a hazard classification module, a processor, a memory and other auxiliary modules. The image processing module, the fireworks judgment module and the hazard classification module are computing programs, and the image processing module, the fireworks judgment module and the hazard classification module are loaded on the processor. The memory is used to store data processed by the processor and log data. The results processed by the processor and the data stored in the memory are compressed and transmitted through other auxiliary modules. The image processing module is established based on image segmentation technology and image optimization technology, and is used to segment and optimize the image data collected by the detection component. The image segmentation technology in the image processing module specifically uses a region-based segmentation algorithm, and the image optimization technology in the image processing module specifically uses an image enhancement technology; It should be noted that the region-based segmentation algorithm is an image segmentation technique that divides an image into several regions. These regions usually have similar characteristics, such as color, texture, brightness, etc., and have continuity within the image. The goal of this segmentation method is to divide the image into a set of regions, each of which is different from the other regions and can be used for further image analysis and processing. Common region-based segmentation methods include region growing method, split and merge method and energy function-based method; Image enhancement technologies mainly include color transformation, histogram equalization, and homomorphic filtering. Color transformation includes brightness adjustment, contrast adjustment, saturation adjustment, and hue adjustment. ‌‌Histogram equalization is used to improve the contrast of an image, especially when the image has low contrast; ‌‌Homomorphic filtering is used to eliminate the impact of uneven illumination and enhance image details.

[0008] The fireworks judgment module is established based on the fireworks detection algorithm, which is used to perform feature recognition on the image data processed by the image processing module, so as to determine whether there are fireworks in the image data and the type of fireworks. The fireworks detection algorithm includes two types: open fire detection algorithm and dark fire detection algorithm. The open fire detection algorithm uses the ‌LntonAIServer fireworks detection algorithm, and the dark fire detection algorithm uses algorithms based on color detection and algorithms based on edge detection. The specific edge detection algorithm uses Sobel edge detection, and the color detection algorithm uses detection methods based on RGB color space, HSV color space and YUV color space. It should be noted that YOLOv8 is a target detection model that can be used for a variety of tasks, including smoke and open flame detection, providing high-speed and high-precision target detection capabilities. Through real-time analysis of input images, it can accurately identify fire hazards such as smoke and open flames; ‌LntonAIServer smoke and fire detection algorithm is an intelligent monitoring solution based on deep learning technology, which aims to detect and identify signs of smoke and fire in real time and improve the accuracy and efficiency of fire warning. ‌The algorithm uses deep neural networks (DNN), especially convolutional neural networks (CNN), to accurately capture smoke and fire features by training a large number of annotated smoke and fire images, and can provide efficient and accurate detection services in various complex scenarios‌; The Sobel edge detection algorithm is a commonly used edge detection method, mainly used in the fields of image processing and computer vision. It detects edges by calculating the gradient strength of each pixel in the image, and is an edge detection method based on first-order gradients; The principle of the Sobel edge detection algorithm; The Sobel operator contains two sets of 3x3 matrices, which are used for horizontal and vertical edge detection respectively. By convolving these two matrices with the image, we can get the approximate brightness difference of the image in the horizontal and vertical directions. These values ​​are used to calculate the gradient strength of each pixel to determine the edge position; ‌Fire color detection algorithms mainly include detection methods based on RGB color space, HSV color space, and YUV color space, among which: Fire detection algorithm based on RGB color space The RGB color space is the most basic color model, which identifies flames by detecting the red and green components in the image. The red and green components of flames are usually large, and the green component is larger than the blue component. This method is simple and intuitive, but it is easily affected by brightness, so it does not work well when dealing with scenes with large brightness changes‌12.

[0009] Fire detection algorithm based on HSV color space The HSV color space represents color using three dimensions: hue, saturation, and brightness. This method is more adaptable to lighting changes than RGB because the HSV space is insensitive to changes in brightness. By setting a suitable threshold, the flame color can be effectively detected‌ 34.

[0010] Fire Detection Algorithm Based on YUV Color Space The YUV color space decomposes color information into brightness and chromaticity. By detecting changes in the Y channel (brightness) and the U and V channels (chromaticity), flames can be effectively identified.

[0011] The hazard classification module is established based on feature extraction technology, and is used to perform secondary judgment on the image data and surrounding environment image data that the fireworks judgment module judges to have fireworks, so as to judge what kind of fire the fireworks are and the level of the fire and fireworks. The hazard classification module is divided into a fire and fireworks range level unit and a fire and fireworks type unit. The fire and fireworks range level unit judges the level of fireworks by the fire and fireworks range formula. The fire and fireworks range formula is as follows:

[0012] Among them, when S≥1, the level of fire smoke is a high - risk level; when 0 < S < 1, the level of fire smoke is a low - risk level. N is the area of the burning region in the image, and M is the maximum safe burning area value set for the actual park; the fire smoke type unit classifies fire types into six types: A, B, C, D, E, and F. The fire smoke type level selects the CLIP polymorphic model, which can match the selected image with the corresponding text to determine what type of fire the fire smoke is generated from.

[0013] The processor module selects one of Intel Core i7 and AMD Ryzen 7, which is used to carry the image processing module, the smoke and fire judgment module, and the risk classification module, and is also used to control the operation of the detection component 1; The memory selects the Netac N530S series solid - state drive, which is used to temporarily store the data processed by the image processing module, the smoke and fire judgment module, the risk classification module, and the log data of the processor; The other auxiliary modules are used for the processor to exchange information with external devices. A smart park smoke and fire detection method based on a vision large - model includes the following steps: S1. Arrange the detection system in the park: Install detection components at different parts according to the park terrain, and arrange switches, main control servers, and warning modules at appropriate locations, and then connect them together through networks and optical cables; S2. Run monitoring and collect data: Under the control of the processor, the detection components will start, the camera will monitor a certain range of the park, and through the operation of the electric push rod and the drive motor, the shooting range of the camera will change; S3. Data processing: The data collected by the camera will be transmitted into the processor. Subsequently, the image processing module will perform segmentation and optimization processing on the image data. Then, the smoke and fire judgment module will perform feature recognition on the segmented and optimized image to determine whether there is smoke and fire in the image data. When smoke and fire are found in the image, the risk classification module will perform a secondary judgment on the image to determine what type of fire the smoke and fire are generated from and the level of the fire smoke. When no smoke and fire are found in the image, the risk classification module will not operate; S4. Response processing: After detecting the generation of smoke and fire and the level of the smoke and fire type, the detection component will transmit the detected information to the main control server through the switch, and the main control server will transmit the information to the warning module, and the warning module will send out an alarm message to remind the workers to extinguish the fire in time.

[0014] In the above technical solution, a smart park smoke and fire detection system and method provided by the present invention have the following beneficial effects: (1) The hazard grading module designed by the present invention is divided into a fire and fireworks range grade unit and a fire and fireworks type unit. The fire and fireworks range grade unit can judge the grade of fireworks, and the fire and fireworks type unit can judge what type of fire the fire and fireworks are. In this way, when the fire and fireworks are detected, the range and type of the fireworks can be obtained, which is convenient for workers to use appropriate fire extinguishing methods to quickly extinguish the fire, thereby improving the response speed of fire fighting, accelerating the speed of fire extinguishing, and reducing the losses in the park.

[0015] (2) The smoke and fire judgment module designed by the present invention can detect the type of smoke and fire through the open fire detection algorithm and the dark fire detection algorithm, and judge whether the smoke and fire is an open fire or a dark fire, thereby improving the detection diversity of the detection system so that when a dark fire is found, workers can be notified in time to extinguish the fire.

[0016] (3) When the detection component designed by the present invention detects the generation of fire and smoke, the high beam headlights will light up under the control of the computing module, and will emit light to illuminate the burning ground, so that workers can find the fire location in time at night to extinguish the fire. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0018] Figure 1 A schematic diagram of the system structure of an embodiment of a smart park smoke and fire detection system and method based on a visual large model of the present invention.

[0019] Figure 2 A schematic diagram of the detection component structure provided for an embodiment of a smart park smoke and fire detection system and method based on a visual large model of the present invention.

[0020] Figure 3 A schematic diagram of the planar structure of an installation tube provided for an embodiment of a smart park smoke and fire detection system and method based on a visual large model of the present invention.

[0021] Figure 4 A schematic diagram of the internal structure of a power box provided in an embodiment of a smart park smoke and fire detection system and method based on a visual large model of the present invention.

[0022] Figure 5 A schematic diagram of the computing module structure provided for an embodiment of a smart park smoke and fire detection system and method based on a visual large model of the present invention.

[0023] Figure 6A method flow chart is provided for an embodiment of a smart park smoke and fire detection system and method based on a visual large model of the present invention.

[0024] Description of reference numerals: 1. Detection component; 2. Switch; 3. Main control server; 4. Warning module; 5. Connecting frame; 6. Power box; 7. Connecting rod; 8. Hinge seat; 9. Mounting tube; 10. Electric push rod; 11. High beam; 12. Glass cover; 13. Camera; 14. Computing module; 15. Threaded cover; 16. Drive motor; 17. Signal module; 18. Battery. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0026] like Figure 1-5 As shown, a smart park smoke and fire detection system based on a visual large model provided by an embodiment of the present invention includes a detection component 1, a switch 2, a main control server 3, and an alarm module 4. The output end of the detection component 1 is communicatively connected to the input end of the switch 2 through a network, the output end of the switch 2 is communicatively connected to the input end of the main control server 3 through a network, and the output end of the main control server 3 is communicatively connected to the input end of the alarm module 4 through a network; The detection assembly 1 includes a mounting tube 9, one end of which is threadedly connected to a glass cover 12, a camera 13 is installed inside the mounting tube 9 by bolts, and the camera 13 is preferably a 1.5mm thick, and a computing module 14 is installed inside the mounting tube 9 by bolts, and a threaded cover 15 is threadedly connected to one end of the mounting tube 9. The computer module 14 can be repaired by unscrewing the threaded cover 15. The detection assembly 1 also includes a connecting frame 5, and an electric box 6 is installed on the outer wall of one side of the connecting frame 5 by bolts, and a connecting rod 7 is provided at the bottom of the electric box 6, and an electric box 6 is installed on the outer wall of one side of the connecting rod 7 by a hinge. The electric push rod 10 is preferably a THK model. When the electric push rod 10 is started, the output end of the electric push rod 10 will pull up or push one end of the installation tube 9 to move, so that the installation tube 9 swings up and down around the connecting rod 7 with the hinge seat 8, so as to expand the monitoring distance of the camera 13. The bottom end of the connecting rod 7 is equipped with a hinge seat 8 through a rotating shaft, and the bottom end of the hinge seat 8 is installed with the installation tube 9 through bolts. The output end of the electric push rod 10 is rotatably connected with the installation tube 9 through a hinge, and the outer wall of one side of the bottom of the installation tube 9 is equipped with a high beam lamp 1 through bolts. 1. The high beam 11 is preferably an RJW7101 explosion-proof searchlight. Under the control of the computing module 14, the high beam 11 will light up and emit light to illuminate the fire ground, so that workers can find the fire location in time at night to extinguish the fire. A drive motor 16 is installed on one side of the bottom of the inner wall of the power box 6 by bolts. The drive motor 16 is preferably 57BYGH650-23. When the drive motor 16 is started, the connecting rod 7 will move the electric push rod 10 and the installation cylinder 9 in a circle, so that the shooting position of the camera 13 changes, so as to better monitor the park. Monitoring is performed, and the output end of the drive motor 16 is fixedly connected to the connecting rod 7 through a coupling. A signal module 17 is installed on one side of the top of the inner wall of the power box 6 by bolts. The signal module 17 uses a local area network signal transmitter, such as a router. A battery 18 is provided on one side of the bottom of the inner wall of the power box 6. The battery 18 is a lithium battery that can provide power for the operation of the detection component 1. The battery 18 is externally connected to a photovoltaic device, and the battery 18 is electrically connected to the signal module 17, the drive motor 16, the high beam 11, and the electric push rod 10 through wires; The computing module 14 includes an image processing module, a fireworks judgment module, a danger grading module, a processor, a memory and other auxiliary modules. The image processing module, the fireworks judgment module and the danger grading module are computing programs, and the image processing module, the fireworks judgment module and the danger grading module are loaded on the processor. The memory is used to store data processed by the processor and log data. The results processed by the processor and the data stored in the memory are compressed and transmitted through other auxiliary modules. The image processing module is established based on image segmentation technology and image optimization technology, and is used to segment and optimize the image data collected by the detection component 1. The image segmentation technology in the image processing module specifically uses a region-based segmentation algorithm, and the image optimization technology in the image processing module specifically uses an image enhancement technology; It should be noted that the region-based segmentation algorithm is an image segmentation technique that divides an image into several regions. These regions usually have similar characteristics, such as color, texture, brightness, etc., and have continuity within the image. The goal of this segmentation method is to divide the image into a set of regions, each of which is different from the other regions and can be used for further image analysis and processing. Common region-based segmentation methods include region growing method, split and merge method and energy function-based method; Image enhancement technologies mainly include color transformation, histogram equalization, and homomorphic filtering. Color transformation includes brightness adjustment, contrast adjustment, saturation adjustment, and hue adjustment. ‌‌Histogram equalization is used to improve the contrast of an image, especially when the image has low contrast; ‌‌Homomorphic filtering is used to eliminate the impact of uneven illumination and enhance image details.

[0027] The fireworks judgment module is established based on the fireworks detection algorithm, which is used to perform feature recognition on the image data processed by the image processing module, so as to determine whether there are fireworks in the image data and the type of fireworks. The fireworks detection algorithm includes two types: open fire detection algorithm and dark fire detection algorithm. The open fire detection algorithm uses the ‌LntonAIServer fireworks detection algorithm, and the dark fire detection algorithm uses algorithms based on color detection and algorithms based on edge detection. The specific edge detection algorithm uses Sobel edge detection, and the color detection algorithm uses detection methods based on RGB color space, HSV color space and YUV color space. It should be noted that YOLOv8 is a target detection model that can be used for a variety of tasks, including smoke and open flame detection, providing high-speed and high-precision target detection capabilities. Through real-time analysis of input images, it can accurately identify fire hazards such as smoke and open flames; ‌LntonAIServer smoke and fire detection algorithm is an intelligent monitoring solution based on deep learning technology, which aims to detect and identify signs of smoke and fire in real time and improve the accuracy and efficiency of fire warning. ‌The algorithm uses deep neural networks (DNN), especially convolutional neural networks (CNN), to accurately capture smoke and fire features by training a large number of annotated smoke and fire images, and can provide efficient and accurate detection services in various complex scenarios‌; The Sobel edge detection algorithm is a commonly used edge detection method, mainly used in the fields of image processing and computer vision. It detects edges by calculating the gradient strength of each pixel in the image, and is an edge detection method based on first-order gradients; The principle of the Sobel edge detection algorithm; The Sobel operator contains two sets of 3x3 matrices, which are used for horizontal and vertical edge detection respectively. By convolving these two matrices with the image, we can get the approximate brightness difference of the image in the horizontal and vertical directions. These values ​​are used to calculate the gradient strength of each pixel to determine the edge position; ‌Fire color detection algorithms mainly include detection methods based on RGB color space, HSV color space, and YUV color space, among which: Fire detection algorithm based on RGB color space The RGB color space is the most basic color model, which identifies flames by detecting the red and green components in the image. The red and green components of flames are usually large, and the green component is larger than the blue component. This method is simple and intuitive, but it is easily affected by brightness, so it does not work well when dealing with scenes with large brightness changes‌12.

[0028] Fire detection algorithm based on HSV color space The HSV color space represents color using three dimensions: hue, saturation, and brightness. This method is more adaptable to lighting changes than RGB because the HSV space is insensitive to changes in brightness. By setting a suitable threshold, the flame color can be effectively detected‌ 34.

[0029] Fire Detection Algorithm Based on YUV Color Space The YUV color space decomposes color information into brightness and chromaticity. By detecting changes in the Y channel (brightness) and the U and V channels (chromaticity), flames can be effectively identified.

[0030] The hazard classification module is established based on feature extraction technology, and is used to perform secondary judgment on the image data and surrounding environment image data that the fireworks judgment module judges to have fireworks, so as to judge what kind of fire the fireworks are and the level of the fire and fireworks. The hazard classification module is divided into a fire and fireworks range level unit and a fire and fireworks type unit. The fire and fireworks range level unit judges the level of fireworks by the fire and fireworks range formula. The fire and fireworks range formula is as follows:

[0031] Among them, when S≥1, the level of fire and smoke is a high - risk level; when 0 < S < 1, the level of fire and smoke is a low - risk level. N is the area of the burning area in the image, and M is the maximum safe burning area value set for the actual park. The fire and smoke type unit classifies fire types into six types: A, B, C, D, E, and F. The fire and smoke type level selects the CLIP polymorphic model, which can match the selected image with the corresponding text to determine what type of fire the fire and smoke are caused by. It should be noted that: Class A fire: It refers to a fire caused by solid substances, such as fires caused by the combustion of organic substances such as wood, coal, cotton, wool, hemp, and paper.

[0032] Class B fire: It refers to a fire caused by liquid or meltable solid substances, such as fires caused by the combustion of kerosene, diesel, methanol, ethanol, asphalt, paraffin, etc.

[0033] Class C fire: It refers to a fire caused by gases, such as fires caused by the combustion of gas, natural gas, methane, ethane, propane, hydrogen, etc.

[0034] Class D fire: It refers to a fire caused by metals, such as fires caused by the combustion of potassium, sodium, magnesium, aluminum - magnesium alloy, etc., mostly caused by getting wet and encountering high temperatures and catching fire naturally.

[0035] Class E fire: It refers to an electrical fire, such as fires in switchboards, transformer rooms, weak - current equipment rooms, etc.

[0036] Class F fire: It refers to a fire of the cooking substances in cooking utensils, such as animal and vegetable oils and fats.

[0037] Fire extinguishing methods Class A fire: Water - type fire extinguishers, foam fire extinguishers, dry - powder fire extinguishers, and halon fire extinguishers can be used.

[0038] Class B fire: Dry - powder fire extinguishers, halon fire extinguishers, and carbon dioxide fire extinguishers can be used.

[0039] Class C fire: First, close the gas - conveying valve or pipeline, cut off the power supply, and then cool and extinguish the fire. Dry - powder fire extinguishers and halon fire extinguishers can be used.

[0040] Class D fire: Dry - sand - type, cast - iron powder, or sodium chloride dry - powder special metal fire extinguishers for metal fires can be used (water, foam, and water - based substances are prohibited, and carbon dioxide and dry - powder fire extinguishers cannot be used either).

[0041] Class E fire: Carbon dioxide, dry - powder, and halon fire extinguishers can be used (water is prohibited).

[0042] Class F fire: It can be extinguished by covering the pot lid or using a foam fire extinguisher (water is strictly prohibited).

[0043] The processor module uses one of Intel Core i7 and AMD Ryzen 7, which is used to carry the image processing module, the fireworks judgment module, the danger classification module, and is also used to control the operation of the detection component 1; The memory uses Netac N530S series solid-state hard disk, which is used to temporarily store the processing data of the image processing module, fireworks judgment module, and hazard classification module, as well as the log data of the processor; Other auxiliary modules are used for the processor to exchange information with external devices.

[0044] A method for detecting fireworks in a smart park based on a large visual model, such as Figure 6 As shown, the following steps are included: S1. Park layout detection system: Install detection components 1 in different parts according to the park terrain, and arrange switches 2, master servers 3 and warning modules 4 at appropriate locations, and then connect them together through the network and optical cables; S2. Operation monitoring and data collection: Under the control of the processor, the detection component 1 will be started, the camera 13 will monitor a certain range of the park, and through the operation of the electric push rod 10 and the drive motor 16, the range of the camera 13 will change; S3. Data processing: The data collected by the camera 13 is transmitted to the processor, and then the image processing module performs segmentation and optimization processing on the image data. The subsequent fireworks judgment module performs feature recognition on the segmented and optimized image to determine whether there is fireworks in the image data. When fireworks are found in the image, the hazard classification module performs a secondary judgment on the image to determine what kind of fire the fireworks are and the level of the fire. When no fireworks are found in the image, the hazard classification module does not operate; S4. Response processing: After detecting the generation of fireworks and the type and level of fireworks, the detection component 1 transmits the detection information to the main control server 3 through the switch 2, and the main control server 3 transmits the information to the warning module 4, and the warning module 4 sends an alarm message to remind the workers to put out the fire in time.

[0045] Working principle: When applying the detection system, workers can install the detection component 1 in different parts according to the terrain of the park, and arrange the switch 2, the main control server 3 and the warning module 4 at the appropriate location, and then connect the detection component 1, the switch 2, the main control server 3 and the warning module 4 together through the network and the optical cable; subsequently, under the control of the processor in the computing module 14, the detection component 1 will be started, wherein the camera 13 will be turned on to monitor a certain range of the park, and the electric push rod 10 will be started, and the output end of the electric push rod 10 will pull up or push one end of the installation tube 9 to move, so that the installation tube 9 with the hinge seat 8 swings up and down around the connecting rod 7, so as to expand the monitoring distance of the camera 13, and the driving motor 16 will be started, and the connecting rod 7 will carry the electric push rod 10 and the installation tube 9 to do a circular motion, so that the shooting position of the camera 13 changes, so as to better monitor the park; during monitoring, the data collected by the camera 13 will be transmitted to the processor, and then the image processing module in the processor will segment and optimize the image data, and the optimization processing After that, the fireworks judgment module will perform feature recognition on the segmented optimized image to determine whether there are fireworks in the image data. When fireworks are found in the image, the hazard classification module will perform a secondary judgment on the image to determine what kind of fire the fireworks are and the level of the fire. After detecting the generation of fireworks and the type and level of fireworks, the processor will transmit the fire information to the signal module 17 through other auxiliary modules. Then the signal module 17 will transmit the information to the switch 2, and the switch 2 will transmit the information to the main control server 3. The main control server 3 will transmit the information to the warning module 4. The warning module 4 will send out an alarm message to remind the workers that there is a fire in the park, so that the workers can extinguish the fire in the park in time. At the same time, when fire and fireworks are found, through the operation of the drive motor 16 and the electric push rod 10, the camera 13 will face the fire position, so that the workers can understand the situation of the fire and fireworks during the process of extinguishing the fire. At the same time, under the control of the computing module 14, the high beam 11 will light up and emit light to illuminate the burning ground, so that the workers can find the fire position in time at night to extinguish the fire.

[0046] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A smart park fire and smoke detection system based on a visual large model, comprising a detection component (1), a switch (2), a main control server (3), and an alarm module (4), characterized in that: The output end of the detection component (1) is communicatively connected to the input end of the switch (2) via a network, the output end of the switch (2) is communicatively connected to the input end of the main control server (3) via a network, and the output end of the main control server (3) is communicatively connected to the input end of the warning module (4) via a network; The detection assembly (1) comprises a mounting tube (9), one end of the mounting tube (9) being threadedly connected to a glass cover (12), a camera (13) being installed inside the mounting tube (9) via bolts, a computing module (14) being installed inside the mounting tube (9) via bolts, and one end of the mounting tube (9) being threadedly connected to a threaded cover (15); The computing module (14) comprises an image processing module, a fireworks judgment module, a hazard grading module, a processor, a memory and other auxiliary modules, wherein the image processing module, the fireworks judgment module and the hazard grading module are computing programs, and the image processing module, the fireworks judgment module and the hazard grading module are loaded on the processor, and the memory is used to store data processed by the processor and log data, and the results of the processing by the processor and the data stored in the memory are compressed and transmitted through other auxiliary modules; The image processing module is established based on image segmentation technology and image optimization technology, and is used to segment and optimize the image data collected by the detection component (1); The fireworks judgment module is established based on the fireworks detection algorithm and is used to perform feature recognition on the image data processed by the image processing module, so as to judge whether fireworks are generated in the image data and the type of fireworks; The hazard classification module is established based on feature extraction technology and is used to perform secondary judgment on the image data and surrounding environment image data judged by the fireworks judgment module to have fireworks, so as to judge what kind of fire the fireworks are and the grade of the fire and fireworks; The processor module is selected from one of Intel Core i7 and AMD Ryzen 7, and is used to carry the image processing module, the fireworks judgment module, the danger classification module, and is also used to control the operation of the detection component (1); The memory uses a Netac N530S series solid-state hard disk to temporarily store the image processing module, the fireworks judgment module, the danger classification module processing data and the processor log data; The other auxiliary modules are used for the processor to exchange information with external devices.

2. According to the smart park smoke and fire detection system based on visual big model according to claim 1, it is characterized in that: The image segmentation technology in the image processing module specifically selects a region-based segmentation algorithm, and the image optimization technology in the image processing module specifically selects an image enhancement technology.

3. According to the smart park smoke and fire detection system based on visual big model as described in claim 1, it is characterized in that: The smoke and fire detection algorithm includes two types: open fire detection algorithm and dark fire detection algorithm. The open fire detection algorithm selects the LntonAIServer smoke and fire detection algorithm, and the dark fire detection algorithm selects an algorithm based on color detection and an algorithm based on edge detection. Specifically, the edge detection-based algorithm selects Sobel edge detection, and the color detection-based algorithm selects a detection method based on RGB color space, HSV color space and YUV color space.

4. According to the smart park smoke and fire detection system based on visual big model as described in claim 1, it is characterized in that: The risk classification module is divided into a range level unit of fire smoke and a type unit of fire smoke. The range level unit of fire smoke judges the level of the smoke through the range formula of fire smoke. The range formula of fire smoke is as follows: Among them, when S≥1, the level of fire smoke is a high-risk level; when 0<S<1, the level of fire smoke is a low-risk level. N is the area of the着火 area in the image, and M is the maximum safe着火 area value set for the actual park.

5. According to the smart park smoke and fire detection system based on visual big model according to claim 4, it is characterized in that: The type unit of fire smoke classifies the types of fires into six types: A, B, C, D, E, and F. The fire smoke type level selects the CLIP polymorphic model, which can match the selected image with the corresponding text to judge what type of fire the fire smoke is generated from.

6. The smart park smoke and fire detection system based on visual big model according to claim 1 is characterized in that: The detection component (1) further includes a connecting frame (5). One outer wall of the connecting frame (5) is installed with a power box (6) through bolts, and a connecting rod (7) is arranged at the bottom of the power box (6). One outer wall of the connecting rod (7) is installed with an electric push rod (10) through a hinge.

7. The smart park smoke and fire detection system based on visual big model according to claim 6 is characterized in that: The bottom end of the connecting rod (7) is installed with a hinge seat (8) through a rotating shaft, and the bottom end of the hinge seat (8) is installed together with the installation cylinder (9) through bolts.

8. The smart park smoke and fire detection system based on visual big model according to claim 7 is characterized in that: The output end of the electric push rod (10) is rotatably connected to the installation cylinder (9) through a hinge. One outer wall of the bottom side of the installation cylinder (9) is installed with a high beam lamp (11) through bolts.

9. The smart park smoke and fire detection system based on visual big model according to claim 8 is characterized in that: One side of the bottom inner wall of the power box (6) is installed with a driving motor (16) through bolts, and the output end of the driving motor (16) is fixedly connected to the connecting rod (7) through a coupling. One side of the top inner wall of the power box (6) is installed with a signal module (17). One side of the bottom inner wall of the power box (6) is provided with a storage battery (18), and the storage battery (18) is electrically connected to the signal module (17), the driving motor (16), the high beam lamp (11), and the electric push rod (10) through wires respectively.

10. A method for detecting fire and smoke in a smart park based on a visual big model, comprising a method for detecting fire and smoke in a smart park based on a visual big model as described in any one of claims 1 to 9, characterized in that: It includes the following steps: S1. Deploy the detection system in the park: Install the detection component (1) at different parts according to the terrain of the park, and deploy the switch (2), the main control server (3), and the warning module (4) at appropriate locations, and then connect them together through the network and optical cable; S2. Run monitoring and collect data: Under the control of the processor, the detection component (1) will be started, and the camera (13) will monitor a certain range of the park. And through the operation of the electric push rod (10) and the driving motor (16), the shooting range of the camera (13) will change; S3. Data processing: The data collected by the camera (13) will be transmitted into the processor. Subsequently, the image processing module will perform segmentation and optimization processing on the image data. Then, the smoke judgment module will perform feature recognition on the segmented and optimized image to judge whether there is smoke in the image data. When it is found that there is smoke in the image, the risk classification module will perform a secondary judgment on the image to judge what type of fire the smoke is generated from and the fire smoke level. When it is found that there is no smoke in the image, the risk classification module will not operate; It should be noted that there is an unclear expression "着火 area" in the original text which should be replaced with a more accurate term. Also, the English translation should be adjusted according to the correct content. S4. Response processing: After detecting the generation of fireworks and the type and level of fireworks, the detection component (1) transmits the detection information to the main control server (3) through the switch (2), and the main control server (3) transmits the information to the warning module (4), and the warning module (4) issues an alarm message to remind workers to extinguish the fire in time.

Images