Mine fire identification and alarm method based on flame and smoke images

By using near-infrared explosion-proof cameras equipped with filters and fill lights in mines, combined with image processing methods, quickly identifying and distinguishing fire and interference sources, the high false alarm rate and missed rate of fire monitoring in the existing technology are solved. The problem of this problem has achieved rapid and accurate fire identification and alarm, ensuring the safety of underground personnel.

CN119028078APending Publication Date: 2024-11-26CHINA UNIV OF MINING & TECH (BEIJING)
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202410895055.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-05
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

The existing mine fire monitoring technology has a high false alarm rate and a low false alarm rate in terms of identification and alarm, and the judgment time is long, making it difficult to quickly and accurately identify the fire situation, affecting timely emergency rescue.

Method used

A near-infrared explosion-proof camera equipped with filters and fill lights is used to perform periodic cyclic fill light. Combined with image processing methods, the characteristics of flame and smoke images are analyzed, and the fire and interference sources are distinguished, background noise interference is reduced, and rapid identification is achieved. and accurate alarm.

Benefits of technology

It effectively shortens the judgment time for fire identification, reduces the false alarm and missed alarm rates, improves detection speed and accuracy, and wins more escape time for underground personnel.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119028078A_ABST
    Figure CN119028078A_ABST
Patent Text Reader

Abstract

The invention discloses a mine fire recognition and alarm method based on flame and smoke images, and the method comprises the steps: collecting and recognizing the features of a suspected fire image in a monitoring region in real time through installing a near-infrared camera underground; and mine fire identification and alarm are carried out by monitoring the characteristics that early smoke and middle and later flame combustion have irregular forms. The fire identification and alarm method fully considers the image features of the early smoldering stage and the middle and later combustion stages of the fire, is rapid, accurate and simple in identification, and can strive for precious rescue and escape time for underground persons in distress.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a mine fire identification and alarm method based on flame and smoke images, and the method relates to the fields of digital image processing technology and communication technology. Background Art

[0002] In the coal industry, accidents such as gas, fire, flood, roof, and coal dust plague coal mine safety production. Data show that among the statistics of major accidents in my country's coal mines, mine fires and explosions caused by fires account for the highest proportion of accidents and are the most serious. Once a mine fire occurs, if the fire cannot be controlled in time, the scope of the impact will expand rapidly, causing a large number of casualties and property losses; it may further serve as a detonation source to cause explosions in areas with high concentrations of gas and coal dust underground, causing secondary trauma to the underground environment. Therefore, it is crucial for coal safety production to quickly identify underground fire disasters, report to the police in a timely manner, and launch emergency plans and emergency rescue in a targeted manner.

[0003] The existing mine fire monitoring technology includes comprehensive monitoring methods using various sensors, such as temperature, smoke, gas and other sensors. Although this type of monitoring has the advantages of high reliability and simple operation, it is complex due to the underground operation scene, and its layout and maintenance workload is large and it is greatly affected by interference sources; the infrared radiation temperature measurement equipment monitoring method can measure the temperature of the fire source, but the temperature measurement accuracy is greatly affected by underground environmental factors and interference sources; the visual feature monitoring method is the current mainstream mine fire monitoring technology, but the existing monitoring technology uses more characteristic variables to determine the fire. Although the information features are rich, they also contain more unstable factors. The comprehensive determination of the fire disaster based on multi-lens and multi-frame images will result in a longer determination time and a higher underreporting rate and false alarm rate.

[0004] Therefore, it is necessary to study new methods for rapid identification of mine fires, shorten the judgment time, and reduce the false alarm rate and missed alarm rate of existing monitoring technologies for mine fire perception. Early detection of mine fires and the ability to alarm fire disasters in the first place are important guarantees for timely emergency rescue and saving the lives of people in distress underground. Summary of the invention

[0005] The purpose of the present invention is to provide a mine fire identification and alarm method based on flame and smoke images, which method fully considers the image characteristics of the smoldering and burning stages of the fire. The main factor affecting fire identification in coal mines is the mine light source, while the contour boundary of the fire flame image is irregular, and the contour boundary of the mine light source image is regular. Therefore, based on the above characteristics, the present invention distinguishes and identifies the mine fire flame and its fire interference light source, and innovatively uses the monitoring device and image processing method provided by the present invention, which can effectively reduce background noise interference, can quickly identify fire and interference sources, accurately and simply, and can buy more rescue and escape time for people in distress underground. The fire flame and smoke image recognition and alarm method includes installing a near-infrared explosion-proof camera equipped with a filter + fill light in the electromechanical chamber, tunnel, coal mining face and excavation face to monitor the monitoring area in real time. The fill light performs periodic cycle fill light on the area monitored by the camera in real time according to the set fill light pulse duration P1 and non-fill light pulse duration P2, wherein N1 and N2 are pulse duration coefficients; T is the camera video cycle, in ms; F is the camera frame rate; within the fill light pulse duration P1, a fire warning is performed based on whether the characteristics of the suspected fire target area on the suspected fire image meet the fire smoke characteristic setting conditions; a level 1 fire alarm is performed based on whether the characteristics of the suspected fire target area on the suspected fire image meet the fire smoke and flame characteristic setting conditions; within the non-fill light pulse duration P2, a level 2 fire alarm is performed based on whether the characteristics of the suspected fire target area on the suspected fire image meet the fire flame characteristic setting conditions; when the number of level 1 fire alarms U1 and the number of level 2 fire alarms U2 that continuously appear within the set time S meet the setting condition of (U1+2U2)≥U, a level 3 fire alarm is performed, where S≥(P1+P2), U≥3.

[0006] 1. The fire identification and alarm method further comprises: the working process of the fire alarm determination comprises:

[0007] Step 1: The fill light performs periodic cycle fill light on the area monitored by the camera in real time according to the set fill light pulse duration P1 and non-fill light pulse duration P2, and then step 2 is further executed;

[0008] Step 2: A near-infrared explosion-proof camera equipped with a filter and a fill light collects video images of the monitoring area in real time, and the monitoring video images are transmitted to the back end of the monitoring server in real time, and the monitoring video is framed, image denoised, and image enhanced in real time to obtain a continuous frame monitoring image, and step 3 is further executed;

[0009] Step 3: When the fill light is filling light for the area monitored by the camera in real time, further execute step 4; otherwise, execute step 9;

[0010] Step 4: Determine the pixel characteristic value and M of the frame image within the fill light pulse duration P1 frame by frame r , when the pixel feature value of the rth frame image and F1≤M r <F2(r≥1), the frame image is determined to be a suspected fire smoke target frame image, and step 5 is executed; when the pixel feature value of the rth frame image and M r ≥F2 and (F2>F1 and F1 and F2 are set thresholds), the frame image is determined to be a suspected fire smoke and flame target frame image, and step 6 is executed; when the above suspected fire image determination conditions are not met, return to step 1 for continuous monitoring;

[0011] Step 5: Select the processed N (N≥5) consecutive odd-numbered frames including the rth frame image to enhance the image of the suspected smoke target area; N Perform binarization operation and further execute step 7;

[0012] Step 6: Select the processed continuous K (K ≥ 3) odd-numbered frames including the rth frame image to enhance the image of the suspected smoke and flame target area; K Perform a binarization operation and further execute step 8;

[0013] Step 7: Determine whether the suspected fire target area feature on the suspected fire image meets the fire smoke feature setting condition. If so, a fire smoke warning is issued and step 11 is further executed. Otherwise, the execution returns to step 1.

[0014] Step 8: Determine whether the characteristics of the suspected fire target area on the suspected fire image meet the fire smoke and flame characteristic setting conditions. If so, a fire level 1 alarm is issued and step 13 is further executed. Otherwise, the execution returns to step 1.

[0015] Step 9: Determine the pixel characteristic value and M of the frame image within the non-fill light pulse duration P2 frame by frame n , when the pixel feature value of the nth frame image and M n ≥F3 (n≥1 and F3 is the set threshold), the frame image is determined to be a suspected fire flame target frame image, and step 10 is executed; when the above suspected fire image determination conditions are not met, return to step 1 for continuous monitoring;

[0016] Step 10: Binarize the suspected fire flame target frame image; determine whether the suspected fire target area feature on the suspected fire image meets the fire flame feature setting conditions. If so, a fire level 2 alarm is issued and step 15 is further executed; otherwise, return to step 1;

[0017] Step 11: Determine whether to respond to the fire warning. If yes, proceed to step 12; otherwise, return to step 1.

[0018] Step 12: Release the fire warning response and return to continuous monitoring;

[0019] Step 13: Determine whether to respond to the fire alarm level 1, if yes, further execute step 14, otherwise return to step 1;

[0020] Step 14: cancel the fire level 1 alarm response, delete the level 1 alarm signal that exceeds the set time, calculate the number of fire level 1 alarms within the set time, and execute step 17;

[0021] Step 15: Determine whether to respond to the fire level 2 alarm. If yes, go to step 16, otherwise go back to step 1.

[0022] Step 16: cancel the fire level 2 alarm response, delete the level 2 alarm signal that exceeds the set time, calculate the number of fire level 2 alarms within the set time, and execute step 17;

[0023] Step 17: Determine whether the continuous occurrence of fire level 1 alarm and fire level 2 alarm meets the setting conditions of fire level 3 alarm. If yes, further execute step 18, otherwise return to execute step 1;

[0024] Step 18: Perform a fire level 3 alarm, and then proceed to step 19;

[0025] Step 19: Determine whether to respond to the fire alarm level 3, if yes, go to step 20, otherwise go back to step 1;

[0026] Step 20: Cancel the fire level 3 alarm response, clear the number of fire level 3 alarms, and return to continuous monitoring.

[0027] 2. The fire identification and alarm method further comprises: whether the suspected fire target area characteristics on the suspected fire image meet the fire smoke characteristic setting conditions, that is, according to the target boundary connected coordinate point group {(X a1 ,Y a1 ),(X a2 ,Y a2 ),···(X ai ,Y aj )}, extract the suspected target image contour, and arbitrarily select a point coordinate (X e ,Y e ), make L through this coordinate point d (d>1 and N + ) straight lines and intersect with the suspected target image contour. When there are multiple straight lines L zWhen the number of intersections between (z≤d) and the boundary of the suspected target image contour is greater than 2, it is determined that the fire smoke feature setting conditions are met, where X ai , Y aj are the horizontal and vertical coordinate point values ​​of the image boundary, respectively; ai and aj are the subscript constants of the horizontal and vertical coordinate points of the image boundary.

[0028] 3. The fire identification and alarm method further comprises: whether the suspected fire target area characteristics on the suspected fire image meet the fire smoke and flame characteristics setting conditions, that is, according to the target boundary connected coordinate point group {(X a1 ,Y a1 ),(X a2 ,Y a2 ),···(X ai ,Y aj )}, extract the suspected target image contour, and arbitrarily select a point coordinate (X v ,Y v ), make L through this coordinate point g (g>1 and N + ) straight lines and intersect with the suspected target image contour. When there are multiple straight lines L b When the number of intersections between (b≤g) and the boundary of the suspected target image contour is greater than 3, it is determined that the fire smoke and flame feature setting conditions are met.

[0029] 4. The fire identification and alarm method further comprises: whether the suspected fire target area characteristics on the suspected fire image meet the fire flame characteristic setting conditions, that is, according to the suspected target image boundary connected coordinate points {(X h1 ,Y h1 ),(X h2 ,Y h2 ),···(X hi ,Y hj )}, extract the suspected target image contour, and arbitrarily select a point coordinate (X u ,Y u ), make L through this coordinate point s (s>1 and N + ) straight lines and intersect with the suspected target image contour. When there are multiple straight lines L w When the number of intersections between (w≤s) and the boundary of the suspected target image contour is greater than or equal to 4, it is determined that the fire flame feature setting conditions are met, where X hi , Y hj are the horizontal and vertical coordinate point values ​​of the image boundary, hi and hj are the subscript constants of the horizontal and vertical coordinate points of the image boundary.

[0030] 5. The fire identification and alarm method further includes: the filter is a long-wave pass filter with a wavelength greater than 780nm; the filter performs real-time and uninterrupted filtering when it is carried out with an explosion-proof camera for monitoring.

[0031] 6. The fire identification and alarm method further includes: the fill light is a fill light device that periodically emits near-infrared band.

[0032] 7. The fire identification and alarm method further includes: the periodic cycle fill light is a fill light pulse duration P1 + a non-fill light pulse duration P2 as a cycle fill light period.

[0033] 8. The fire identification and alarm method further includes: selecting the processed N (N≥5) consecutive odd-numbered frames including the rth frame image in step 5 to perform image enhancement of the suspected smoke target area; the image enhancement of the suspected smoke target area is to perform equidistant difference frame and superposition operations on the selected consecutive multiple frame images.

[0034] 9. The fire identification and alarm method further includes: selecting the processed continuous K (K≥3) odd-numbered frames including the rth frame image in step 6 to perform image enhancement of the suspected smoke and flame target area; the suspected smoke and flame target area image enhancement is to perform equidistant difference frame and superposition operations on the selected continuous multiple frame images.

[0035] 10. The fire identification and alarm method further comprises: in step 5, the enhanced target frame image Q N Perform binarization operation, that is, when the single pixel value M of the target frame image (i,j) If it is greater than the set threshold Q1, the pixel value M (i,j) =1, otherwise M (i,j) =0, where (i, j) is the pixel coordinate of the frame image.

[0036] 11. The fire identification and alarm method further comprises: in step 6, the enhanced target frame image Q K Perform binarization operation, that is, when the single pixel value R of the target frame image (i,j) If it is greater than the set threshold Q2 (Q2>Q1), then R (i,j) =1, otherwise R (i,j) =0.

[0037] 12. The fire identification and alarm method further comprises: in step 10, the suspected fire flame target frame image is binarized, that is, when the single point pixel value E (i,j) If it is greater than the set threshold Q3, then it is E (i,j) =1, otherwise E (i,j) =0.

[0038] 13. The fire identification and alarm method further comprises: the coordinates (X e ,Y e ), the point coordinates (X e ,Y e ) must meet the following conditions Where X min , X max Y is the minimum and maximum value of the horizontal coordinate points among the connected coordinate points on the boundary of the suspected target image. min , Y max is the minimum and maximum value of the ordinate points among the connected coordinate points on the boundary of the suspected target image; there are multiple straight lines L z When the number of intersection points with the boundary of the suspected target image contour is greater than 2, it is determined that the fire smoke feature setting condition is met. To meet the fire smoke feature setting condition, the number of straight lines z must meet the condition z≥5.

[0039] 14. The fire identification and alarm method further comprises: the coordinates (X v ,Y v ), the point coordinates (X v ,Y v ) must meet the following conditions Where X min , X max Y is the minimum and maximum value of the horizontal coordinate points among the connected coordinate points on the boundary of the suspected target image. min , Y max is the minimum and maximum value of the ordinate points among the connected coordinate points on the boundary of the suspected target image; there are multiple straight lines L b When the number of intersections with the boundary of the suspected target image contour is greater than 3, it is determined that the fire smoke and flame feature setting conditions are met. To meet the fire smoke and flame feature setting conditions, the number of straight lines b must meet the condition b≥4.

[0040] 15. The fire identification and alarm method further comprises: the coordinates (X u ,Y u ), the point coordinates (X u ,Y u ) must meet the following conditions Where X min , X max Y is the minimum and maximum value of the horizontal coordinate points among the connected coordinate points on the boundary of the suspected target image. min , Y max is the minimum and maximum value of the ordinate points among the connected coordinate points on the boundary of the suspected target image; there are multiple straight lines L sWhen the number of intersection points with the boundary of the suspected target image contour is greater than or equal to 4, it is determined that the fire flame feature setting condition is met. To meet the fire flame feature setting condition, the number of straight lines w must meet the condition w≥3.

[0041] 16. The fire identification and alarm method further comprises: performing equidistant difference frame and superposition operation on the selected continuous multiple frame images, the equidistant difference frame and superposition operation is to perform non-interval frame difference operation, one-frame difference operation, two-frame difference operation and three-frame difference operation on the multiple frame images respectively, and finally superimpose the results of the four frame image difference operations, and superimpose and merge them with the r frame image again to form a new smoke frame image;

[0042] The example of selecting consecutive odd-numbered frames of images with N=5 is as follows: when r≥3, the rth frame image is taken as the middle frame, and 5 consecutive frames of images are selected to enhance the suspected smoke target image. The target image Q N Requirements:

[0043]

[0044] When r=1, 5 consecutive frames are selected with r+2 as the intermediate frame to enhance the suspected smoke target image. N Requirements:

[0045]

[0046] When r=2, take r+1 as the intermediate frame and select 5 consecutive frames to enhance the suspected smoke target image; the target image Q N Requirements:

[0047]

[0048] 17. The fire identification and alarm method further comprises: performing equidistant difference frame and superposition operation on the selected continuous multiple frame images, the equidistant difference frame and superposition operation is to perform non-interval frame difference operation and one-frame difference operation on the multiple frame images respectively, and finally superimpose the results of the two frame image difference operations, and then superimpose and merge them with the r frame image again to form a new flame frame image;

[0049] The example of selecting consecutive odd-numbered frames of images with K=3 is as follows: when r≥2, the rth frame image is taken as the middle frame, and three consecutive frames of images are selected to enhance the suspected flame target image. The target image Q K Requirements:

[0050] Q K =[[r-(r-1)]+[(r+1)-r]+[(r+1)-(r-1)]+r];

[0051] When r=1, three consecutive frames are selected with r+1 as the intermediate frame to enhance the suspected smoke target image. The target image Q K Requirements:

[0052] Q K =[[(r+1)-r]+[(r+2)-(r+1)]+[(r+2)-r]+r].

[0053] The mine fire identification and alarm method based on flame and smoke images has the following characteristics:

[0054] 1. The monitoring device of the present invention has obvious advantages: a near-infrared explosion-proof camera equipped with filtering + fill light is used to collect underground monitoring video images, and the fill light is used for periodic cycle fill light. When identifying fire smoke and flames, the fill light (near-infrared band radiation) is added, which can significantly enhance the brightness of smoke and flame images, effectively filter out the interference of background noise, and improve the detection speed and accuracy; secondly, since the flame itself has strong near-infrared radiation, it is not necessary to add a fill light in identifying the flame. Further, based on other devices of the present invention and the image processing method used in the present invention, the flame image can also be efficiently identified.

[0055] 2. The fire smoke recognition method of the present invention fully analyzes the image contour features of the smoldering and burning stages of the fire, and fully analyzes the image contour features of the interference sources of suspected fires underground. The method performs fire judgment based on the coordinate information of the intersection of the image contour and a straight line passing through any point in the contour, which is more conducive to quickly identifying and eliminating the interference sources of suspected fires, and can reduce false alarms and omissions of fire disasters.

[0056] 3. The method of the present invention not only efficiently utilizes the characteristic information of the fire image, but also the algorithm of the image processing part is concise and efficient, and a single-frame image can quickly determine the fire information. The feature extraction of the method has high recognition, few variables, and little background interference, thereby shortening the time for distinguishing suspected fire smoke, and further buying more escape time for trapped and affected people underground. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 Schematic diagram of the implementation scheme of the mine fire identification and alarm method based on flame and smoke images.

[0058] Figure 2 Schematic diagram of the structure of fire monitoring and alarm equipment.

[0059] Figure 3 Schematic diagram of the fire determination process of the mine fire recognition and alarm method based on flame and smoke images. DETAILED DESCRIPTION

[0060] Figure 1This is an example of an implementation plan for a mine fire identification and alarm method based on flame and smoke images. The main components include:

[0061] 1. A storage server (101), wherein the storage server is connected in communication with an image acquisition camera (105) and is responsible for storing and forwarding real-time video image data of a monitoring area provided by the camera; the storage server is connected in communication with a fire monitoring alarm device (106) and is responsible for storing and forwarding fire smoke alarm data provided by the fire monitoring alarm device; and provides a monitoring terminal (102) with a service for viewing on-site monitoring data of a monitored area or retrieving historical monitoring image data.

[0062] 2. A monitoring terminal (102) is installed on the well and is used to receive the alarm signal from the fire monitoring alarm device (106); it is responsible for providing underground environment monitoring data display services, and the storage server (101) provides real-time, historical data and fire alarm data, and has sound and light alarm functions; production management personnel can retrieve and query the historical data stored in the storage server (101) through the monitoring terminal.

[0063] 3. Core switch (103), the core management and switching device of mining Ethernet, is responsible for the management and data exchange of all devices connected to mining Ethernet, has routing function, and is connected to the Internet.

[0064] 4. Ring network switch (104), underground switching equipment of mining Ethernet, installed underground, multiple ring network switches are connected in a ring network manner.

[0065] 5. An explosion-proof camera (105), an image acquisition device, is installed in underground electromechanical chambers, tunnels, coal mining faces and excavation faces, and is responsible for acquiring real-time video images of fire-prone areas such as underground tunnels and faces; the camera comprises a near-infrared explosion-proof camera equipped with a filter + fill light; the camera has network output and analog video output functions; the network interface is directly connected to the ring network switch (104), and the real-time acquired video image data of the monitoring area is transmitted to the storage server (101); the analog video output port is connected to the fire monitoring alarm device (106).

[0066] 6. Fire monitoring and alarm equipment (106), which is responsible for receiving and processing the monitoring video image data collected by the camera (105), has built-in graphic image processing software, and when the processed suspected fire image meets the set fire alarm conditions, it sends fire alarm data to the storage server (101); and sends a fire smoke alarm signal to the monitoring terminal (102); it has wired and wireless communication functions.

[0067] 7. Suspected fire area: The camera monitors the suspected smoke or fire points in the area in real time. The image data is monitored and collected and uploaded by the camera in real time.

[0068] Figure 2 It is a schematic diagram of the structure of a fire monitoring and alarm device, the main components of which are:

[0069] Core processor, graphics processor, storage unit, clock unit, power supply unit, USB interface unit, video image acquisition module, wireless communication unit, network interface unit, SD card interface unit.

[0070] 1. Core processor (201), using Broadcom BCM2837 processor, using ARM Cortex-A53 architecture, 64-bit quad-core 1.2GHz.

[0071] 2. Graphics processor (202), which adopts Dual Core VideoCore IV GPU processor.

[0072] 3. Storage unit (203), using 1GB LPDD2 memory.

[0073] 4. Clock unit (204), using 19.2MHz crystal oscillator.

[0074] 5. The power supply unit (205) adopts an AC / DC module, with an input of 100V~240VAC and an output of 12VDC, which is used to power the equipment.

[0075] 6. USB interface unit (206), supporting 4 USB interfaces.

[0076] 7. The video image acquisition module (207) converts the analog video signal into digital video data, inputs the analog video output port connected to the camera (105), transmits the digital video data to the core processor (201) through the USB port, and supports multi-channel video acquisition.

[0077] 8. Communication module (208), responsible for converting the RS485 communication interface into a USB communication interface to connect other data monitoring devices.

[0078] 9. A wireless communication unit (209), supporting 802.11b / g / n protocol, and used for wireless communication with a monitoring device supporting wireless communication.

[0079] 10. The network interface unit (210) is responsible for accessing the mining Ethernet and connecting to the ring network switch (104).

[0080] 11. SD card unit (211), used to store system files, library files, monitoring program files, etc., using Linux system management, built-in OpenCV library for video data processing, using a Micro SD card of no less than 4GB.

[0081] An example of the fire determination process of the mine fire recognition and alarm method based on flame and smoke images is as follows: Figure 3 As shown, the process includes:

[0082] 1. (301) Cameras are installed in underground electromechanical chambers, tunnels, coal mining faces and excavation faces to collect images of the monitored area in real time and upload them to fire monitoring and alarm equipment for image recognition processing, and further execute step (302).

[0083] 2. (302) The fire monitoring alarm device pre-processes the monitoring image, and the monitoring image is subjected to image denoising and image enhancement to obtain a continuous frame monitoring image. Execute step (303).

[0084] 3. (303) Determine whether to supplement the camera monitoring area with near-infrared light. If so, execute step (304); otherwise, execute step (322).

[0085] 4. (304) Determine the pixel feature value and M of the frame image frame by frame r , when the pixel feature value of the rth frame image and F1≤M r <F2(r≥1), then the frame image is determined to be a suspected fire smoke target frame image, and step (405) is executed; otherwise, step (312) is executed;

[0086] 5. (305) Select the processed N (N≥5) consecutive odd-numbered frames including the rth frame image to enhance the image of the suspected smoke target area; N Perform a binarization operation and execute step (306).

[0087] 6. (306) According to the connected coordinate points (X ai ,Y aj ), extract the suspected target image contour, and arbitrarily select a point coordinate (X e ,Y e ), make L through this coordinate point d (d is a positive integer) straight lines, execute step (307).

[0088] 7.(307) Find the straight line L d If the number of straight lines z whose intersections with the image contour boundary are greater than 2, execute step (308).

[0089] 8. (308) Determine the straight line L d Whether the number z of straight lines whose number of intersection points with the image contour boundary is greater than 2 satisfies the condition z≥5, if so, execute step (309), otherwise execute step (301).

[0090] 9. (309) Issue a fire warning and execute step (322).

[0091] 10. (310) Determine whether the fire warning response is activated. If it is activated, execute step (311), otherwise return to execute step (301).

[0092] 11. (311) Release the fire warning response and return to continuous monitoring.

[0093] 12.(312) Determine the pixel feature value and M of the frame image frame by frame r , when the pixel feature value of the rth frame image and M r ≥F2(r≥1), then the frame image is determined to be a suspected fire smoke and flame target frame image, and step (313) is executed; otherwise, step (301) is executed;

[0094] 13. (313) Select the processed continuous K (K ≥ 3) odd-numbered frames including the rth frame image to enhance the image of the suspected smoke target area; K Perform a binarization operation and execute step (314).

[0095] 14.(314) According to the connected coordinate points (X ai ,Y aj ), extract the suspected target image contour, and arbitrarily select a point coordinate (X v ,Y v ), make L through this coordinate point g (g is a positive integer) straight lines, execute step (315).

[0096] 15.(315) Find the straight line L g If the number of straight lines b whose intersections with the image contour boundary are greater than 3, execute step (316).

[0097] 16.(316) Determine the straight line L g Whether the number b of straight lines whose intersections with the image contour boundary are greater than 3 satisfies the condition b≥4, if so, execute step (317), otherwise execute step (301).

[0098] 17. (317) Conduct a Level 1 fire alarm and execute step (318).

[0099] 18. (318) Determine whether to initiate a Level 1 fire alarm response. If so, execute step (319); otherwise, return to execute step (301).

[0100] 19. (319) Cancel the fire level 1 alarm response and execute step (320).

[0101] 20. (320) Delete the fire level 1 alarm response that exceeds time S, calculate the number of level 1 alarms within time S, and execute step (321).

[0102] 21. (321) Determine whether the continuous occurrence of fire level 1 alarm and fire level 2 alarm satisfies the setting condition of fire level 3 alarm (U1+2U2)≥U. If so, execute step (331); otherwise, return to execute step (301).

[0103] 22.(322) Determine the pixel feature value and M of the frame image frame by frame n , when the pixel feature value of the nth frame image and M n ≥F3(n≥1), then the frame image is determined to be a suspected fire flame target frame image, and step (323) is executed; otherwise, step (301) is executed;

[0104] 23. (323) Perform a binarization operation on the suspected fire flame target frame image and execute step (324).

[0105] 24.(324) According to the connected coordinate points (X hi ,Y hj ), extract the suspected target image contour, and arbitrarily select a point coordinate (X u ,Y u ), make L through this coordinate point s (s is a positive integer) straight lines, execute step (325).

[0106] 25.(325) Find the straight line L s The number of straight lines w whose intersection points with the image contour boundary is greater than or equal to 4 is determined, and step (326) is executed.

[0107] 26.(326) Determine the straight line L s Whether the number w of straight lines whose number of intersection points with the image contour boundary is greater than or equal to 4 satisfies the condition w≥3, if so, execute step (327), otherwise execute step (301).

[0108] 27. (327) Conduct a Level 2 fire alarm and execute step (328).

[0109] 28. (328) Determine whether to initiate a Level 2 fire alarm response. If so, proceed to step (329); otherwise, return to step (301).

[0110] 29. (329) Cancel the fire level 2 alarm response and execute step (330).

[0111] 30. (330) Delete the fire level 2 alarm response that exceeds time S, calculate the number of level 2 alarms within time S, and execute step (321).

[0112] 31. (331) Conduct a Level 3 fire alarm and execute step (332).

[0113] 32. (332) Determine whether to initiate a Level 3 fire alarm response. If so, proceed to step (333). Otherwise, return to step (301).

[0114] 33. (333) Release the Level 3 fire alarm response, clear the number of Level 3 fire alarms, and return to continuous monitoring.

Claims

1. A mine fire identification and alarm method based on flame and smoke images, characterized in that: Near-infrared explosion-proof cameras equipped with filters and fill-in lights are installed in electromechanical chambers, tunnels, coal mining faces and excavation faces to monitor the surveillance area in real time. The fill-in lights perform periodic cycle fill-in lighting on the area monitored by the camera in real time according to the set fill-in light pulse duration P1 and non-fill-in light pulse duration P2. N1 and N2 are pulse duration coefficients; T is the camera video cycle, in ms; F is the camera frame rate; within the fill light pulse duration P1, a fire warning is performed based on whether the characteristics of the suspected fire target area on the suspected fire image meet the fire smoke characteristic setting conditions; a level 1 fire alarm is performed based on whether the characteristics of the suspected fire target area on the suspected fire image meet the fire smoke and flame characteristic setting conditions; within the non-fill light pulse duration P2, a level 2 fire alarm is performed based on whether the characteristics of the suspected fire target area on the suspected fire image meet the fire flame characteristic setting conditions; when the number of level 1 fire alarms U1 and the number of level 2 fire alarms U2 that continuously appear within the set time S meet the setting condition of (U1+2U2)≥U, a level 3 fire alarm is performed, where S≥(P1+P2), U≥3.

2. A method for identifying and alarming a mine fire based on flame and smoke images as claimed in claim 1, characterized in that: The working process of fire alarm determination includes: Step 1: The fill light performs periodic cycle fill light on the area monitored by the camera in real time according to the set fill light pulse duration P1 and non-fill light pulse duration P2, and then step 2 is further executed; Step 2: A near-infrared explosion-proof camera equipped with a filter and a fill light collects video images of the monitoring area in real time, and the monitoring video images are transmitted to the back end of the monitoring server in real time, and the monitoring video is framed, image denoised, and image enhanced in real time to obtain a continuous frame monitoring image, and step 3 is further executed; Step 3: When the fill light is filling light for the area monitored by the camera in real time, further execute step 4; otherwise, execute step 9; Step 4: Determine the pixel characteristic value and M of the frame image within the fill light pulse duration P1 frame by frame r , when the pixel feature value of the rth frame image and F1≤M r <F2(r≥1), the frame image is determined to be a suspected fire smoke target frame image, and step 5 is executed; when the pixel feature value of the rth frame image and M r ≥F2 and (F2>F1 and F1 and F2 are set thresholds), the frame image is determined to be a suspected fire smoke and flame target frame image, and step 6 is executed; when the above suspected fire image determination conditions are not met, return to step 1 for continuous monitoring; Step 5: Select the processed N (N≥5) consecutive odd-numbered frames including the rth frame image to enhance the image of the suspected smoke target area; N Perform binarization operation and further execute step 7; Step 6: Select the processed continuous K (K ≥ 3) odd-numbered frames including the rth frame image to enhance the image of the suspected smoke and flame target area; K Perform a binarization operation and further execute step 8; Step 7: Determine whether the suspected fire target area feature on the suspected fire image meets the fire smoke feature setting condition. If so, a fire smoke warning is issued and step 11 is further executed. Otherwise, the execution returns to step 1. Step 8: Determine whether the characteristics of the suspected fire target area on the suspected fire image meet the fire smoke and flame characteristic setting conditions. If so, a fire level 1 alarm is issued and step 13 is further executed. Otherwise, the execution returns to step 1. Step 9: Determine the pixel characteristic value and M of the frame image within the non-fill light pulse duration P2 frame by frame n , when the pixel feature value of the nth frame image and M n ≥F3 (n≥1 and F3 is the set threshold), the frame image is determined to be a suspected fire flame target frame image, and step 10 is executed; when the above suspected fire image determination conditions are not met, return to step 1 for continuous monitoring; Step 10: Binarize the suspected fire flame target frame image; determine whether the suspected fire target area feature on the suspected fire image meets the fire flame feature setting conditions. If so, a fire level 2 alarm is issued and step 15 is further executed; otherwise, return to step 1; Step 11: Determine whether to respond to the fire warning. If yes, proceed to step 12; otherwise, return to step 1. Step 12: Release the fire warning response and return to continuous monitoring; Step 13: Determine whether to respond to the fire alarm level 1, if yes, further execute step 14, otherwise return to step 1; Step 14: cancel the fire level 1 alarm response, delete the level 1 alarm signal that exceeds the set time, calculate the number of fire level 1 alarms within the set time, and execute step 17; Step 15: Determine whether to respond to the fire level 2 alarm. If yes, go to step 16, otherwise go back to step 1. Step 16: cancel the fire level 2 alarm response, delete the level 2 alarm signal that exceeds the set time, calculate the number of fire level 2 alarms within the set time, and execute step 17; Step 17: Determine whether the continuous occurrence of fire level 1 alarm and fire level 2 alarm meets the setting conditions of fire level 3 alarm. If yes, further execute step 18, otherwise return to execute step 1; Step 18: Perform a fire level 3 alarm, and then proceed to step 19; Step 19: Determine whether to respond to the fire alarm level 3, if yes, go to step 20, otherwise go back to step 1; Step 20: Cancel the fire level 3 alarm response, clear the number of fire level 3 alarms, and return to continuous monitoring.

3. A method for identifying and alarming a mine fire based on flame and smoke images as claimed in claim 1, characterized in that: Whether the suspected fire target area characteristics on the suspected fire image meet the fire smoke characteristic setting conditions is based on the target boundary connected coordinate point group {(X a1 ,Y a1 ),(X a2 ,Y a2 ),···(X ai ,Y aj )}, extract the suspected target image contour, and arbitrarily select a point coordinate (X e ,Y e ), make L through this coordinate point d (d>1 and N + ) straight lines and intersect with the suspected target image contour. When there are multiple straight lines L z When the number of intersections between (z≤d) and the boundary of the suspected target image contour is greater than 2, it is determined that the fire smoke feature setting conditions are met, where X ai , Y aj are the horizontal and vertical coordinate point values ​​of the image boundary, respectively; ai and aj are the subscript constants of the horizontal and vertical coordinate points of the image boundary.

4. A method for identifying and alarming a mine fire based on flame and smoke images as claimed in claim 1, characterized in that: Whether the suspected fire target area characteristics on the suspected fire image meet the fire smoke and flame characteristics setting conditions is based on the target boundary connected coordinate point group {(X a1 ,Y a1 ),(X a2 ,Y a2 ),···(X ai ,Y aj )}, extract the suspected target image contour, and arbitrarily select a point coordinate (X v ,Y v ), make L through this coordinate point g (g>1 and N + ) straight lines and intersect with the suspected target image contour. When there are multiple straight lines L b When the number of intersections between (b≤g) and the boundary of the suspected target image contour is greater than 3, it is determined that the fire smoke and flame feature setting conditions are met.

5. A method for identifying and alarming a mine fire based on flame and smoke images as claimed in claim 1, characterized in that: Whether the suspected fire target area characteristics on the suspected fire image meet the fire flame characteristic setting conditions is based on the suspected target image boundary connected coordinate points {(X h1 ,Y h1 ),(X h2 ,Y h2 ),···(X hi ,Y hj )}, extract the suspected target image contour, and arbitrarily select a point coordinate (X u ,Y u ), make L through this coordinate point s (s>1 and N + ) straight lines and intersect with the suspected target image contour. When there are multiple straight lines L w When the number of intersections between (w≤s) and the boundary of the suspected target image contour is greater than or equal to 4, it is determined that the fire flame feature setting conditions are met, where X hi , Y hj are the horizontal and vertical coordinate point values ​​of the image boundary, hi and hj are the subscript constants of the horizontal and vertical coordinate points of the image boundary.

6. A method for identifying and alarming a mine fire based on flame and smoke images as claimed in claim 1, characterized in that: The optical filter is a long-wave pass filter with a wavelength greater than 780nm; the optical filter performs real-time and uninterrupted filtering when it is carried out with an explosion-proof camera for monitoring.

7. A method for identifying and alarming a mine fire based on flame and smoke images as claimed in claim 1, characterized in that: The fill light is a fill light device that periodically emits near-infrared waves.

8. A method for identifying and alarming a mine fire based on flame and smoke images as claimed in claim 1, characterized in that: The periodic cycle fill light is a fill light pulse duration P1 + a non-fill light pulse duration P2, which is a cycle fill light period.

9. A method for identifying and alarming a mine fire based on flame and smoke images as claimed in claim 2, characterized in that: In the step 5, N (N≥5) consecutive odd-numbered frames of images including the r-th frame image are selected after processing to perform image enhancement of the suspected smoke target area; The image enhancement of the suspected smoke target area is to perform equidistant difference frame and superposition operations on the selected continuous multiple frame images.

10. A method for identifying and alarming a mine fire based on flame and smoke images as claimed in claim 2, characterized in that: In step 6, K (K≥3) consecutive odd-numbered frames including the rth frame image are selected after processing to perform image enhancement of the suspected smoke and flame target area; the image enhancement of the suspected smoke and flame target area is to perform equidistant difference frame and superposition operations on the selected consecutive multiple frames of images.

11. A method for identifying and alarming a mine fire based on flame and smoke images as claimed in claim 2, characterized in that: In step 5, the enhanced target frame image Q N Perform binarization operation, that is, when the single pixel value M of the target frame image (i,j) If it is greater than the set threshold Q1, the pixel value M (i,j) =1, otherwise M (i,j) =0, where (i, j) is the pixel coordinate of the frame image.

12. A method for identifying and alarming a mine fire based on flame and smoke images as claimed in claim 2, characterized in that: In step 6, the enhanced target frame image Q K Perform binarization operation, that is, when the single pixel value R of the target frame image (i,j) If it is greater than the set threshold Q2 (Q2>Q1), then R (i,j) =1, otherwise R (i,j) =0.

13. A method for identifying and alarming a mine fire based on flame and smoke images as claimed in claim 2, characterized in that: In step 10, the suspected fire flame target frame image is binarized, that is, when the single point pixel value E (i,j) If it is greater than the set threshold Q3, then it is E (i,j) =1, otherwise E (i,j) =0.

14. A method for identifying and alarming a mine fire based on flame and smoke images as claimed in claim 3, characterized in that: The coordinates (X) of a point in the contour of the arbitrarily selected target image e ,Y e ), the point coordinates (X e ,Y e ) must meet the following conditions Where X min , X max Y is the minimum and maximum value of the horizontal coordinate points among the connected coordinate points on the boundary of the suspected target image. min , Y max is the minimum and maximum value of the ordinate points among the connected coordinate points on the boundary of the suspected target image; there are multiple straight lines L z When the number of intersection points with the boundary of the suspected target image contour is greater than 2, it is determined that the fire smoke feature setting condition is met. To meet the fire smoke feature setting condition, the number of straight lines z must meet the condition z≥5.

15. A method for identifying and alarming a mine fire based on flame and smoke images as claimed in claim 4, characterized in that: The coordinates (X) of a point in the contour of the arbitrarily selected target image v ,Y v ), the point coordinates (X v ,Y v ) must meet the following conditions Where X min , X max Y is the minimum and maximum value of the horizontal coordinate points among the connected coordinate points on the boundary of the suspected target image. min , Y max is the minimum and maximum value of the ordinate points among the connected coordinate points on the boundary of the suspected target image; there are multiple straight lines L b When the number of intersections with the boundary of the suspected target image contour is greater than 3, it is determined that the fire smoke and flame feature setting conditions are met. To meet the fire smoke and flame feature setting conditions, the number of straight lines b must meet the condition b≥4.

16. A method for identifying and alarming a mine fire based on flame and smoke images as claimed in claim 5, characterized in that: The coordinates (X) of a point in the contour of the arbitrarily selected target image u ,Y u ), the point coordinates (X u ,Y u ) must meet the following conditions Where X min , X max Y is the minimum and maximum value of the horizontal coordinate points among the connected coordinate points on the boundary of the suspected target image. min , Y max is the minimum and maximum value of the ordinate points among the connected coordinate points on the boundary of the suspected target image; there are multiple straight lines L s When the number of intersection points with the boundary of the suspected target image contour is greater than or equal to 4, it is determined that the fire flame feature setting condition is met. To meet the fire flame feature setting condition, the number of straight lines w must meet the condition w≥3.

17. A method for identifying and alarming a mine fire based on flame and smoke images as claimed in claim 9, characterized in that: Performing equidistant difference frame and superposition operations on the selected continuous multiple frame images, the equidistant difference frame and superposition operations are respectively performing no-interval frame difference operations, one-frame difference operations, two-frame difference operations and three-frame difference operations on the multiple frame images, and finally superimposing the results of the four frame image difference operations and superimposing and merging them with the r frame image again to form a new smoke frame image; The example of selecting consecutive odd-numbered frames of images with N=5 is as follows: when r≥3, the rth frame image is taken as the middle frame, and 5 consecutive frames of images are selected to enhance the suspected smoke target image. The target image Q N Requirements: When r=1, 5 consecutive frames are selected with r+2 as the intermediate frame to enhance the suspected smoke target image. N Requirements: When r=2, take r+1 as the intermediate frame and select 5 consecutive frames to enhance the suspected smoke target image; the target image Q N Requirements:

18. A method for identifying and alarming a mine fire based on flame and smoke images as claimed in claim 10, characterized in that: Performing equidistant difference frame and superposition operation on the selected continuous multiple frame images, the equidistant difference frame and superposition operation is to perform non-interval frame difference operation and one-frame difference operation on the multiple frame images respectively, and finally superimposing the results of the two frame image difference operations, and superimposing and merging with the r frame image again to form a new flame frame image; The example of selecting consecutive odd-numbered frames of images with K=3 is as follows: when r≥2, the rth frame image is taken as the middle frame, and three consecutive frames of images are selected to enhance the suspected flame target image. The target image Q K Requirements: Q K =[[r-(r-1)]+[(r+1)-r]+[(r+1)-(r-1)]+r]; When r=1, three consecutive frames are selected with r+1 as the intermediate frame to enhance the suspected smoke target image. K Requirements: Q K =[[(r+1)-r]+[(r+2)-(r+1)]+[(r+2)-r]+r]。

Citation Information

Cited By

  • Mining fire monitoring method and system based on sound-light alarm

    CN119992806A

  • Fire identification and early warning method based on weather radar

    CN120220315A