A fire identification method for cable tunnels and a track-type fire-fighting robot

Abstract

This invention discloses a fire identification method and a track-mounted fire-fighting robot for cable tunnels. The fire identification method comprises: collecting images of locations within the cable tunnel where ultraviolet signals are emitted; inputting the collected images into a constructed fire identification model, and outputting fire identification results. The invention can automatically identify fires within cable tunnels and, if a fire occurs, automatically extinguish it using the track-mounted fire-fighting robot, effectively reducing economic losses and casualties caused by fires.

Description

Technical Field

[0001] The present invention belongs to the technical field of fire protection, and in particular relates to a fire identification method for a cable tunnel and a track-type fire-fighting robot. Background Art

[0002] As urban electricity demand increases, the need for and utilization of above-ground space becomes increasingly urgent. To address both the shortage of electricity demand and the lack of above-ground space, the vigorous development and construction of urban underground cable tunnels can help resolve these issues. However, as cable tunnels proliferate throughout urban underground spaces, the safety hazards and risks associated with cable tunnel operations also increase. To effectively improve cable operation safety, regular tunnel inspections are necessary. Currently, intelligent cable tunnel monitoring systems and intelligent cable tunnel robots are gradually replacing manual inspections and becoming the primary development trend for future cable tunnel inspections. In response to State Grid Corporation of China's needs, intelligent cable tunnel monitoring systems and intelligent cable tunnel robot inspections are becoming increasingly common. A key research direction is how to enable back-end operations personnel to detect and address emergencies immediately.

[0003] Research on cable tunnel inspection robots can be categorized into two types: top-of-tunnel rail inspection robots and ground-based wheeled inspection robots. Due to height differences, tunnel wall deformation, and ground waterlogging along cable tunnel inspection routes, inspection tasks differ from those for substation inspections. Substations have more meters, while cable tunnels have almost no meters. However, cable tunnels require additional inspection tasks, such as hazardous gases, ground waterlogging, tunnel wall deformation, and cable bracket corrosion. Based on these factors, research trends in tunnel inspection robots are primarily focused on top-of-tunnel rail inspections. Currently, cable tunnel robots can essentially perform automated inspections and monitoring. However, if a fire is detected, they can only send an alarm signal to the inspectors and cannot take any countermeasures. If the initial fire is not serious, the cable robot can effectively and promptly control the fire, potentially saving significant economic losses and minimizing casualties. Summary of the Invention

[0004] To address the deficiencies in the prior art, the present invention provides a fire identification method and a track-type fire-fighting robot for cable tunnels, which can automatically identify fires in cable tunnels and automatically extinguish fires after they occur, effectively reducing economic losses and personal injuries caused by fires.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] In a first aspect, a fire identification method for a cable tunnel is provided, comprising: collecting images of locations in the cable tunnel where ultraviolet signals are emitted; inputting the collected images into a constructed fire identification model, and outputting fire identification results.

[0007] Furthermore, the fire recognition model is based on the YOLOv5 recognition model, and uses the K-means algorithm to improve the dark channel image defogging algorithm in the YOLOv5 recognition model, and uses the improved dark channel image defogging algorithm to defog the collected image data.

[0008] Furthermore, the improved dark channel image defogging algorithm includes:

[0009] The pixels of the image are regarded as a sample set X = {X1, X2, ..., X n}, X i is the i-th pixel in the sample set X, 1≤i≤n, n is a positive integer, representing the total number of pixels in the image;

[0010] According to the relationship between the number of categories K and the clustering error S of the sample set, the sample set is initialized to K cluster centers. The relationship between the S value and the K value is as follows:

[0011]

[0012] Among them, C j is the jth cluster center, 1≤j≤K; p is C j Pixels in L j is the centroid of the jth cluster center;

[0013] Calculate the Euclidean distance between each pixel and each cluster center, and assign each pixel to the cluster center closest to it. i To cluster center C j The Euclidean distance dis(X i , C j )for:

[0014]

[0015] Among them, a is the attribute of the pixel, 1≤a≤k; X ia is the ath attribute of the i-th pixel; C ja is the ath attribute of the jth cluster center;

[0016] After all pixels are assigned, the pixel values in each cluster are averaged to obtain the new cluster center point;

[0017] Recalculate the distance from each pixel to the cluster center and classify it again, repeating this cycle until the center point change meets the set threshold;

[0018] Calculate the average value of the pixel values in each cluster to obtain K pixel values, and then take the weighted average of the K pixel values to obtain the A value.

[0019] According to a second aspect, a track-type fire-fighting robot for a cable tunnel is provided, comprising: a robot body; a driving mechanism installed on the robot body, the driving mechanism being used to drive the robot body to move along a track installed in the cable tunnel; a fire sprinkler device installed on the robot body; an ultraviolet flame sensor installed on the robot body, the ultraviolet flame sensor being used to collect ultraviolet signals emitted in the cable tunnel; a flame identifier installed on the robot body, the flame identifier being used to control a flame recognition camera lens to collect images of the position where the ultraviolet signal is emitted in the cable tunnel based on the ultraviolet signal emitted in the cable tunnel collected by the ultraviolet flame sensor; a fire recognition module being used to receive images collected by the flame recognition camera lens, perform fire recognition according to the fire recognition method for cable tunnels described in the first aspect, and output recognition results; and a controller being used to control the driving mechanism to move the fire-fighting robot to the location where the fire occurs, and to control the fire sprinkler device to spray and extinguish the fire, and at the same time, to sound a fire alarm based on the recognition results output by the fire recognition module.

[0020] Furthermore, the driving mechanism includes a fourth motor, a transmission shaft and a driving roller. The fourth motor is installed inside the robot body. The fourth motor is connected to the driving roller through the transmission shaft, and the driving roller is rollingly arranged on the track surface.

[0021] Furthermore, the fire sprinkler device includes: a fifth motor installed inside the robot body; a rotating drum that is transmission-connected to the output shaft of the fifth motor; a swivel connected to the rotating drum through a fixed rod and rotating synchronously with the rotating drum; a spray arm installed on the swivel, the spray arm being provided with a plurality of spray holes for horizontal fire extinguishing of cables in the cable tunnel, and the plurality of the spray holes being connected to the pipe water injection connection port provided on the robot body through a pipe.

[0022] Furthermore, the fire sprinkler device also includes: a mounting plate installed on the rotating drum; a fire sprinkler head for longitudinally extinguishing fires on cables in a cable tunnel, the fire sprinkler head being mounted on the mounting plate and connected to the pipe water injection connection port, a first motor being mounted on the mounting plate, the first motor being used to drive the fire sprinkler head to swing within a set range.

[0023] Furthermore, the fire sprinkler device also includes: a third motor installed in the rotating drum, the third motor is connected to a screw rod on both sides of the third motor, the screw rod is threadedly connected to the internal thread block, and the internal thread block is fixedly arranged on one end of the square slide rod, and the other end of the square slide rod is slidably connected to the rotating ring; the other end of the square slide rod is provided with a second motor, the output shaft of the second motor is connected to one end of the pneumatic telescopic rod, and the other end of the pneumatic telescopic rod is connected to the spray arm; the rotating ring is provided with a clamping sleeve for accommodating the pneumatic telescopic rod.

[0024] Furthermore, the flame identifier and the flame identification shooting lens are respectively mounted on the mounting plate.

[0025] Furthermore, a smoke sensor and a signal transmitter are provided inside the flame identifier.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] (1) The present invention collects images of the direction where ultraviolet signals are emitted in the cable tunnel; inputs the collected images into the constructed fire recognition model, and outputs the fire recognition results; the present invention can automatically identify the fire in the cable tunnel and automatically extinguish the fire after the fire occurs, which can effectively reduce the economic losses and personal injuries caused by the fire; the fire recognition model uses YOLOv5 as the recognition model, adopts the K-means algorithm to improve the dark channel image defogging algorithm, and uses the improved algorithm to defog the collected flame image, thereby improving the fire video image recognition accuracy. The flame identifier is equipped with a smoke sensor and a signal transmitter, which can promptly warn the outside world of the fire;

[0028] (2) In the present invention, the driving mechanism drives the robot body to move on the track to the vicinity of the fire point, and the flame identifier controls the flame recognition camera lens to capture images of the position where the ultraviolet signal is emitted in the cable tunnel based on the ultraviolet signal emitted in the cable tunnel collected by the ultraviolet flame sensor; the fire recognition module receives the image captured by the flame recognition camera lens, performs fire recognition, and outputs the recognition result, thereby accurately locating the fire location; the fire sprinkler device sprays and extinguishes the fire based on the recognition result output by the fire recognition module, extinguishes the fire in the early stage of the fire to avoid causing greater losses, and at the same time, issues a fire alarm. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is a schematic diagram of the overall structure of a rail-type fire-fighting robot for cable tunnels provided by an embodiment of the present invention;

[0030] Figure 2 Schematic diagram of the installation positions of the fourth motor and the fifth motor in an embodiment of the present invention;

[0031] Figure 3 This is a schematic structural diagram of a swivel in an embodiment of the present invention;

[0032] Figure 4 2 is a schematic diagram of the relationship curve between S value and K value in an embodiment of the present invention;

[0033] Figure 5 Schematic diagram of the structure of the YOLOv5 recognition model in an embodiment of the present invention;

[0034] Figure 6 1 is a flow chart of fire extinguishing in an embodiment of the present invention, wherein (a) is the fire identification process and (b) is the fire extinguishing process;

[0035] Figure 7 The following is a comparison of the flame image defogging effects. (a) is the flame image identified before improvement, and (b) is the flame image identified after using the K-means improved dark channel defogging algorithm.

[0036] In the figure: 1. Robot body; 2. UV flame sensor; 3. Rotating drum; 4. Rotating ring; 5. First motor; 6. Flame identifier; 7. Flame identification camera lens; 8. Fire sprinkler head; 9. Spray hole; 10. Spray arm; 11. Pneumatic telescopic rod; 12. Mounting plate; 13. Clamping sleeve; 14. Driving roller; 15. Pipe water injection connection port; 16. Driving mechanism; 17. Internal thread block; 18. Screw; 19. Square slide rod; 20. Second motor; 21. Third motor; 22. Fixed rod; 23. Fourth motor; 24. Transmission shaft; 25. Fifth motor. DETAILED DESCRIPTION

[0037] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention.

[0038] Example 1:

[0039] A fire identification method for a cable tunnel comprises: collecting images of locations in the cable tunnel where ultraviolet signals are emitted; inputting the collected images into a constructed fire identification model, and outputting fire identification results.

[0040] In this embodiment, the fire recognition model is based on the YOLOv5 recognition model. The K-means algorithm is used to improve the dark channel image defogging algorithm in the YOLOv5 recognition model, and the improved dark channel image defogging algorithm is used to defog the collected image data. The structure of the YOLOv5 recognition model is as follows: Figure 5 shown.

[0041] The flame target detection process in the fire recognition model is as follows:

[0042] (1) The YOLOv5 algorithm is used to divide the input flame image into N×N cells. Each cell generates a priori boxes for targets of different scales, such as large, medium, and small. If the center of the identified target falls within a certain grid, the priori box of that grid is responsible for tracking and identifying the target. YOLOv5 uses the confidence level c to represent the probability of target classification and the performance of matching the target in the priori box.

[0043] c=PH (1)

[0044] Among them, P is the probability of the target in the predicted box, which is 0 if there is no target in the predicted box, otherwise it is 1; H is the intersection-union ratio between the predicted box and the true box.

[0045] (2) The divided flame images are normalized and the normalized flame dataset is sent to the lower-layer feature extraction network for feature extraction.

[0046] (3) The prediction box is set through K-means clustering and divided into boxes of different sizes. For different detection targets, the prediction box position, that is, the center point coordinates, is calculated.

[0047] (4) Based on the offset value of the predicted coordinates, calculate the target center point position and the prediction box width and height [10-11].

[0048] (5) Output target recognition results.

[0049] Dark channel dehazing algorithm

[0050] The dark channel dehazing algorithm reduces the impact of uneven lighting distribution, dust, moisture, and other factors on the flame image, increases the flame image details, and improves the cable tunnel fire recognition rate. The dark channel dehazing model formula is as follows:

[0051]

[0052] I(x)=J(x)t(x)+A[1-t(x)] (3)

[0053] Where t(x) is the transmittance, x is the pixel value; q is the defogging parameter, which is generally set to 0.95; y is the pixel index value; Ω(x) is the window centered on x; I(y) is the input image; ε is the pixel channel; A is the global atmospheric light component; I(x) is the flame image to be defogged; and J(x) is the flame image after processing without fog.

[0054] The calculation of t(x) is related to the A value, which is calculated using the foggy image. First, the top 0.1% of pixels in the dark channel image are selected by brightness. Then, the point with the highest brightness at the corresponding position in the original foggy image is found, which is the A value.

[0055] Improved dark channel dehazing algorithm based on K-means

[0056] Although the dark channel dehazing algorithm achieves good results in dehazing most foggy images, some pixels are missed when calculating the A value, resulting in incomplete dehazing and affecting the accuracy of subsequent fire video image recognition. To this end, K-means is used to improve the dark channel dehazing algorithm.

[0057] The pixels of the flame image are regarded as the sample set X = {X1, X2, ..., X n}, X i is the i-th pixel in the sample set X, 1≤i≤n, n is a positive integer representing the total number of pixels in the image; the K-means clustering algorithm is used to cluster pixels in the same area into one class, so that all pixels in the entire image can be taken when calculating the value of A. The specific steps of the K-means improved dark channel dehazing algorithm are as follows:

[0058] (1) According to the relationship between the number of categories K and the clustering error S of the sample set, the sample set is initialized to K cluster centers. The relationship curve between S value and K value is as follows: Figure 4 As shown, the relationship between S value and K value is as follows:

[0059]

[0060] Among them, C j is the jth cluster center, 1≤j≤K; p is C j Pixels in L j is the centroid of the jth cluster center;

[0061] From the relationship curve between S value and K value, it can be seen that the S value decreases rapidly as the K value increases. When K=5, there is an obvious inflection point and the S value tends to be flat, that is, the K value of the cluster center is 5.

[0062] (2) Calculate the Euclidean distance between each pixel and each cluster center, and assign each pixel to the cluster center closest to it. i To cluster center C j The Euclidean distance dis(X i , C j )for:

[0063]

[0064] Among them, a is the attribute of the pixel, 1≤a≤k; X ia is the ath attribute of the i-th pixel; C ja is the ath attribute of the jth cluster center.

[0065] (3) After all pixels are assigned, the pixel values in each cluster are averaged to obtain the new cluster center.

[0066] (4) Recalculate the distance from each pixel to the cluster center and classify it again, and repeat this cycle until the center point changes very little.

[0067] (5) Calculate the average value of the pixel values in each cluster to obtain K pixel values, and then take the weighted average of the K pixel values, which is the A value.

[0068] The defogging effect of flame image before and after the K-means improved dark channel defogging algorithm is as follows Figure 7 As shown in the figure, it can be seen that the improved dehazed image has more significant color and richer feature information.

[0069] The process of building the cable tunnel fire identification algorithm model is as follows:

[0070] (1) The K-means improved dark channel dehazing algorithm is used to dehaze the flame image, thereby improving the image details and removing interference.

[0071] (2) The frame difference method and the mixed Gaussian model fusion algorithm are used to extract features from the dynamically evolving flame image to reduce the influence of static background on fire identification.

[0072] (3) Label and name the processed dataset. Generate an XML file after labeling, including the coordinates and width and height information of the label box.

[0073] (4) Configure YOLOv5 algorithm model parameters and conduct model training and testing.

[0074] Example 2:

[0075] Based on the fire identification method for cable tunnels described in Example 1, this embodiment provides a track-type fire-fighting robot for cable tunnels, such as Figures 1 to 3 As shown, it includes: a robot body 1, a driving mechanism 16 installed on the robot body 1, a fire sprinkler device, an ultraviolet flame sensor 2 and a flame identifier 6; the driving mechanism 16 is used to drive the robot body 1 to move along a track installed in the cable tunnel, and the ultraviolet flame sensor 2 is used to collect ultraviolet signals emitted in the cable tunnel; the flame identifier 6 is used to control the flame recognition camera lens 7 to collect images of the direction where the ultraviolet signal is emitted in the cable tunnel according to the ultraviolet signal emitted in the cable tunnel collected by the ultraviolet flame sensor 2; the fire recognition module receives the image collected by the flame recognition camera lens 7, and performs fire recognition according to the fire recognition method for cable tunnels described in Example 1, and outputs the recognition result; the controller controls the driving mechanism 16 to move the fire extinguishing robot to the fire location according to the recognition result output by the fire recognition module, and controls the fire sprinkler device to spray and extinguish the fire, and at the same time, issues a fire alarm.

[0076] The driving mechanism 16 includes a fourth motor 23, a transmission shaft 24 and a driving roller 14. The fourth motor 23 is installed inside the robot body 1. The fourth motor 23 is connected to the driving roller 14 through the transmission shaft 24, and the driving roller 14 is rollingly arranged on the track surface.

[0077] The fire sprinkler system includes a fifth motor 25 mounted within the robot body 1, a drum 3 drivingly connected to the output shaft of the fifth motor 25, a swivel 4 connected to the drum 3 via a fixed rod 22 and rotating synchronously with the drum 3, and a spray arm 10 mounted on the swivel 4. The spray arm 10 is provided with a plurality of spray holes 9 for horizontal fire extinguishing of cables within the cable tunnel. The plurality of spray holes 9 are connected to a pipe water injection connection 15 mounted on the robot body 1 via a pipe. The pipe water injection connection 15 is used to connect to an external firefighting pipe, thereby obtaining water for firefighting. The spray holes 9 are arranged on the side of the spray arm 10, allowing firefighting cables to be extinguished from the side.

[0078] The anti-spray device also includes a mounting plate 12 mounted on the rotating drum 3 and a fire sprinkler head 8 for longitudinally extinguishing fires on cables in the cable tunnel. The fire sprinkler head 8 is mounted on the mounting plate 12 and is connected to the pipe water injection connection port 15. The fire sprinkler head 8 cooperates with the spray hole 9 to form a multi-directional spray fire extinguishing system for the fire point, thereby achieving the purpose of improving fire extinguishing efficiency and saving fire water. Based on this, the present invention can also set a small water storage tank inside the robot body 1. The water storage tank obtains fire water supply from the external fire pipe through the pipe water injection connection port 15 and then disconnects. The fire sprinkler head 8 and the spray hole 9 are respectively connected to the water storage tank through pipes. When extinguishing a fire, the fire is extinguished by relying on the stored water in the water storage tank, thereby freeing the fire extinguishing robot from its dependence on the external fire pipe and improving the movement speed of the fire extinguishing robot.

[0079] The fire sprinkler device also includes a third motor 21 installed in the rotating drum 3. The third motor 21 is transmission-connected with a screw rod 18 on both sides. The screw rod 18 is threadedly connected to the internal thread block 17, and the internal thread block 17 is fixedly set at one end of the square slide rod 19. The other end of the square slide rod 19 is slidingly connected to the swivel 4; the other end of the square slide rod 19 is provided with a second motor 20, and the output shaft of the second motor 20 is connected to one end of the pneumatic telescopic rod 11, and the other end of the pneumatic telescopic rod 11 is connected to the spray arm 10; a clamping sleeve 13 for accommodating the pneumatic telescopic rod 11 is provided on the swivel 4.

[0080] The ultraviolet flame sensor 2 is installed on one side surface of the robot body 1, the fifth motor 25 is arranged on the inner side of the robot body 1, the rotating drum 3 is transmitted and arranged at the lower end of the fifth motor 25, and several fixed rods 22 are fixedly arranged on the outer surface of the rotating drum 3. The pneumatic telescopic rod 11 is driven by an air pump, and a spray arm 10 is fixedly set at one end of the pneumatic telescopic rod 11. The third motor 21 is arranged on the inner side of the rotating drum 3. The mounting plate 12 is fixed to the lower end of the rotating drum 3. A first motor 5 is fixed on one side of the mounting plate 12. The first motor 5 is used to drive the flame identifier 6. The flame identification shooting lens 7 is fixedly set on the surface of the flame identifier 6, and the fire sprinkler head 8 is set at the lower end of the flame identifier 6.

[0081] Clamping sleeves 13 are fixed on both sides of the swivel 4. These consist of an upper open ring and an X-shaped opening fixed below the open ring. Clamping sleeves 13 are made of plastic, and their inner diameter is compatible with the pneumatic telescopic rod 11. Clamping sleeves 13 are used to store the pneumatic telescopic rod 11 when not in use, facilitating the movement of the fire-fighting robot.

[0082] A plurality of spray holes 9 are fixedly provided on the inner surface of the spray arm 10 , and the water inlet of the spray arm 10 is connected to the drain outlet of the robot body 1 through a hose.

[0083] The screw rods 18 on both sides of the third motor 21 are arranged with opposite thread squares.

[0084] The flame identifier 6 is electrically connected to the ultraviolet flame sensor 2 and the flame identification camera lens 7. A smoke sensor and a signal transmitter are provided inside the flame identifier 6 to warn the outside world of a fire.

[0085] In the present invention, a flame identifier senses the fire point through an ultraviolet flame sensor, and a flame recognition camera takes photos. YOLOv5 is used as the recognition model, and the K-means algorithm is adopted to improve the traditional dark channel image defogging algorithm. The improved algorithm is used to defog the collected flame image, thereby improving the fire video image recognition accuracy. A smoke sensor and a signal transmitter are provided inside the flame identifier, which can promptly alert the outside world to a fire.

[0086] In the present invention, the robot body starts the fourth motor to drive the driving roller at one end of the transmission shaft to rotate, driving the robot body to move on the track to the vicinity of the fire point, starts the first motor to drive the flame identifier to adjust the pitch angle, so as to accurately identify the fire position through the flame identification camera lens, starts the second motor to drive the pneumatic telescopic rod to rotate out from the inside of the clamping sleeve vertically downward, and at the same time starts the pneumatic telescopic rod to drive the spray arm to extend and align with the fire point, starts the third motor to drive the screw rod to rotate, thereby driving the internal thread block to move to both sides, indirectly driving the square slide rod to slide outward inside the rotating ring to adjust the spacing of the spray arms, starts the fifth motor to drive the rotating drum to rotate, thereby indirectly driving the spray arm and the flame identifier to swing back and forth, at this time the water pipe is filled with water through the pipe water injection connection port, and the water is sprayed out through the fire sprinkler head and the spray holes on the surface of the spray arm to extinguish the fire, so that the fire is extinguished in the early stage to avoid causing greater losses.

[0087] Example 3:

[0088] Based on the fire identification method for cable tunnels described in Example 1 and the rail-type fire-fighting robot for cable tunnels described in Example 2, this embodiment provides a method for extinguishing a fire using the rail-type fire-fighting robot for cable tunnels described in Example 2, such as Figure 6 As shown, the following steps are included.

[0089] Step 1: The flame identifier 6 senses the fire point through the ultraviolet flame sensor 2, and the flame recognition camera 7 takes a photo. YOLOv5 is used as the recognition model, and the K-means algorithm is used to improve the traditional dark channel image defogging algorithm. The improved algorithm is used to defog the collected flame image to improve the accuracy of fire video image recognition.

[0090] Step 2: The robot body 1 starts the fourth motor 23 to drive the driving roller 14 at one end of the transmission shaft 24 to rotate, driving the robot body 1 to move on the track to the vicinity of the fire point, and starts the first motor 5 to drive the flame identifier 6 to adjust the pitch angle, so as to accurately identify the fire position through the flame recognition camera 7;

[0091] Step 3: Start the second motor 20 to drive the pneumatic telescopic rod 11 to rotate out of the clamping sleeve 13 and vertically downward. At the same time, start the pneumatic telescopic rod 11 to extend the spray arm 10 to align with the fire point. Start the third motor 21 to drive the screw rod 18 to rotate, thereby driving the internal thread block 17 to move to both sides, indirectly driving the square slide rod 19 to slide outward inside the swivel 4 to adjust the spacing of the spray arms 10.

[0092] Step 4: Start the fifth motor 25 to drive the drum 3 to rotate, thereby indirectly driving the spray arm 10 and the flame identifier 6 to swing back and forth. At this time, water is injected into the water pipe through the pipe water injection connection port 15, and water is sprayed out through the fire sprinkler head 8 and the spray hole 9 on the surface of the spray arm 10 to extinguish the fire.

[0093] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A fire identification method for a cable tunnel, characterized in that: include: Collect images of locations within the cable tunnel where UV signals are emitted; Input the collected images into the constructed fire recognition model and output the fire recognition results; The fire recognition model is based on the YOLOv5 recognition model, and uses the K-means algorithm to improve the dark channel image defogging algorithm in the YOLOv5 recognition model, and uses the improved dark channel image defogging algorithm to defog the collected image data; The improved dark channel image defogging algorithm includes: The pixels of the image are regarded as a sample set X = {X1, X2, ..., X n }, X i is the i-th pixel in the sample set X, 1≤i≤n, n is a positive integer, representing the total number of pixels in the image; According to the number of categories K Clustering error with the sample set S The relationship between the sample set is initialized as K cluster centers, S Value and K The relationship between the values is as follows: （4） in, For the j Cluster centers, ; p for Pixels in ; For the j The centroid of the cluster centers; Calculate the Euclidean distance between each pixel and each cluster center, assign each pixel to the cluster center closest to it, and the pixel To the cluster center Euclidean distance between for: （5） in, is the attribute of the pixel, ; For the i The first pixel Attributes; For the j The first cluster center attributes; After all pixels are assigned, the pixel values in each cluster are averaged to obtain the new cluster center point; Recalculate the distance from each pixel to the cluster center and classify it again, repeating this cycle until the center point change meets the set threshold; Calculate the average value of the pixel values in each cluster to obtain K pixel values, and then take the weighted average of the K pixel values to obtain the A value.

2. A rail-type fire-fighting robot for cable tunnels, characterized in that: include: Robot body (1); a driving mechanism (16) mounted on the robot body (1), the driving mechanism (16) being used to drive the robot body (1) to move along a track mounted in the cable tunnel; A fire sprinkler device installed on the robot body (1); an ultraviolet flame sensor (2) mounted on the robot body (1), the ultraviolet flame sensor (2) being used to collect ultraviolet signals emitted in the cable tunnel; A flame identifier (6) mounted on the robot body (1), the flame identifier (6) being used to control a flame identification camera lens (7) to capture an image of a position in the cable tunnel where the ultraviolet signal is emitted, based on the ultraviolet signal emitted in the cable tunnel collected by the ultraviolet flame sensor (2); a fire identification module, configured to receive images captured by the flame identification camera lens (7), perform fire identification according to the fire identification method for cable tunnels according to claim 1, and output an identification result; The controller is used to control the driving mechanism (16) to move the fire extinguishing robot to the fire location according to the recognition result output by the fire recognition module, and control the fire sprinkler device to spray and extinguish the fire, and at the same time, issue a fire alarm.

3. The rail-type fire-fighting robot for cable tunnels according to claim 2, characterized in that: The driving mechanism (16) includes a fourth motor (23), a transmission shaft (24) and a driving roller (14). The fourth motor (23) is installed inside the robot body (1). The fourth motor (23) is connected to the driving roller (14) through the transmission shaft (24), and the driving roller (14) is rollingly arranged on the track surface.

4. The rail-type fire-fighting robot for cable tunnels according to claim 2, characterized in that: The fire sprinkler device comprises: a fifth motor (25) mounted inside the robot body (1); A rotating drum (3) drivingly connected to the output shaft of the fifth motor (25); a rotating ring (4) connected to the rotating drum (3) via a fixed rod (22) and rotating synchronously with the rotating drum (3); A spray arm (10) is mounted on the rotating ring (4), and the spray arm (10) is provided with a plurality of spray holes (9) for horizontally extinguishing fires on cables in a cable tunnel. The plurality of spray holes (9) are connected to a pipe water injection connection port (15) provided on the robot body (1) through a pipe.

5. The rail-type fire-fighting robot for cable tunnels according to claim 4, characterized in that: The fire sprinkler system further comprises: a mounting plate (12) mounted on the rotating drum (3); A fire sprinkler head (8) for longitudinally extinguishing a fire on a cable in a cable tunnel, wherein the fire sprinkler head (8) is mounted on the mounting plate (12) and is in communication with the pipe water injection connection port (15), and a first motor (5) is mounted on the mounting plate (12), wherein the first motor (5) is used to drive the fire sprinkler head (8) to swing within a set range.

6. The rail-type fire-fighting robot for cable tunnels according to claim 4, characterized in that: The fire sprinkler system further comprises: A third motor (21) is installed in the rotating drum (3), and screw rods (18) are connected to both sides of the third motor (21) in a transmission manner. The screw rod (18) is threadedly connected to the internal thread block (17), and the internal thread block (17) is fixedly arranged on one end of a square slide rod (19), and the other end of the square slide rod (19) is slidably connected to the rotating ring (4); a second motor (20) is provided at the other end of the square slide rod (19), and the output shaft of the second motor (20) is connected to one end of the pneumatic telescopic rod (11), and the other end of the pneumatic telescopic rod (11) is connected to the spray arm (10); a clamping sleeve (13) for accommodating the pneumatic telescopic rod (11) is provided on the rotating ring (4).

7. The rail-type fire-fighting robot for cable tunnels according to claim 5, characterized in that: The flame identifier (6) and the flame identification shooting lens (7) are respectively mounted on the mounting plate (12).

8. The rail-type fire-fighting robot for cable tunnels according to claim 2, characterized in that: The flame identifier (6) is internally provided with a smoke sensor and a signal transmitter.

Images