CN118022219A

CN118022219A - Method for extinguishing forest fire by using parachute fire extinguishing bullet

Info

Publication number: CN118022219A
Application number: CN202410340065.4A
Authority: CN
Inventors: 李开源; 张高上; 肖飞; 唐海林
Original assignee: Qingdao Research Institute Of Wuhan University Of Technology
Current assignee: Qingdao Research Institute Of Wuhan University Of Technology
Priority date: 2024-03-25
Filing date: 2024-03-25
Publication date: 2024-05-14

Abstract

The invention provides a method for extinguishing forest fire by using a parachuting fire extinguishing bomb, which belongs to the technical field of parachuting fire extinguishing bombs and comprises the following steps: acquiring an infrared remote sensing image with forest fire extinguishing function and a monitoring image shot by a camera; the monitoring image is divided into a plurality of grids, wherein the size of each grid is inscribed square size of the effective fire extinguishing range of the parachuting fire extinguishing bomb; fusing the monitor images after grid division and the aligned infrared remote sensing images to obtain a fused image; calculating a three-dimensional temperature vector set by adopting a pre-trained fire three-dimensional temperature model; according to the three-dimensional temperature vector set, calculating the optimal parachute opening height of the parachute fire extinguishing bomb by utilizing aerodynamics; calculating the optimal fuze time length of the parachute fire extinguishing bullet according to the optimal parachute opening height and the safe flying height of the unmanned aerial vehicle for throwing the parachute fire extinguishing bullet; and based on the optimal fuze duration, after the fuze of the parachute fire extinguishing bomb is adjusted, the unmanned aerial vehicle bomb is used and put in.

Description

Method for extinguishing forest fire by using parachute fire extinguishing bullet

Technical Field

The invention belongs to the technical field of parachuting fire extinguishing bombs, and particularly relates to a method for extinguishing forest fires by using parachuting fire extinguishing bombs.

Background

Forest fires are a natural disaster with serious harm, and huge ecological environment loss and economic loss are brought to the world each year. According to statistics, the direct economic loss from forest fires is up to billions of dollars annually worldwide. Moreover, forest fires can directly burn a large amount of trees and wild animals and plants, and can also generate a large amount of toxic smoke, thereby bringing serious threat to the atmospheric environment and resident health in the surrounding area.

Therefore, the forest fire can be effectively and timely extinguished, and the influence of the forest fire on the ecological environment and the safety of human lives and property is always a global important subject. Currently, forest fire extinguishing means commonly adopted in various countries in the world mainly comprise manual extinguishing, mechanical extinguishing, unmanned plane bomb extinguishing and the like.

The manual rescue is to form a small team by professional firefighters or forest workers, carry simple tools such as shovels and buckets, and rush to the scene for rescue in the first time when forest fires occur. The manual rescue has the advantage of flexibility, but has obvious defects of limited personnel, high safety risk and the like.

The mechanical fire fighting is to use large-scale mechanical equipment such as bulldozers, forklift and the like to open fire-proof wires around forest fires, thereby preventing the fire from spreading. The method is suitable for flat areas, has wide effective range, is difficult to penetrate into complex terrain areas, and has heavy equipment and high cost.

In contrast, the unmanned aerial vehicle-mounted bomb can be used for avoiding the problem when being put out, and specifically, the unmanned aerial vehicle is used for putting out the parachute fire extinguishing bomb at a certain height above a fire disaster, and after the arrival time of a fire extinguishing bomb fuze, the fire retardant is sprayed out by explosion, so that the fire extinguishing is realized.

However, the unmanned aerial vehicle is adopted to put in the parachute fire extinguishing bullet to put out Lin Huoshi, and the unmanned aerial vehicle cannot be put in the optimal height because the flying height of the unmanned aerial vehicle is limited by the fire scene environment. The time length of the parachute fire extinguishing bomb fuze for mass production is fixed and cannot be dynamically adjusted according to the actual height. The explosion height of the parachute fire extinguishing bomb deviates from the optimal explosion height, so that the fire extinguishing range is not accurately covered, the fire extinguishing efficiency is difficult to fully exert, and the overall extinguishing effect is reduced.

Disclosure of Invention

In view of the above, the invention provides a method for extinguishing forest fires by using a parachute fire extinguishing bomb, which can solve the technical problems that in the prior art, the unmanned aerial vehicle is used for putting the parachute fire extinguishing bomb to extinguish Lin Huoshi, the explosion height of the parachute fire extinguishing bomb is deviated from the optimal explosion height, the fire extinguishing range is not covered accurately, the fire extinguishing efficiency is difficult to fully exert, and the overall extinguishing effect is reduced.

The invention is realized in the following way:

the first aspect of the invention provides a method for extinguishing forest fires with a parachuting fire extinguishing bomb, comprising the steps of:

s10, acquiring an infrared remote sensing image with forest fire extinguishing function and a monitoring image shot by a camera;

S20, based on the monitoring image, aligning the infrared remote sensing image, and dividing the monitoring image into a plurality of grids, wherein the size of each grid is the inscribed square size of the effective extinguishing range of the parachute extinguishing bomb;

s30, fusing the monitor images after grid division and the aligned infrared remote sensing images to obtain a fused image;

S40, inputting the fusion image by adopting a pre-trained fire three-dimensional temperature model to obtain a three-dimensional temperature vector set, wherein the three-dimensional temperature vector set comprises temperature vectors of each grid, and each temperature vector is used for describing temperatures of different heights;

s50, calculating the optimal parachute opening height of the parachute fire extinguishing bomb by utilizing aerodynamics according to the three-dimensional temperature vector set;

s60, calculating the optimal fuze duration of the parachute fire extinguishing bullet according to the optimal parachute opening height and the safe flying height of the unmanned aerial vehicle for throwing the parachute fire extinguishing bullet;

And S70, after the fuze of the parachute fire extinguishing bomb is adjusted based on the optimal fuze duration, the unmanned aerial vehicle bomb is used and put in.

Based on the technical scheme, the method for extinguishing forest fires by adopting the parachuting fire extinguishing bullet can be further improved as follows:

The specific steps of the step S20 include: aligning the infrared remote sensing image with the monitoring image; extracting feature points of the two images by using a SIFT algorithm or a SURF algorithm by adopting a registration method based on the feature points, solving an affine transformation matrix by matching the feature points, and aligning the infrared image with the monitoring image; dividing the aligned monitoring image into a plurality of grids, wherein the side length of each grid is 2 times of the root number of the radius of the effective fire extinguishing range of the parachute fire extinguishing bomb, and the overlapping degree of the adjacent grids is 10-30%.

The specific steps of the step S30 include: the method comprises the steps of adopting a Laplacian pyramid fusion method to fuse a gridded monitoring image with an aligned infrared remote sensing image to obtain a fused image; respectively decomposing the monitoring image and the infrared image into a high-frequency component and a low-frequency component with different scales; selecting a low-frequency component of the monitoring image for the low-frequency component; for the high-frequency components, adopting a weighted average method to fuse the high-frequency components of the two images; reconstructing to obtain a fused image; in the fusion process, the brightness and contrast of the two images are normalized.

The specific steps of the step S40 include: the three-dimensional temperature model based on deep learning is adopted, the fusion image is taken as input, and a convolutional neural network and a cyclic neural network are utilized to extract a feature map and context information; restoring a high-dimensional temperature vector of each grid through the full connection layer and the up-sampling layer; the dimension of each temperature vector corresponds to different heights, the length is 10 to 20, and the temperature range is 0 to 1200 ℃; in the training stage, supervised learning is adopted, real scene temperature data is used as a label, and the difference between model output and the label is minimized.

The specific steps of the step S50 include: traversing each temperature vector according to the three-dimensional temperature vector set, and finding a first height which is lower than a 500 ℃ threshold value to be used as an umbrella opening height candidate value of the grid; calculating the maximum horizontal drift distance of the height by using a drag formula by taking the grid center as a coordinate source point; if the drift distance is less than half of the side length of the grid, the height is a feasible solution, otherwise, the umbrella opening height is increased; traversing all grids, and taking the minimum height of a feasible solution as the optimal umbrella opening height; and weather parameters and numerical simulation are introduced to improve the drift prediction accuracy.

The specific steps of the step S60 include: determining the safe flying height H of the unmanned aerial vehicle, and calculating the difference d between the H and the optimal parachute opening height H, wherein the value range of d is 50-100 meters; the kinetic formula used is:

h＝0.5×g×t²+v0×t；

Wherein, the fuze duration t, wherein g is gravity acceleration, v0 is initial velocity of the projectile body; designing the allowance of t; and monitoring the height and attitude parameters of the unmanned aerial vehicle in real time, and dynamically calculating the correction value of the fuze duration.

The specific steps of the step S70 include: according to the optimal fuze time length, correspondingly setting fire extinguishing bombs, and programming and writing the fire extinguishing bombs into a bomb control circuit; the unmanned aerial vehicle carries a plurality of fire extinguishing bombs to go to a forest fire target area; arriving at the gathering route of the operation area, decelerating and flying, and automatically searching a safe delivery point; the projectile bodies are thrown one by one, or an automatic throwing mode is started; and monitoring the fuse delay and the flight state, and performing first-time supplementary feeding or re-operation on the misfeeding or missing area.

The three-dimensional temperature model structure of the fire disaster is an encoder-decoder structure, and the encoder comprises a convolutional neural network layer and a long-term and short-term memory network layer and is used for extracting characteristics and capturing context information from an input fusion image; the decoder comprises a deconvolution layer and an up-sampling layer, and restores the characteristic diagram output by the encoder into a three-dimensional temperature field; the training data set of the model comprises real fire scene acquisition data, simulated fire scene experimental data and data subjected to data augmentation treatment, and all the data are preprocessed to enable the data formats to be consistent.

Furthermore, the three-dimensional temperature model of the fire disaster introduces an attention mechanism module and a loss function term based on physical constraint in the training process so as to improve the reconstruction precision of the temperature field and meet the thermodynamic rule.

Further, the parachuting fire extinguishing bomb comprises a shell, an explosion device and a flame-retardant parachute, wherein the explosion device is arranged in the shell and comprises a detonator, a safety device group and an explosive part, the duration of the detonator, the safety device group and the explosive part can be adjusted, the shell is filled with a water-based halogen-free liquid flame retardant, and the top of the shell is fixedly connected with the flame-retardant parachute.

Compared with the prior art, the method for extinguishing forest fires by adopting the parachuting fire extinguishing bullet has the beneficial effects that: the method is based on infrared thermal imaging and visual monitoring, and can accurately acquire the three-dimensional temperature field distribution information of the fire scene; and by combining with a computer model, the optimal parachute opening height and the optimal fuze duration of the parachute fire extinguishing bullet are optimally calculated, so that accurate throwing is realized, and the fire extinguishing bullet can uniformly and efficiently cover the whole fire scene. Greatly improves the accuracy and efficiency of the rescue and avoids waste. The technical problems that in the prior art, the unmanned aerial vehicle is used for throwing the parachute fire extinguishing bomb to extinguish Lin Huoshi, the explosion height of the parachute fire extinguishing bomb deviates from the optimal explosion height, the fire extinguishing range is not covered accurately, the fire extinguishing efficiency is difficult to fully exert, and the overall fire extinguishing effect is reduced are solved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments of the present invention will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a flow chart of a method provided by the present invention;

FIG. 2 is a schematic diagram of the overall structure of the parachuting fire extinguishing bomb in the embodiment;

FIG. 3 is an internal system structure of a parachute fire extinguishing bullet core body in the embodiment;

FIG. 4 is a schematic illustration of a parachute fire extinguishing bomb for fire extinguishing a fire line;

Wherein reference numerals are denoted as: 1. the fire-fighting device comprises an umbrella body, an umbrella body storage box, a shell, a water-based halogen-free liquid flame retardant, a core, a shell fixer, a safety device, a remote control ejection device, a timing fuse, a flame-out fuse, an explosive 11, an explosive 12, a collision fuse, a firing area, a non-firing area, a fire extinguishing bomb 15 and a firing line 16.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

As shown in fig. 1, the invention provides a method flow chart for extinguishing forest fires by using parachute fire extinguishing bomb, which comprises the following steps: the method comprises the following steps of:

s40, inputting a fusion image by adopting a pre-trained fire three-dimensional temperature model to obtain a three-dimensional temperature vector set, wherein the three-dimensional temperature vector set comprises temperature vectors of each grid, and each temperature vector is used for describing temperatures of different heights;

and S70, after the fuze of the parachute fire extinguishing bomb is adjusted based on the optimal fuze duration, using the unmanned aerial vehicle to launch the fire extinguishing bomb.

The following describes in detail the specific embodiments of the above steps:

in step S10, an infrared remote sensing image with forest fire extinguishing function and a monitoring image shot by a camera are obtained. The method aims to acquire a heat source distribution image and a visual image of a forest fire scene, and prepare for subsequent processing. The infrared remote sensing image can be obtained by using the technologies such as aerial remote sensing or satellite remote sensing, and the thermal infrared band with the wavelength of 8-14 microns can be used for effectively detecting the surface temperature distribution. The monitoring image can be obtained by shooting through an unmanned aerial vehicle, an aircraft or a ground monitoring camera, and visual information of the scene is provided.

The obtained infrared remote sensing image and the monitoring image are required to meet certain quality requirements. The resolution of the infrared remote sensing image should be not lower than 10 meters, and the temperature measurement accuracy should be not lower than 1 ℃. The resolution of the monitor image should be not lower than 1080P, and the frame rate should be not lower than 24fps. The time stamp error of both images should be controlled to be within 10 seconds. If the time stamp error is too large, a time correction is required. In addition, significant cloud shielding, smoke interference, etc. should be avoided to ensure image quality.

The specific implementation manner of step S20 is: based on the monitoring image, the infrared remote sensing image is aligned, and the monitoring image is divided into a plurality of grids, wherein the size of each grid is the inscribed square size of the effective fire extinguishing range of the parachute fire extinguishing bomb.

First, registration and alignment of the infrared remote sensing image and the monitoring image is required. Because the shooting angles and the field of view ranges of the two images are different, a common reference point or characteristic needs to be found for registration. Common image registration methods include feature point-based registration, phase correlation-based registration, and the like. If feature point registration is adopted, feature points of the two images can be extracted by using SIFT, SURF and other algorithms, an affine transformation matrix is solved through matching of the feature points, and an infrared image is aligned with a monitoring image.

After alignment, the monitoring image needs to be divided into several grids. The size of the grid refers to the radius r of the effective fire extinguishing range of the parachute fire extinguishing bomb, the effective fire extinguishing range is inscribed in a square, and the side length of the grid is the root number of 2 times r. The value range of r is 10-20 m, and can be determined according to the specific fire extinguishing bomb model. If fire extinguishing bombs with a flowering radius of 15 meters are used, the side length of the grid is about 21 meters.

In dividing the grid, attention should be paid to the degree of overlapping of the grids. In order to avoid fire extinguishing dead zones, adjacent grids should have a certain overlapping range, and the overlapping degree can be set to be 10% -30%. In addition, the number of grids is not excessive, so that the subsequent calculation complexity is not increased, and the number is usually controlled to be less than 1000.

The specific implementation manner of step S30 is: and fusing the monitor images after grid division and the aligned infrared remote sensing images to obtain a fused image.

The purpose of the fusion image is to organically combine the infrared heat source information and the visual information, and provide input for the subsequent three-dimensional temperature modeling. The fusion method has various choices, and commonly used methods comprise wavelet transform domain fusion, laplacian pyramid fusion and the like.

Taking Laplacian pyramid fusion as an example, the monitoring image and the infrared image can be respectively decomposed into high-frequency components (details such as edges and textures) and low-frequency components (integral outlines) with different scales. Selecting a low-frequency component of the monitoring image for the low-frequency component; for the high frequency components, a certain rule (such as weighted average, maximum value taking, etc.) is adopted to fuse the high frequency components of the two images. And finally, obtaining a fused image through reconstruction.

In the fusion process, proper normalization processing is needed for brightness, contrast and the like of the two images, so that the influence of over-obvious information of a certain mode on the fusion effect is avoided. In addition, the fusion parameters can be dynamically adjusted according to the change of the temperature field, so that the fusion quality is improved.

The specific implementation manner of step S40 is: and (3) inputting a fusion image by adopting a pre-trained fire three-dimensional temperature model to obtain a three-dimensional temperature vector set, wherein each temperature vector is used for describing the temperatures of different heights.

The task of the three-dimensional temperature model is to infer the temperature distribution at different heights within each grid from the fused image. This requires the model to have sufficient generalization capability to capture both the morphological features of the flame and accurately predict the temperature in the invisible region.

The three-dimensional temperature model based on deep learning can be adopted, the image characteristics are extracted by utilizing structures such as a convolutional neural network and a cyclic neural network, and the three-dimensional modeling is carried out on the temperature field by combining the three-dimensional temperature model with a known thermophysical model. The model needs to be pre-trained on a large amount of labeling data (such as infrared images, thermocouple temperature measurement data and the like) so as to fully learn priori knowledge of fire temperature distribution.

Specifically, the fusion image can be used as input, the feature map is extracted through operations such as convolution and pooling, then the feature map is sent into a circulating network such as LSTM to capture context information, and finally, the high-dimensional temperature vector of each grid is restored through a full connection layer, an up-sampling layer and the like. The dimensions of each temperature vector correspond to different heights, and the length can be set to 10-20, and the temperature range is 0-1200 ℃.

For the training process, supervised learning can be adopted, real scene temperature measurement data is used as a label, and the difference between model output and the label is minimized, such as using a mean square error Loss. In addition, priori knowledge can be introduced, such as strengthening the attention of the model to the flame region through an attention mechanism, adding thermophysical constraint and the like, so that the prediction accuracy is improved.

In the deployment stage, the trained model is input into a fusion image, and then a three-dimensional temperature vector set of each grid can be obtained. The dimension of the temperature vector set is related to the number of grids, and if the number of grids is N and the temperature vector length is M, the dimension of the vector set is NxM.

The specific implementation manner of step S50 is as follows: according to the three-dimensional temperature vector set, calculating the optimal parachute opening height of the parachute fire extinguishing bomb by utilizing aerodynamics.

After the parachute extinguishing bomb is put in, the parachute kit needs to be opened at a proper height so that the parachute kit slowly descends and is uniformly distributed in a fire scene. The choice of the height of the umbrella is related to the effective coverage area and the dropping precision of the fire extinguishing bomb. Too high can increase the influence of wind power diffusion, and too low can not fully utilize the deceleration effect of the umbrella.

Therefore, the optimal parachute opening height needs to be calculated by combining the temperature field distribution and the aerodynamic model. Specifically, the first height below the 500 degrees celsius threshold may be found as an umbrella opening height candidate for the grid by calculating each temperature vector. Then, the maximum horizontal drift distance of the height is calculated by using a drag formula by taking the grid center as a coordinate source point:

where d is the drift distance, u is the wind speed, t is the descent time, w0 is the initial descent speed, w1 is the final constant speed, and g is the gravitational acceleration. If the drift distance is less than half the grid side length, then this height is a viable solution, otherwise the umbrella opening height needs to be increased.

And after traversing and calculating all grids, taking the minimum height of a feasible solution as the optimal umbrella opening height. If all heights do not meet the requirements, the temperature threshold may be suitably relaxed, such as by increasing it to 600 degrees celsius.

In the calculation process, meteorological parameters such as wind speed, wind direction and the like need to be obtained in real time, and numerical simulation methods such as CFD and the like can also be introduced, so that accuracy of drift prediction is improved. In addition, for complex fire environments, the temperature field can be discretized into a plurality of areas, the optimal umbrella opening height is solved respectively, and finally, the average value or the weighted average value is obtained.

The specific implementation manner of step S60 is: and calculating the optimal fuze duration of the parachute fire extinguishing bullet according to the optimal parachute opening height and the safe flying height of the unmanned aerial vehicle for throwing the parachute fire extinguishing bullet.

The fuze duration determines the time interval from the unmanned aerial vehicle to the throwing of the fire extinguishing bomb. If the fuze is too long, the projectile body can land to cause harm; if too short, the umbrella may not fully deploy and begin to descend. Therefore, the fuze duration needs to be reasonably set according to the optimal parachute opening height and flying height.

First, it is necessary to determine the safe flight level limit H of the unmanned aerial vehicle, typically not lower than 120 meters.

Then, calculating the difference d between the unmanned aerial vehicle and the optimal parachute opening height h:

d＝H-h

Because the fire extinguishing bomb has a certain sagging distance in the air, enough height needs to be reserved, and the value range of d can be set to be 50-100 meters.

Next, the required fuze duration t is calculated using the kinetic formula:

h＝0.5×g×t²+v0×t

Wherein h is the height of the parachute, g is the gravity acceleration, v0 is the initial falling speed of the projectile body, and t is the fuze duration. Substituting d and the known parameters can obtain an approximate analytic solution of t. The numerical method can also be utilized to traverse the solution to find the minimum positive root t with h=d as the optimal fuze duration.

For example, assuming d=80 meters, v0=10 meters/second, substituting the formula yields t of about 4.1 seconds. In practical application, t can be appropriately designed to ensure safety and reliability.

In addition, if the unmanned aerial vehicle is influenced by air flow, the flying height is unstable, and a real-time correction mode can be adopted. And continuously monitoring parameters such as the height, the speed, the gesture and the like of the unmanned aerial vehicle, and dynamically calculating a correction value of the fuze duration. The state estimation is performed, for example, by using a Kalman filtering method, and the fuze duration is adjusted accordingly.

The specific implementation manner of step S70 is: and based on the optimal fuze duration, after the fuze of the parachute fire extinguishing bomb is adjusted, the unmanned aerial vehicle bomb is used and put in.

Firstly, according to the calculated optimal fuze duration, the fire extinguishing bomb is correspondingly set. The setting of the fuze duration can be programmed into a control circuit within the projectile. When unmanned plane carries the bullet, still need pay attention to the magazine reasonable in design, avoid the bullet body impaired.

The unmanned aerial vehicle is required to have certain load capacity to carry a plurality of extinguishing bomb to put in. The multi-rotor unmanned aerial vehicle or the fixed-wing unmanned aerial vehicle can meet the requirements of maximum voyage, endurance time and the like.

After the unmanned aerial vehicle takes off, the unmanned aerial vehicle goes to a target area according to the forest fire coordinates. Because the unmanned aerial vehicle operation range is generally large, a plurality of airlines or waypoints can be set for planning. After reaching the operation area, the route is folded, the cruising speed is reduced, and the preparation is made for throwing.

During throwing, the ground obstacle avoidance function can be started first, a safe throwing point can be independently found, and the distance between the throwing unmanned aerial vehicle and the ground obstacle is kept more than 30 meters. Then the projectile bodies are thrown one by one, or an automatic throwing mode is started.

When in delivery, the fuze delay and the flight state need to be monitored at any time. If an accident occurs, the operation can be immediately stopped. For the misthrow or missing area, the first time tissue is supplemented or reworked. After all the operations are completed, the unmanned aerial vehicle can return independently or can be controlled manually. In the whole process, the ground command center can monitor the position and the state of the unmanned aerial vehicle in real time, record, play back and evaluate each round of throwing process, and are used for the later optimization analysis.

Wherein, the structural design of fire three-dimensional temperature model:

The model is based on a deep learning framework, the overall architecture being an encoder-decoder structure. The encoder part is formed by stacking a Convolutional Neural Network (CNN) and a long-short-term memory network (LSTM) and is used for extracting characteristics and capturing context information from an input fusion image; the decoder part consists of a deconvolution layer and an up-sampling layer, and restores the characteristic diagram output by the encoder into a three-dimensional temperature field.

In particular, the encoder consists of 6 convolutional layers and 2 LSTM layers. The convolution layer is responsible for extracting low-level and high-level semantic features of the image, and a convolution kernel of 3x3 is used to activate the function as ReLU. The LSTM layer then captures the sequence context, facilitating modeling of the temperature field. The decoder end has 4 deconvolution layers and 2 up-sampling layers, the feature map is restored to high-dimensional temperature voxels through up-sampling and deconvolution operations, and the final output shape is (N, C, D, H and W), wherein N is the batch size, C is the number of temperature channels, D is the height dimension, and H and W are rows and columns respectively.

In addition, a attention mechanism module is introduced into the model and is used for strengthening the attention to the flame area and improving the reconstruction accuracy. Meanwhile, some loss function items based on physical constraint are added, so that the output temperature field is ensured to accord with a certain thermodynamic rule.

Training data set acquisition:

Acquisition of a training dataset is an extremely important and difficult process requiring a large amount of measured data. Mainly comprises the following parts:

1. And the real fire scene data acquisition comprises on-site infrared remote sensing images, monitoring videos and a large amount of temperature point measurement data. The temperature measurement can be acquired in different areas and different heights of a fire scene by means of a remote control robot, a thermocouple and the like. The obtained data need to be marked with information such as temperature values, position coordinates and the like.

2. And simulating the experimental data of the fire scene, wherein the standard training set can be obtained through a large-scale simulation experiment due to the difficulty in obtaining the real fire scene data and high risk. And setting controllable combustion scenes of different scales indoors or at an experimental base, arranging equipment such as a temperature sensor, a camera and the like, and acquiring accurate temperature distribution and imaging data.

3. And the data is amplified, namely the problem of uneven data distribution possibly exists only by actually measured data due to the large difference of fire scenes. It is therefore necessary to expand the diversity of the training set by some data augmentation techniques, such as flipping, rotation, cropping, etc.

4. And data preprocessing, namely preprocessing the acquired multi-source data such as format conversion, noise reduction, registration and the like, and ensuring the consistency of the data input into the model.

Examples

The embodiment discloses a method for extinguishing forest fires by using parachute fire extinguishing bullet, which aims to realize accurate, efficient and unmanned forest fire extinguishing. The method integrates the multidisciplinary theories such as remote sensing technology, computer vision technology, artificial intelligence technology, aerodynamics and the like, can intelligently analyze the fire scene environment, reasonably plan the throwing scheme, and accurately put in the special parachuting fire extinguishing bullet through the unmanned aerial vehicle to carry out accurate fire extinguishing on forest fires.

As shown in fig. 2, the fire extinguishing bomb with an umbrella body adopted in the embodiment is assembled by three main components of the umbrella body 1, the shell 3 and the core body 5. The shell 3 is provided with a water-based halogen-free liquid flame retardant 4, and the core 5 is provided with an explosive 11 and a detonation device. The umbrella body 1 is made of flame-retardant nylon, and has enough toughness to be required by materials, and cannot be damaged by wind scraping. When putting in fire extinguishing bomb, the umbrella body 1 can be automatically ejected and unfolded from the umbrella body storage box 2. The housing 3 and the core 5 are made of a flame retardant polymer material (rigid polyvinyl chloride resin), and a material having sufficient strength is required. The shell fixer 6 and the shell 3 are made of the same material and are used for fixing the shell and the core.

As shown in fig. 3, the core 5 is internally provided with various detonating devices, including a fuse 7, a remote control ejection device 8, a timing fuse 9, a pull-explosion fuse 10 and a collision fuse 12. A safety 7 is connected to the umbrella body 1 for preventing accidental explosions during the delivery process. The remote control ejection device 8 can be activated and ejected through a remote control of the unmanned aerial vehicle, so that the explosion-pulling fuse 10 is pulled to ignite the explosive 11. The timing fuze 9 then automatically detonates the explosive 11 after a preset time has been reached. The collision fuse 12 is activated when the extinguishing bomb lands, as a backup detonator. The explosive 11 is a safe and environment-friendly ammonium nitrate explosive.

The method for extinguishing forest fires by adopting the parachuting fire extinguishing bullet in the embodiment comprises the following steps:

step one: acquiring an infrared remote sensing image with forest fire extinguishing function and a monitoring image shot by a camera

Firstly, remote sensing monitoring is carried out on forest fire sites by using devices such as unmanned aerial vehicles, aircrafts and the like, and image data of two different modes are obtained. The infrared remote sensing image can reflect the three-dimensional temperature distribution condition of the fire scene, and the monitoring image provides visual context information of the scene.

In order to ensure the image quality, the resolution of the acquired infrared remote sensing image is not lower than 10 meters, and the temperature measurement precision is not lower than 1 ℃; the resolution of the monitoring image is not lower than 1080P, and the frame rate is not lower than 24fps. The time stamp errors of the two images are controlled within 10 seconds, so that obvious cloud and fog shielding and smoke interference are avoided.

Step two: based on the monitoring image, the infrared remote sensing image is aligned, and the monitoring image is divided into a plurality of grids

And performing spatial alignment on the infrared remote sensing image and the monitoring image, and taking the monitoring image as a reference. And extracting characteristic points of the two images by using a SIFT algorithm or a SURF algorithm by adopting a registration method based on the characteristic points, solving an affine transformation matrix by matching the characteristic points, and aligning the infrared image with the monitoring image.

After alignment, the monitoring image is divided into a plurality of square grids, the side length of each grid is 2 times of the radius of the effective fire extinguishing range of the parachute fire extinguishing bomb, and if the fire extinguishing bomb with the flowering radius of 15 meters is adopted, the side length of each grid is about 21 meters. The overlapping degree of adjacent grids is controlled to be 10% -30% so as to avoid fire extinguishing blind areas. Optionally, the number of grids is controlled within 1000, so that the subsequent calculation is prevented from being too complex.

Step three: fusing the monitor images after grid division and the aligned infrared remote sensing images to obtain a fused image

And carrying out pixel-level fusion on the monitoring image and the infrared image by adopting a Laplacian pyramid fusion method to obtain a fusion image. The method comprises the steps of respectively decomposing two images into a low-frequency component and a high-frequency component, selecting low-frequency overall contour information of a monitoring image, fusing the high-frequency components of the two images by adopting weighted average on high-frequency detail information, and finally reconstructing to obtain a fused image.

In the fusion process, the input monitoring image and the input infrared image are required to be preprocessed in brightness, contrast and the like, so that the influence of the fusion effect caused by the fact that certain mode information is too obvious is avoided.

Step four: a pre-trained fire three-dimensional temperature model is adopted, and a fusion image is input to obtain a three-dimensional temperature vector set

The fusion image is input into a three-dimensional temperature model based on deep learning, characteristic information and context information of the fusion image are extracted by utilizing structures such as a convolutional neural network and a cyclic neural network, and a priori thermophysical knowledge is combined to restore a high-dimensional temperature vector of each grid.

The model is pre-trained on a large amount of labeling data (such as infrared images, thermocouple temperature data and the like), and can accurately predict the temperature distribution of different heights of each area of a fire scene. The dimension of each temperature vector corresponds to different heights, the vector length is 10-20, and the temperature range is 0-1200 ℃.

In the training process, a supervised learning mode is adopted, actually measured temperature data is used as a label, and the difference between model prediction and the label is minimized. And strategies such as a attention mechanism, physical constraint and the like are introduced, so that the temperature field reconstruction capability of the model is improved.

Step five: according to the three-dimensional temperature vector set, calculating the optimal parachute opening height of the parachute fire extinguishing bomb by utilizing aerodynamics

Traversing the temperature vector set, and finding the first height below the threshold of 500 ℃ for each temperature vector as an umbrella opening height candidate of the grid. And calculating the maximum horizontal drift distance of the height by using a drag formula by taking the center of the grid as a coordinate source point. If the drift distance is less than half of the grid side length, the height is a feasible solution, otherwise, the umbrella opening height is increased until the precision requirement is met.

The drag formula is: d= (u+0.5 (w0+w1) t 2)/g

Where d is the drift distance, u is the wind speed, t is the descent time, w0 is the initial descent speed, w1 is the final constant speed, and g is the gravitational acceleration.

After traversing all grids, taking the minimum height in the feasible solution as the optimal umbrella opening height. If no solution is available, the temperature threshold is relaxed appropriately. And a real-time meteorological data and numerical simulation method is introduced, so that the drift prediction precision is improved. For complex fire fields, the average value is obtained after the regional solution.

Step six: calculating the optimal fuze time length of the parachute fire extinguishing bomb according to the optimal parachute opening height and the unmanned aerial vehicle flight height

Firstly, determining the maximum safe flying height H of the unmanned plane, and calculating the difference d between the H and the optimal umbrella opening height H, wherein the value range of d is 50-100 meters. Solving the fuze duration t by using a dynamic formula:

h＝0.5*g*t^2+v0*t

wherein g is gravity acceleration, v0 is initial velocity of the projectile body, and d and known parameters are substituted to obtain t. In practical application, t is appropriately designed for margin to ensure reliability.

Meanwhile, state parameters such as the height, the speed and the attitude of the unmanned aerial vehicle are required to be obtained in real time, state estimation is carried out by using methods such as Kalman filtering, and the duration of the fuze is dynamically corrected so as to cope with the influence of a complex flight environment.

Step seven: according to the optimal fuze time length, the fuze of the parachute fire extinguishing bomb is adjusted, and the unmanned aerial vehicle bomb is used and put in

And D, writing the calculated fuze duration into a control circuit in the bomb body in a programming mode to set parameters of the detonating device of the fire extinguishing bomb. Meanwhile, the design of the bullet magazine needs to be paid attention to, and the bullet bodies are prevented from being damaged in the transportation process.

The unmanned aerial vehicle is required to have certain load and endurance to carry a plurality of fire extinguishing bombs to go to a target area for putting out the fire. Multiple rotor or fixed wing unmanned aerial vehicle can be adopted, but the index requirements of maximum range, maximum effective load and the like are met.

After the unmanned aerial vehicle reaches a forest fire operation area, the unmanned aerial vehicle needs to retract a route, reduce cruising speed and prepare for throwing. By utilizing the ground obstacle avoidance function, a safe throwing point is independently searched, and a safe distance of more than 30 meters is kept between the safe throwing point and ground obstacles during throwing.

Then the unmanned plane starts to throw the parachute fire extinguishing bullet one by one or automatically. In the throwing process, ground commander monitors the position, state, fuze delay and other data of the unmanned aerial vehicle in real time, and the unmanned aerial vehicle is guaranteed to be free of loss. If an unexpected situation occurs, the operation may be terminated immediately.

For the miscast or missing area, the first time tissue is supplemented or reworked. After all the operations are completed, the unmanned aerial vehicle returns to the home position or is controlled to return manually. The whole operation process is monitored and recorded by a ground command center and is used for subsequent analysis and optimization.

Finally, a specific extinguishing process of the parachute fire extinguishing bomb on the forest fire site is introduced, and the specific extinguishing process is shown in fig. 4:

after the umbrella fire extinguishing bomb 15 is thrown to the upper air of a forest fire site through an unmanned aerial vehicle, firstly the safety device 7 and the umbrella body 1 are ejected together, the umbrella body 1 is automatically unfolded, and the fire extinguishing bomb starts to slide down in a uniform descending posture. When reaching the preset optimal umbrella opening height, the remote control ejection device 8 is activated, and the explosion-pulling fuse 10 is pulled to ignite the explosive 11. After explosive deflagration, the high-pressure liquid flame retardant 4 is sprayed into spherical water mist, the spherical water mist radially covers the ignition area 13, and precise fire extinguishing is carried out on the periphery of the fire wire 16.

If the pull-explosion fuse 10 fails, the extinguishing bomb will continue to fall to the ground. Upon impact with the ground, the impact fuse 12 is activated to detonate the explosive 11, producing the same water mist fire suppression effect. If the former two detonation modes fail, the timing fuze 9 finally detonates automatically after reaching the preset time. The triple insurance design ensures the fire extinguishing process without loss.

Through above-mentioned step, this embodiment is through the accurate purpose-made parachuting fire extinguishing bomb of throwing in of unmanned aerial vehicle, has realized the cover of whole forest fire scene to put out the rescue in high efficiency, and the whole process need not fire fighter directly to go deep into the scene of a fire, greatly reduced the operation risk, really realized intelligent, unmanned accurate high-efficient put out.

Specifically, the principle of the invention is as follows: the multidisciplinary theory such as computer vision, artificial intelligence, aerodynamics and the like is organically combined with forest fire extinguishing practice, so that the unmanned accurate extinguishing of the whole-flow intellectualization is realized.

Firstly, through fusion processing of an infrared remote sensing image and a visual monitoring image, fire scene temperature field information and background environment information are fully acquired. The infrared image can reflect three-dimensional temperature distribution, the monitoring image provides visual context information, and the infrared image and the monitoring image are complementary, so that more accurate characterization of a scene of a fire scene is facilitated.

Then, a three-dimensional temperature field modeling technology based on deep learning is introduced, the fusion image is taken as input, characteristic information is extracted by using model structures such as a convolutional neural network and a cyclic neural network, and then a priori thermophysical knowledge is combined, so that a high-dimensional temperature vector of each region is finally predicted. The method overcomes the limitation that the traditional model is difficult to reconstruct the three-dimensional temperature field with high precision, and lays a foundation for the next step of rescue planning.

The optimal parachute opening height calculation of the parachute extinguishing bomb is realized by comprehensively utilizing temperature field information and aerodynamic theory. Specifically, by analyzing each temperature vector, finding the first height below 500 degrees celsius as a viable parachute opening height candidate for the area; and then, carrying out drift analysis on each height by utilizing a drag formula, and selecting the minimum height meeting the precision requirement as the optimal parachute opening height. In this way, it can be ensured that the fire extinguishing bomb can still hit the target area accurately under the influence of wind power.

And calculating the optimal fuze duration, wherein a dynamics principle is mainly applied. The time interval for enabling the fire extinguishing bomb to accurately open the umbrella is solved by analyzing the difference between the flight height of the unmanned aerial vehicle and the optimal umbrella opening height and substituting the difference into a kinematic formula. Meanwhile, parameters such as the height and the attitude of the unmanned aerial vehicle are required to be acquired in real time, and the fuze duration is dynamically corrected so as to cope with a complex flight environment.

Finally, through programming the parachute landing fire extinguishing bomb, the optimal parachute opening height and the fuze duration parameters are written into a bomb control circuit, and the unmanned aerial vehicle is utilized to realize remote autonomous throwing, so that the whole accurate unmanned fire extinguishing process is completed.

The foregoing is merely illustrative of the present invention, and the present invention is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present invention.

Claims

1. A method for extinguishing forest fires by using a parachuting fire extinguishing bomb, which is characterized by comprising the following steps:

2. A method for extinguishing forest fires using parachuting fire extinguishing bullet according to claim 1, characterized in that the specific steps of step S20 include: aligning the infrared remote sensing image with the monitoring image; extracting feature points of the two images by using a SIFT algorithm or a SURF algorithm by adopting a registration method based on the feature points, solving an affine transformation matrix by matching the feature points, and aligning the infrared image with the monitoring image; dividing the aligned monitoring image into a plurality of grids, wherein the side length of each grid is 2 times of the root number of the radius of the effective fire extinguishing range of the parachute fire extinguishing bomb, and the overlapping degree of the adjacent grids is 10-30%.

3. A method for extinguishing forest fires using parachuting fire extinguishing bullet according to claim 1, characterized in that the specific steps of step S30 include: the method comprises the steps of adopting a Laplacian pyramid fusion method to fuse a gridded monitoring image with an aligned infrared remote sensing image to obtain a fused image; respectively decomposing the monitoring image and the infrared image into a high-frequency component and a low-frequency component with different scales; selecting a low-frequency component of the monitoring image for the low-frequency component; for the high-frequency components, adopting a weighted average method to fuse the high-frequency components of the two images; reconstructing to obtain a fused image; in the fusion process, the brightness and contrast of the two images are normalized.

4. A method for extinguishing forest fires using parachuting fire extinguishing bullet according to claim 1, characterized in that the specific steps of step S40 include: the three-dimensional temperature model based on deep learning is adopted, the fusion image is taken as input, and a convolutional neural network and a cyclic neural network are utilized to extract a feature map and context information; restoring a high-dimensional temperature vector of each grid through the full connection layer and the up-sampling layer; the dimension of each temperature vector corresponds to different heights, the length is 10 to 20, and the temperature range is 0 to 1200 ℃; in the training stage, supervised learning is adopted, real scene temperature data is used as a label, and the difference between model output and the label is minimized.

5. A method for extinguishing forest fires using parachuting fire extinguishing bullet according to claim 1, characterized in that the specific steps of step S50 include: traversing each temperature vector according to the three-dimensional temperature vector set, and finding a first height which is lower than a 500 ℃ threshold value to be used as an umbrella opening height candidate value of the grid; calculating the maximum horizontal drift distance of the height by using a drag formula by taking the grid center as a coordinate source point; if the drift distance is less than half of the side length of the grid, the height is a feasible solution, otherwise, the umbrella opening height is increased; traversing all grids, and taking the minimum height of a feasible solution as the optimal umbrella opening height; and weather parameters and numerical simulation are introduced to improve the drift prediction accuracy.

6. A method for extinguishing forest fires using parachuting fire extinguishing bullet according to claim 1, characterized in that the specific steps of step S60 include: determining the safe flying height H of the unmanned aerial vehicle, and calculating the difference d between the H and the optimal parachute opening height H, wherein the value range of d is 50-100 meters; the kinetic formula used is:

h＝0.5×g×t²+v0×t；

7. A method for extinguishing forest fires using parachuting fire extinguishing bullet according to claim 1, characterized in that the specific steps of step S70 include: according to the optimal fuze time length, correspondingly setting fire extinguishing bombs, and programming and writing the fire extinguishing bombs into a bomb control circuit; the unmanned aerial vehicle carries a plurality of fire extinguishing bombs to go to a forest fire target area; arriving at the gathering route of the operation area, decelerating and flying, and automatically searching a safe delivery point; the projectile bodies are thrown one by one, or an automatic throwing mode is started; and monitoring the fuse delay and the flight state, and performing first-time supplementary feeding or re-operation on the misfeeding or missing area.

8. The method for extinguishing forest fires with parachuting fire extinguishing bullet as claimed in claim 1, wherein the three-dimensional temperature model structure of the fire is an encoder-decoder structure, and the encoder comprises a convolutional neural network layer and a long-short-term memory network layer for extracting features and capturing context information for the input fusion image; the decoder comprises a deconvolution layer and an up-sampling layer, and restores the characteristic diagram output by the encoder into a three-dimensional temperature field; the training data set of the model comprises real fire scene acquisition data, simulated fire scene experimental data and data subjected to data augmentation treatment, and all the data are preprocessed to enable the data formats to be consistent.

9. The method for extinguishing forest fires by using parachuting fire extinguishing bullet as claimed in claim 8, wherein the three-dimensional temperature model of fire introduces a attentiveness mechanism module and a loss function term based on physical constraint in the training process to improve the reconstruction accuracy of the temperature field and satisfy thermodynamic rules.

10. A method of extinguishing forest fires using a parachuting fire extinguishing bomb according to any of claims 1 to 9, characterised in that the parachuting fire extinguishing bomb comprises a housing, an explosive device, a fire retardant parachute, wherein the explosive device is arranged in the housing, the explosive device comprises a detonator, a fuse set and an explosive part of adjustable duration of ignition, the housing is filled with a water-based halogen-free liquid fire retardant, and the fire retardant parachute is fixedly connected to the top of the housing.