CN111882812A - Fire monitoring system and method - Google Patents

Abstract

The invention discloses a fire monitoring system and method. The system includes: a central processing unit, a signal conditioning module and a motor driving module connected with the central processing unit, an infrared sensor connected with the signal conditioning module, a horizontal motor connected with the motor driving module for driving the rotating head, and a horizontal motor connected with the motor driving module. Optical processing device with mechanically connected rotating head, the optical processing device is used to collect light by scanning horizontally in the range of ³⁶⁰⁰ degrees and scanning in the vertical plane under the driving of the rotating head, and reflecting, focusing, After filtering, the spectrum of multiple bands including fire characteristic bands is projected onto the target surface of the infrared sensor in turn. The invention expands the fire detection range, improves the detection sensitivity and precision, eliminates the influence of sunlight, artificial light source and general heat source spectrum, and reduces the false alarm rate.

Description

A fire monitoring system and method

technical field

The invention belongs to the technical field of fire detection and early warning, and in particular relates to a fire monitoring system and method based on infrared detection.

Background technique

At present, the fire detection technology inside and outside the building is relatively mature. In the prior art, whether there is a fire is determined mainly by detecting the combustion products of the fire (such as gas, smoke, temperature, flame, etc.). These technologies all face a big problem, the detection distance is very limited, usually 100m is a standard distance, and some devices can reach 200-300m.

In practical applications, some scenarios have strict requirements on the detection distance and coverage area. For example, for the detection of forest and grassland fires, the detection distance is often required to reach several kilometers or tens of kilometers. Therefore, the general traditional detection methods can no longer meet the requirements. In the past two decades, with the development of image analysis technology, people began to analyze smoke and flames for the images of large-scale variable field of view cameras. However, due to the many sources of false alarms in nature, the application of this technology has been greatly affected. limits. The most typical representative technology at this stage is the German FireWatch system. FireWatch uses a BW camera with a specific band (weak vegetation reflection), which can obtain relatively good smoke images, which improves the detection ability to a certain extent. In recent years, the most widely used technology in China is the binocular camera detection technology based on infrared thermal imaging cameras and visible light cameras. The imaging camera obtains parameters related to temperature radiation, such as temperature or imaging gray value, and combines the analysis of the two images to realize fire detection. However, these technologies all have the problems of short detection distance and high false alarm rate.

SUMMARY OF THE INVENTION

In order to solve the above problems existing in the prior art, the present invention provides a fire monitoring system and method.

To achieve the above object, the present invention adopts the following technical solutions:

The invention provides a fire monitoring system, which comprises: a central processing unit, a signal conditioning module and a motor driving module connected with the central processing unit, an infrared sensor connected with the signal conditioning module, and a driving module connected with the motor driving module for driving The horizontal motor of the rotating head also includes a light processing device mechanically connected to the rotating head, and the light processing device is used to collect light by scanning horizontally in a 360° range and scanning in a vertical plane under the driving of the rotating head , and after the light is reflected, focused and filtered, the spectrum of multiple bands including the fire characteristic band is projected on the target surface of the infrared sensor in turn.

The present invention also provides a method for applying the system for fire detection, comprising the following steps:

Step 1, in each scanning period, collect the voltage signal amplified and transformed by the signal conditioning module output by the infrared sensor in real time, and save the value of each data point;

Step 2, in the scanning period of the first filter operation, if the value of a certain data point is found to be greater than the set threshold, the data point is a suspected point; A suspected area composed of adjacent suspected points is calculated, and the average value M _n of all the suspected point values in the suspected area is calculated;

Step 3, traverse the data points of the previous scan period of the scan period, generate a suspected area corresponding to the suspected area of the scan period, and calculate the average value M _n-1 of all the suspected point values in the suspected area, if no suspected area is generated. area, set Mn _-1 =0;

Step 4, according to the size of _Mn , Mn _-1 and Mn _-2 , fire detection is performed according to the following steps, where Mn _-2 is the average value of the suspected area in the first two scan periods of the current scan period:

Step 4.1, if L _fire = 0, and M _n /M _n-1 ≥ C ₁ , Mn /M _{n -2} ≥ C ₂ , it is judged that a fire has occurred, and L _fire = 1; L _fire is a fire sign, L _fire = 0 means no fire; L _fire = 1 means fire occurs, C ₁ and C ₂ are the set thresholds;

Step 4.2, if L _fire = 0, and Mn _-1 = Mn _-2 =0, and _Mn >0, it is judged that a fire has occurred, and L _fire =1;

Step 4.3, if L _fire =0, and Mn _-2 =0, _Mn /Mn _-1 ≥C ₁ , it is judged that a fire has occurred, and L _fire =1;

Step 4.4, if L _fire =0, and Mn _-1 =0, _Mn /Mn _-2 ≥C ₂ , it is judged that a fire has occurred, and L _fire =1;

Step 4.5, if L _fire =1 and _Mn /Mn _-1 ≥C ₁ , it is judged that a fire has occurred, and the fire flag L _fire =1 is maintained.

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

In the present invention, a central processing unit, a signal conditioning module, an infrared sensor, a motor driving module, a horizontal motor and a light processing device are provided. The straight plane scans to collect the light, and after the light is reflected, focused and filtered, the spectrum of multiple bands including the fire characteristic band is projected on the target surface of the infrared sensor in turn, realizing the automatic detection of fire. Due to the installation of a light processing device capable of reflecting, focusing, filtering and other processing of light, multi-band infrared spectrum detection based on fire characteristic bands is carried out, which eliminates the influence of sunlight, artificial light source and general heat source spectrum, and improves the fire detection range. and accuracy, reducing the false positive rate.

Description of drawings

1 is a block diagram showing the composition of a fire monitoring system according to an embodiment of the present invention;

FIG. 2 is a schematic structural diagram of a light processing device.

In the figure: 1-central processing unit, 2-signal conditioning module, 3-infrared sensor, 4-light processing device, 41-main mirror, 42-optical pod, 43-switching motor, 44-vertical motor, 45 -1st filter, 46-second filter, 5-motor drive module, 6-horizontal motor, 7-rotating head.

Detailed ways

The present invention will be further described in detail below in conjunction with the accompanying drawings.

A fire monitoring system according to an embodiment of the present invention has a block diagram as shown in FIG. 1 . The system includes: a central processing unit 1 , a signal conditioning module 2 and a motor drive module 5 connected to the central processing unit 1 , and a signal conditioning module 2 . The connected infrared sensor 3, the horizontal motor 6 connected with the motor drive module 5 and used to drive the rotating head 7, also includes a light processing device 4 mechanically connected with the rotating head 7, and the light processing device 4 is used in the rotating head 7. Driven, the light is collected by scanning horizontally in a 360° range and scanning in the vertical plane at the same time, and after the light is reflected, focused, and filtered, the spectrum of multiple bands including the fire characteristic band is projected to the infrared sensor in turn. 3 on the target surface.

In this embodiment, the system is mainly composed of a central processing unit 1 , a signal conditioning module 2 , an infrared sensor 3 and a light processing device 4 . The central processing unit 1 is respectively connected with the signal conditioning module 2 and the motor driving module 5 , and the infrared sensor 3 is connected with the signal conditioning module 2 . in,

The central processing unit 1 is the data processing and control center of the system, and is mainly used for data processing and coordinating the work of each module by outputting control signals. The central processing unit 1 is generally composed of a microprocessor and a data memory.

The infrared sensor 3 is mainly used to convert the sensed infrared radiation of different bands into weak electrical signals, which are amplified and transformed by the signal conditioning module 2 and then sent to the central processing unit 1. The magnitude of the electrical signal is used to determine whether a fire has occurred according to a certain algorithm. In order to reduce the influence of high temperature in the environment, the infrared sensor 3 generally adopts a cooling type infrared sensor. When the refrigerated infrared sensor is working, its own refrigerator can reduce its own temperature. Compared with the uncooled infrared sensor, it has the advantages of higher sensitivity, higher accuracy, and wider detection temperature range. Of course, the disadvantage is that it consumes more energy and costs more.

The signal conditioning module 2 is mainly used for amplifying, high-speed data acquisition (analog-to-digital conversion), filtering and other processing of the weak electrical signal output by the infrared sensor 3 , and sending the output digital signal to the central processing unit 1 .

The light processing device 4 is mainly used for effectively collecting the light on the scene and irradiating it on the target surface of the infrared sensor 3 . Since the present embodiment performs fire detection based on the infrared spectrum emitted by the flame, the light processing device 4 should be able to effectively receive the light on the scene in a wide range. To this end, the light processing device 4 is designed to be able to perform horizontal scanning within a 360° range centered on its installation position, and to perform scanning within a vertical plane, so as to significantly increase the monitoring range. The so-called scanning is a visual description of the movement trajectory of the optical axis of the light-receiving component (such as a mirror) in the light processing device 4 when it rotates. The light processing device 4 can also perform processing such as reflection, focusing, filtering (or filtering) on the collected light, which not only improves the intensity of the light irradiated on the infrared sensor 3, but also obtains a wide range of wavelengths including fire characteristics through filtering processing. band spectrum. The central processing unit 1 determines whether a fire has occurred by comparing the detected signal of the fire characteristic band with a signal of a band different from the band. In this embodiment, the multi-band infrared detection technology including fire characteristic bands is adopted, which improves the fire detection accuracy; meanwhile, due to the filtering process, the interference of the spectrum of sunlight, artificial light sources and general heat sources can be eliminated, and the false alarm rate can be reduced.

As an optional embodiment, the signal conditioning module 2 includes a preamplifier and a data acquisition circuit that are connected to each other.

This embodiment provides a technical solution of the signal conditioning module 2 . The signal conditioning module 2 is mainly composed of a preamplifier and a data acquisition circuit. The preamplifier is used to amplify the weak signal output by the infrared sensor 3, generally including a linear amplifier and a logarithmic amplifier. The linear amplifier is mainly used to meet the gain requirements, and the logarithmic amplifier is used to improve the dynamic range (improve small signals to suppress large signals) . The data acquisition circuit is mainly used to convert analog signals into digital signals under the action of the central processing unit 1, and is generally composed of ADC chips.

As an optional embodiment, the light processing device 4 includes: a switching motor 43 and a vertical motor 44 connected to the motor driving module 5 , and also includes a main reflector 41 mounted on the rotating head 7 and located on the rotating head 7 . The lower part is an optical pod 42 mainly composed of a plurality of mirrors, a first filter 45 and a second filter 46 installed on the turntable below the optical pod 42, and the light passes through the main reflector 41 and the optical pod 42. The plurality of mirrors inside are then irradiated on the first filter 45 or the second filter 46; the main mirror 41 is rotated horizontally by the rotating head 7, and is also driven by the vertical motor 44. The speed of the horizontal rotation speed rotates in the vertical plane; the horizontal rotation of the main mirror 41 is one or half a scan cycle, and at the end of each scan cycle, the switching motor 43 drives the turntable to rotate horizontally, and the optical pod 42 is aligned with the horizontal rotation. The first filter 45 or the second filter 46 is switched to the second filter 46 or the first filter 45; the first filter 45 can transmit the spectrum of the fire characteristic wavelength band, and the second filter 46 A spectrum other than the fire characteristic band can be transmitted.

This embodiment presents a technical solution of the light processing device 4 . The light processing device 4 is composed of three parts: optical components, electrical components and mechanical components. The optical components mainly include the main reflector 41, the optical pod 42, the first filter 45 and the second filter 46; the electrical components mainly include the switching motor 43 and the vertical motor 44; the mechanical components mainly include the rotating head 7, the turntable And the transmission mechanism between the motor and the driven parts. The on-site light is received by the main reflector 41, and after being reflected and focused by multiple reflectors in the optical pod 42, it is irradiated on the first filter 45 or the second filter 46 for filtering, so that the first filter 45 or the second filter 46 is filtered. A filter 45 only transmits the spectrum of the fire characteristic wavelength band, and the second filter 46 transmits the spectrum different from the fire characteristic wavelength band. The rotating head 7 is driven by the horizontal motor 6 to rotate horizontally, which drives the main mirror 41 to also rotate horizontally, so that it can receive light in a range of 360°, that is, to achieve 360° scanning in the horizontal direction. Since the maximum viewing angle that the main mirror 41 can effectively receive light is limited, the horizontal rotation can only ensure the maximum monitoring range in the horizontal direction, and cannot guarantee that there is no dead angle for monitoring in the vertical direction. Therefore, in this embodiment, a vertical motor 44 is provided to drive the main mirror 41 to rotate at a high speed in the vertical plane (up to 20 revolutions per second), that is, its optical axis scans in the vertical plane, which greatly improves the main reflection The monitoring range of the mirror 41 in the vertical direction. When the main mirror 41 scans vertically, it can collect data within a certain pitch angle range toward the front and the back. Therefore, the main mirror 41 can be rotated horizontally for half a circle, that is, 180 degrees, to complete a 360-degree range of data. collection. In this case, a half rotation of the main mirror 41 horizontally is one scanning period. Of course, it is also possible to collect data only when the main mirror 41 faces the front. In this case, one horizontal rotation of the main mirror 41 is one scanning period. At the end of each scanning cycle, the switch motor 43 drives the turntable to rotate horizontally to switch the filters aligned with the optical pod 42 . That is, the two filters work alternately in different scan cycles. The vertical motor 44 and its transmission mechanism and the main reflector 41 are all installed on the rotating head 7; the switching motor 43, the two optical filters and the turntable for fixing them are all installed in the optical pod 42; Bearings and electric slip rings are used for load-bearing connection between the cabins 42, and the optical pod 42 does not move when the rotating head 7 rotates. It is worth noting that this embodiment only provides a preferred implementation. For example, although only two filters are set to obtain spectra of two wavelength bands, it is not limited to two, and three or even more can be set. More filters to obtain more wavelengths of the spectrum.

As an optional embodiment, the system further includes: a camera connected to the central processing unit 1, a tilt motor connected to the motor drive module 5, the camera and the tilt motor are installed on the rotating head 7, and the rotating head 7 drives the camera to level Rotating, the pitching motor can drive the camera to rotate in the vertical plane through the pitching mechanism; the central processing unit 1 judges whether there is a fire by performing image processing on the video signal input by the camera.

This embodiment presents a technical solution for detecting fire based on video image processing. The video detection in this embodiment is a supplement to the infrared detection. When the infrared detection result is a fire, the video detection can be used for further confirmation, so as to improve the accuracy of the fire detection and reduce the false alarm rate. In addition, it can also be used to form a panoramic color map of spectral remote sensing for real-time image monitoring. The video detection part is mainly composed of a camera connected to the central processing unit 1 and a pitch motor connected to the motor drive module 5 . The video camera is mainly used to shoot video images of the scene, and a digital video camera is generally used to directly output digital video signals to the central processing unit 1 . in,

The camera is installed on the rotating head 7, and the rotating head 7 drives the camera to rotate horizontally, so that the camera can shoot video images in a 360° range around it in real time.

The tilt motor is also installed on the rotating head 7, and is mechanically connected with the camera through the tilt mechanism. The pitching mechanism is generally composed of worm gears or gears. The pitching motor can drive the camera to rotate in the vertical plane through the pitching mechanism, that is, pitch up or pitch down. By changing the pitch angle of the camera (the angle between the optical axis and the horizontal plane), the monitoring range in the vertical direction can be extended. For example, after the camera rotates horizontally at a fixed pitch angle for one or several weeks, step the pitch angle up or down an angle and continue to rotate horizontally until it reaches the upper limit position or the lower limit position, and then change the pitch angle in the opposite direction. . In this way, panoramic coverage in the vertical direction can be achieved. The pitch motor generally adopts a stepper motor, which is conducive to precise control. As an improved technical solution, the tilt motor can be eliminated, and a plurality of fixed field of view cameras can be installed in the vertical direction. For example, a plurality of cameras can be mounted on the cylindrical side of the rotating head 7, with the optical axes of all the cameras lying in the same vertical plane and with different pitch angles. The number of cameras is determined by the vertical coverage angle and the vertical field of view of a single camera. Assuming that you need to cover 70° in the vertical direction and use a camera with a 25° vertical field of view, at least 3 cameras are required. Under normal circumstances, the field of view of the camera overlaps a ΔΦ angle, and ΔΦ generally takes 1° to 2°. The technical solution before the improvement requires many rotations to obtain a large field of view in the vertical direction; in the improved technical solution, since multiple cameras with different pitch angles work at the same time, the image stitching algorithm can be used in one scan cycle to obtain a large field of view in the vertical direction. A large field of view is obtained in the straight direction.

As an optional embodiment, the method for judging whether a fire has occurred includes the following steps:

Real-time acquisition of the image f _t (θ _i ) taken when the camera rotates horizontally, f _t (θ _i ) represents a frame of video image taken when the current scanning period, that is, the t-th scanning period, and the azimuth angle of the camera’s optical axis direction is θ _i , i = 1, 2, 3, ...;

Differentiate the video images obtained in the current scan period and the previous scan period: Δf _t (θ _i )=f _t (θ _i )-f _t-1 (θ _i );

Input Δf _t (θ _i ) into the convolutional neural network, if the output of the convolutional neural network is greater than the set threshold, it is considered that a fire has occurred.

This embodiment provides a video image-based fire detection method. When the camera is rotated horizontally, the video images of the scene are generally captured in real time at equal angular intervals. A frame of video image is composed of many pixels, and the pixel data is generally expressed in the form of R, G, B or Y, U, V. For the sake of simplicity, in this embodiment, f _t (θ _i ) is used to represent a frame of video image captured by the camera in the current scanning period when the azimuth angle is θ _i . In the prior art, fire detection is generally realized by directly performing feature extraction on captured video images. However, the practice shows that this method is used for fire detection, and the detection sensitivity is low. In view of this, in this embodiment, the judgment is made by calculating the difference Δf _t (θ _i ) of the video images captured in two adjacent scanning periods. The detection principle is that the video images before and after the fire are greatly different. Inputting the difference Δf _t (θ _i ) into a trained convolutional neural network for convolution operation is equivalent to weighted summation of the values corresponding to different pixel points. After the convolutional neural network, a value V _NN is output. The larger the value of V _NN , the greater the possibility of fire. The V _NN is compared with the set threshold, and if it exceeds the set threshold, it is considered that a fire has occurred. The difference between the video images of two adjacent scanning periods is obtained under the condition that the camera pitch angle remains unchanged, which ensures that the video images of the two scanning periods belong to the same airspace. For the aforementioned improved technical solution, that is, when a plurality of fixed field of view cameras are installed in the vertical direction, the above-mentioned same processing is performed on the video images output by each camera. As long as one detects that a fire has occurred, it is considered that a fire has occurred. .

As an optional embodiment, the system further includes an angle sensor and a limit switch connected to the central processing unit 1 .

In this embodiment, in order to realize the precise control of the tilt angle of the camera, an angle sensor connected to the central processing unit 1 is installed on the tilt mechanism to measure the angle rotated by the camera in the vertical direction; in order to simplify the control, in the tilt mechanism A plurality of limit switches connected to the central processing unit 1 are installed on it.

The present invention also provides an embodiment of a method for fire detection using the system, comprising the following steps:

S101, in each scanning period, collect in real time the voltage signal amplified and transformed by the infrared sensor 3 and output by the signal conditioning module 2, and save the value of each data point;

S102. During the scan cycle of the first filter 45, if the value of a certain data point is found to be greater than the set threshold, the data point is a suspected point; For the suspected area formed by adjacent suspected points, calculate the average value M _n of all the suspected point values in the suspected area;

S103, traverse the data points of the previous scan period of the scan period, generate a suspected area corresponding to the suspected area of the scan period, and calculate the average value M _n-1 of all the suspected point values in the suspected area, if no suspected area is generated , set Mn _-1 =0;

S104, according to the size of _Mn , Mn _-1 and Mn _-2 , perform fire detection according to the following steps, where Mn _-2 is the average value of the suspected area in the first two scan periods of the current scan period;

S1041. If L _fire = 0, and M _n /M _n-1 ≥ C ₁ , Mn /M _{n -2} ≥ C ₂ , it is judged that a fire has occurred, and L _fire = 1; L _fire is a fire sign, and L _fire = 0 means no fire; L _fire = 1 means fire occurs, C ₁ and C ₂ are the set thresholds;

S1042. If L _fire = 0, and Mn _-1 = Mn _-2 =0, and _Mn >0, it is judged that a fire has occurred, and L _fire =1 is set;

S1043. If L _fire = 0, and Mn _-2 = 0, _Mn /Mn _-1 ≥ C ₁ , it is judged that a fire has occurred, and L _fire = 1;

S1044. If L _fire = 0, and Mn _-1 = 0, and _Mn /Mn _-2 ≥ C ₂ , it is judged that a fire has occurred, and L _fire = 1;

S1045 , if L _fire =1, and _Mn /Mn _-1 ≥C ₁ , it is determined that a fire has occurred, and the fire flag L _fire =1 is maintained.

In this embodiment, step S101 is mainly used for data collection. The weak voltage signal output by the infrared sensor 3 is amplified to a certain amplitude by the preamplifier of the signal conditioning module 2, and then converted into a digital signal by the data acquisition circuit under the action of the central processing unit 1, and then read in by the central processing unit 1 and processed. save.

In this embodiment, step S102 is mainly used to generate a suspected area. As mentioned above, a horizontal rotation of the main mirror 41 is a scanning period, and the filters are switched once at the end of each scanning period. characteristic band spectrum. In order to improve the detection sensitivity, in this embodiment, fire identification is performed by using the fire characteristic band spectrum that is sensitive to flame, that is, the infrared spectrum passing through the first filter 45 . In order to further improve the detection efficiency, a multi-band infrared spectrum detection technology (two wavelength bands in this embodiment) is also adopted, and whether a fire occurs is identified by comparing the spectrum data of different wavelength bands. First, the data of the scanning period in which the first optical filter 45 works is processed. If the value of a data point is found to be greater than the set threshold, the data point is considered as a suspected point. In order to eliminate the influence of noise, the size of the threshold can be appropriately increased. Taking the suspected point as the root, other data points are searched in the surrounding area to determine whether it is a suspected point, and a suspected area composed of a plurality of adjacent suspected points is obtained. In order to eliminate the influence of noise, the minimum value of the number of suspected points in the suspected area can be limited, that is, a few suspected points are regarded as noise points. After obtaining the suspected area, calculate the average value of all the suspected point values in the suspected area, and then turn to S103 to judge the fire.

In this embodiment, step S103 is mainly used to process data points of a scan period before the current scan period to generate a suspected area. The scanning period is the scanning period during which the second optical filter 46 works, and the wavelength band corresponding to the data point is a fire characteristic wavelength band different from the working of the first optical filter 45 . Follow the same method as the previous step to generate the suspected area corresponding to the suspected area obtained in the previous step. The so-called correspondence means that the data points of the suspected region of the two scanning periods correspond to the same airspace.

In this embodiment, step S104 is mainly used to perform fire detection according to the average value of the suspected area in three scanning periods. This embodiment provides the determination methods for 5 cases. For the situation described in S1042, in order to reduce false alarms, it is also possible to wait until the next scan period, and calculate the average value Mn ₊₁ of the suspected area corresponding to the next scan period. If _Mn /Mn ₊₁ ≥C ₂ , then A fire is determined to have occurred; otherwise, no fire has occurred. If Mn ₊₁ = 0 also occurs, it is determined that a fire has occurred.

Claims (7)

1. a fire monitoring system is characterized in that, comprising: central processing unit, the signal conditioning module and the motor drive module that are connected to the central processing unit, the infrared sensor that is connected to the signal conditioning module, the one that is connected to the motor drive module for driving The horizontal motor of the rotating head also includes a light processing device mechanically connected to the rotating head, and the light processing device is used to collect light by scanning horizontally in a 360° range and scanning in a vertical plane under the driving of the rotating head , and after the light is reflected, focused and filtered, the spectrum of multiple bands including the fire characteristic band is projected on the target surface of the infrared sensor in turn. 2 . The fire monitoring system according to claim 1 , wherein the signal conditioning module comprises a preamplifier and a data acquisition circuit that are connected to each other. 3 . 3 . The fire monitoring system according to claim 2 , wherein the light processing device comprises: a switching motor and a vertical motor connected to the motor drive module, and a main reflector mounted on the rotating head, located in the rotating head. 4 . The optical pod mainly composed of a plurality of reflectors under the rotating head, the first optical filter and the second optical filter installed on the turntable under the optical pod, the light passes through the main reflector and the multiple optical pods. The reflector is then irradiated onto the first filter or the second filter; the main reflector is driven by the rotating head to rotate horizontally, and is also driven by the vertical motor at a speed significantly higher than the horizontal rotation speed in the vertical plane. Rotation; the main mirror rotates horizontally for one or half a scan cycle. At the end of each scan cycle, the switching motor drives the turntable to rotate horizontally, switching the first filter or the second filter aligned with the optical pod. is the second filter or the first filter; the first filter can transmit the spectrum of the fire characteristic wavelength band, and the second filter can transmit the spectrum different from the fire characteristic wavelength band. 4. The fire monitoring system according to claim 1, wherein the system further comprises: a camera connected to the central processing unit, a pitch motor connected to the motor drive module, the camera and the pitch motor are mounted on the rotating head, The rotating head drives the camera to rotate horizontally, and the tilting motor can drive the camera to rotate in the vertical plane through the tilting mechanism; the central processing unit judges whether a fire occurs by processing the video signal input by the camera. 5. The fire monitoring system according to claim 4, wherein the method for judging whether a fire occurs comprises the following steps: Real-time acquisition of the image f _t (θ _i ) taken when the camera rotates horizontally, f _t (θ _i ) represents a frame of video image taken when the current scanning period, that is, the t-th scanning period, and the azimuth angle of the camera’s optical axis direction is θ _i , i = 1, 2, 3, ...; Differentiate the video images obtained in the current scan period and the previous scan period: Δf _t (θ _i )=f _t (θ _i )-f _t-1 (θ _i ); Input Δf _t (θ _i ) into the convolutional neural network, if the output of the convolutional neural network is greater than the set threshold, it is considered that a fire has occurred. 6. The fire monitoring system according to claim 5, wherein the system further comprises an angle sensor and a limit switch connected to the central processing unit. 7. A method of applying the system of claim 3 to carry out fire detection, characterized in that, comprising the steps of: Step 1, in each scanning period, collect the voltage signal amplified and transformed by the signal conditioning module output by the infrared sensor in real time, and save the value of each data point; Step 2, in the scanning period of the first filter operation, if the value of a certain data point is found to be greater than the set threshold, the data point is a suspected point; A suspected area composed of adjacent suspected points is calculated, and the average value M _n of all the suspected point values in the suspected area is calculated; Step 3, traverse the data points of the previous scan period of the scan period, generate a suspected area corresponding to the suspected area of the scan period, and calculate the average value M _n-1 of all the suspected point values in the suspected area, if no suspected area is generated. area, set Mn _-1 =0; Step 4, according to the size of _Mn , Mn _-1 and Mn _-2 , fire detection is performed according to the following steps, where Mn _-2 is the average value of the suspected area in the first two scan periods of the current scan period: Step 4.1, if L _fire = 0, and M _n /M _n-1 ≥ C ₁ , Mn /M _{n -2} ≥ C ₂ , it is judged that a fire has occurred, and L _fire = 1; L _fire is a fire sign, L _fire = 0 means no fire; L _fire = 1 means fire occurs, C ₁ and C ₂ are the set thresholds; Step 4.2, if L _fire = 0, and Mn _-1 = Mn _-2 =0, and _Mn >0, it is judged that a fire has occurred, and L _fire =1; Step 4.3, if L _fire =0, and Mn _-2 =0, _Mn /Mn _-1 ≥C ₁ , it is judged that a fire has occurred, and L _fire =1; Step 4.4, if L _fire =0, and Mn _-1 =0, _Mn /Mn _-2 ≥C ₂ , it is judged that a fire has occurred, and L _fire =1; Step 4.5, if L _fire =1 and _Mn /Mn _-1 ≥C ₁ , it is judged that a fire has occurred, and the fire flag L _fire =1 is maintained.

Images