CN108538011A - A kind of laser radar fire detection method - Google Patents
A kind of laser radar fire detection method Download PDFInfo
- Publication number
- CN108538011A CN108538011A CN201810184097.4A CN201810184097A CN108538011A CN 108538011 A CN108538011 A CN 108538011A CN 201810184097 A CN201810184097 A CN 201810184097A CN 108538011 A CN108538011 A CN 108538011A
- Authority
- CN
- China
- Prior art keywords
- frequency
- spectrogram
- layer
- disturbance
- obtains
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B17/00—Fire alarms; Alarms responsive to explosion
- G08B17/10—Actuation by presence of smoke or gases, e.g. automatic alarm devices for analysing flowing fluid materials by the use of optical means
- G08B17/103—Actuation by presence of smoke or gases, e.g. automatic alarm devices for analysing flowing fluid materials by the use of optical means using a light emitting and receiving device
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/88—Lidar systems specially adapted for specific applications
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Emergency Management (AREA)
- Chemical & Material Sciences (AREA)
- Business, Economics & Management (AREA)
- Analytical Chemistry (AREA)
- Computer Networks & Wireless Communication (AREA)
- Electromagnetism (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Photometry And Measurement Of Optical Pulse Characteristics (AREA)
- Fire-Detection Mechanisms (AREA)
Abstract
The present invention proposes a kind of laser radar fire detection method, this method acquires photo-signal by laser radar first, Short Time Fourier Transform is carried out to signal, obtain the time-frequency distributions of signal, feature extraction is carried out to time-frequency distributions result again, the micro-disturbance feature of air-flow is obtained, then depth convolutional neural networks is used to carry out fire identification.Beneficial effects of the present invention are as follows：Take full advantage of the air-flow micro-disturbance feature generated when flame combustion, signal is acquired by laser radar, generate time frequency signal spectrogram, and then Classification and Identification is carried out to air-flow micro-disturbance spectrogram by using depth convolutional neural networks, the new method for realizing fire detection, improves the accuracy rate and fire detection speed of fire identification.
Description
Technical field
The present invention relates to fire detection technical fields, particularly relate to a kind of laser radar fire detection method.
Background technology
Fire incident is generally sudden accident, has high risks to personal safety and property safety.Fire on road
Calamity is typically to be led to fuel leakage by vehicle spontaneous combustion, car crass and caused；In building fire then mainly by electrical equipment,
The reasons such as circuit aging or other dangerous material and cause.It is especially on road after fire occurs in tunnel, serious friendship can be caused
Pass blocking plug, safety dredging are difficult；Fire occurs in building, is easy to cause evacuating personnel difficulty.Various harm based on fire,
It is thought that by installing fire detection and alarm system, realizes the discovery as early as possible of fire, takes measures as early as possible, to mitigate fire
Harm.
In recent years, fire detection is rapidly developed, fire detection method is mainly the following at present：
One is distributed fiber optic temperature detection techniques, are connected by optical fiber between host and measurand, by electricity
The distribution curve that signal carries out processing and comparing calculation goes out temperature along optical fiber, the major defect of this method is that the reaction time is long,
Service life is short, position inaccurate of alarming, and needs welding optical cable again after detecting fire behavior, leads to accuracy decline, then add
Upper system price is high, so hardly resulting in large-scale promotion.
One is dual wavelength flame automatic measurement techniques, by the specific wavelength and spectral range that detect radiant light in fire
To judge fire.Two sensors that can receive different-waveband are arranged in system, using variation in combustion frequency and spatial distribution this two
A feature recognition flame combustion realizes fire detection.This method reaction speed is fast, wrong report is few；But its major defect is to first
Phase flame is insensitive, affected by environment larger, and price is high, maintenance is big.
One is optical fiber and grating sensing temperature Detection Techniques, using the light sensitivity of fiber optic materials, when the temperature change of fiber grating
When, linear change occurs for the centre wavelength for the narrow band light being reflected back, to measure the temperature of corresponding monitoring point.This method is fixed
Level is true, affected by environment small.The disadvantage is that autgmentability is poor, demodulated equipment is complicated, expensive.
One is video flame monitoring technology, shoot video image with video camera, then collect and carry out number in computer
Image procossing identifies flame, to judge fire according to the characteristic feature of flame.This method can directly utilize original regard
Frequency monitoring image, the disadvantage is that being disturbed very greatly, being easy wrong report and failing to report.
The above fire detection method respectively has advantage and disadvantage, has certain limitation in practical applications, therefore there is an urgent need to one
The kind fire detection method that detection speed is fast, accuracy rate is high.
Invention content
The present invention proposes a kind of laser radar fire detection method, solves fire detection technology in the prior art and uses office
Sex-limited big problem.
The technical proposal of the invention is realized in this way：
A kind of laser radar fire detection method, method and step are as follows：
(1) laser radar continuous acquisition photo-signal is used, and intercepts the signal data x [k] of a period of time sequence in order；
(2) calculating formula is used to signal data x [k]
Short Time Fourier Transform is carried out, obtains time-frequency matrix, wherein X (m, f) is time-frequency matrix, and w [k] is window function, and k is corresponding defeated
Enter the time of signal, m is the sliding position of window function, and f is frequency；
(3) modulus is carried out to time-frequency matrix X (m, f), obtains spectrogram X1(m, f), to spectrogram X1(m, f) is filtered out
Zero-frequency processing, i.e., to X1The value tax zero of each four points Y (m, -2) in (m, f) and its left and right, Y (m, -1), Y (m, 1) and Y (m, 2), obtains
To spectrogram X2(m, f)；
(4) to spectrogram X2(m, f) carries out analyzing processing, after filtering out aperiodic Doppler frequency principal component, obtains new
Reflection air-flow micro-disturbance Doppler spectrum X3(m, f)；
(5) a depth convolutional neural networks are established, network structure is followed successively by input layer, convolutional layer, pond layer, convolution
Layer, pond layer, flat layer and output layer；Wherein input layer number is the points of micro-disturbance spectrogram, and output layer is a section
Point, when it is 1 to export node layer output, judgement has fire；Each layer activation primitive of the network structure is relu functions, output
Activation primitive is softmax；Utilize the Doppler spectrum X3(m, f) is divided into the depth of training sample and test sample to foundation
Degree convolutional neural networks are trained study and test, obtain trained depth convolutional neural networks；
(6) in fire detection, real-time acquisition is calculated into the air-flow micro-disturbance spectrogram X that analysis obtains3(m, f) is inputed to
Trained depth convolutional neural networks are classified in step (5), identify currently whether there is fire.
Preferably, the specific algorithm of the step (4) is as follows：
I, using calculating formulaTo X2(m, f) center of gravity in a frequency direction, obtains
To center of gravity sequence W (m)；
II, fitting of a polynomial is carried out using least square method to W (m), obtains cubic curve formula ξ (t)=a3t3+a2t2+
a1t+a0；
III, to spectrogram X2(m, f) carries out frequency spectrum shift, removes Doppler frequency principal component, retains reflection flame and causes
The Doppler frequency of air-flow micro-disturbance presses offset ξ (t)=a to the frequency spectrum at point m3m3+a2m2+a1m+a0It is moved, is obtained
New air-flow micro-disturbance spectrogram X3(m, f).
The realization principle of the present invention is as follows：
When flame combustion, the frequency of flicker is typically in the low frequency range of 10~20Hz, since flashing for flame can be to surrounding
Air forms micro-disturbance.Disturbed air forms the air-flow with some cycles pulsation rule fine motion speed, using laser thunder
Up to the Doppler frequency that can detect air-flow, after filtering out zero-frequency and aperiodic frequency principal component, reflection air-flow perturbation is obtained
Dynamic Doppler frequency spectrum, flame can be judged by being identified by the feature to the Doppler frequency spectrum.
Beneficial effects of the present invention are：
Classification and Identification is carried out to air-flow micro-disturbance spectrogram by using depth convolutional neural networks, improves fire identification
Accuracy rate and fire detection speed.
Specific implementation mode
Below in conjunction with the embodiment of the present invention, technical scheme in the embodiment of the invention is clearly and completely described,
Obviously, described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.Based in the present invention
Embodiment, the every other embodiment that those of ordinary skill in the art are obtained without creative efforts, all
Belong to the scope of protection of the invention.
Embodiment
A kind of laser radar fire detection method based on the study of micro-disturbance depths of features, is as follows：
(1) laser radar continuous acquisition photo-signal is used, and intercepts the signal data x of a period of time sequence in order
[k], it is 2 seconds that the length of time window is taken in the present embodiment, in data of the acquisition for training, needs to be provided with flame and nothing
The scene of flame, gathered data as much as possible；
(2) this segment signal data x [k] collected to step (1) uses calculating formulaShort Time Fourier Transform is carried out, time-frequency is obtained
Matrix, wherein X (m, f) is time-frequency matrix, and w [k] is window function, and k is the time of corresponding input signal, and m is the sliding of window function
Position, f are frequencies；
(3) modulus is carried out to time-frequency matrix X (m, f), obtains spectrogram X1(m, f) takes window function in the present embodiment
Length of window is 2048, spectrogram X1The row of (m, f) represents the time, row represent frequency, amplitude represents the real number of Doppler's intensity
Two-dimensional array；
To spectrogram X1(m, f) carries out filtering out zero-frequency processing, i.e., to X1Each four points Y (m, -2) in (m, f) and its left and right, Y
The value of (m, -1), Y (m, 1) and Y (m, 2) assign zero, obtain spectrogram X2(m, f)；By filtering out zero-frequency processing, signal can be removed
The influence of middle fixed object and the intrinsic zero-frequency of system.
(4) the spectrogram X that step (3) is obtained2(m, f) carries out analyzing processing, filters out aperiodic Doppler frequency master
After component, the Doppler frequency spectrum of reflection air-flow micro-disturbance is obtained, specific algorithm is as follows：
I, using calculating formulaTo X2(m, f) center of gravity in a frequency direction, obtains
To center of gravity sequence W (m)；
II, fitting of a polynomial is carried out using least square method to W (m), obtains cubic curve formula ξ (t)=a3t3+a2t2+
a1t+a0；
III, to spectrogram X2(m, f) carries out frequency spectrum shift, removes Doppler frequency principal component, retains reflection flame and causes
The Doppler frequency of air-flow micro-disturbance presses offset ξ (t)=a to the frequency spectrum at point m3m3+a2m2+a1m+a0It carries out movement and (works as ξ
(m) when ＞ 0, moving direction is negative；As ξ (m) ＜ 0, moving direction is just；The vacant locations of removal are 0), to obtain new gas
Flow micro-disturbance spectrogram X3(m, f)；
(5) a depth convolutional neural networks are established, network structure is followed successively by input layer, convolutional layer, pond layer, convolution
Layer, pond layer, flat layer and output layer；Wherein input layer number is the points of micro-disturbance spectrogram, and output layer is a section
Point, when the output of the node of output layer is 1, judgement has fire；Other each layers of above-mentioned network structure in addition to last layer swash
Function living is relu functions, and output activation primitive is softmax；The air-flow micro-disturbance frequency obtained using step (4) gathered data
Spectrogram X3(m, f) is divided into training sample and test sample and is trained study and test to the depth convolutional neural networks of foundation,
Obtain trained depth convolutional neural networks, wherein there is the training data enrolled when flame and test data to be labeled as 1, nothing
The training data and test data enrolled when flame are labeled as 0.
(6) in fire detection, real-time acquisition is calculated into the air-flow micro-disturbance spectrogram X that analysis obtains3(m, f) is inputed to
Trained depth convolutional neural networks are classified in step (5), identify currently whether there is fire；If Current neural net
When network output result is 1, expression currently has fire.
In conclusion the method for the invention takes full advantage of the air-flow micro-disturbance feature generated when flame combustion, pass through
Laser radar acquires signal, generates time frequency signal spectrogram, and then by using depth convolutional neural networks to air-flow micro-disturbance
Spectrogram carries out Classification and Identification, realizes the new method of fire detection, improves the accuracy rate and fire detection speed of fire identification.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
With within principle, any modification, equivalent replacement, improvement and so on should all be included in the protection scope of the present invention god.
Claims (2)
1. a kind of laser radar fire detection method, which is characterized in that its method and step is as follows：
(1) laser radar continuous acquisition photo-signal is used, and intercepts the signal data x [k] of a period of time sequence in order；
(2) calculating formula is used to signal data x [k]
Short Time Fourier Transform is carried out, obtains time-frequency matrix, wherein X (m, f) is time-frequency matrix, and w [k] is window function, and k is corresponding defeated
Enter the time of signal, m is the sliding position of window function, and f is frequency；
(3) modulus is carried out to time-frequency matrix X (m, f), obtains spectrogram X1(m, f), to spectrogram X1(m, f) carries out filtering out zero-frequency
Processing, i.e., to X1The value tax zero of each four points Y (m, -2) in (m, f) and its left and right, Y (m, -1), Y (m, 1) and Y (m, 2), obtains frequency
Spectrogram X2(m, f)；
(4) to spectrogram X2(m, f) carries out analyzing processing and obtains new reflection after filtering out aperiodic Doppler frequency principal component
The Doppler spectrum X of air-flow micro-disturbance3(m, f)；
(5) a depth convolutional neural networks are established, network structure is followed successively by input layer, convolutional layer, pond layer, convolutional layer, pond
Change layer, flat layer and output layer；Wherein input layer number is the points of micro-disturbance spectrogram, and output layer is a node, when
When output node layer output is 1, judgement has fire；Each layer activation primitive of the network structure is relu functions, output activation
Function is softmax；Utilize the Doppler spectrum X3(m, f) is divided into training sample and test sample and is rolled up to the depth of foundation
Product neural network is trained study and test, obtains trained depth convolutional neural networks；
(6) in fire detection, real-time acquisition is calculated into the air-flow micro-disturbance spectrogram X that analysis obtains3(m, f) inputs to step
(5) trained depth convolutional neural networks are classified in, identify currently whether there is fire.
2. a kind of laser radar fire detection method according to claim 1, which is characterized in that the tool of the step (4)
Body algorithm is as follows：
I, using calculating formulaTo X2(m, f) center of gravity in a frequency direction obtains weight
Heart sequence W (m)；
II, fitting of a polynomial is carried out using least square method to W (m), obtains cubic curve formula ξ (t)=a3t3+a2t2+a1t+
a0；
III, to spectrogram X2(m, f) carries out frequency spectrum shift, removes Doppler frequency principal component, retains reflection flame and causes air-flow
The Doppler frequency of micro-disturbance presses offset ξ (t)=a to the frequency spectrum at point m3m3+a2m2+a1m+a0It is moved, is obtained new
Air-flow micro-disturbance spectrogram X3(m, f).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810184097.4A CN108538011B (en) | 2018-03-07 | 2018-03-07 | Laser radar fire detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810184097.4A CN108538011B (en) | 2018-03-07 | 2018-03-07 | Laser radar fire detection method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108538011A true CN108538011A (en) | 2018-09-14 |
| CN108538011B CN108538011B (en) | 2020-12-15 |
Family
ID=63486785
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810184097.4A Active CN108538011B (en) | 2018-03-07 | 2018-03-07 | Laser radar fire detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108538011B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109379311A (en) * | 2018-09-30 | 2019-02-22 | 中国人民解放军战略支援部队信息工程大学 | Ultrashort wave signal specific recognition methods based on convolutional neural networks |
| CN112200788A (en) * | 2020-10-16 | 2021-01-08 | 清华大学 | High-temperature deformation measuring device and method |
| CN113643503A (en) * | 2021-07-23 | 2021-11-12 | 上海嘉筠通信技术有限公司 | Adopt 77G radar module to realize smoke alarm of amortization function |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040239912A1 (en) * | 2001-05-30 | 2004-12-02 | Rui Mario Correia Da Silva Vilar | Lidar system controlled by computer for smoke identification applied, in particular, to early stage forest fire detection |
| CN202711408U (en) * | 2012-06-18 | 2013-01-30 | 中国南方电网有限责任公司超高压输电公司 | Ion spatial electric current density-based direct current transmission line mountain fire monitoring device |
| CN204143592U (en) * | 2014-11-04 | 2015-02-04 | 无锡北斗星通信息科技有限公司 | Based on the forest fire detection platform of satellite remote sensing images |
| CN104933821A (en) * | 2015-06-15 | 2015-09-23 | 华南理工大学 | Method used for calculating transmission line corridor forest fire smoke concentration alarm threshold |
| CN105683729A (en) * | 2014-01-23 | 2016-06-15 | 通用显示器公司 | Multi-spectral flame detector with radiant energy estimation |
| CN106910309A (en) * | 2017-04-18 | 2017-06-30 | 南昌航空大学 | Forest fire detecting system based on unmanned plane during flying platform |
-
2018
- 2018-03-07 CN CN201810184097.4A patent/CN108538011B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040239912A1 (en) * | 2001-05-30 | 2004-12-02 | Rui Mario Correia Da Silva Vilar | Lidar system controlled by computer for smoke identification applied, in particular, to early stage forest fire detection |
| CN202711408U (en) * | 2012-06-18 | 2013-01-30 | 中国南方电网有限责任公司超高压输电公司 | Ion spatial electric current density-based direct current transmission line mountain fire monitoring device |
| CN105683729A (en) * | 2014-01-23 | 2016-06-15 | 通用显示器公司 | Multi-spectral flame detector with radiant energy estimation |
| CN204143592U (en) * | 2014-11-04 | 2015-02-04 | 无锡北斗星通信息科技有限公司 | Based on the forest fire detection platform of satellite remote sensing images |
| CN104933821A (en) * | 2015-06-15 | 2015-09-23 | 华南理工大学 | Method used for calculating transmission line corridor forest fire smoke concentration alarm threshold |
| CN106910309A (en) * | 2017-04-18 | 2017-06-30 | 南昌航空大学 | Forest fire detecting system based on unmanned plane during flying platform |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109379311A (en) * | 2018-09-30 | 2019-02-22 | 中国人民解放军战略支援部队信息工程大学 | Ultrashort wave signal specific recognition methods based on convolutional neural networks |
| CN109379311B (en) * | 2018-09-30 | 2021-08-17 | 中国人民解放军战略支援部队信息工程大学 | Ultra-short wave specific signal identification method based on convolutional neural network |
| CN112200788A (en) * | 2020-10-16 | 2021-01-08 | 清华大学 | High-temperature deformation measuring device and method |
| CN113643503A (en) * | 2021-07-23 | 2021-11-12 | 上海嘉筠通信技术有限公司 | Adopt 77G radar module to realize smoke alarm of amortization function |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108538011B (en) | 2020-12-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101556724B (en) | Safety management system of optical fiber perimeter and pattern recognition method thereof | |
| CN108538011A (en) | A kind of laser radar fire detection method | |
| US10467764B2 (en) | Method for counting and characterization of particles in a fluid in movement | |
| EP1389331B1 (en) | Self-aspirating fire detection system | |
| Fouda et al. | Pattern recognition of optical fiber vibration signal of the submarine cable for its safety | |
| EP3321906B1 (en) | High sensitivity fiber optic based detection | |
| EP3321905B1 (en) | High sensitivity fiber optic based detection | |
| CN105765361A (en) | Bearing monitoring apparatus and method | |
| CN112525201B (en) | Underwater target tracking method based on electromagnetic field characteristic multi-information fusion | |
| CN113362560B (en) | Photoelectric smoke sensing detection method for accurately identifying fire smoke | |
| CN109489721A (en) | A kind of water environment sensitizing range remote sensing recognition system | |
| CN109120336B (en) | False alarm prevention and false alarm prevention method based on phase sensitive optical time domain reflection sensor | |
| CN103399363A (en) | Online observation device and method for weather phenomena based on light attenuation and scattering theory | |
| CN111986436A (en) | Comprehensive flame detection method based on ultraviolet and deep neural networks | |
| CN116989913A (en) | Cable line temperature monitoring method and system based on optical fiber sensing | |
| CN107193013B (en) | method and system for measuring marine environment parameters | |
| US11361643B2 (en) | High sensitivity fiber optic based detection system | |
| CN106092843A (en) | A kind of PM2.5 real-time detecting system based on Fibre Optical Sensor | |
| CN107545684A (en) | A kind of multipoint disturbance event alarming method in distributed optical fiber vibration sensing system | |
| CN202383294U (en) | Multiple meteorological parameter synchronous measurement laser radar | |
| US12073700B2 (en) | Chambered high sensitivity fiber optic smoke detection | |
| CN107784261B (en) | Road surface state detection method and device | |
| CN104198348A (en) | System and method for PM2.5 concentration detection based on photoelectric integration | |
| CN1038622C (en) | Fire-alarm system for the early detection of fires | |
| CN115615332A (en) | State on-line monitoring system of optical fiber composite overhead ground wire |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |