CN2516968Y - Suction laser picture smoke identification fire detection device - Google Patents

Abstract

The utility model discloses an air-breathing laser image smoke-detecting fire accident detecting device, which is characterized in that: fume or gasoloid under test is sucked by an aspirator pump into a fume detection room; dispersion image produced under laser function from fume particle is filmed by an CCD video camera before transferred into digital signal by an image capture card and sent to an image signal processor for analyzing and handling; the strength, the projection property, the Eu-number, the complexity, the grain energy measuring, directional measuring and granularity of the dispersion image for the fume particle are inputted into the fume identifying neural network for classifying and judging and identifying whether fire accident fume or non-fire accident factor; alarm is controlled; the controller drives a fire accident linkage alarm equipment. The utility model has the advantages of avoiding interruption from environmental factor, accurately classifying and identifying fire accident fume and non-fire accident gasoloid, being suitable for early, flexibly, quickly and reliably automatically detecting and alarming of fire accident from clean space to bad and complex environment site.

Description

A kind of aspiration type laser image smoke sensing fire hazard sniffer

Technical field:

The utility model belongs to the fire detection technology field, particularly relates to the laser image smoke sensing fire hazard sniffer.

Background technology:

Smog is one of early sign of fire, and it is a key areas of detection that smoke is surveyed.That spot-type smoke detector has is simple in structure, with low cost, application quantity big, characteristics such as wide distribute, it is one of main sniffer in the present automatic fire alarm system, spot-type smoke detector comprises two types of optical detector of fire smoke and ion smoke detectors, and optical detector of fire smoke comprises two kinds on scattered-light type and dim light type again." application laser " the 2nd phase of the calendar year 2001 (Vol.21 that sponsors according to Chinese optics association, No.2, pp.79-81) introduce, there is following defective in spot-type smoke detector owing to the limitation that is subjected on its detection mechanism and the structural design aspect detection sensitivity and the reliability:

1. the anti-interference in air flow of ion smoke detector is indifferent, and is subject to the influence of environment temperature, humidity and air pressure change, owing to adopted radiomaterial in its ionization chamber, can have a negative impact to environment and human body simultaneously, and its application is subjected to certain restriction;

2. optical detector of fire smoke is to the bulky grain smoke particle sensitivity of the generation of glowing, the granule smoke particle sensitivity that ion smoke detector then produces naked light; When detector is set in a certain grade of sensitivity, could report to the police when then granule smog will reach higher concentration, easily cause failing to report; If sensitivity is transferred to higher, then can increase the probability of wrong report;

3. the color of smog has the difference of bright (greyish white) and dark (black), and the scattered-light type optical detector of fire smoke is to canescence cigarette sensitivity, and insensitive to black smoke; Obscuration-type photoelectric smoke detector then has response preferably to black smoke, and insensitive to the canescence cigarette; Add the differences such as composition, grain size of fire hazard aerosol fog, existing some type smoke smoke detection technology is inconsistent to the sensitivity of the naked light of different comburants or identical comburant and the smog that glows;

4. in actual applications, spot-type smoke detector is subject to the interference of non-pyric factors such as dust, electromagnetism, easily reports by mistake.

Summary of the invention:

The utility model provides a kind of naked light to different comburants or identical comburant and smog sensitivity is more consistent and reliability the is higher aspiration type laser image smoke sensing fire hazard sniffer of glowing to overcome the above-mentioned defective of existing fire detector.

This aspiration type laser image smoke sensing fire hazard sniffer, it is characterized in that comprising that laser instrument 3, attenuator 5, smoke detection chamber 7, ccd video camera 11, image pick-up card 13, image-signal processor 14 constitute, one transparency window 2 is arranged on the wall of smoke detection chamber 7, laser instrument 3 is contained in the outside of transparency window 2, the top of smoke detection chamber 7 and a draft tube 1 link, the bottom of smoke detection chamber 7 has an exhausr port and aspiration pump 12 to join, the outlet of aspiration pump 12 is connected to outside the cabinet by gas outlet 9, and ccd video camera 11 is located at the bottom of smoke detection chamber 7 bottom view windows 10; The laser beam light path has an aperture 4 that leads to attenuator 5 through the place on the anterior wall of smoke detection chamber 7; The installation site of described laser instrument 3 makes laser through after the collimated, and is perpendicular with the flow of flue gas direction, and enters attenuator 5 from aperture 4; Image pick-up card 13 is inserted on the slot of image-signal processor 14, and the signal output part of ccd video camera 11 links to each other with the input port of image pick-up card 13, and the output signal of image-signal processor 14 is connected to the input end of alarm controller 15.

Can be provided with a cone 6 in the described attenuator 5, this cone 6 and laser beams coaxial, aperture 4 is pointed in its vertex of a cone tip 8, make through the laser beam after the collimated and enter the tip 8 of directive cones 6 in the attenuator 5 from aperture 4, laser in attenuator 5 through reflecting many times, decay until being absorbed fully; Described attenuator 5 also can be the hollow ball shape structure that has a little ingate.

During use, start aspiration pump smog to be measured or gasoloid are sucked in the smoke detection chamber, laser is incided in the smoke detection chamber, pass smog, reenter to be mapped to and decay in the attenuator until being absorbed fully; When in the smoke detection chamber smoke particle being arranged, smoke particle produces scattered light under laser action, and ccd video camera is taken the dispersion image of smoke particle, is converted to digital signal through image pick-up card, sends into image-signal processor analysis, handles; Image-signal processor is with the input as smog identification neural network of the intensity in the dispersion image of smoke particle, projection properties, Euler's numbers, complexity, texture energy tolerance, directivity tolerance and fineness degree, by smog identification neural network the smog aerosol in the smoke detection chamber is classified, judged, identify when being " fire hazard aerosol fog ", alarm controller drives fire interlink warning equipment; Then do not report to the police when being judged as " non-pyric factor ".

The utility model is a kind of based on the laser image smoke sensing fire hazard Detection Techniques of gasoloid being carried out Classification and Identification; Owing to adopted laser sense cigarette and the analytical technology of initiatively bleeding, improved the sensitivity of surveying, and time of fire alarming is shifted to an earlier date; Owing to adopted CCD imaging and image analysis processing technology, the different comburants or the naked light of identical comburant and the dispersion image feature of the smoke particle that glows become possibility in the feasible extraction smoke detection chamber; Owing to adopted smog identification neural net method, can carry out Classification and Identification to fire hazard aerosol fog and non-fire gasoloid exactly, avoided the interference of environmental factor, have very high reliability; Thereby realized the naked light and the more consistent target of smog sensitivity of glowing, can accurately identify the interference of non-pyric factors such as dust, reached reliable and stable fire detecting and alarm different comburants or identical comburant.The utility model fire detecting arrangement is applicable to early stage, sensitive, quick, the reliable automatic fire detection-alarm from clean room to abominable complex environment place.

The major parameter of the dispersion image of smog identification neural network input has: the temporal correlation of the projection properties on (gradation of image) analysis of the intensity of smoke particle dispersion image and correlation analysis, the laser beam vertical direction, the Euler's numbers of particle picture and complexity and sequence image.

The intensity of smoke particle dispersion image (gradation of image) is analyzed according to rice (Mie) scattering theory, projection is calculated the dispersion image of two dimension is transformed to one-dimension array, projection according to dispersion image is carried out rude classification to image, and provides the particle diameter of bulky grain particles such as dust; The Euler's numbers of particle picture are meant the poor of white region number and hole number, and whether the distribution density of small-particle, particle etc. is arranged during the complexity computational analysis in combining image zone is published picture and looked like; With the correlativity that the seasonal effect in time series dispersion image has, analyze the concentration and the moving situation of smoke particle.

CALCULATION OF PARAMETERS is as follows in the program:

Euler's numbers:

EUL4＝V-(E1+E2)+R (1)

In EUL8=V-(E1+E2)-(D1+D2)+T-R (2) formula, EUL4 is 4 neighborhood Euler's numbers; EUL8 is 8 neighborhood Euler's numbers; V is the analyzed area sum of all pixels; E1 is 2 and crosses the horizontal linking number of threshold pixels; E2 is 2 and crosses the vertical linking number of threshold pixels; D1 is 2 and crosses threshold pixels diagonal angle linking number; D2 is 2 and crosses threshold pixels opposition angle linking number; T is 3 and crosses threshold pixels L type linking number; R be 4 cross threshold pixels 2 * 2 in abutting connection with number.

Complexity:

CMP=L ²In/S (3) formula, CMP is a complexity; L is that the profile in zone is long; S is the area in zone.

Texture energy tolerance: $\mathrm{ENERGY}=\underset{u=0}{\overset{M-1}{\mathrm{\Σ}}}\underset{v=0}{\overset{M-1}{\mathrm{\Σ}}}\left|F(u,v)\right|......\left(4\right)$ In the formula, ENERGY is a texture energy tolerance, the general status of texture variations in its reflection window; (u v) is that (promptly (u represents v) in the spectrogram that coordinate is (u, value v) to F for x, the frequency spectrum that y) obtains behind fourier transform for the intensity f of two dimensional image to F; M is the length of square image blocks, and promptly the image block size is M * M.

Directivity tolerance: $\mathrm{DIR}=\underset{u=0}{\overset{M-1}{\mathrm{\Σ}}}\underset{v=0}{\overset{M-1}{\mathrm{\Σ}}}{\tan }^{-1}(v/u)\left|F(u,v)\right|/\mathrm{ENERGY}......\left(5\right)$ In the formula, DIR is a directivity tolerance, and it is the weighted sum at grain direction angle; (u v) is the frequency spectrum of two dimensional image to F; U, v are coordinate position; ENERGY is a texture energy tolerance; M is the length of square image blocks.

Fineness degree: $\mathrm{FIN}=\frac{r_4-r_3}{r_2-r_1}\·\frac{\underset{\mathrm{\θ}=0}{\overset{\mathrm{\π}/2}{\mathrm{\Σ}}}\underset{r=r_3}{\overset{r_4}{\mathrm{\Σ}}}\left|F(r,\mathrm{\θ})\right|}{\underset{\mathrm{\θ}=0}{\overset{\mathrm{\π}/2}{\mathrm{\Σ}}}\underset{r=r_1}{\overset{r_2}{\mathrm{\Σ}}}\left|F(r,\mathrm{\θ})\right|}....\left(4\right)$ In the formula, FIN is a fineness degree, and it reflects grain size; (r θ) is F (u, polar form v) to F; r ₁, r ₂Be the judge utmost point footpath of scope of polar coordinates; r ₃, r ₄Judge the utmost point footpath of particle for desire under the polar coordinates; θ is the polar angle under the polar coordinates.

The utility model adopts Artificial Neural Network, intensity in the dispersion image of the smoke particle that Flame Image Process is calculated, projection properties, Euler's numbers, complexity, texture energy tolerance, directivity tolerance and fineness degree are as the input of smog identification feedforward neural network, and dissimilar fire hazard aerosol fogs all is arranged to a certain sensitivity, neural network is by after carrying out learning training to sample set, to the judgement of classifying of the smog aerosol in the smoke detection chamber, identify " fire hazard aerosol fog " or " non-pyric factor ", thereby realize the naked light of different comburants or identical comburant and the smog sensitivity of glowing than the higher fire detecting and alarm of reliability of making peace.

Description of drawings:

Accompanying drawing 1 is the principle of compositionality figure of system of a kind of aspiration type laser image smoke sensing fire hazard sniffer of the utility model.

Accompanying drawing 2 is the structure diagram of a kind of attenuator of designing in the utility model.

Accompanying drawing 3 is smog identification feedforward neural network structural representation.

Accompanying drawing 4 is picture signal processing of the present utility model, judgement and control program process flow diagram.

Embodiment:

Embodiment 1

Present embodiment aspiration type laser image smoke sensing fire hazard sniffer, by comprising semiconductor laser 3, attenuator 5, smoke detection chamber 7, colourful CCD video camera 11, image pick-up card 13, image-signal processor 14 constitutes at interior parts, one transparency window 2 is arranged on the wall of smoke detection chamber 7, semiconductor laser 3 is contained in the outside of transparency window 2, the top of smoke detection chamber 7 and a draft tube 1 link, bottom one exhausr port and the aspiration pump 12 of smoke detection chamber 7 join, the outlet of aspiration pump 12 is connected to outside the cabinet by gas outlet 9, the inside of smoke detection chamber 7 is through oxidation processes, form one deck black matt film, to reduce inner diffuse reflection, colourful CCD video camera 11 is located at the bottom of smoke detection chamber 7 bottom view windows 10, and transparency window 2 and view window 10 all adopt optical glass to process; The laser beam light path has an aperture 4 that leads to attenuator 5 through the place on the anterior wall of smoke detection chamber 7, and hole diameter is 5mm; The wavelength of described semiconductor laser 3 is 670nm, and laser is through after the collimated, the laser beam that forming a beam diameter is 2mm, the angle of divergence is very little, and its installation site makes laser beam and flow of flue gas direction perpendicular, and enters the attenuator 5 from aperture 4; Image pick-up card 13 is inserted on the slot of image-signal processor 14, and the signal output part of colourful CCD video camera 11 links to each other with the input port of image pick-up card 13, and the output signal of image-signal processor 14 is connected to the input end of alarm controller 15.

Be provided with a cone 6 in the attenuator 5 that present embodiment adopts, this cone 6 and laser beams coaxial, aperture 4 is pointed in its tip 8, the projection that on smoke detection chamber 7, is connected with attenuator 5, its inwall be one with cone 6 coaxial, equidirectional recessed circular conical surfaces, adopt between the projection of attenuator 5 and smoke detection chamber 7 to be threaded, and with O-ring seals 16 sealings; The laser that laser instrument 3 sends is through after the collimated, pass smoke detection chamber 7, enter in the attenuator 5 from aperture 4, the tip 8 of directive cone 6, laser in attenuator 5 cone 6 and the recessed circular conical surface in the connection projection of smoke detection chamber 7 between through reflecting many times, decay until being absorbed fully.

During work, smog to be measured or gasoloid are sucked in the smoke detection chamber by aspiration pump, semiconductor laser sends the laser that a branch of wavelength is 670nm, incide in the smoke detection chamber, pass smog, incide the tip of coming the cone of light direction in the attenuator with the sensing of laser beams coaxial, make laser in attenuator through reflecting many times, decay until being absorbed fully; When in the smoke detection chamber smoke particle being arranged, smoke particle produces scattered light under the effect of laser, and colourful CCD video camera is shot the dispersion image of smoke particle, is converted to digital signal through image pick-up card, sends into image-signal processor analysis, handles; Image-signal processor is with the input as smog identification neural network of the intensity in the dispersion image of smoke particle, projection properties, Euler's numbers, complexity, texture energy tolerance, directivity tolerance and fineness degree, by smog identification neural network the smog aerosol in the smoke detection chamber is classified, judged, identify when being " fire hazard aerosol fog ", alarm controller drives fire interlink warning equipment; When being judged as " non-pyric factor ", then do not drive fire interlink warning equipment.

Present embodiment has adopted typical three layers of feedforward neural network to discern neural network model as fire hazard aerosol fog, the input I of network ₁, I ₂, I ₃, I ₄,---, I _i, the intensity in the dispersion image of corresponding smoke particle, projection properties, Euler's numbers, complexity, texture energy tolerance, directivity are measured and fineness degree respectively, and input value all normalizes to [0,1]; Be output as O ₁And O ₂, respectively corresponding " fire hazard aerosol fog " and " non-pyric factor ", the output valve scope also is [0,1]; Middle hidden layer has j neuron (M ₁, M ₂..., M _j), i * j bar line is arranged between input layer and hidden layer, its weights are w1 _Ij, 2 * j bar line is arranged between hidden layer and output layer, its weights are w2 _JkWeight w 1 between input layer and the hidden layer _IjAnd between hidden layer and the output layer weight w 2 _Jk, behind network training, determine by sample set.

Claims (3)

1, a kind of aspiration type laser image smoke sensing fire hazard sniffer, it is characterized in that comprising laser instrument (3), attenuator (5), smoke detection chamber (7), ccd video camera (11), image pick-up card (13), image-signal processor (14) constitutes, one transparency window (2) is arranged on the wall of smoke detection chamber (7), laser instrument (3) is contained in the outside of transparency window (2), the top of smoke detection chamber (7) and a draft tube (1) link, the bottom of smoke detection chamber (7) has an exhausr port and aspiration pump (12) to join, the outlet of aspiration pump (12) is connected to outside the cabinet by gas outlet (9), and ccd video camera (11) is located at the bottom of smoke detection chamber (7) bottom view window (10); The laser beam light path has an aperture (4) that leads to attenuator (5) through the place on the anterior wall of smoke detection chamber (7); The installation site of described laser instrument (3) makes laser through after the collimated, and is perpendicular with the flow of flue gas direction, and enters attenuator (5) from aperture (4); Image pick-up card (13) is inserted on the slot of image-signal processor (14), the signal output part of ccd video camera (11) links to each other with the input port of image pick-up card (13), and the output signal of image-signal processor (14) is connected to the input end of alarm controller (15).

2, aspiration type laser image smoke sensing fire hazard sniffer as claimed in claim 1, be characterised in that and be provided with a cone (6) in the described attenuator (5), this cone (6) and laser beams coaxial, aperture (4) is pointed in its vertex of a cone tip (8), makes the tip (8) that enters the interior directive cone of attenuator (5) (6) through the laser beam after the collimated from aperture (4).

3, aspiration type laser image smoke sensing fire hazard sniffer as claimed in claim 1 is characterised in that described attenuator (5) is the hollow ball shape structure that has a little ingate.

Images