CN101373553A - Early-stage smog video detecting method capable of immunizing false alarm in dynamic scene - Google Patents
Early-stage smog video detecting method capable of immunizing false alarm in dynamic scene Download PDFInfo
- Publication number
- CN101373553A CN101373553A CNA2008101217195A CN200810121719A CN101373553A CN 101373553 A CN101373553 A CN 101373553A CN A2008101217195 A CNA2008101217195 A CN A2008101217195A CN 200810121719 A CN200810121719 A CN 200810121719A CN 101373553 A CN101373553 A CN 101373553A
- Authority
- CN
- China
- Prior art keywords
- smog
- video
- background
- scene
- early
- 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
- 238000000034 method Methods 0.000 title claims abstract description 38
- 230000003053 immunization Effects 0.000 title claims description 12
- 238000004458 analytical method Methods 0.000 claims abstract description 27
- 239000000779 smoke Substances 0.000 claims abstract description 26
- 238000001514 detection method Methods 0.000 claims abstract description 21
- 238000012544 monitoring process Methods 0.000 claims abstract description 9
- 230000000007 visual effect Effects 0.000 claims abstract description 6
- 230000001186 cumulative effect Effects 0.000 claims abstract description 5
- 235000019504 cigarettes Nutrition 0.000 claims description 43
- 241000208125 Nicotiana Species 0.000 claims description 17
- 235000002637 Nicotiana tabacum Nutrition 0.000 claims description 17
- 238000009825 accumulation Methods 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 13
- 230000009466 transformation Effects 0.000 claims description 9
- 238000012423 maintenance Methods 0.000 claims description 8
- 230000006835 compression Effects 0.000 claims description 6
- 238000007906 compression Methods 0.000 claims description 6
- 230000003068 static effect Effects 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 239000003086 colorant Substances 0.000 claims description 3
- 230000004069 differentiation Effects 0.000 claims description 3
- 230000036039 immunity Effects 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 238000004422 calculation algorithm Methods 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 claims description 2
- 230000008030 elimination Effects 0.000 claims description 2
- 238000003379 elimination reaction Methods 0.000 claims description 2
- 230000007547 defect Effects 0.000 abstract description 2
- 230000001960 triggered effect Effects 0.000 abstract 1
- 238000005516 engineering process Methods 0.000 description 14
- 230000035508 accumulation Effects 0.000 description 9
- 230000008859 change Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 230000008447 perception Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 235000002918 Fraxinus excelsior Nutrition 0.000 description 1
- 244000283207 Indigofera tinctoria Species 0.000 description 1
- 230000001154 acute effect Effects 0.000 description 1
- 239000002956 ash Substances 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 239000003344 environmental pollutant Substances 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 231100000614 poison Toxicity 0.000 description 1
- 230000007096 poisonous effect Effects 0.000 description 1
- 231100000719 pollutant Toxicity 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000005728 strengthening Methods 0.000 description 1
- 230000008719 thickening Effects 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses an early-stage smoke video detecting method which is immune to the false alarm in a dynamic scene. A turntable drives an IP web cam to intermittently rotate along the horizontal direction, and temporarily stops after rotating for each IP web cam visual angle; the video image data of the monitored scene are collected by the IP web cam, and are converted and transmitted to a monitoring microcomputer through an optical-electrical converter and an Ethernet; the monitoring microcomputer carries out the study and the cumulative evidence analysis to the video image data, monitors the dynamic background of the monitored scene, and judges and analyzes if the smoke is truly available in the monitored scene; and an alarm is triggered if the smoke is determined to be available. The invention overcomes the defects of the prior art that the false alarm rate under the opened dynamic scene is high, the detection robustness is low, and the like, and can correctly detect out the early-stage smoke within 4-6 seconds.
Description
Technical field
The invention belongs to the computer digital image process field, relate to video image and obtain technology, video image enhancement techniques, video image segmentation technology, Image Compression, the multithreading treatment technology of video image; Be particularly related to a kind of early-stage smog video detecting method that can immunizing false alarm in dynamic scene.
Background technology
For preventing that fire from taking place and dangerous chemical leakage, prior art roughly is divided into: the single-point type analog prober that the energy-sensitive smoke detector of air suction type smoke detector, single-point type smoke detector (ion type smog detector, photoelectric smoke detector), projected bundle's formula smoke detector, reflecting bundle formula smoke detector, air sampling smoke detector, radiation (ultraviolet smoke detector, infrared ray smoke detector, spark ashes smoke detector), image are right etc.Its detection principle is the variation according to physical quantitys such as smog, gas, temperature, proposes Rapid Alarm.Yet above-mentioned smoke detection Technology Need could brought into play the detection effect quickly and efficiently from smog generation closer place, source.In space length and all relatively large place (as: power house, grain depot, oil depot, hangar, big warehouse, ancient buildings, tunnel, railway station, shopping center etc.) of floor area, or exist the place of air blast, above-mentioned smoke detector is installed then can not be played a role well, the early warning of its fire prevention, anti-leak is a very difficult problem all the time.
Video smoke by optical imagery is surveyed, and its operating distance is far away relatively, need not wait until smog near or arrive the detecting device mounting points, whether existence that just can the perception fact to be.Existing method is then based on Wavelet Detection, and small echo subimage LH, HL and HH have comprised the level of background image, the high-frequency information at vertical and diagonal angle, and its contour edge produces local extremum in the small echo subimage.Contour edge in the background image can be thickened by the smog covering, and may disappear over time because of the smog thickening, causes wavelet coefficient values to reduce because of visibility reduces, thereby realizes detecting.
Yet there is the rate of false alarm height in existing video detection technology, detects critical defects such as robustness is low in dynamic scene.Therefore, the dynamic background in big disturbance scene is learnt automatically, and to the static state of similar smog, the automatic rejecting of dynamic object, is the key of early-stage smog feature video detection technique.
Summary of the invention
The present invention is primarily aimed at the deficiency that the conventional smoke Fog detector exists, and a kind of early-stage smog video detecting method that can immunizing false alarm in dynamic scene is provided.
The objective of the invention is to be achieved through the following technical solutions: a kind of early-stage smog video detecting method that can immunizing false alarm in dynamic scene, rotary head drives IP network video camera clearance-type ground along continuous straight runs and rotates back and forth, the The Cloud Terrace revolution suspends after crossing an IP network video camera visual angle, by the monitored scene video view data of IP network camera acquisition, and be transferred to control microcomputer by the conversion of photoelectric commutator and Ethernet, control microcomputer is learnt and the evidence cumulative analysis vedio data, obtain the dynamic background in the monitoring scene, and whether there is real smog in the monitored scene of discriminatory analysis, if confirm to have smog, then trigger and report to the police; Wherein, described vedio data is learnt to comprise following concrete steps with the process of evidence cumulative analysis:
(1) off-line learning of the RGB components operation of early-stage smog combination:, obtain the best RGB color component computing combination of cutting apart early stage grey cigarette, blue or green cigarette, tobacco by the off-line learning of color gamut.
(2) frame of video being carried out smog based on gamut compression cuts apart: during detection, create multithreading earlier, make up by the best RGB color component computing that the smog colour feature study is obtained, divide 3 the tunnel color video frame carried out RGB components operation and bit mask thereof, cut apart in real time obtain have grey cigarette, the similar smog district of blue or green cigarette, tobacco color characteristic.
(3) study of the background in the frame of video DYNAMIC COMPLEX scene and maintenance: in the testing process, carry out the dynamic learning and the maintenance update of background simultaneously, obtain the dynamic background of video, safeguard that with background background subtraction down eliminates static class in the scene like the smog district, make up with the high frequency of wavelet transformation again and eliminate dynamic similar smog district interference in the scene.
(4) the long sequence smog accumulation of evidence analysis of frame of video and the connection analysis of two field picture: carry out the connected domain analysis of the accumulation of evidence and the frame of multi-frame video, by analyzing sub-thread result's exclusive disjunction with grey cigarette, blue or green cigarette, tobacco, realize the wrong report immunity in the dynamic scene, differentiation marks real smog.
The invention has the beneficial effects as follows: the present invention has overcome the rate of false alarm height that exists in the prior art, has detected shortcomings such as robustness is low under open dynamic scene; Can get rid of automatically and cause rigidity or non-rigid mobile object, sunshine or the pollutant reported by mistake, and discern slow cloud layer automatically and change with smog and change; The operating distance that video smoke is surveyed is far away relatively, can reach more than the 100m.
Description of drawings
Fig. 1 is the former figure of dynamic scene that comprises blue or green cigarette and similar blue or green cigarette district,
Fig. 2 is to the circulation study of Fig. 1 with RGB colour gamut feature, obtains the exemplary graph as a result in similar smog district in conjunction with the FloodFill method that contains error band,
Fig. 3 be representation class among Fig. 2 like the template figure of smog region of interest,
Fig. 4 uses Fig. 1
Components operation and bit mask be figure as a result,
Fig. 5 extracts the figure as a result that former figure shown in Figure 1 obtains similar smog with Fig. 4,
Fig. 6 is the trace plot (initial background is the 1st frame, comprises present frame, interim background, permanent these 3 curves of background) of the background study maintenance process of certain image block in the video sequence for example,
Fig. 7 is the dynamic scene frame of video figure that comprises smog and similar smog,
Fig. 8 is RGB components operation and wavelet transformation analysis figure thereof, and its upper left corner is the high fdrequency component synoptic diagram after radio-frequency component subgraph HL, LH, HH synthesize,
Fig. 9 is the binaryzation figure as a result that Fig. 8 is carried out obtaining after last 7 bit masks very noisy,
Figure 10 (a) and (b), (c), (d) are the former figure of dynamic video frame and the tracking piecemeal locations drawing thereof for example,
Figure 10 (e) is the tracking oscillogram that each the frame image block wavelet energy value of dynamic video to Figure 10 (a)-(d) give an example changes,
Figure 11 is to the Smoke Detection of dynamic scene video sequence 1 exemplary graph as a result,
Figure 12 is to the Smoke Detection of dynamic scene video sequence 2 exemplary graph as a result,
Figure 13 is the process flow diagram of the overall technology path of implementing,
Figure 14 is the process flow diagram of the technology path implemented in the sub-thread.
Embodiment
Detection principle of the present invention is to adopt computer vision technique and the immune operation rule of wrong report, the regional change that photographs in the image sequence is judged through a series of filtrations, when finding that motion feature relevant with the smog behavior and color characteristic occur, then judging automatically at short notice has smog episode to take place.No matter rig camera is placed on 10m or 100m place far away, does not need the detecting device of smog arrival by the time, just can detect its characteristic signal immediately at the smog initial stage that produces, and reaches the preventive effect of early stage monitoring.This technology is particularly useful for the fire hazard monitoring in the aforementioned large-scale place, and is used in poisonous, harmful place the automatic monitoring of carrying out dangerous chemical leakage etc. based on visual theory.
The early-stage smog video detecting method of the present invention's energy immunizing false alarm in dynamic scene realizes that on the smog video detection system this smog video detection system comprises IP network video camera, rotary head, Ethernet transmission system, photoelectric commutator and microcomputer; Be installed in the IP network camera acquisition vedio data of rotary head, change and be transferred to microcomputer by photoelectric commutator and Ethernet and carry out video analysis.Specifically, adopt IP network video camera (but important place is at additional far thermal camera at night) as sensor, by video image processing technology and wrong report immune detection method of discrimination, automatically discern the special diffusion motion of smog and the pattern feature of early-stage smog, many evidence accumulations with spatial domain and time domain strengthen the generation of judging smog episode, and with monitoring result with color mark and automatically the alarm communication system mode export, in time give the alarm to the system manager.This detection principle does not need by the time that smog arrives detecting device, and whether existence that just can the perception fact, therefore, can monitor potential disaster event in the very first time.Adopt the TCP/IP Network Video Transmission, can realize that telemonitoring controls in the different location, adopt LAN (Local Area Network) optical fiber to transmit the video image that the IP video camera monitors, the information transmission distance can reach more than several kilometers.All detect the data resource unification and are managed by database, and detection system is the time and the content of recording events generation automatically, so that the tracking enquiry measured target.System can realize importing, the derivation of view data library information easily, and shares other data.
Method of the present invention is: rotary head drives IP network video camera clearance-type ground along continuous straight runs and rotates back and forth, the The Cloud Terrace revolution stops after crossing an IP network video camera visual angle, the monitored scene video view data of IP network camera acquisition is transferred to control microcomputer by photoelectric commutator and Ethernet conversion, microcomputer is learnt vedio data and is analyzed, obtain the background in the monitoring scene, whether there is real smog in the monitored scene of discriminatory analysis then, if confirm to have smog, then trigger and report to the police; Wherein, vedio data is learnt and the process analyzed comprises following concrete steps: the off-line learning of the RGB of (1) early-stage smog (reddish yellow indigo plant) components operation combination:, obtain the best RGB color component computing combination of cutting apart early stage grey cigarette, blue or green cigarette, tobacco by the off-line learning of color gamut; (2) frame of video being carried out smog based on gamut compression cuts apart: during detection, create multithreading earlier, make up by the best RGB color component computing that the smog colour feature study is obtained, divide 3 the tunnel color video frame carried out RGB components operation and bit mask thereof, cut apart the smog district that obtains class ashy cigarette, blue or green cigarette, tobacco color characteristic in real time; (3) study of the background in the frame of video DYNAMIC COMPLEX scene and maintenance: in the testing process, carry out the dynamic learning and the maintenance update of background simultaneously, obtain the dynamic background of video, safeguard that with background background subtraction down eliminates static class in the scene like the smog district, the high frequency with wavelet transformation makes up the seemingly smog district interference of eliminating in the scene of dynamic class again; (4) the long sequence smog accumulation of evidence analysis of frame of video and the connection analysis of two field picture: carry out the connected domain analysis of the accumulation of evidence and the frame of multi-frame video, by analyzing sub-thread result's exclusive disjunction with grey cigarette, blue or green cigarette, tobacco, realize the wrong report immunity in the dynamic scene, differentiation marks real smog.
Below in conjunction with accompanying drawing core technology of the present invention is described in further detail, it is more obvious that purpose of the present invention and effect will become.
1, the off-line learning of the RGB components operation of early-stage smog combination:
By the situation that investigation fire early-stage smog takes place, find that fire generation usual earlier exists grey cigarette, blue or green cigarette, these three kinds of typical smog of tobacco, can satisfy fire alarm under most situations substantially to the detection of this three classes early-stage smog.This method adopts multithreading to create three and analyzes sub-thread, according to the color gamut feature of grey cigarette, blue or green cigarette, tobacco, simultaneously the color video two field picture is carried out dividing processing respectively.Concrete steps are as follows:
(a) with the circulation study of RGB colour gamut feature,, obtain smoke characteristics core region of interest template (ROI) in conjunction with containing error band FloodFill algorithm.The result is for example shown in Fig. 2,3.
(b) according to the computing formula of RGB components operation
T(i，j)＝{rR(i，j)+gG(i，j)+bB(i，j)|r，g，b∈[-3，3]} (1)
Obtain the feature gray-scale map after full images is strengthened by color.
T (i, j) expression is to image i, and j ranks place pixel carries out the resulting feature gray-scale map of color component combinatorial operation.R (i, j), G (i, j), B (i, j) presentation video i, the rgb value of j ranks place pixel, r, g, b for corresponding R respectively (i, j), G (i, j), B (i, RGB component combination parameter j), exhaustive automatically [3.00,3.00] in the RGB combination coefficient of 0.01 step-length, and ask for characteristic pattern by formula (1), itself and region of interest are learnt the absolute value that template is asked the binary map difference, add up the plain value of this result images and be 1 quantity, judge results of learning to search for its minimum value, and terminating point is learnt in control.This method is represented the likelihood score of the two coupling with ρ:
In the formula: R, C are expressed as the row, column number of matched image, P respectively
j(i, c), (i c) represents the binary map of the j time combined result and region of interest study template respectively to Q.After study is finished, get and make the r of correspondence when ρ is minimum value, g, b are optimum RGB component combination parameter.Figure 4 shows that the segmentation effect figure that adopts this components operation and bit mask method that the exemplary graph 1 of the dynamic scene that comprises blue or green cigarette and similar blue or green cigarette district is obtained.
2, frame of video is cut apart based on the smog of gamut compression:
(a) color compressed of bit mask.
The color compressed of bit mask (colour bits of conductively-closed are put 0) is a kind of color compressed method that is similar to uniform sampling in the three-dimensional true color of RGB space.(that is: the pixel component value and 11111000 that scans is asked and computing as 3 colors in end in 8 of: each components of shielding RGB color
), the image after the compression changes on visual effect not quite, is greatly reduced but calculate the search volume.
(b) with the RGB combination parameter of the grey cigarette that obtains with RGB components operation off-line learning in the step 1, blue or green cigarette, tobacco, calculate characteristic pattern by formula (1), reduce in 8 colors last 7 with 10000000 bit masks and computing then, obtain similar smog district by extracting former figure again.Figure 5 shows that from the former figure of dynamic scene (Fig. 1) that comprises blue or green cigarette and similar blue or green cigarette district and cut apart the similar smog district that obtains.
3, study of the background in the frame of video DYNAMIC COMPLEX scene and maintenance
At the above similar smog interference problem (as: sky, fog) of handling the back existence, with the dynamic background of present frame with the study gained, after carrying out RGB components operation and bit mask thereof separately, do difference again, remove the static similar smog that retains substantially and disturb.
Spatial domain small echo denoising: take one deck discrete wavelet to decompose to frame of video, comprise after the decomposition that a width of cloth low-frequency component subgraph LL becomes Molecular Graphs HL, LH, HH frequently with three panel heights.HL, LH, HH are respectively vertically, level, to the edge and the texture information of angular direction.The filling the air of smog, translucent performance makes image edge, the reduction of the sharp property of texture, causes the HFS energy in the small echo subgraph to reduce, and can eliminate dynamic class smog district by the energy variation that detects the small echo subgraph and disturb.Based on this fact, definition:
w
n(x，y)＝|LH
n(x，y)|
2+|HL
n(x，y)|
2+|HH
n(x，y)|
2 (3)
In the formula, w
n(x, y) the former figure of expression n frame (x, the y) energy of position, LH
n(x, y), HL
n(x, y), HH
n(x y) represents n frame (x, y) the one-level wavelet coefficient of the corresponding radio-frequency component in position respectively.For reducing noise effect and improving operation efficiency, to former figure piecemeal, the size of piece is (k
1, k
2), can be taken as 2 pixel *, 2 pixels.Per minute piece (k, wavelet energy E (l k)
1, l
2) computing method are:
Handle image at yuv space, can do wavelet transformation Y component (brightness); Handle at rgb space, generally select the R component.The present invention utilizing color to form gray-scale map afterwards, carries out wavelet transformation to this gray-scale map with the components operation of rgb space then.Fig. 8 is RGB components operation+wavelet transformation figure, and Fig. 9 is the very noisy position analysis figure that components operation+wavelet analysis+bit mask obtains, and Fig. 8 transformation results is carried out binaryzation design sketch after 7 bit masks that is:.Through the data of components operation and wavelet transformation,, just can isolate very noisy zone (seeing the upper left corner figure of Fig. 9) again through bit mask.Adopt the method for this RGB components operation+wavelet analysis+bit mask, judge the common region of Fig. 9 very noisy zone and similar smog gauge point by contrast, eliminate the very noisy point that dynamically " background subtraction " carries over, obtain clean, dynamic similar smog district.
Background study of the present invention and maintaining method comprise automatic extraction and upgrade interim background and permanent background two parts.In the process of upgrading background, upgrade interim background earlier, when interim change of background amount runs up to a certain degree, upgrade permanent background again.Introduce interim background as the benefit of upgrading buffer zone, it is contaminated to be that the permanent background of learning to obtain is difficult in complex scene, can be updated adaptively along with the variation of dynamic scene again.
This step is specific as follows:
(a) according to the mobile object criterion, the identification moving target.
(b), in interim background, upgrade its corresponding image block to non-moving target area.
(c) if the difference of the interim background of image block and permanent background is accumulated when surpassing threshold value, then upgrade its permanent background:
Wherein, the mobile object criterion of above-mentioned step (a) is: | I
n(x, y)-B (x, y) | T, I
n(x, y)-the n two field picture, B (x, y)-forever (x, y) block of pixels, T-discrimination threshold on the background.B in the formula (3)
Tmp(x y) is interim background, and its α is iteration control parameter (being about 0.95), controls the renewal speed of interim background; B
Perm(x y) is permanent background, and its β is the renewal speed controlled variable (being about 0.7) of permanent background.Fig. 6 employing formula (5) of having given an example carries out temporarily-the renewal graph of a relation of permanent background study certain image block of certain frame of video, and from Fig. 6 as seen: permanent background can be updated adaptively, and is difficult for disturbed pollution.
4, the long sequence smog accumulation of evidence analysis of frame of video and the connection analysis of two field picture
Fill the air edge and the texture that blocks background gradually along with smog produces also, the small echo subgraph energy value of background reduces gradually.After smog sheltered from background fully, the small echo subgraph energy of background went to zero.This is that the energy of its HFS is lower because the texture of smog is level and smooth relatively.From the wavelet energy wave form varies figure of Figure 10 (e) as seen: the wavelet energy change procedure that smog covers the background performance slowly changes along a smooth curve.When entering surveillance zone, blocked background as the chaff interference (as: personage) of similar smog color, from sequence image as seen: acute variation can take place in the small echo subgraph energy of background, and energy value increases suddenly.Based on the energy variation stationarity, can further judge the true and false in smog district.The concrete steps that the long sequence smog accumulation of evidence analysis of this frame of video and the connection of two field picture are analyzed are:
(a) the time domain window by the accumulation of 40～120 frame smog evidences comes the analysis video sequence.
(b) the every blocks of pixels of statistics is as the times N of smog candidate district appearance
1 n(l
1, l
2), and the times N that occurs as the smog candidate regions continuously of front and back two frames
2 n(l
1, l
2).
(c) judge: if block of pixels (l
1, l
2) N that in this time domain window, adds up
1 n(l
1, l
2) and N
2 n(l
1, l
2) above preset threshold (T
1, T
2), then be judged to real smog zone; Otherwise then be chaff interference, have the interference of the mobile thing of similar smog color with further elimination.
(d) these three of grey cigarettes, blue or green cigarette, tobacco are analyzed sub-threads and the accumulation of evidence result of same frame of video is carried out or operate,, show that then there is smog really in scene if the result is very, otherwise, then there is not smog.
(e) analyze by the area size of each piecemeal that is judged to smog being carried out connected domain, filter out the mark zone of area, further eliminate little assorted point and disturb, obtain clean early-stage smog nucleus mark less than setting threshold (as: 20 pixels).
Figure 13 is the overall technology path process flow diagram of implementing of the early-stage smog video detecting method of this energy immunizing false alarm in dynamic scene, and it can judge the existence of the grey cigarette in the dynamic scene, blue or green cigarette, tobacco simultaneously; Figure 14 is when judging in the dynamic scene grey cigarette, blue or green cigarette, tobacco, the Smoke Detection technology path process flow diagram of implementing in the sub-thread of a certain judgement.
At last, the noise removal function as the connected domain of strengthening the final real smog judge mark of picture frame is analyzed can further omit the operation steps of the relevant spatial domain small echo denoising in the step 3, and it is embodiment of the present invention two.The rotary head that drives the IP network video camera also can remain on fixed-direction, and does not rotate back and forth.
Advantage of the present invention and remarkable result are for having multiple static state, dynamic disturbance thing in the dynamic scene In the situation, can accurately, stably detect grey cigarette, blue or green cigarette, Huang at fire early stage (in 4～6 seconds) This 3 quasi-representative smog of cigarette provides timely early warning. This technology has overcome the conventional smoke Fog detector open big empty Between upper restriction, and the high shortcoming of rate of false alarm that exists of conventional images type flame/smoke detection system.
Claims (5)
- One kind can immunizing false alarm in dynamic scene early-stage smog video detecting method, it is characterized in that, rotary head drives IP network video camera clearance-type ground along continuous straight runs and rotates back and forth, the The Cloud Terrace revolution suspends after crossing an IP network video camera visual angle, by the monitored scene video view data of IP network camera acquisition, and be transferred to control microcomputer by the conversion of photoelectric commutator and Ethernet, control microcomputer is learnt and the evidence cumulative analysis vedio data, obtain the dynamic background in the monitoring scene, and whether there is real smog in the monitored scene of discriminatory analysis, if confirm to have smog, then trigger and report to the police.Wherein, described vedio data is learnt to comprise following concrete steps with the process of evidence cumulative analysis:(1) off-line learning of the RGB components operation of early-stage smog combination:, obtain the best RGB color component computing combination of cutting apart early stage grey cigarette, blue or green cigarette, tobacco by the off-line learning of color gamut.(2) frame of video being carried out smog based on gamut compression cuts apart: during detection, create multithreading earlier, make up by the best RGB color component computing that the smog colour feature study is obtained, divide 3 the tunnel color video frame carried out RGB components operation and bit mask thereof, cut apart in real time obtain have grey cigarette, the similar smog district of blue or green cigarette, tobacco color characteristic.(3) study of the background in the frame of video DYNAMIC COMPLEX scene and maintenance: in the testing process, carry out the dynamic learning and the maintenance update of background simultaneously, obtain the dynamic background of video, safeguard that with background background subtraction down eliminates static class in the scene like the smog district, make up with the high frequency of wavelet transformation again and eliminate dynamic similar smog district interference in the scene.(4) the long sequence smog accumulation of evidence analysis of frame of video and the connection analysis of two field picture: carry out the connected domain analysis of the accumulation of evidence and the frame of multi-frame video, by analyzing sub-thread result's exclusive disjunction with grey cigarette, blue or green cigarette, tobacco, realize the wrong report immunity in the dynamic scene, differentiation marks real smog.
- 2. early-stage smog video detecting method that can immunizing false alarm in dynamic scene according to claim 1 is characterized in that described step (1) is specially:(a) with the circulation study of RGB colour gamut feature,, obtain smoke characteristics core region of interest template in conjunction with containing error band FloodFill algorithm.(b) according to the computing formula of RGB components operationT(i，j)＝{rR(i，j)+gG(i，j)+bB(i，j)|r，g，b∈[-3，3]}Obtain the feature gray-scale map after picture frame is strengthened by color; In the formula, and T (i, j) expression is to image i, j ranks place pixel carries out the resulting feature gray-scale map of color component combinatorial operation, and R (i, j), G (i, j), B (i, j) difference presentation video i, the rgb value of j ranks place pixel, r, g, b for corresponding R respectively (i, j), G (i, j), B (i, RGB component combination parameter j).
- 3. early-stage smog video detecting method that can immunizing false alarm in dynamic scene according to claim 1 is characterized in that described step (2) is specially:(a) color compressed of bit mask.(b) in the step (1), the RGB combination parameter calculation of the grey cigarette that obtains with RGB components operation off-line learning, blue or green cigarette, tobacco obtains the feature gray-scale map, reduce in 8 colors last 7 with 10000000 bit masks and computing then, obtain similar smog district by extracting former figure again.
- 4. early-stage smog video detecting method that can immunizing false alarm in dynamic scene according to claim 1 is characterized in that described step (3) is specially:(a) according to the mobile object criterion, the identification moving target; The mobile object criterion is: | I n(x, y)-B (x, y) |〉T, wherein, I n(x y) is the n two field picture, and (x is that (T is a discrimination threshold for x, y) block of pixels on the permanent background y) to B.(b), in interim background, upgrade its corresponding image block to non-moving target area.(c) if the difference of the interim background of image block and permanent background is accumulated when surpassing threshold value, then upgrade its permanent background:In the formula, B Tmp(x y) is interim background, and its α is the iteration control parameter, controls the renewal speed of interim background; B Perm(x y) is permanent background, and its β is the renewal speed controlled variable of permanent background.
- 5. early-stage smog video detecting method that can immunizing false alarm in dynamic scene according to claim 1 is characterized in that described step (4) is specially:(a) the time domain window by the accumulation of 40～120 frame smog evidences comes the analysis video sequence.(b) the every blocks of pixels of statistics is as the times N of smog candidate district appearance -1 n(l 1, l 2), and the times N that occurs as the smog candidate regions continuously of front and back two frames 2 n(l 1, l 2).(c) judge: if block of pixels (l 1, l 2) N that in this time domain window, adds up -1 n(l 1, l 2) and N 2 n(l 1, l 2) above preset threshold (T 1, T 2), then be judged to real smog zone; Otherwise then be chaff interference, have the interference of the mobile thing of similar smog color with further elimination.(d) these three of grey cigarettes, blue or green cigarette, tobacco are analyzed sub-threads and the accumulation of evidence result of same frame of video is carried out or operate,, show that then there is smog really in scene if the result is very, otherwise, then there is not smog.(e) analyze by the area size of each piecemeal that is judged to smog being carried out connected domain, filter out the mark zone of area, further eliminate little assorted point and disturb, obtain clean early-stage smog nucleus mark less than setting threshold.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008101217195A CN101373553B (en) | 2008-10-23 | 2008-10-23 | Early-stage smog video detecting method capable of immunizing false alarm in dynamic scene |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008101217195A CN101373553B (en) | 2008-10-23 | 2008-10-23 | Early-stage smog video detecting method capable of immunizing false alarm in dynamic scene |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101373553A true CN101373553A (en) | 2009-02-25 |
| CN101373553B CN101373553B (en) | 2010-06-16 |
Family
ID=40447703
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2008101217195A Expired - Fee Related CN101373553B (en) | 2008-10-23 | 2008-10-23 | Early-stage smog video detecting method capable of immunizing false alarm in dynamic scene |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101373553B (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101833838A (en) * | 2010-05-27 | 2010-09-15 | 王巍 | Large-range fire disaster analyzing and early warning system |
| CN102074091A (en) * | 2011-01-26 | 2011-05-25 | 南京大学 | Distributed optical fiber temperature sensing wavelet analysis alarm method and system for tunnel fire |
| CN101789124B (en) * | 2010-02-02 | 2011-12-07 | 浙江大学 | Segmentation method for space-time consistency of video sequence of parameter and depth information of known video camera |
| CN102567722A (en) * | 2012-01-17 | 2012-07-11 | 大连民族学院 | Early-stage smoke detection method based on codebook model and multiple features |
| CN102621772A (en) * | 2012-03-19 | 2012-08-01 | 北京汉邦高科数字技术股份有限公司 | Method for manually focusing high definition serial digital interface (HD-SDI) camera or internet protocol (IP) network camera |
| CN102646312A (en) * | 2012-05-11 | 2012-08-22 | 武汉大学 | Forest smoke-fire monitoring and recognizing method suitable for distributed type parallel processing |
| CN101587622B (en) * | 2009-06-18 | 2012-09-05 | 李秋华 | Forest rocket detecting and identifying method and apparatus based on video image intelligent analysis |
| CN105741471A (en) * | 2016-04-22 | 2016-07-06 | 天津天利航联火灾报警设备有限公司 | Compound detector for detecting smoke and temperature of vehicle and detection method thereof |
| CN105915840A (en) * | 2016-04-05 | 2016-08-31 | 三峡大学 | Factory smoke emission automatic monitoring method based on video signal |
| CN106023199A (en) * | 2016-05-19 | 2016-10-12 | 重庆扬讯软件技术有限公司 | Image analysis technology-based fume blackness intelligent detection method |
| CN108419043A (en) * | 2017-06-30 | 2018-08-17 | 唐山钢铁集团微尔自动化有限公司 | A kind of mist detecting device based on intelligent video analysis |
| CN108921215A (en) * | 2018-06-29 | 2018-11-30 | 重庆邮电大学 | A kind of Smoke Detection based on local extremum Symbiotic Model and energy spectrometer |
| CN109204106A (en) * | 2018-08-27 | 2019-01-15 | 浙江大丰实业股份有限公司 | Stage equipment mobile system |
| CN109272688A (en) * | 2018-09-25 | 2019-01-25 | 武汉理工光科股份有限公司 | The tactful automatic adjusting method of fiber grating perimeter security system alarm and system |
| CN109289347A (en) * | 2018-12-03 | 2019-02-01 | 宁波锦觅网络科技有限公司 | Specific aim purifying and recovering system platform |
| CN109447999A (en) * | 2018-09-29 | 2019-03-08 | 佛山市云米电器科技有限公司 | Oil smoke image processing method based on dynamic background modeling |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU2002220440B2 (en) * | 2000-12-28 | 2007-08-23 | Siemens Schweiz Ag | Video smoke detection system |
| CN1325903C (en) * | 2001-06-22 | 2007-07-11 | 瑞士西门子有限公司 | Video-smoke detection system and method for examining the same |
-
2008
- 2008-10-23 CN CN2008101217195A patent/CN101373553B/en not_active Expired - Fee Related
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101587622B (en) * | 2009-06-18 | 2012-09-05 | 李秋华 | Forest rocket detecting and identifying method and apparatus based on video image intelligent analysis |
| CN101789124B (en) * | 2010-02-02 | 2011-12-07 | 浙江大学 | Segmentation method for space-time consistency of video sequence of parameter and depth information of known video camera |
| CN101833838A (en) * | 2010-05-27 | 2010-09-15 | 王巍 | Large-range fire disaster analyzing and early warning system |
| CN101833838B (en) * | 2010-05-27 | 2012-06-06 | 王巍 | Large-range fire disaster analyzing and early warning system |
| CN102074091A (en) * | 2011-01-26 | 2011-05-25 | 南京大学 | Distributed optical fiber temperature sensing wavelet analysis alarm method and system for tunnel fire |
| CN102074091B (en) * | 2011-01-26 | 2012-04-25 | 南京大学 | Distributed optical fiber temperature sensing wavelet analysis alarm method and system for tunnel fire |
| CN102567722A (en) * | 2012-01-17 | 2012-07-11 | 大连民族学院 | Early-stage smoke detection method based on codebook model and multiple features |
| CN102621772A (en) * | 2012-03-19 | 2012-08-01 | 北京汉邦高科数字技术股份有限公司 | Method for manually focusing high definition serial digital interface (HD-SDI) camera or internet protocol (IP) network camera |
| CN102646312A (en) * | 2012-05-11 | 2012-08-22 | 武汉大学 | Forest smoke-fire monitoring and recognizing method suitable for distributed type parallel processing |
| CN105915840A (en) * | 2016-04-05 | 2016-08-31 | 三峡大学 | Factory smoke emission automatic monitoring method based on video signal |
| CN105915840B (en) * | 2016-04-05 | 2019-04-30 | 三峡大学 | A method of the factory smoke discharge based on vision signal monitors automatically |
| CN105741471A (en) * | 2016-04-22 | 2016-07-06 | 天津天利航联火灾报警设备有限公司 | Compound detector for detecting smoke and temperature of vehicle and detection method thereof |
| CN106023199A (en) * | 2016-05-19 | 2016-10-12 | 重庆扬讯软件技术有限公司 | Image analysis technology-based fume blackness intelligent detection method |
| CN106023199B (en) * | 2016-05-19 | 2019-02-01 | 重庆扬讯软件技术股份有限公司 | A kind of flue gas blackness intelligent detecting method based on image analysis technology |
| CN108419043A (en) * | 2017-06-30 | 2018-08-17 | 唐山钢铁集团微尔自动化有限公司 | A kind of mist detecting device based on intelligent video analysis |
| CN108921215A (en) * | 2018-06-29 | 2018-11-30 | 重庆邮电大学 | A kind of Smoke Detection based on local extremum Symbiotic Model and energy spectrometer |
| CN109204106A (en) * | 2018-08-27 | 2019-01-15 | 浙江大丰实业股份有限公司 | Stage equipment mobile system |
| CN109204106B (en) * | 2018-08-27 | 2020-08-07 | 浙江大丰实业股份有限公司 | Stage equipment moving system |
| CN109272688A (en) * | 2018-09-25 | 2019-01-25 | 武汉理工光科股份有限公司 | The tactful automatic adjusting method of fiber grating perimeter security system alarm and system |
| CN109447999A (en) * | 2018-09-29 | 2019-03-08 | 佛山市云米电器科技有限公司 | Oil smoke image processing method based on dynamic background modeling |
| CN109289347A (en) * | 2018-12-03 | 2019-02-01 | 宁波锦觅网络科技有限公司 | Specific aim purifying and recovering system platform |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101373553B (en) | 2010-06-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101373553B (en) | Early-stage smog video detecting method capable of immunizing false alarm in dynamic scene | |
| CN106910203B (en) | The quick determination method of moving target in a kind of video surveillance | |
| CN102663869B (en) | Indoor fire detection method based on video monitoring platform | |
| CN110874592B (en) | Forest fire smoke image detection method based on total bounded variation | |
| CN103714325B (en) | Left object and lost object real-time detection method based on embedded system | |
| CN101441771B (en) | Video fire hazard smoke detecting method based on color saturation degree and movement mode | |
| CN109447168A (en) | A kind of safety cap wearing detection method detected based on depth characteristic and video object | |
| CN101739686B (en) | Moving object tracking method and system thereof | |
| CN101908141B (en) | Video smoke detection method based on mixed Gaussian model and morphological characteristics | |
| CN110516609A (en) | A kind of fire video detection and method for early warning based on image multiple features fusion | |
| CN112800860B (en) | High-speed object scattering detection method and system with coordination of event camera and visual camera | |
| CN107229894A (en) | Intelligent video monitoring method and system based on computer vision analysis technology | |
| CN108022249B (en) | Automatic extraction method for target region of interest of remote sensing video satellite moving vehicle | |
| CN103986906B (en) | Door opening and closing detection method based on monitoring videos | |
| CN108765453B (en) | Expressway agglomerate fog identification method based on video stream data | |
| CA2818579A1 (en) | Calibration device and method for use in a surveillance system for event detection | |
| CN104134222A (en) | Traffic flow monitoring image detecting and tracking system and method based on multi-feature fusion | |
| US9995843B2 (en) | Methods and system for blasting video analysis | |
| CN115620212B (en) | Behavior identification method and system based on monitoring video | |
| CN104866827A (en) | Method for detecting people crossing behavior based on video monitoring platform | |
| CN105915840A (en) | Factory smoke emission automatic monitoring method based on video signal | |
| CN115331264B (en) | Mine worker safety identification management method based on neural network | |
| Lian et al. | A novel method on moving-objects detection based on background subtraction and three frames differencing | |
| CN109948474A (en) | AI thermal imaging all-weather intelligent monitoring method | |
| CN114202646A (en) | Infrared image smoking detection method and system based on deep learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20100616 Termination date: 20121023 |