CN101751744A - Detection and early warning method of smoke - Google Patents
Detection and early warning method of smoke Download PDFInfo
- Publication number
- CN101751744A CN101751744A CN200810239324A CN200810239324A CN101751744A CN 101751744 A CN101751744 A CN 101751744A CN 200810239324 A CN200810239324 A CN 200810239324A CN 200810239324 A CN200810239324 A CN 200810239324A CN 101751744 A CN101751744 A CN 101751744A
- Authority
- CN
- China
- Prior art keywords
- smog
- time
- space
- video image
- module
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 65
- 238000000034 method Methods 0.000 title claims abstract description 47
- 239000000779 smoke Substances 0.000 title claims abstract description 37
- 238000004458 analytical method Methods 0.000 claims abstract description 12
- 238000000605 extraction Methods 0.000 claims abstract description 10
- 238000012549 training Methods 0.000 claims abstract description 8
- 238000003066 decision tree Methods 0.000 claims description 23
- 239000002245 particle Substances 0.000 claims description 22
- 230000008569 process Effects 0.000 claims description 10
- 230000000750 progressive effect Effects 0.000 claims description 6
- 230000014509 gene expression Effects 0.000 claims description 3
- 239000000203 mixture Substances 0.000 claims description 3
- 230000006399 behavior Effects 0.000 claims description 2
- 230000008859 change Effects 0.000 claims description 2
- 238000004090 dissolution Methods 0.000 claims description 2
- 230000005764 inhibitory process Effects 0.000 claims description 2
- 238000004321 preservation Methods 0.000 claims 1
- 238000012545 processing Methods 0.000 abstract description 9
- 230000000007 visual effect Effects 0.000 abstract description 6
- 238000003909 pattern recognition Methods 0.000 abstract description 5
- 238000012544 monitoring process Methods 0.000 abstract description 4
- 238000007619 statistical method Methods 0.000 abstract 1
- 238000005516 engineering process Methods 0.000 description 5
- 239000000443 aerosol Substances 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000002485 combustion reaction Methods 0.000 description 3
- 239000002360 explosive Substances 0.000 description 3
- 230000033001 locomotion Effects 0.000 description 3
- CURLTUGMZLYLDI-UHFFFAOYSA-N Carbon dioxide Chemical compound O=C=O CURLTUGMZLYLDI-UHFFFAOYSA-N 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 239000007789 gas Substances 0.000 description 2
- 238000009434 installation Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 239000003595 mist Substances 0.000 description 2
- 231100000614 poison Toxicity 0.000 description 2
- 230000007096 poisonous effect Effects 0.000 description 2
- 230000002123 temporal effect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- UGFAIRIUMAVXCW-UHFFFAOYSA-N Carbon monoxide Chemical compound [O+]#[C-] UGFAIRIUMAVXCW-UHFFFAOYSA-N 0.000 description 1
- 238000010521 absorption reaction Methods 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 229910002092 carbon dioxide Inorganic materials 0.000 description 1
- 239000001569 carbon dioxide Substances 0.000 description 1
- 229910002091 carbon monoxide Inorganic materials 0.000 description 1
- 238000007596 consolidation process Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000012535 impurity Substances 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000000197 pyrolysis Methods 0.000 description 1
- 238000010223 real-time analysis Methods 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 239000002341 toxic gas Substances 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
- 229910001868 water Inorganic materials 0.000 description 1
Images
Abstract
The invention relates to a detection and early warning method of smoke, which comprises the following steps of: collecting a monitoring video by utilizing the vision of a computer at real time; scanning the video by using a window with variable size according to a rule along an X axis, a Y axis and a time axis T to obtain subimage sequences; carrying out characteristic analysis and extraction on the subimage sequences by utilizing pattern recognition and signal processing and adopting characteristics to reflect spatial distribution, time distribution and frequency characteristics; carrying out pattern classification on each subimage sequence by adopting statistical analysis and determining whether a fire occurs in positions corresponding to the subimage sequence or not; and carrying out region merging and eliminating operation on the detection results of all the subimage sequences to obtain the detection result of the space of the entire visual field and giving out the results of the position, the size, the reliability and the like of fire detection; and when the method is applied, the method is divided into a scene and smoke pattern characteristic training and learning stage and a real-time running detection and warning stage. Different characteristics of the smoke are automatically recognised by a wide-angle detection and interference algorithm, the fire detection is accurately finished at real time, and warning information is sent out in time.
Description
Technical field
The present invention relates to computer vision, signal Processing and area of pattern recognition, particularly based on the Smoke Detection and the method for early warning of computer vision.
Background technology
For a long time, at large space, or have the occasion of high velocity air, especially the incipient fire smoke detection under Outdoor Scene such as stockyard, forest worldwide all is a difficult problem.Because under this class environment, exist many factors that influence detection, mainly comprise: detection mode, spatial altitude, thermal barrier, coverage, gas velocity, explosive/toxic gas, acceptable rate of false alarm, warning information management and remote signal transmission or the like.Traditional detection means has often lost effect in such environment.In this case, because the image-type fire detection technology has the characteristics of contactless detection for detection, be not subjected to the restriction of spatial altitude, thermal boundary, explosive/environmental baseline such as poisonous, make this technology become the effective means of carrying out detection at large spaces such as integrated mill, warehouse, forest and outdoor open space.
Not only has ability based on the Smoke Detection technology of vision in indoor large space and outdoor open space detection of fires, and can reach the early stage effect of protecting of early fire alarming by the early sign of analyzing fire image, promptly just can trigger fire alarm, and important informations such as fire alarm occurrence positions and fire condition are provided according to the early stage image of smog at the smog initial stage that produces.
The generation of smog generally includes combustion product and solid-state high-temperature product:
1. the combustion product principal ingredient is water, carbon monoxide and carbon dioxide.The typical physical of combustion product is a gas characteristic spectrum.
2. solid-state high-temperature product derives from the formed material of combustible pyrolysis under the impurity in the combustible and the condition of high temperature, and particle diameter is at 0.025 micron to 100 microns, and the physical features that shows usually has the scattering and the absorption of pair light.So showing as in the visible light field, smog has certain color, the feature of texture and shape.
Visible smoke detection system (VSD, Visual Smoke Detection) is based on visual pattern type fire detecting system to the fire hazard aerosol fog image analysing computer, only need analyze by the video image that common closed-circuit TV camera (CCTV) is collected, utilize high-performance computer and advanced image processing techniques, wide-angle to survey and special algorithm of interference, automatically discern the different characteristic of smog, finish fire detection fast and accurately, and in time give the alarm and notify the person on duty.
On detectivity, compare with traditional fire detecting system based on the VSD system of smog visual identity and to have significant advantage.The VSD system has realized the utmost point early detection to the fire initial stage, and the direct detection burning things which may cause a fire disaster can detect the trickle smoke particle that human eye be can't see, and can detect the smog of all kinds.The VSD system is present unique outdoor smoke detection solution, be not subjected to the influence of high velocity air motion, and for Conventional detectors, smog can be dispelled, and can't arrive detector at all, says nothing of and finishes detection by high velocity air.The direct benefit that the VSD system brings is exactly to make the operator on duty see the image of accurate position, scene of fire in the very first time, thereby takes corresponding measure fast.The generation of wrong report has been eliminated in advanced smog movement pattern analysis algorithm and visual alarm checking largely, and eliminates the false alarm a great problem in the conventional fire system just.In explosive and poisonous environment, the VSD system can also be by adjusting video camera and lens parameters, just can finish effective detection to interior zone smog at environmental externality, avoids adopting the flame proof detector of costliness.
The characteristics of VSD system maximum can be utilized most of existing Close Circuit Television (CCTV) system exactly, as safety-protection system, traffic control system etc., do not need to increase extra field apparatus and wiring just can be built fire alarm control system easily, saved system cost greatly, reduced the complicacy of system's installation and maintenance.
Relevant VSD system is 99814715.X and 02812437.5 patented technology for example, these smog detection methods based on video mainly be utilize smog movement, edge, block, the features such as frequency of pixel intensity between geometric profile and picture frame, obtain suspicious smog zone, utilize Bayes's criterion or rule-based method of discrimination then, finally determine the smog zone in the image.Its shortcoming is: the product price height, the formation of algorithm is by cumulative rises, the form of the algorithm function module that designs in order to remove a certain special interference source in the video monitoring environment is finished, cause the unordered and complicated of various piece in the algorithm, and system all needs to carry out according to concrete application scenarios the process of a large amount of parameter adjustments before operation.Traditional smoke detector is subjected to the restriction of detection principle, just can finish detection when often needing very near mist source, and can not provide specifying informations such as fire location, size, burning degree, these inadequate natural endowments make them can not be applied to outdoor large space, wait in the fire monitoring system of special occasions at a distance.
Summary of the invention
In order to solve the problem that the prior art fire detection is reported to the police, the objective of the invention is to based on computer vision technique, image processing techniques, pattern-recognition, and provide a kind of mist source that can detect in real time, at a distance, early-stage smog detects and reports to the police reliably, condition of a fire size, burning degree, positional information can be provided, can be applied to the fire hazard monitoring of outdoor large space, special occasions such as remote, the fire hazard aerosol fog based on computer vision that is particularly useful for using in stockyard, forest detects and method for early warning.
In order to reach above-mentioned purpose, the invention provides a kind of Smoke Detection and method for early warning based on computer vision, its Smoke Detection and early warning comprise:
Step 1: scene and smog pattern feature training study stage: at first gather video image sample; And mark contains the positive sample of smog and does not contain the negative sample of smog; From image pattern, extract the feature that is used to reflect space distribution, time distribution, frequency characteristic; By the statistical learning of above feature being determined statistics model and particle smog decision tree;
Step 2: detect and warning stage: gather video image at first in real time and keep the buffer memory video image of some; Video image is carried out pre-service; Utilize statistics model and particle smog decision tree the zone of non-smog in the video image to be rejected by progressive mode, stay suspicious region, by local space rule constrain module, the comprehensive ruling of local time's constraints module stays suspicious smog zone and marks Smoke Detection results area and warning then.
Beneficial effect of the present invention: the present invention adopts the method for statistical study, key algorithm is by using with methods such as global rule constraint, local space rule constrain, local time's constraint, statistics model, particle smog decision tree and characteristic extracting module, each subimage sequence is carried out pattern classification, determine whether this subimage sequence correspondence position has fire to take place.Utilize high-performance computer and advanced image processing techniques, wide-angle to survey and special algorithm of interference, discern the different characteristic of smog automatically, accurate, the real-time fire detection of finishing, and in time send warning message.The present invention can utilize most of existing Close Circuit Television (CCTV) system, as safety-protection system, traffic control system etc., do not need to increase extra field apparatus and wiring just can be built fire alarm control system easily, saved system cost greatly, reduced the complicacy of system's installation and maintenance.
Description of drawings
Fig. 1 is Smoke Detection and the method for early warning overview flow chart that the present invention is based on computer vision;
Fig. 2 is a key algorithm frame construction drawing of the present invention;
Fig. 3 is a decision-making decision tree synoptic diagram of the present invention;
Fig. 4 is that subimage sequence of the present invention is obtained synoptic diagram.
Embodiment
Specific embodiments of the invention are described with reference to the accompanying drawings, are noted that wherein described embodiment only is for illustrative purposes, rather than limit the invention.
The computer vision methods of utilizing of the present invention is gathered in real time to the smog video; Utilize pattern-recognition and method for processing signals that the smog video image that collects is carried out real-time analysis and detection; By the scope of smog, density, rate of propagation etc. are analyzed, judge the situation of point of origin and in time give the alarm, comprise 5 key steps:
1) at first, utilize computer vision methods that monitor video is gathered in real time;
2) with the window of variable size according to certain rule along X-axis, Y-axis, the T axle obtains a series of subimage sequence for the time shaft scan video;
3) utilize pattern-recognition and method for processing signals that the subimage sequence that scanning obtains is carried out signature analysis and extraction, select for use feature can reflect space distribution, time distribution, frequency characteristic;
4) method that adopts counting statistics to analyze is carried out pattern classification to each subimage sequence, determines whether this subimage sequence correspondence position has fire to take place;
5) all subimage sequence testing results are carried out operations such as zone merging, rejecting, obtain the testing result in space, whole visual field, provide the result such as position, size, confidence level of fire detection.
See also Fig. 1, be the Smoke Detection that the present invention is based on computer vision and the algorithm flow chart of method for early warning, embodiments of the invention shown in the figure mainly are made up of following eight parts: wherein 1-4 part is scene and smog pattern feature training study stage, and the real time execution that the several sections of back belongs to system detects and warning stage.
S1. gather sample.At first to set up the fire hazard aerosol fog video database of exterior space scenes such as comprising stockyard, forest.On the basis of improving the fire hazard aerosol fog video database, utilize the associated video image in the database, use sample collection with the statistics model learning.Each frame of video image is divided into rectangular tiles, the big or small minimum of fritter is 4 * 4 to be 32 * 32 to the maximum, will form rectangular block sequence according to rule along X-axis, Y-axis and T time shaft scan video image with the window of variable size like this is subimage sequence, and these subimages have just formed sample space.As shown in Figure 5.
S2. marking sample, is to align sample (sample that contains smog) and negative sample (sample that does not contain smog) marks.
S3. feature extraction, it is each pixel at the rectangular tiles of subimage sequence, ask the gray scale and the tone component of pixel, like this on time shaft, each pixel can form two pixel sequence, each pixel sequence is averaged and variance, as the value of this point, so just obtained four number of sub images, the value of its pixel is respectively the average of gray-scale pixels sequence, the variance of gray-scale pixels sequence, the average of tone pixel sequence, the variance of tone pixel sequence.With our feature structure of structure the rectangular tiles that obtains subimage sequence is carried out the eigenwert that this point was searched for and tried to achieve to feature space.Eigenwert is made up of three parts, is respectively the type code of the position in type code, search characteristics space of institute's use characteristic structure and size, pixel sequence.
S4. training sample, be respectively positive sample that contains smog and the negative sample that does not contain smog to be trained, estimate the similarity degree of measuring between sample image with the divergence conduct, and utilize this similarity to construct a plurality of Weak Classifiers, then Weak Classifier is weighted and promotes, obtain the strong strong classifier group of classification capacity, described strong classifier group comprises: statistics model and particle smog decision tree.
S5. the buffer memory of video image, be to utilize computer vision that monitor video is gathered in real time, when the video image of real-time collection is carried out Smoke Detection, system preserves current image frame and preceding 2-4 two field picture thereof, be the 3-5 frame video image that system cache comprises current image frame, be used for Smoke Detection and warning local time constraints module and cross the continuity method that the time of above continuous buffer memory video image judgement smog upward changes.
S6. the pre-service of video image is that video signal image is strengthened.Mainly comprise and adopt in the selectively outstanding video image of some technological means (comprising algorithm) some unwanted feature in interested feature or inhibition (covering) image, image after the improvement satisfies the needs of some special analysis more than original image, is more suitable in the recognition system of machine vision characteristic and machine.And in whole image enhancement processing process, can not increase the information of original image.
S7. mark the Smoke Detection result, at first object to be analyzed in the video image is decomposed into serial subimage, and these subimages overlap, be the video image that collects in real time by camera system to be carried out the smog analysis by sorter, mark being judged as the zone of containing smog, mark is not done in the zone of not containing smog handled, can on video image, produce the smog zone that much is marked like this.Utilize image processing method that dissolution process is carried out in the smog zone, remove small cavity, form whole smog area marking.
S8. sending a warning, is scale and the time-continuing process of judging the smog zone, provides the information such as position, size, confidence level of fire detection and sends smog alert.After smog no longer takes place, and the smog zone disappears, then stop to send warning.
Smoke detection algorithm frame design of the present invention, adopt global rule constraints module, local space rule constrain module, local time's constraints module, statistics model, particle smog decision tree and characteristic extracting module, utilize high-performance computer and advanced image processing techniques, wide-angle to survey and special algorithm of interference, automatically discern the different characteristic of smog, accurately, the real-time fire detection of finishing, and in time send warning message.
See also accompanying drawing 2, be divided into for the algorithm logic framework shown in the figure: described detection and warning are the algorithm logic frameworks that adopts characteristic extracting module, statistics model, particle smog decision tree, local space rule constrain module, local time's constraints module and global rule constraint; Wherein:
Described detection and warning are the algorithm logic frameworks that adopts characteristic extracting module, statistics model, particle smog decision tree, local space rule constrain module, local time's constraints module and global rule constraint; Wherein:
At the bottom is characteristic extracting module, be used for extracting the characteristic of smog on space, time and frequency from the projection that the gray scale and the tone space of subimage sequence produce, the characteristic of described smog on space, time and frequency is determined according to the statistics that aligns negative sample, and the composition characteristic space.Try to achieve the eigenwert of this sequence by subimage sequence being carried out the feature space search;
Whether statistics model and particle smog decision tree module comprise smog in order to judge video image, or uncertain be smog; This statistics model is connected with characteristic extracting module, be used for receiving and the smog of statistical nature extraction module output in the space, the eigenwert in time and frequecy characteristic space, be used for judging whether video image comprises smog; This particle smog decision tree carries out statistical learning acquisition strong classifier to a large amount of smog and non-smog image space, time and frequency characteristic, by the mode that strong classifier is progressive the image of non-smog is rejected away, and what stay is exactly suspicious video image;
The comprehensive ruling of local space rule constrain module and local time-constrain module stays suspicious smog zone; Local space rule constrain module is connected with the statistics model with particle smog decision tree, considers continuity on the smog space by local space rule constrain module analysis smog zone position and area change spatially;
Local time's constraints module is connected with the statistics model with particle smog decision tree, and the time that whether occurs consideration smog by smog zone in analysis a period of time is continuously gone up the continuity that changes;
The global rule constraints module to the analysis of smog zone expression behaviour on room and time, is got rid of non-smog zone according to the distribution character of smog comprehensive local space rule constrain module and local time-constrain module, mark smog zone, and send warning message.
Among Fig. 2 from lower to upper, the bottom is a feature extraction layer, feature produces projection from gray scale and tone space, feature structure has taken into full account the characteristic of smog on room and time, these characteristics are to establish according to the statistics of sample, all feature base composition characteristic space expressions sample, for computing machine, this is the mode that its observes video sample, and this is equivalent to video sample is mapped as a point on the feature space.
In the Smoke Detection and each step of alarm method based on machine vision of the present invention, at first to set up video database, gather sample.Gather the video data occur the outdoor environment of smog scene in a large number and consolidation form under the outdoor environment of smog scene not occur, each frame in the video is divided into rectangular tiles, will form the rectangular block sequence along time shaft is subimage sequence, and these subimage sequence have just formed sample space such as accompanying drawing 4 is that subimage sequence is obtained synoptic diagram among the present invention.
Among Fig. 4, comprise the transverse axis X and the Z-axis Y of corresponding two field picture 1,2,3, i, n, time shaft T, image in the video sequence sample.Scheme medium and small square frame and partly be the window of variable size.Utilize the window of variable size to obtain a series of subimage sequence along transverse axis X, Z-axis Y, time shaft T scan video in the subimage acquisition process according to certain rule.
Then the subimage sequence sample that obtains is marked, align sample (sample that contains smog) and negative sample (sample that does not contain smog) marks.
In characteristic extraction procedure, each pixel at subimage, ask its gray scale and tone component, like this on time shaft, each pixel can form two pixel sequence, each pixel sequence is averaged and variance, as the value of this point, so just obtained four number of sub images, the value of its pixel is respectively the average of gray-scale pixels sequence, the variance of gray-scale pixels sequence, the average of tone pixel sequence, the variance of tone pixel sequence.
Next want computation of characteristic values.With our feature structure of structure the subimage that obtains in the top step is carried out the eigenwert that this point was searched for and tried to achieve to feature space.Eigenwert is made up of three parts, is respectively the type code of the type code of institute's use characteristic structure, the position of carrying out the feature space search and size, pixel sequence.
Align sample (sample that contains smog) respectively and negative sample (sample that does not contain smog) carries out training study, construct a plurality of Weak Classifiers, then Weak Classifier is made up and promotes, obtain the strong strong classifier group of classification capacity.
In above scene and in the pattern feature training study stage, finished among the present invention feature extraction layer based on the bottom in the Smoke Detection of computer vision and the method for early warning.
Through feature extraction layer, feature produces projection from gray scale and tone space, feature structure has taken into full account the characteristic of smog on room and time, these characteristics are to establish according to the statistics of sample, the feature space that all feature bases support has been expressed sample, for computing machine, this is the mode that it observes video sample, this is equivalent to video sample is mapped as a point on the feature space, the feature structure that we construct as shown in Figure 2, wherein dark zone can obtain this regional time changing value, the poor representation space changing value of peripheral white portion and dark areas.Through after above each step, set up relevant statistics model.
In the real time execution detection and warning stage of Smoke Detection of the present invention and early warning system, real time video image is through after the characteristic extracting module, video image is mapped as a point on the feature space, smog is arranged or do not have smog with definite this sample according to this distributing position on feature space, or it is uncertain, this part work is finished by statistics model and two modules of particle smog decision tree, and the statistics model provides foundation for decision-making.As shown in Figure 3, it is decision-making decision tree synoptic diagram among the present invention, the decision-making decision tree comprises: video sample, strong classifier group 1,2,3...N represent each step of assorting process, behind certain strong classifier 1,2,3...N, be judged as the sample F that mistake is non-smog, after strong classifier N step, remaining sample is considered to detect the suspicious sample in back.We reject away the sample of non-smog by progressive like this mode, and what stay is exactly suspicious sample, for the overall situation, and suspicious region just.By local space rule constrain module, the comprehensive ruling of local time's constraints module stays suspicious smog zone then.Local space rule constrain module mainly is to consider continuity on the smog space, local time's constraints module is mainly considered the continuity of the temporal variation of smog, the global rule constraints module considers that then the distribution character of smog is to get rid of the zone of non-smog, smog is arranged or do not have smog with definite this sample according to this distributing position on feature space, or uncertain.This part work is finished by statistics model and two modules of particle smog decision tree, and the statistics model provides foundation for decision-making.
Utilize progressive mode that the sample of non-smog is rejected away, what stay is exactly suspicious sample.And for the overall situation, suspicious region just.Comprehensive ruling by local space rule constrain module, local time's constraints module stays suspicious smog zone then.Local space rule constrain module mainly is to consider continuity on the smog space, and local time's constraints module is mainly considered the continuity of the temporal variation of smog, and the global rule constraints module considers that then the distribution character of smog is to get rid of the zone of non-smog.Judge the scale and the time-continuing process in smog zone at last and send smog alert.After smog no longer takes place, and the smog zone disappears, then stop to give the alarm.
Describing above is to be used to illustrate of the present invention, and it only is an example, and those of ordinary skills can adopt the implementation of various deformation according to actual conditions, all should belong within protection scope of the present invention.Therefore, scope of the present invention should not described by this and limit.It should be appreciated by those skilled in the art,, all belong to claim of the present invention and come restricted portion in any modification or partial replacement that does not depart from the scope of the present invention.
Claims (9)
1. Smoke Detection and method for early warning, it is characterized in that: the method comprising the steps of:
Step 1: scene and smog pattern feature training study stage: at first gather video image sample; And mark contains the positive sample of smog and does not contain the negative sample of smog; From image pattern, extract the feature that is used to reflect space distribution, time distribution, frequency characteristic; By the statistical learning of above feature being determined statistics model and particle smog decision tree;
Step 2: detect and warning stage: gather video image at first in real time and keep the buffer memory video image of some; Video image is carried out pre-service; Utilize statistics model and particle smog decision tree the zone of non-smog in the video image to be rejected by progressive mode, stay suspicious region, by local space rule constrain module, the comprehensive ruling of local time's constraints module stays suspicious smog zone and marks Smoke Detection results area and warning then.
2. Smoke Detection according to claim 1 and method for early warning, it is characterized in that described detection and warning are the algorithm logic frameworks that adopts characteristic extracting module, statistics model, particle smog decision tree, local space rule constrain module, local time's constraints module and global rule constraint; Wherein:
At the bottom is characteristic extracting module, be used for extracting the characteristic of smog on space, time and frequency from the projection that the gray scale and the tone space of subimage sequence produce, the characteristic of described smog on space, time and frequency is determined according to the statistics that aligns negative sample, and the composition characteristic space.Try to achieve the eigenwert of this sequence by subimage sequence being carried out the feature space search;
Whether statistics model and particle smog decision tree module comprise smog in order to judge video image, or uncertain be smog; This statistics model is connected with characteristic extracting module, be used for receiving and the smog of statistical nature extraction module output in the space, the eigenwert in time and frequecy characteristic space, be used for judging whether video image comprises smog; This particle smog decision tree carries out statistical learning acquisition strong classifier to a large amount of smog and non-smog image space, time and frequency characteristic, by the mode that strong classifier is progressive the image of non-smog is rejected away, and what stay is exactly suspicious video image;
The comprehensive ruling of local space rule constrain module and local time-constrain module stays suspicious smog zone; Local space rule constrain module is connected with the statistics model with particle smog decision tree, considers continuity on the smog space by local space rule constrain module analysis smog zone position and area change spatially;
Local time's constraints module is connected with the statistics model with particle smog decision tree, and the time that whether occurs consideration smog by smog zone in analysis a period of time is continuously gone up the continuity that changes;
The global rule constraints module to the analysis of smog zone expression behaviour on room and time, is got rid of non-smog zone according to the distribution character of smog comprehensive local space rule constrain module and local time-constrain module, mark smog zone, and send warning message.
3. Smoke Detection according to claim 1 and method for early warning, it is characterized in that, gathering sample is that each two field picture in the video image is divided into rectangular tiles, window with variable size obtains serial subimage sequence according to rule along X-axis, Y-axis and T time shaft scan video image, and these subimage sequence form sample space.
4. Smoke Detection according to claim 1 and method for early warning, it is characterized in that, described feature extraction is each pixel at subimage sequence, ask the gray scale and the tone component of pixel, like this on time shaft, each pixel can form two pixel sequence, each pixel sequence is averaged and variance, value as this point, obtain four rectangular tiles, the value of its pixel is respectively the average of gray-scale pixels sequence, the variance of gray-scale pixels sequence, the average of tone pixel sequence, the variance of tone pixel sequence.
5. Smoke Detection according to claim 1 and method for early warning, it is characterized in that, the described statistical learning training sample that utilizes, be respectively positive sample that contains smog and the negative sample that does not contain smog to be trained, estimate the similarity degree of measuring between sample image with the divergence conduct, and utilize this similarity structure Weak Classifier, and at last Weak Classifier is weighted and promotes, obtain the strong strong classifier group of classification capacity.
6. according to the Smoke Detection and the method for early warning of claim 1, it is characterized in that, described buffer memory video image is to utilize computer vision that monitor video is gathered in real time, and preservation current image frame and preceding 2-4 two field picture thereof, buffer memory comprises the 3-5 frame video image of current image frame, is used for Smoke Detection and warning local time constraints module and goes up the continuity that changes by the time of above continuous buffer memory video image judgement smog.
7. according to the Smoke Detection and the method for early warning of claim 1, it is characterized in that the pre-service of described video image is that vision signal is strengthened, unwanted feature in the image is covered in interested feature or inhibition in the selectively outstanding video image.
8. according to the Smoke Detection and the method for early warning of claim 1, it is characterized in that, described mark Smoke Detection result adopts subimage, at first object to be analyzed in the video image is decomposed into serial subimage, and these subimages overlap, and by sorter subimage are carried out the smog analysis then, mark being judged as the zone of containing smog, mark is not done in the zone of not containing smog handled, can on video image, produce the smog zone that much is marked like this; Utilize Flame Image Process that dissolution process is carried out in the smog zone, remove small cavity, form whole smog area marking.
9. according to the Smoke Detection and the method for early warning of claim 1, it is characterized in that described warning is scale and a time-continuing process of judging the smog zone, provide the information such as position, size, confidence level of fire detection and send smog alert; After smog no longer takes place, and the smog zone disappears, then stop to send warning.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008102393245A CN101751744B (en) | 2008-12-10 | 2008-12-10 | Detection and early warning method of smoke |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008102393245A CN101751744B (en) | 2008-12-10 | 2008-12-10 | Detection and early warning method of smoke |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101751744A true CN101751744A (en) | 2010-06-23 |
| CN101751744B CN101751744B (en) | 2011-08-31 |
Family
ID=42478673
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2008102393245A Expired - Fee Related CN101751744B (en) | 2008-12-10 | 2008-12-10 | Detection and early warning method of smoke |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101751744B (en) |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102121906A (en) * | 2011-01-31 | 2011-07-13 | 无锡中星微电子有限公司 | Smoke detection system as well as intelligent control system and intelligent control method of range hood |
| CN102163360A (en) * | 2011-03-24 | 2011-08-24 | 杭州海康威视系统技术有限公司 | Tunnel smoke video detecting method and device |
| CN102663350A (en) * | 2012-03-23 | 2012-09-12 | 长安大学 | Road tunnel fire detection method based on video |
| CN103248878A (en) * | 2013-05-23 | 2013-08-14 | 南车株洲电力机车有限公司 | Pattern recognition method, device and system of abnormal situation of fully mechanized coal mining face |
| CN103903008A (en) * | 2014-03-26 | 2014-07-02 | 国家电网公司 | Power transmission line fog level recognition method and system based on images |
| CN103983574A (en) * | 2014-06-03 | 2014-08-13 | 上海安维尔信息科技有限公司 | Smoke detection method |
| CN104715559A (en) * | 2015-03-06 | 2015-06-17 | 温州大学 | Smoke detection and fire alarm method based on trajectory identification |
| CN106339657A (en) * | 2015-07-09 | 2017-01-18 | 张�杰 | Straw incineration monitoring method and device based on monitoring video |
| CN106462737A (en) * | 2014-05-20 | 2017-02-22 | 高通股份有限公司 | Systems and methods for haziness detection |
| CN106781210A (en) * | 2016-12-23 | 2017-05-31 | 安徽工程大学机电学院 | The recognition methods of cigarette in fire based on spatial closure |
| CN108363992A (en) * | 2018-03-15 | 2018-08-03 | 南京邮电大学 | A kind of fire behavior method for early warning monitoring video image smog based on machine learning |
| CN109165575A (en) * | 2018-08-06 | 2019-01-08 | 天津艾思科尔科技有限公司 | A kind of pyrotechnics recognizer based on SSD frame |
| CN109544852A (en) * | 2018-12-27 | 2019-03-29 | 秒针信息技术有限公司 | The monitoring method and device of dining room fire |
| CN109946209A (en) * | 2019-03-13 | 2019-06-28 | 北京英视睿达科技有限公司 | A kind of method and device of atmosphere pollution alarm |
| CN111127814A (en) * | 2019-12-19 | 2020-05-08 | 浙江大华技术股份有限公司 | Fire alarm identification method and related device |
| CN111814689A (en) * | 2020-07-09 | 2020-10-23 | 浙江大华技术股份有限公司 | Fire recognition network model training method, fire recognition method and related equipment |
| CN112507925A (en) * | 2020-12-16 | 2021-03-16 | 安徽建筑大学 | Fire detection method based on slow characteristic analysis |
| CN112614165A (en) * | 2020-12-04 | 2021-04-06 | 浙江大华技术股份有限公司 | Smoke and fire monitoring method and device, camera, electronic device and storage medium |
| CN112634577A (en) * | 2019-09-24 | 2021-04-09 | 中科智云科技有限公司 | Method and apparatus for smoke alarm |
| CN113139500A (en) * | 2021-05-10 | 2021-07-20 | 重庆中科云从科技有限公司 | Smoke detection method, system, medium and device |
| CN113505758A (en) * | 2021-09-07 | 2021-10-15 | 广州汇图计算机信息技术有限公司 | Smog early warning system based on visual detection |
| CN114973110A (en) * | 2022-07-27 | 2022-08-30 | 四川九通智路科技有限公司 | On-line monitoring method and system for highway weather |
| CN115086614A (en) * | 2022-06-15 | 2022-09-20 | 广州中辉科技有限公司 | Fire control monitoring camera device with smog alarming function |
| CN116486307A (en) * | 2023-05-05 | 2023-07-25 | 江苏三棱智慧物联发展股份有限公司 | Chemical industry park safety early warning method and system based on video analysis technology |
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 |
| WO2003001187A1 (en) * | 2001-06-22 | 2003-01-03 | Siemens Building Technologies Ag | Video smoke detection system and method for examining the same |
-
2008
- 2008-12-10 CN CN2008102393245A patent/CN101751744B/en not_active Expired - Fee Related
Cited By (36)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102121906A (en) * | 2011-01-31 | 2011-07-13 | 无锡中星微电子有限公司 | Smoke detection system as well as intelligent control system and intelligent control method of range hood |
| CN102163360A (en) * | 2011-03-24 | 2011-08-24 | 杭州海康威视系统技术有限公司 | Tunnel smoke video detecting method and device |
| CN102163360B (en) * | 2011-03-24 | 2013-07-31 | 杭州海康威视系统技术有限公司 | Tunnel smoke video detecting method and device |
| CN102663350A (en) * | 2012-03-23 | 2012-09-12 | 长安大学 | Road tunnel fire detection method based on video |
| CN103248878A (en) * | 2013-05-23 | 2013-08-14 | 南车株洲电力机车有限公司 | Pattern recognition method, device and system of abnormal situation of fully mechanized coal mining face |
| CN103248878B (en) * | 2013-05-23 | 2016-03-09 | 南车株洲电力机车有限公司 | A kind of mode identification method, equipment and system of fully-mechanized mining working unusual condition |
| CN103903008A (en) * | 2014-03-26 | 2014-07-02 | 国家电网公司 | Power transmission line fog level recognition method and system based on images |
| CN103903008B (en) * | 2014-03-26 | 2018-01-19 | 国家电网公司 | A kind of method and system of the mist grade based on image recognition transmission line of electricity |
| CN106462737A (en) * | 2014-05-20 | 2017-02-22 | 高通股份有限公司 | Systems and methods for haziness detection |
| CN103983574A (en) * | 2014-06-03 | 2014-08-13 | 上海安维尔信息科技有限公司 | Smoke detection method |
| CN104715559A (en) * | 2015-03-06 | 2015-06-17 | 温州大学 | Smoke detection and fire alarm method based on trajectory identification |
| CN106339657A (en) * | 2015-07-09 | 2017-01-18 | 张�杰 | Straw incineration monitoring method and device based on monitoring video |
| CN106339657B (en) * | 2015-07-09 | 2019-02-22 | 张�杰 | Crop straw burning monitoring method based on monitor video, device |
| CN106781210A (en) * | 2016-12-23 | 2017-05-31 | 安徽工程大学机电学院 | The recognition methods of cigarette in fire based on spatial closure |
| CN106781210B (en) * | 2016-12-23 | 2019-05-10 | 安徽信息工程学院 | The recognition methods of cigarette in fire based on spatial closure |
| CN108363992B (en) * | 2018-03-15 | 2021-12-14 | 南京钜力智能制造技术研究院有限公司 | Fire early warning method for monitoring video image smoke based on machine learning |
| CN108363992A (en) * | 2018-03-15 | 2018-08-03 | 南京邮电大学 | A kind of fire behavior method for early warning monitoring video image smog based on machine learning |
| CN109165575A (en) * | 2018-08-06 | 2019-01-08 | 天津艾思科尔科技有限公司 | A kind of pyrotechnics recognizer based on SSD frame |
| CN109165575B (en) * | 2018-08-06 | 2024-02-20 | 天津艾思科尔科技有限公司 | Pyrotechnic recognition algorithm based on SSD frame |
| CN109544852A (en) * | 2018-12-27 | 2019-03-29 | 秒针信息技术有限公司 | The monitoring method and device of dining room fire |
| CN109544852B (en) * | 2018-12-27 | 2021-05-25 | 秒针信息技术有限公司 | Restaurant fire monitoring method and device |
| CN109946209A (en) * | 2019-03-13 | 2019-06-28 | 北京英视睿达科技有限公司 | A kind of method and device of atmosphere pollution alarm |
| CN112634577A (en) * | 2019-09-24 | 2021-04-09 | 中科智云科技有限公司 | Method and apparatus for smoke alarm |
| CN111127814A (en) * | 2019-12-19 | 2020-05-08 | 浙江大华技术股份有限公司 | Fire alarm identification method and related device |
| CN111814689A (en) * | 2020-07-09 | 2020-10-23 | 浙江大华技术股份有限公司 | Fire recognition network model training method, fire recognition method and related equipment |
| CN112614165A (en) * | 2020-12-04 | 2021-04-06 | 浙江大华技术股份有限公司 | Smoke and fire monitoring method and device, camera, electronic device and storage medium |
| CN112614165B (en) * | 2020-12-04 | 2024-03-15 | 浙江大华技术股份有限公司 | Firework monitoring method, device, camera, electronic device and storage medium |
| CN112507925A (en) * | 2020-12-16 | 2021-03-16 | 安徽建筑大学 | Fire detection method based on slow characteristic analysis |
| CN113139500A (en) * | 2021-05-10 | 2021-07-20 | 重庆中科云从科技有限公司 | Smoke detection method, system, medium and device |
| CN113139500B (en) * | 2021-05-10 | 2023-10-20 | 重庆中科云从科技有限公司 | Smoke detection method, system, medium and equipment |
| CN113505758A (en) * | 2021-09-07 | 2021-10-15 | 广州汇图计算机信息技术有限公司 | Smog early warning system based on visual detection |
| CN115086614A (en) * | 2022-06-15 | 2022-09-20 | 广州中辉科技有限公司 | Fire control monitoring camera device with smog alarming function |
| CN114973110B (en) * | 2022-07-27 | 2022-11-01 | 四川九通智路科技有限公司 | On-line monitoring method and system for highway weather |
| CN114973110A (en) * | 2022-07-27 | 2022-08-30 | 四川九通智路科技有限公司 | On-line monitoring method and system for highway weather |
| CN116486307A (en) * | 2023-05-05 | 2023-07-25 | 江苏三棱智慧物联发展股份有限公司 | Chemical industry park safety early warning method and system based on video analysis technology |
| CN116486307B (en) * | 2023-05-05 | 2023-11-21 | 江苏三棱智慧物联发展股份有限公司 | Chemical industry park safety early warning method and system based on video analysis technology |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101751744B (en) | 2011-08-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101751744B (en) | Detection and early warning method of smoke | |
| CN105809679B (en) | Mountain railway side slope rockfall detection method based on visual analysis | |
| CN110738127B (en) | Helmet identification method based on unsupervised deep learning neural network algorithm | |
| CN101334924B (en) | Fire hazard probe system and its fire hazard detection method | |
| CN101465033B (en) | Automatic tracking recognition system and method | |
| CN111462451B (en) | Straw burning detection alarm system based on video information | |
| CN108319926A (en) | A kind of the safety cap wearing detecting system and detection method of building-site | |
| CN201278180Y (en) | Automatic tracking recognition system | |
| CN106128022B (en) | A kind of wisdom gold eyeball identification violent action alarm method | |
| CN104971458B (en) | Based on the more fire source recognition methods for automatically tracking positioning jet stream extinguishing device | |
| CN106128053A (en) | A kind of wisdom gold eyeball identification personnel stay hover alarm method and device | |
| CN111222478A (en) | Construction site safety protection detection method and system | |
| CN106210634A (en) | A kind of wisdom gold eyeball identification personnel fall down to the ground alarm method and device | |
| CN105976365A (en) | Nocturnal fire disaster video detection method | |
| CN108038424A (en) | A kind of vision automated detection method suitable for working at height | |
| CN109911550A (en) | Scratch board conveyor protective device based on infrared thermal imaging and visible light video analysis | |
| CN112183472A (en) | Method for detecting whether test field personnel wear work clothes or not based on improved RetinaNet | |
| CN106127814A (en) | A kind of wisdom gold eyeball identification gathering of people is fought alarm method and device | |
| CN112184773A (en) | Helmet wearing detection method and system based on deep learning | |
| CN201091014Y (en) | Fire detecting device | |
| CN109410497A (en) | A kind of monitoring of bridge opening space safety and alarm system based on deep learning | |
| CN115083088A (en) | Railway perimeter intrusion early warning method | |
| CN116846059A (en) | Edge detection system for power grid inspection and monitoring | |
| KR101196678B1 (en) | Real-time fire detection device and method | |
| CN111553305A (en) | Violation video identification system and method |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20110831 Termination date: 20171210 |