CN108108695A - Fire defector recognition methods based on Infrared video image - Google Patents
Fire defector recognition methods based on Infrared video image Download PDFInfo
- Publication number
- CN108108695A CN108108695A CN201711401582.4A CN201711401582A CN108108695A CN 108108695 A CN108108695 A CN 108108695A CN 201711401582 A CN201711401582 A CN 201711401582A CN 108108695 A CN108108695 A CN 108108695A
- Authority
- CN
- China
- Prior art keywords
- msub
- mrow
- target
- infrared light
- light image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Transforming Light Signals Into Electric Signals (AREA)
- Fire-Detection Mechanisms (AREA)
- Closed-Circuit Television Systems (AREA)
Abstract
The invention discloses a kind of fire defector recognition methods based on Infrared video image, this method is first with infrared camera scan Infrared video image, former frame and this frame infrared light image are obtained, then extracts the doubtful flame region of adjacent two field pictures respectively using luminance threshold side；Object matching is carried out to the doubtful flame region of adjacent two frame again；Objective degrees of confidence and the correlation of adjacent two frames same target that target deletes judgement and utilizes setting are carried out to object matching result again and carries out flame object identification.This method can effectively and rapidly identify flame object, also can guarantee the identification to flame object under night-environment, can reduce the consumption in algorithm operational process to calculator memory in identification process to the dynamic deletion of target memory.
Description
Technical field
The present invention relates to technical field of computer vision, particularly a kind of fire defector identification based on Infrared video image
Method.
Background technology
In recent years, Large Space Building Fires, mine fire, forest fire, tunnel fire hazard etc. frequently occur not only to make the mankind
Life and property by massive losses, great destruction is also caused to human ecological environment.With the continuous progress of science, people
Start to be conceived to the research of intelligent video analysis to take precautions against fire, based on image/video analysis fire defector and recognition methods
Quick popularization is obtained.
Fire defector and recognition methods based on image/video analysis normally comprise extraction and the fire of doubtful flame region
Flame identifies.The extraction of doubtful flame region is the premise of flame identification, and it is that fire is visited that flame image is separated from background
The basis of survey is related to the accuracy of the reliability and target identification of subsequent characteristics extraction.Using the difference based on background model
The doubtful conflagration area extraction of method, the algorithm is by realizing effective detection to flame, calmodulin binding domain CaM cluster based on background model
The algorithm of growth, finally realizes extracted region.These methods assume that initial background does not include the training sequence of moving target,
Limit the use condition of background model.
Also some algorithms often carry out interference source exclusion with reference to the motion detection of flame color model, and sensitivity is by Image Acquisition
The limitation of equipment quality and motion detection algorithm quality, and the flame color model of general visible images is just for partially red, partially yellow
Flame color design, limitation is larger.
Fire defector based on video image can utilize the features such as movement, color, the time-frequency of flame to realize flame identification.Its
In, merely with the flame identification method of the static natures such as color, be easily subject to it is similar disturbed with flame color scenery, affect and be
The robustness of system.Phollips et al. carries out flame knowledge using the grey level histogram intensity of flame and the time change of consecutive frame
Not, relatively good detection environment (the nonflame light interference of less movement) is also needed, and its experimental data is oneself
What subjectivity determined, if input data changes, effect can have a greatly reduced quality.Yamagishi et al. proposes a kind of based on god
Flame detecting method through network, the calculation amount of algorithm are bigger.Meanwhile some nights are affected, and can not be ensured complete
Weather monitors.
The content of the invention
To overcome the deficiencies in the prior art, the present invention provides a kind of fire defector based on Infrared video image and knows
Other method, this method is by extracting the object matching of doubtful flame region and the adjacent doubtful flame region of two frames, further according to setting
Fixed objective degrees of confidence and the correlation in adjacent two frames same target region carry out flame object, can effectively and rapidly identify
Flame object can reduce in algorithm operational process the dynamic deletion of target memory in identification process and disappear to calculator memory
Consumption, while make use of the physical characteristic of infrared light image so that flame object also can be successfully identified under night-environment.
To realize above-mentioned technical purpose, the present invention adopts the following technical scheme that：
A kind of fire defector recognition methods based on Infrared video image, comprises the following steps：
S1 utilizes infrared camera scan Infrared video image, obtains former frame and this frame infrared light image.In order to timely
Flame object is identified, it is necessary to gather image while handling image according to the methods below.
S2 extracts the doubtful flame region in former frame and this frame infrared light image respectively；
S3 carries out object matching to the doubtful flame region of former frame and this frame infrared light image；
The deletion judgement of each target of the doubtful flame region of S4 this frame infrared light images and flame object identification；
S5 inputs the infrared light image of new next frame, continues with step S2-S4 to complete the flame object of a new frame
Identification.
Split in S2 of the present invention by carrying out the image based on brightness of image to the infrared light image gathered in S1, so as to carry
Take out doubtful flame region.Specifically include following steps：
S2.1 assumes that infrared light image maximum gradation value is Max, then choosesMax is the seed point of region growing.
The average gray value of pixel in neighborhood centered on S2.2 solution seed points, by each picture in seed neighborhood of a point
The absolute value of the difference of the gray value of vegetarian refreshments and this average gray value is compared with the threshold value T set.
If S2.3 is less than the threshold value T of setting, which is merged with seed point, and as new seed point；If
More than or equal to the threshold value T of setting, then give up；Continue to judge next pixel.
S2.4 is repeated in step S2.2 and S2.3, meets step S2.2 and step S2.3 until all in infrared light image
The pixel of middle requirement is all merged into some region, and this region is exactly doubtful flame region.
Doubtful flame region in former frame and this frame infrared light image is extracted using the method in S2.1 to S2.4 respectively.
S3 of the present invention comprises the following steps：
S3.1 marks each target of the doubtful flame region of adjacent two field pictures by 8 neighborhood connected component labeling methods respectively
And obtain the boundary rectangle of each target.
S3.2 records the upper left point A of the boundary rectangle of each target in adjacent two frames infrared light image1, lower-left point A2, upper right
Point A3, lower-right most point A4And the central point O of boundary rectangle.
If the point A of the boundary rectangle of some target in S3.3 this frame infrared light images1, point A2, point A3, point A4And point O
This five points in, there are some point or certain some target of several points in former frame infrared light image boundary rectangle in,
With regard to initial decision, the two targets of this adjacent two frames infrared light image are same target.
If the target in S3.4 this frame infrared light images exists with more than two targets in former frame infrared light image
There are multiple destination matches, then need to carry out secondary in the situation in step S3.3, i.e., adjacent two frames infrared light image
Matching；The principle of Secondary Match is：It is included according to above-mentioned five points of the boundary rectangle of the target in this frame infrared light image
The center of the boundary rectangle of number and two targets in the boundary rectangle of another target in former frame infrared light image
The distance of point judged, i.e., the number comprising point at most and the distance between two central points recently, then illustrate the two mesh
It is designated as same target.Here, the distance of two central points uses Euclidean distance, if O1(x1,y1) and O2(x2,y2) it is respectively two targets
The central point of boundary rectangle, then the Euclidean distance computational methods of two central points be：
D=sqrt ((x1-x2)2+(y1-y2)2) (1)
If the point A of the boundary rectangle of either objective in S3.5 this frame infrared light images1, point A2, point A3, point A4And point O
This five points in, there is no in the boundary rectangle of either objective in former frame infrared light image, i.e. this frame infrared light
Each target of doubtful flame region all mismatches in each target of doubtful flame region and former frame infrared light image in image, says
This bright target is emerging target, then records the position of the target, gives goal-setting existence confidence level believe, initially
30 and setting flame confidence level are turned to, is initialized as 0, doubtful flame object number adds 1.
S4 of the present invention comprises the following steps：
S4.1 is obtained the same target of former frame infrared light image and this frame infrared light image by the object matching of step S3,
If the target has been labeled as flame object before former frame infrared light image or former frame infrared light image, to the mesh
Mark carries out delete operation.
S4.1.1 is obtained the identical mesh of former frame infrared light image and this frame infrared light image by the object matching of step S3
Mark, for having the target area of same target in former frame infrared light image and this frame infrared light image, if the target exists
It has been determined as flame object before former frame infrared light image or former frame infrared light image, has judged this frame infrared light image
In also whether there are bright areas for the target area；Wherein bright areas refers to that the gray value in this region there are pixel is big
In or be equal toMax, then the region is bright areas.
If the target area for having same target in the S4.1.2 former frame infrared light images and this frame infrared light image is deposited
In bright areas, then the existence confidence level believe of the target area adds 2, then judges whether its confidence level believe that survives is big
In 200 or less than 0, if so, deleting this target, the memory of the target is discharged；Otherwise, continue to retain this target.
If there is the target area of same target not in the S4.1.3 former frame infrared light images and this frame infrared light image
There are bright areas, then the existence confidence level believe of the target area subtracts 2, then whether judges its existence confidence level believe
More than 200 or less than 0, if so, deleting this target, the memory of the target is discharged；Otherwise continue to retain this target.
S4.2 is obtained the same target of former frame infrared light image and this frame infrared light image by the object matching of step S3,
If the target is not denoted as flame object before former frame infrared light image and former frame infrared light image, to its into
Row flame object identifies.
S4.2.1 is obtained the identical mesh of former frame infrared light image and this frame infrared light image by the object matching of step S3
Mark, if being not flagged as flame object and the mesh before the same target of former frame infrared light image and this frame infrared light image
The fresh target that the not doubtful flame region for being designated as judging in step S3 occurs, according to the target in this frame infrared light image
The doubtful flame information of location updating, and calculate same target in the doubtful flame region of two adjacent two frames infrared light images
Correlation, the correlation calculations formula of same target are：
In above formula, X1(n1,n2) for certain point (n of some target in former frame infrared light image1,n2) pixel value, X2
(n1,n2) be this frame infrared light image identical point pixel value；M1And M2For the horizontal and vertical pixel of single frames infrared light image
Points；n1And n2For the coordinate that certain in image is put, m1And m2For offset.
The influence of noise may make two frames false matches phenomenon occur, in order to eliminate this influence, to C (m1,m2) make normalizing
Change is handled, and is positioned as after normalization
Wherein,WithRepresent that the same target region of former frame infrared light image and this frame infrared light image is (i.e. previous
There is the target area of same target in frame infrared light image and this frame infrared light image) average pixel value intensity, i.e.,：
If S4.2.2 related coefficientsMore than the threshold value T of setting, then this target existence confidence level believe subtracts
1 and flame confidence level score subtracts 1；Whether the existence confidence level believe of target is more than 200 or less than 0, if then deleting
Except this target, the memory of the target is discharged.
If S4.2.3 related coefficientsLess than the threshold value T of setting, then confidence level of surviving score adds 1,
The value of believe adds 1, and updates flame information, judges whether score at this time is more than 10, if so, judging that this target is
Flame object；If score is not more than 10, judge whether the existence confidence level believe of target is more than 200 or less than 0, if
It is to delete this target, discharges the memory of the target.
Compared with prior art, the invention has the advantages that：
1) present invention respectively carries out adjacent two frames infrared image processing using luminance threshold method and obtains two images each
Doubtful flame region, each connected domain in doubtful flame region is marked by connected component labeling method, then passes through phase
The doubtful flame region of adjacent two frames carries out object matching, the successful match to adjacent two frame and have been marked as the target of flame into
Row, which is deleted, to be judged, i.e., when whether target existence confidence level believe is more than 200 or less than 0, if then deleting this target,
The memory of the target is discharged, can so reduce the occupancy to calculator memory of algorithm in the process of running.
2) present invention is during the doubtful flame region of adjacent two frame carries out object matching, it is contemplated that multiple targets
Similar situation has carried out Secondary Match and has obtained two most preferably close targets.
3) present invention judges that this method passes through red using the correlation and confidence level memory flame of adjacent two frames target area
Outer video image can effectively and rapidly identify flame object, make use of the physical characteristic of infrared image so that in night ring
Also flame object, while dynamic delete target memory during flame identification can be successfully identified under border, algorithm can be reduced
Consumption of the operational process to calculator memory.
Description of the drawings
Fig. 1 is the flow chart of the present invention；
Fig. 2 is the object matching flow chart of the doubtful flame region of adjacent two frame of the embodiment of the present invention；
Fig. 3 is that the target of the embodiment of the present invention deletes decision flow chart；
Fig. 4 is the flame object identification process figure of the embodiment of the present invention.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with attached drawing to embodiment party of the present invention
Formula is described in further detail.
Referring to Fig. 1, for the present invention is based on the flow chart of the fire defector recognition methods of Infrared video image, the present invention includes
Following steps：
A kind of fire defector recognition methods based on Infrared video image, comprises the following steps：
S1 utilizes infrared camera scan Infrared video image, obtains former frame and this frame infrared light image.In order to timely
Flame object is identified, it is necessary to gather image while handling image according to the methods below.
S2 extracts the doubtful flame region in former frame and this frame infrared light image respectively；
Infrared radiation is a kind of most commonly used electromagenetic wave radiation existing for nature, it is normal based on any object
Can all generate molecule and the random movement of atom of itself under rule environment, not stop eradiation and go out thermal infrared energy, molecule and
Atomic motion is more violent, and the energy of radiation is bigger, otherwise the energy of radiation is smaller.Thermal camera is according to object emission or reflection
Infrared imaging, the flame of different flame colors is respectively provided with higher brightness in Infrared video image.The process that flame occurs is adjoint
It is luminous, fever the phenomenon that, the temperature of flame generation area will be apparently higher than the temperature of ambient enviroment.
In view of the Luminance Distribution of its thermography is directly proportional to temperature height, the present invention passes through the infrared light to collecting
Image carries out the image segmentation based on brightness of image, so as to tentatively extract high temp objects region, high temp objects region, that is, doubtful
Flame region.It is as follows：
S2.1 assumes that infrared light image maximum gradation value is Max, then choosesMax is the seed point of region growing；
The average gray value of neighborhood (the window neighborhood as being 3 × 3) interior pixel centered on S2.2 solution seed points, will plant
The gray value of each pixel and the absolute value of the difference of this average gray value in sub- neighborhood of a point are compared with the threshold value T set
Compared with；
If S2.3 is less than the threshold value T of setting, which is merged with seed point, and as new seed point；If
More than or equal to the threshold value T of setting, then give up；Continue to judge next pixel.
S2.4 is repeated in step S2.2 and S2.3 until all in infrared light image meet the requirements (meets step S2.2
Required in step S2.3) pixel be all merged into some region, and this region is exactly doubtful flame region.
Doubtful flame region in former frame and this frame infrared light image is extracted using the method in S2.1 to S2.4 respectively.
S3 carries out object matching to the doubtful flame region of former frame and this frame infrared light image；It is adjacent two with reference to Fig. 2
The object matching flow chart of the doubtful flame region of frame infrared light image, comprises the following steps：
S3.1 marks each target of the doubtful flame region of adjacent two field pictures by 8 neighborhood connected component labeling methods respectively
And obtain the boundary rectangle of each target；
S3.2 records the upper left point A of the boundary rectangle of each target in adjacent two frames infrared light image1, lower-left point A2, upper right
Point A3, lower-right most point A4And the central point O of boundary rectangle；
If the point A of the boundary rectangle of some target in S3.3 this frame infrared light images1, point A2, point A3, point A4And point O
This five points in, there are some point or certain some target of several points in former frame infrared light image boundary rectangle in,
Can the two targets of initial decision this adjacent two frames infrared light image be same target；
If the target in S3.4 this frame infrared light images exists with more than two targets in former frame infrared light image
There are multiple destination matches, then need to carry out secondary in the situation in step S3.3, i.e., adjacent two frames infrared light image
Matching.The thought of Secondary Match is：According to above-mentioned five points of the boundary rectangle of the target in this frame infrared light image (target
The upper left point A of boundary rectangle1, lower-left point A2, upper right point A3, lower-right most point A4And the central point O of boundary rectangle) included in previous
The central point of the boundary rectangle of number and two targets in the boundary rectangle of another target in frame infrared light image
Distance judged, i.e., the number comprising point is more and the distance between central point is nearer, then it is phase to illustrate the two targets
Same target.
Calculate at the distance between 2 points using Euclidean distance.Assuming that O1(x1,y1) and O2(x2,y2) it is respectively that two targets are external
The central point of rectangle, then Euclidean distance be：
D=sqrt ((x1-x2)2+(y1-y2)2) (1)
If some target of S3.5 former frame infrared light images is according to each mesh in method in step S3.3 and former frame
Mark matched, there is no it is occurring in step S3.3 as a result, i.e. in this frame doubtful flame region each target and former frame
In each target of doubtful flame region all mismatch, illustrate that this target for emerging target, then records the position of the target, gives
Goal-setting existence confidence level believe is initialized as 30 and setting flame confidence level, is initialized as 0, doubtful flame
Target numbers add 1.
The deletion judgement of each target of the doubtful flame region of S4 this frame infrared light images and flame object identification；
S4.1 is obtained the same target of former frame infrared light image and this frame infrared light image by the object matching of step S3,
If the target (i.e. the same target of former frame infrared light image and this frame infrared light image) in former frame infrared light image or
Flame object is had been labeled as before former frame infrared light image, then delete operation is carried out to the target；Fig. 3 sentences for target deletion
Determine operational flowchart.
S4.1.1 is obtained the identical mesh of former frame infrared light image and this frame infrared light image by the object matching of step S3
Mark, for having the target area of same target in former frame infrared light image and this frame infrared light image, if the target exists
It has been determined as flame object before former frame infrared light image or former frame infrared light image, has judged this frame infrared light image
In also whether there are bright areas for the target area；Wherein bright areas refers to that the gray value in this region there are pixel is big
In or be equal toMax, then the region is bright areas；
If the target area for having same target in the S4.1.2 former frame infrared light images and this frame infrared light image is deposited
In bright areas, then the existence confidence level believe of the target area adds 2, then judges whether its confidence level believe that survives is big
In 200 or less than 0, if so, deleting this target, the memory of the target is discharged；Otherwise, continue to retain this target.
If there is the target area of same target not in the S4.1.3 former frame infrared light images and this frame infrared light image
There are bright areas, then the existence confidence level believe of the target area subtracts 2, then whether judges its existence confidence level believe
More than 200 or less than 0, if so, deleting this target, the memory of the target is discharged；Otherwise continue to retain this target.
S4.2 is obtained the same target of former frame infrared light image and this frame infrared light image by the object matching of step S3,
If the target is not denoted as flame object before former frame infrared light image and former frame infrared light image, to its into
Row flame object identifies；
Judge with the presence or absence of flame object in doubtful flame region, so as to carry out alarm response.Due to flame so by when
The volume of gas plume inhales the influence of characteristic and air flow, and burned flame shows ceaselessly oscillating characteristic, and this vibration is special
Property can realize flame fire identification using the correlation of image.Fig. 4 is flame object identification process figure, present invention employs
The flame determination method of correlation and confidence level based on target is carried, realizes that process is as follows：
S4.2.1 is obtained the identical mesh of former frame infrared light image and this frame infrared light image by the object matching of step S3
Mark, if being not flagged as flame object and the mesh before the same target of former frame infrared light image and this frame infrared light image
The fresh target that the not doubtful flame region for being designated as judging in step S3 occurs, according to the target in this frame infrared light image
The doubtful flame information of location updating, and calculate same target in the doubtful flame region of two adjacent two frames infrared light images
Correlation, the correlation calculations formula of same target are：
In above formula, X1(n1,n2) for certain point (n of some target in former frame infrared light image1,n2) pixel value, X2
(n1,n2) be this frame infrared light image identical point pixel value；M1And M2For the horizontal and vertical pixel of single frames infrared light image
Points；n1And n2For the coordinate that certain in image is put, m1And m2For offset.
The influence of noise may make two frames false matches phenomenon occur, in order to eliminate this influence, to C (m1,m2) make normalizing
Change is handled, and is positioned as after normalization
Wherein,WithRepresent former frame infrared light image and the same target region of this frame infrared light image (before i.e.
There is the target area of same target in one frame infrared light image and this frame infrared light image) average pixel value intensity, i.e.,：
If S4.2.2 related coefficientsMore than the threshold value T of setting, then this target existence confidence level believe subtracts
1 and flame confidence level score subtracts 1；Whether the existence confidence level believe of target is more than 200 or less than 0, if then deleting
Except this target, the memory of the target is discharged；
If S4.2.3 related coefficientsLess than the threshold value T of setting, then confidence level of surviving score adds 1,
The value of believe adds 1, and updates flame information, judges whether score at this time is more than 10, if so, judging that this target is
Flame object；If score is not more than 10, then judge whether the existence confidence level believe of target is more than 200 or less than 0,
If then deleting this target, the memory of the target is discharged.
S5 inputs the infrared light image of new next frame, continues with step S2-S4 to complete the flame object of a new frame
Identification.
The above is only the preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art
For member, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications also should
It is considered as protection scope of the present invention.
Claims (8)
1. a kind of fire defector recognition methods based on Infrared video image, which is characterized in that comprise the following steps：
S1 utilizes infrared camera scan Infrared video image, obtains former frame and this frame infrared light image；
S2 extracts the doubtful flame region in former frame and this frame infrared light image respectively；
S3 carries out object matching to the doubtful flame region of former frame and this frame infrared light image；
The deletion judgement of each target of the doubtful flame region of S4 this frame infrared light images and flame object identification；
S5 inputs the infrared light image of new next frame, continues with step S2-S4 to complete the identification of the flame object of a new frame.
2. the fire defector recognition methods according to claim 1 based on Infrared video image, which is characterized in that lead in S2
It crosses and the image segmentation based on brightness of image is carried out to the infrared light image gathered in S1, so as to extract doubtful flame region.
3. the fire defector recognition methods according to claim 2 based on Infrared video image, which is characterized in that S2 includes
Following steps：
S2.1 assumes that infrared light image maximum gradation value is Max, then choosesFor the seed point of region growing；
The average gray value of pixel in neighborhood centered on S2.2 solution seed points, by each pixel in seed neighborhood of a point
Gray value and this average gray value absolute value of the difference with set threshold value T compared with；
If S2.3 is less than the threshold value T of setting, which is merged with seed point, and as new seed point；If more than
Or the threshold value T equal to setting, then give up；Continue to judge next pixel；
S2.4 is repeated in step S2.2 and S2.3, and meet in step S2.2 and step S2.3 will until all in infrared light image
The pixel asked all is merged into some region, and this region is exactly doubtful flame region；
Doubtful flame region in former frame and this frame infrared light image is extracted using the method in S2.1 to S2.4 respectively.
4. the fire defector recognition methods according to claim 3 based on Infrared video image, which is characterized in that S3 includes
Following steps：
S3.1 respectively by 8 neighborhood connected component labeling methods come mark each target of the doubtful flame region of adjacent two field pictures and
Obtain the boundary rectangle of each target；
S3.2 records the upper left point A of the boundary rectangle of each target in adjacent two frames infrared light image1, lower-left point A2, upper right point
A3, lower-right most point A4And the central point O of boundary rectangle；
If the point A of the boundary rectangle of some target in S3.3 this frame infrared light images1, point A2, point A3, point A4And point O this
In five points, there are some point or certain some target of several points in former frame infrared light image boundary rectangle in, just just
Beginning judges the two targets of this adjacent two frames infrared light image for same target；
If there are steps with more than two targets in former frame infrared light image for the target in S3.4 this frame infrared light images
There are multiple destination matches in the situation in S3.3, i.e., adjacent two frames infrared light image, then need to carry out Secondary Match；
The principle of Secondary Match is：Former frame is included according to above-mentioned five points of the boundary rectangle of the target in this frame infrared light image
The central point of the boundary rectangle of number and two targets in the boundary rectangle of another target in infrared light image away from
From being judged, i.e., the number comprising point at most and the distance between two central points recently, then it is phase to illustrate the two targets
Same target；
If the point A of the boundary rectangle of either objective in S3.5 this frame infrared light images1, point A2, point A3, point A4And point O this
In five points, it is not present in the boundary rectangle of the either objective in former frame infrared light image, i.e. this frame infrared light image
In in each target of doubtful flame region and former frame infrared light image each target of doubtful flame region all mismatch, illustrate this
Target is emerging target, then records the position of the target, gives goal-setting existence confidence level believe, is initialized as
30 and setting flame confidence level, be initialized as 0, doubtful flame object number adds 1.
5. the fire defector recognition methods according to claim 4 based on Infrared video image, which is characterized in that S3.4
In, the distance of two central points uses Euclidean distance, if O1(x1,y1) and O2(x2,y2) be respectively two target boundary rectangles center
Point, then the Euclidean distance computational methods of two central points be：
D=sqrt ((x1-x2)2+(y1-y2)2) (1)。
6. the fire defector recognition methods according to claim 4 or 5 based on Infrared video image, which is characterized in that S4
Comprise the following steps：
S4.1 is obtained the same target of former frame infrared light image and this frame infrared light image by the object matching of step S3, if should
Target has been labeled as flame object before former frame infrared light image or former frame infrared light image, then to the target into
Row delete operation；
S4.2 is obtained the same target of former frame infrared light image and this frame infrared light image by the object matching of step S3, if should
Target is not denoted as flame object before former frame infrared light image and former frame infrared light image, then carries out fire to it
Flame target identification.
7. the fire defector recognition methods according to claim 6 based on Infrared video image, which is characterized in that S4.1 bags
Include following steps：
S4.1.1 is obtained the same target of former frame infrared light image and this frame infrared light image by the object matching of step S3, right
There is the target area of same target in former frame infrared light image and this frame infrared light image, if the target is in former frame
It has been determined as flame object before infrared light image or former frame infrared light image, has judged the mesh in this frame infrared light image
Marking region, also whether there are bright areas；Wherein bright areas refers to that the gray value in this region there are pixel is more than or waits
InThen the region is bright areas；
If there are bright for the target area with same target in the S4.1.2 former frame infrared light images and this frame infrared light image
Bright area, then the existence confidence level believe of the target area adds 2, then judges whether its confidence level believe that survives is more than
200 or less than 0, if so, deleting this target, discharge the memory of the target；Otherwise, continue to retain this target；
If the target area for having same target in the S4.1.3 former frame infrared light images and this frame infrared light image is not present
Bright areas, then the existence confidence level believe of the target area subtracts 2, then judges whether its confidence level believe that survives is more than
200 or less than 0, if so, deleting this target, discharge the memory of the target；Otherwise continue to retain this target.
8. the fire defector recognition methods according to claim 6 based on Infrared video image, which is characterized in that S4.2 bags
Include following steps：
S4.2.1 is obtained the same target of former frame infrared light image and this frame infrared light image by the object matching of step S3, if
Flame object is not flagged as before the same target of former frame infrared light image and this frame infrared light image and the target is
The fresh target that the not doubtful flame region judged in step S3 occurs, according to the position of the target in this frame infrared light image
Doubtful flame information is updated, and calculates the correlation of same target in the doubtful flame region of two adjacent two frames infrared light images
Property, the correlation calculations formula of same target is：
<mrow>
<mi>C</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>m</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>m</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<msub>
<mi>n</mi>
<mn>1</mn>
</msub>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mrow>
<msub>
<mi>M</mi>
<mn>1</mn>
</msub>
<mo>-</mo>
<mn>1</mn>
</mrow>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<msub>
<mi>n</mi>
<mn>2</mn>
</msub>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mrow>
<msub>
<mi>M</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<mn>1</mn>
</mrow>
</munderover>
<msub>
<mi>X</mi>
<mn>1</mn>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>n</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>n</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
<msub>
<mi>X</mi>
<mn>2</mn>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>n</mi>
<mn>1</mn>
</msub>
<mo>+</mo>
<msub>
<mi>m</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>n</mi>
<mn>2</mn>
</msub>
<mo>+</mo>
<msub>
<mi>m</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
</mrow>
In above formula, X1(n1,n2) for certain point (n of some target in former frame infrared light image1,n2) pixel value, X2(n1,n2)
For the pixel value of this frame infrared light image identical point；M1And M2For the points of the horizontal and vertical pixel of single frames infrared light image；n1
And n2For the coordinate that certain in image is put, m1And m2For offset；
The influence of noise may make two frames false matches phenomenon occur, in order to eliminate this influence, to C (m1,m2) make at normalization
Reason, is positioned as after normalization
<mrow>
<mover>
<mi>C</mi>
<mo>&OverBar;</mo>
</mover>
<mrow>
<mo>(</mo>
<msub>
<mi>m</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>m</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<msub>
<mi>n</mi>
<mn>1</mn>
</msub>
<mo>=</mo>
<mn>0</mn>
</mrow>
<msub>
<mi>M</mi>
<mn>1</mn>
</msub>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<msub>
<mi>n</mi>
<mn>2</mn>
</msub>
<mo>=</mo>
<mn>0</mn>
</mrow>
<msub>
<mi>M</mi>
<mn>2</mn>
</msub>
</munderover>
<mo>&lsqb;</mo>
<msub>
<mi>X</mi>
<mn>1</mn>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>n</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>n</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mover>
<msub>
<mi>X</mi>
<mn>1</mn>
</msub>
<mo>&OverBar;</mo>
</mover>
<mo>)</mo>
<mo>&rsqb;</mo>
<mo>&CenterDot;</mo>
<mo>&lsqb;</mo>
<msub>
<mi>X</mi>
<mn>2</mn>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>n</mi>
<mn>1</mn>
</msub>
<mo>+</mo>
<msub>
<mi>m</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>n</mi>
<mn>2</mn>
</msub>
<mo>+</mo>
<msub>
<mi>m</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mover>
<msub>
<mi>X</mi>
<mn>2</mn>
</msub>
<mo>&OverBar;</mo>
</mover>
<mo>&rsqb;</mo>
</mrow>
<mrow>
<msqrt>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<msub>
<mi>n</mi>
<mn>1</mn>
</msub>
<mo>=</mo>
<mn>0</mn>
</mrow>
<msub>
<mi>M</mi>
<mn>1</mn>
</msub>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<msub>
<mi>n</mi>
<mn>2</mn>
</msub>
<mo>=</mo>
<mn>0</mn>
</mrow>
<msub>
<mi>M</mi>
<mn>2</mn>
</msub>
</munderover>
<mo>&lsqb;</mo>
<msub>
<mi>X</mi>
<mn>1</mn>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>n</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>n</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mover>
<msub>
<mi>X</mi>
<mn>1</mn>
</msub>
<mo>&OverBar;</mo>
</mover>
<mo>)</mo>
<msup>
<mo>&rsqb;</mo>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
<mo>&CenterDot;</mo>
<msqrt>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<msub>
<mi>n</mi>
<mn>1</mn>
</msub>
<mo>=</mo>
<mn>0</mn>
</mrow>
<msub>
<mi>M</mi>
<mn>1</mn>
</msub>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<msub>
<mi>n</mi>
<mn>2</mn>
</msub>
<mo>=</mo>
<mn>0</mn>
</mrow>
<msub>
<mi>M</mi>
<mn>2</mn>
</msub>
</munderover>
<mo>&lsqb;</mo>
<msub>
<mi>X</mi>
<mn>2</mn>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>n</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>n</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mover>
<msub>
<mi>X</mi>
<mn>2</mn>
</msub>
<mo>&OverBar;</mo>
</mover>
<mo>)</mo>
<msup>
<mo>&rsqb;</mo>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>3</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein,WithBeing averaged for the same target region of former frame infrared light image and this frame infrared light image is represented respectively
Pixel intensity value, i.e.,：
<mrow>
<mtable>
<mtr>
<mtd>
<mrow>
<mover>
<msub>
<mi>X</mi>
<mn>1</mn>
</msub>
<mo>&OverBar;</mo>
</mover>
<mo>=</mo>
<mfrac>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<msub>
<mi>n</mi>
<mn>1</mn>
</msub>
<mo>=</mo>
<mn>0</mn>
</mrow>
<msub>
<mi>M</mi>
<mn>1</mn>
</msub>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<msub>
<mi>n</mi>
<mn>2</mn>
</msub>
<mo>=</mo>
<mn>0</mn>
</mrow>
<msub>
<mi>M</mi>
<mn>2</mn>
</msub>
</munderover>
<msub>
<mi>X</mi>
<mn>1</mn>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>n</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>n</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>M</mi>
<mn>1</mn>
</msub>
<mo>&times;</mo>
<msub>
<mi>M</mi>
<mn>2</mn>
</msub>
</mrow>
</mfrac>
</mrow>
</mtd>
<mtd>
<mrow>
<mover>
<msub>
<mi>X</mi>
<mn>2</mn>
</msub>
<mo>&OverBar;</mo>
</mover>
<mo>=</mo>
<mfrac>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<msub>
<mi>n</mi>
<mn>1</mn>
</msub>
<mo>=</mo>
<mn>0</mn>
</mrow>
<msub>
<mi>M</mi>
<mn>1</mn>
</msub>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<msub>
<mi>n</mi>
<mn>2</mn>
</msub>
<mo>=</mo>
<mn>0</mn>
</mrow>
<msub>
<mi>M</mi>
<mn>2</mn>
</msub>
</munderover>
<msub>
<mi>X</mi>
<mn>2</mn>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>n</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>n</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>M</mi>
<mn>1</mn>
</msub>
<mo>&times;</mo>
<msub>
<mi>M</mi>
<mn>2</mn>
</msub>
</mrow>
</mfrac>
</mrow>
</mtd>
</mtr>
</mtable>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>4</mn>
<mo>)</mo>
</mrow>
</mrow>
If S4.2.2 related coefficientsMore than the threshold value T of setting, then this target existence confidence level believe subtracts 1, with
And flame confidence level score subtracts 1；Whether the existence confidence level believe of target is more than 200 or less than 0, if then deleting this
Target discharges the memory of the target；
If S4.2.3 related coefficientsLess than the threshold value T of setting, then confidence level of surviving score adds 1, believe's
Value plus 1, and flame information is updated, judge whether score at this time is more than 10, if so, judging this target for flame object；
If score is not more than 10, judge whether the existence confidence level believe of target is more than 200 or less than 0, if then deleting
This target discharges the memory of the target.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711401582.4A CN108108695B (en) | 2017-12-22 | 2017-12-22 | Fire defector recognition methods based on Infrared video image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711401582.4A CN108108695B (en) | 2017-12-22 | 2017-12-22 | Fire defector recognition methods based on Infrared video image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108108695A true CN108108695A (en) | 2018-06-01 |
| CN108108695B CN108108695B (en) | 2019-11-19 |
Family
ID=62212043
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711401582.4A Active CN108108695B (en) | 2017-12-22 | 2017-12-22 | Fire defector recognition methods based on Infrared video image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108108695B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109147254A (en) * | 2018-07-18 | 2019-01-04 | 武汉大学 | A kind of video outdoor fire disaster smog real-time detection method based on convolutional neural networks |
| CN110648490A (en) * | 2019-09-26 | 2020-01-03 | 华南师范大学 | Multi-factor flame identification method suitable for embedded platform |
| CN110717419A (en) * | 2019-09-25 | 2020-01-21 | 浙江万胜智能科技股份有限公司 | Method for extracting flame characteristics from video image |
| WO2020151453A1 (en) * | 2019-01-22 | 2020-07-30 | 杭州海康微影传感科技有限公司 | Open fire detection method and device, and storage medium |
| CN112465852A (en) * | 2020-12-03 | 2021-03-09 | 国网山西省电力公司晋城供电公司 | Improved region growing method for infrared image segmentation of power equipment |
| WO2022121129A1 (en) * | 2020-12-12 | 2022-06-16 | 南方电网调峰调频发电有限公司 | Fire recognition method and apparatus, and computer device and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102509078A (en) * | 2011-10-28 | 2012-06-20 | 北京安控科技股份有限公司 | Fire detection device based on video analysis |
| CN102693603A (en) * | 2012-06-26 | 2012-09-26 | 山东神戎电子股份有限公司 | Dual spectrum based intelligent monitoring system for forest fire prevention |
| CN102760230A (en) * | 2012-06-19 | 2012-10-31 | 华中科技大学 | Flame detection method based on multi-dimensional time domain characteristics |
| CN102819735A (en) * | 2012-08-17 | 2012-12-12 | 深圳辉锐天眼科技有限公司 | Flame detection method based on video frame image |
| CN103473788A (en) * | 2013-07-31 | 2013-12-25 | 中国电子科技集团公司第三十八研究所 | Indoor fire and flame detection method based on high-definition video images |
-
2017
- 2017-12-22 CN CN201711401582.4A patent/CN108108695B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102509078A (en) * | 2011-10-28 | 2012-06-20 | 北京安控科技股份有限公司 | Fire detection device based on video analysis |
| CN102760230A (en) * | 2012-06-19 | 2012-10-31 | 华中科技大学 | Flame detection method based on multi-dimensional time domain characteristics |
| CN102693603A (en) * | 2012-06-26 | 2012-09-26 | 山东神戎电子股份有限公司 | Dual spectrum based intelligent monitoring system for forest fire prevention |
| CN102819735A (en) * | 2012-08-17 | 2012-12-12 | 深圳辉锐天眼科技有限公司 | Flame detection method based on video frame image |
| CN103473788A (en) * | 2013-07-31 | 2013-12-25 | 中国电子科技集团公司第三十八研究所 | Indoor fire and flame detection method based on high-definition video images |
Non-Patent Citations (5)
| Title |
|---|
| R GOKBERK CINBIS等: ""fire detection in infrared video using wavelet analysis"", 《OPTICAL ENGINEERING》 * |
| 冷春艳: ""基于红外图像处理的森林火灾识别关键技术研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 沈诗林等: ""一种基于视频图像相关性的火灾火焰识别方法"", 《安全与环境学报》 * |
| 王腾等: ""一种基于连续帧图像相似度的火焰识别方法"", 《海军工程大学学报》 * |
| 田佳霖: ""基于火焰特性分析的视频火灾检测"", 《信息技术与信息化》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109147254A (en) * | 2018-07-18 | 2019-01-04 | 武汉大学 | A kind of video outdoor fire disaster smog real-time detection method based on convolutional neural networks |
| CN109147254B (en) * | 2018-07-18 | 2021-05-18 | 武汉大学 | Video field fire smoke real-time detection method based on convolutional neural network |
| WO2020151453A1 (en) * | 2019-01-22 | 2020-07-30 | 杭州海康微影传感科技有限公司 | Open fire detection method and device, and storage medium |
| CN110717419A (en) * | 2019-09-25 | 2020-01-21 | 浙江万胜智能科技股份有限公司 | Method for extracting flame characteristics from video image |
| CN110648490A (en) * | 2019-09-26 | 2020-01-03 | 华南师范大学 | Multi-factor flame identification method suitable for embedded platform |
| CN112465852A (en) * | 2020-12-03 | 2021-03-09 | 国网山西省电力公司晋城供电公司 | Improved region growing method for infrared image segmentation of power equipment |
| CN112465852B (en) * | 2020-12-03 | 2024-01-30 | 国网山西省电力公司晋城供电公司 | Improved region growing method for infrared image segmentation of power equipment |
| WO2022121129A1 (en) * | 2020-12-12 | 2022-06-16 | 南方电网调峰调频发电有限公司 | Fire recognition method and apparatus, and computer device and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108108695B (en) | 2019-11-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108108695B (en) | Fire defector recognition methods based on Infrared video image | |
| CN103440484B (en) | A kind of flame detecting method adapting to outdoor large space | |
| CN107609470B (en) | Method for detecting early smoke of field fire by video | |
| CN106600888B (en) | Automatic forest fire detection method and system | |
| Zhao et al. | SVM based forest fire detection using static and dynamic features | |
| Qi et al. | A computer vision based method for fire detection in color videos | |
| CN103208126B (en) | Moving object monitoring method under a kind of physical environment | |
| CN111091072A (en) | YOLOv 3-based flame and dense smoke detection method | |
| CN106204646A (en) | Multiple mobile object tracking based on BP neutral net | |
| CN106683091A (en) | Target classification and attitude detection method based on depth convolution neural network | |
| CN107909033A (en) | Suspect's fast track method based on monitor video | |
| CN106250845A (en) | Flame detecting method based on convolutional neural networks and device | |
| CN109460764A (en) | A kind of satellite video ship monitoring method of combination brightness and improvement frame differential method | |
| CN108090495A (en) | A kind of doubtful flame region extracting method based on infrared light and visible images | |
| CN112270331A (en) | Improved billboard detection method based on YOLOV5 | |
| CN110334660A (en) | A kind of forest fire monitoring method based on machine vision under the conditions of greasy weather | |
| CN107463954A (en) | A kind of template matches recognition methods for obscuring different spectrogram picture | |
| CN114202646A (en) | Infrared image smoking detection method and system based on deep learning | |
| CN109389185A (en) | Use the video smoke recognition methods of Three dimensional convolution neural network | |
| CN108921215A (en) | A kind of Smoke Detection based on local extremum Symbiotic Model and energy spectrometer | |
| CN112699801A (en) | Fire identification method and system based on video image | |
| CN111951250A (en) | Image-based fire detection method | |
| CN106650638A (en) | Abandoned object detection method | |
| CN110135347A (en) | A kind of flame identification method based on video image | |
| CN110852303A (en) | Eating behavior identification method based on OpenPose |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |