CN107274374B - Smoke monitoring method based on computer vision technology - Google Patents
Smoke monitoring method based on computer vision technology Download PDFInfo
- Publication number
- CN107274374B CN107274374B CN201710531954.9A CN201710531954A CN107274374B CN 107274374 B CN107274374 B CN 107274374B CN 201710531954 A CN201710531954 A CN 201710531954A CN 107274374 B CN107274374 B CN 107274374B
- Authority
- CN
- China
- Prior art keywords
- smoke
- background
- sample set
- color
- foreground
- 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.)
- Active
Links
- 239000000779 smoke Substances 0.000 title claims abstract description 137
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000012544 monitoring process Methods 0.000 title claims abstract description 37
- 238000005516 engineering process Methods 0.000 title claims abstract description 30
- 230000033001 locomotion Effects 0.000 claims abstract description 24
- 239000000126 substance Substances 0.000 claims abstract description 16
- 238000000605 extraction Methods 0.000 claims abstract description 7
- 238000004364 calculation method Methods 0.000 claims abstract description 5
- 238000001514 detection method Methods 0.000 claims description 15
- 238000009499 grossing Methods 0.000 claims description 5
- 238000011410 subtraction method Methods 0.000 claims description 5
- 239000003086 colorant Substances 0.000 claims description 4
- 238000002485 combustion reaction Methods 0.000 claims description 4
- 230000000295 complement effect Effects 0.000 claims description 3
- 230000007797 corrosion Effects 0.000 claims description 3
- 238000005260 corrosion Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000011049 filling Methods 0.000 claims description 2
- 239000002245 particle Substances 0.000 claims description 2
- 239000007789 gas Substances 0.000 description 6
- 238000012545 processing Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 3
- MGWGWNFMUOTEHG-UHFFFAOYSA-N 4-(3,5-dimethylphenyl)-1,3-thiazol-2-amine Chemical compound CC1=CC(C)=CC(C=2N=C(N)SC=2)=C1 MGWGWNFMUOTEHG-UHFFFAOYSA-N 0.000 description 2
- ZAMOUSCENKQFHK-UHFFFAOYSA-N Chlorine atom Chemical compound [Cl] ZAMOUSCENKQFHK-UHFFFAOYSA-N 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 239000000460 chlorine Substances 0.000 description 2
- 229910052801 chlorine Inorganic materials 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 230000000877 morphologic effect Effects 0.000 description 2
- JCXJVPUVTGWSNB-UHFFFAOYSA-N nitrogen dioxide Inorganic materials O=[N]=O JCXJVPUVTGWSNB-UHFFFAOYSA-N 0.000 description 2
- 239000011800 void material Substances 0.000 description 2
- 238000003915 air pollution Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 239000000428 dust Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000007667 floating Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000001681 protective effect Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
Abstract
The invention provides a smoke monitoring method based on a computer vision technology, which is used for detecting smoke according to the color characteristics and the motion characteristics of the smoke: sequentially comprises three stages of foreground extraction based on color, image background removal based on motion and intersection operation of the foreground extracted by the two stages; performing intersection operation on the smoke target image extracted through the color information and the smoke target image extracted through the motion characteristic, wherein the operation result is the finally extracted smoke target; and simultaneously, positioning calculation is carried out on the smoke source. The method sets the color characteristics of the smoke according to the reasons for generating the smoke and the difference of the physical and chemical characteristics of the smoke, adopts single-frame initialization for constructing a background model, adopts a background removing method based on a sample set mode, improves the efficiency, and meets the requirement of algorithm instantaneity; aiming at a large space, the position of the smoke is positioned by more than two monitoring cameras, so that the timeliness and the accuracy of subsequent work are ensured.
Description
Technical Field
The invention relates to the technical field of smoke monitoring and positioning, in particular to a smoke monitoring method based on a computer vision technology.
Background
Smoke generally begins to diffuse before a flame is generated and therefore can be an important feature in determining flame burning; meanwhile, the random emission of smoke is also an important source of air pollution, so that the effective monitoring of the smoke pollution source becomes an important task of environmental protection; in addition, the accidental leakage of certain chemical drugs can also generate smog-like substances, and certain visual basis is provided for monitoring and distinguishing.
The traditional smoke detection device mainly adopts a contact sensor technology, but has the insurmountable defect, under the condition of large indoor space, when the smoke gradually spreads to reach the vicinity of a detector, the smoke becomes very low in concentration and cannot reach the alarm threshold of the traditional detector, even if the smoke reaches the alarm threshold and causes alarm, the scale of harm develops and forms, and the aim of early alarm cannot be fulfilled; in addition, under outdoor large space conditions, such as forests, meadows and the like, the cost for arranging a large number of sensors is too high, the realization is also difficult, and if the sensors are too close to smoke emission sources, such as chimney outlets and the like, the sensors often fail due to severe environments such as high temperature, dust and the like, and the situations of false alarm, false alarm and the like occur.
Due to the limitation of the traditional contact type smoke detection technology in large space detection, the smoke monitoring technology based on image acquisition and image processing technology and computer vision technology is developed. Firstly, a camera inputs a monitored video sequence into a computer through an acquisition card, then, the image content is analyzed by computer software, and the smoke detection is finished according to the image characteristics of the smoke.
At present, the smoke detection based on image technology mainly comprises: 1. based on the chromaticity information of the smoke pixels, only the discrimination of gray smoke and cyan smoke is carried out, and the smoke color generated by the leakage of chemical substances is not considered; 2. extracting the characteristic information of the smoke by using a wavelet transform technology, but the judgment of the smoke based on the wavelet transform is not accurate; 3. the smoke detection algorithm based on the parting idea for coding is time-consuming and cannot meet the requirement of real-time performance; 4. the method is a method for determining the relation of inter-frame motion areas by adopting inter-frame target association and comprehensively analyzing the attributes of a motion target to judge whether the motion target is smoke or not, and the detection of the motion target by adopting a frame difference method is sensitive to environmental noise, so that the selection of a threshold value is important, the method is only suitable for specific occasions and has poor stability; 5. the method is based on the comprehensive weighting analysis of static characteristics such as smoke morphological characteristics and attenuation characteristics of high-frequency signals relative to the background, dynamic characteristics such as change and change frequency of the smoke interior and color characteristics of smoke relative to background color attenuation, but the algorithm has high time and space complexity and is not beneficial to real-time monitoring, in addition, different parameter settings are required under different scenes, and the algorithm has poor robustness and intelligence. 6. Other methods for performing characteristic separation and background modeling on smoke by applying theories such as a Bayes classifier, a mixed Gaussian model and an artificial neuron network exist, but the application has certain limitation, and the method is only suitable for specific occasions and has poor stability.
In addition, there is no way to implement tracking and localization of smoke sources.
Disclosure of Invention
In order to solve the problems in the background art, the invention provides a smoke monitoring method based on a computer vision technology, although the reasons for generating smoke and the physical and chemical characteristics of the smoke are different, the method can set the color characteristics of the smoke through one-time initialization, and the initialization, the setting and the operation are all on the same platform, so that the operation is simple and the maintenance is easy; the background model is constructed by adopting single-frame initialization and adopting a background removal method based on a sample set mode, so that the efficiency is improved, and the real-time requirement of the algorithm is met; aiming at a large space, the position of the smoke is positioned by more than two monitoring cameras, so that the timeliness and the accuracy of subsequent work are ensured.
In order to achieve the purpose, the invention adopts the following technical scheme:
a smoke monitoring method based on computer vision technology comprises the following steps:
step one, video image preprocessing: the noise is filtered through a smoothing function, and then the smoothed image is enhanced, so that the contrast of the image is improved;
step two, smoke detection is carried out according to the color characteristics and the motion characteristics of the smoke: sequentially comprises three stages of 1) foreground extraction based on color, 2) image background removal based on motion, and 3) intersection operation of the foreground extracted by the two stages; performing intersection operation on the smoke target image extracted through the color information and the smoke target image extracted through the motion characteristic, wherein the operation result is the finally extracted smoke target;
step three, positioning calculation of the smoke source: and tracking and positioning the smoke source of the smoke detected in the second step.
The smoke color characteristic model generated by the combustion of the substances in the step two is as follows:
where N is the number of consecutive frames selected, typically 20<N<35，Ii,jIi，jFor each color component, mean, of a pixel point (i, j) in the RGB color modeli,jFor the mean value of the gray levels of the individual channels, Center is the Center of the gray level of the pixel of smoke, which is typically 160, with a parameter of 400,1＝135，2＝195。
the smoke color feature model generated by the leaked chemical substances in the step two is as follows:
where N is the number of successive frames selected, XkIs a certain color component at a pixel point (i, j) in the RGB color model, where k ∈ (R, G, B),
is XkThe complement of (a) is to be added,
and ω is chosen according to the different colors of the gases, for example: chlorine is a yellow-green gas, then
And ω 50, nitrogen dioxide is a reddish brown gas, then
And ω 60.
The image background removal based on the movement in the second step adopts a high-efficiency background reconstruction and foreground extraction method, and the algorithm sequentially comprises the following steps: a. establishing a background sample set, b, judging background pixel points, c, updating the background sample set:
a. establishing a background sample set: the establishment of the background Sample set can be completed by one frame, the algorithm adopts the pixel point neighborhood with similar space-time distribution characteristics, the pixel neighborhood value is randomly selected to fill the model Sample set, the method has high calculation speed, and the establishment of the Sample set Sample (k, i, j) based on all the pixel pointsM×NWherein M, N is the number of pixels in horizontal and vertical directions, i is 0 ≦ M-1, j is 0 ≦ N-1, K is 1 ≦ K ≦ K, where K is usually 2n，n＝3，4，5…。
b. And (3) distinguishing background pixel points: pixel point I in a new framei，jComparing size with elements in the sample set, provided that pixel Ii，jIf the number of the Euclidean distances from the elements in the sample set to be less than R is greater than # minDISTANCE, the point is considered as a background point;
c. updating of background sample set: if any pixel in the current frame is considered to be a background point as long as the number of pixels which are less than R away from the element in the element Sample set is greater than # minDISTANCE, the pixel needs to be replaced by the pixel (k, i, j)M×NThe replacement of the pixels is performed by a random method.
And (c) dynamically updating the background by circulating the step (b) and the step (c), and then extracting the foreground by using a background subtraction method, wherein a void phenomenon occurs when the background is poor because the motion of smoke is generally slow, so that a mathematical morphology method is adopted for expansion corrosion, the void is filled, noise is removed, and a complete foreground target is extracted.
In the third step, the smoke source tracking and positioning adopts a method based on the fitting of a target segment of a connected region and continuous multiframes to mark foreground targets, records the information of the position, the area and the like of each foreground, and simultaneously establishes a tracking linked list for each appearing foreground target, wherein the length of the linked list is L (20)<＝L<30), the smoke characteristics of successive L frames are storedThe position (m, n) of the smoke target mass point is calculated, the smoke target is moved to the central point (m) of the monitoring camera screen by the rotation angle theta of the cameracenter,ncenter) Finally, calculating the position of the generated smoke according to the positions of more than two cameras;
wherein (x, y) is the coordinate position of the smoke source, (x)1，y1) Is the coordinate position of the camera 1, theta1Is the rotation angle of the camera 1 for positioning the smoke source, (x)2，y2) Is the coordinate position of the camera 2, theta2Is the angle of rotation at which the camera 2 locates the smoke source.
Compared with the prior art, the invention has the beneficial effects that:
1. according to the smoke monitoring method based on the computer vision technology, although the reasons for generating smoke and the physical and chemical characteristics of the smoke are different, the color characteristics of the smoke can be set through one-time initialization, and the initialization, the setting and the operation are all on the same platform, so that the smoke monitoring method is simple to operate and easy to maintain.
2. According to the smoke monitoring method based on the computer vision technology, single-frame initialization is adopted for constructing the background model, and a background removing method based on a sample set mode is adopted, so that the efficiency is improved, and the requirement of algorithm instantaneity is met.
3. According to the smoke monitoring method based on the computer vision technology, a management mode of multiple cameras and multiple tasks is adopted for a large space, so that equipment is installed and a coordinate system is constructed on a unified management platform according to the particularity of a monitoring area.
4. According to the smoke monitoring method based on the computer vision technology, the background model is dynamically updated in real time according to the characteristic of the dynamic change of the background in the outdoor large-scale environment, and the influence on the smoke monitoring accuracy due to weather or light is avoided.
5. According to the smoke monitoring method based on the computer vision technology, the positions of smoke generation are positioned through more than two monitoring cameras, and timeliness and accuracy of subsequent work are guaranteed.
6. The smoke monitoring method based on the computer vision technology adopts the computer technology, the multimedia technology and the wireless network communication technology, and the cost of project equipment is low.
Drawings
Fig. 1 is a system overall structure diagram of a smoke monitoring method based on computer vision technology of the invention;
FIG. 2 is a general flow diagram of a smoke monitoring method based on computer vision technology according to the present invention;
fig. 3 is a schematic block diagram of a background subtraction method of the smoke monitoring method based on computer vision technology according to the present invention.
Detailed Description
The following detailed description of the present invention will be made with reference to the accompanying drawings.
As shown in fig. 1, a smoke monitoring system based on computer vision technology comprises: the system comprises an image data acquisition module 1 (a lens, a camera, a protective cover and a tripod head; a spherical camera; a network spherical camera; a video encoder), an IP network transmission module 2 and a monitoring center data processing and analyzing module 3 (a remote monitoring user; a video server; a monitoring client; and data receiving, processing, analyzing and decision-making software).
The method comprises the steps that firstly, a camera of front-end equipment collects a field video image, digital video information is obtained through A/D conversion and converted into an IP data packet, the video data packet is sent to a monitoring center through a data special line, the monitoring center judges whether smoke exists in the image through image processing and analyzing technology, if yes, the position of a smoke generation place is calculated and stored, and meanwhile, early warning information is sent out.
As shown in fig. 2, a smoke monitoring method based on computer vision technology includes the following steps:
step one, video image preprocessing: the noise is filtered through a smoothing function, and then the smoothed image is enhanced, so that the contrast of the image is improved;
step two, smoke detection is carried out according to the color characteristics and the motion characteristics of the smoke: sequentially comprises three stages of 1) foreground extraction based on color, 2) image background removal based on motion, and 3) intersection operation of the foreground extracted by the two stages; performing intersection operation on the smoke target image extracted through the color information and the smoke target image extracted through the motion characteristic, wherein the operation result is the finally extracted smoke target;
step three, positioning calculation of the smoke source: and tracking and positioning the smoke source of the smoke detected in the second step.
The first step is specifically as follows:
1.1, smoothing and filtering the obtained digital video data by adopting a Gaussian smoothing filter;
d (u, v) is the distance from the Fourier transform in place, and σ represents the degree of Gaussian curvature expansion and can be taken to be 3,5, 10.
1.2, enhancing the image of the smoothed result by adopting a histogram equalization method based on partial differential equation;
evolution equation of input image I (I, j, t):
I0(i, j) is an input image, AΩIs the area of the defined domain Ω, A (-) represents the area function, and the equation has a unique steady state solution:
Dmaxis the image gray maximum, and h (I) is the cumulative histogram of image I.
The second step is specifically as follows:
2.1, extracting the smoke color characteristics by adopting a smoke color characteristic model, and finding that for most combustion materials, when the temperature is lower, the color of smoke is transited from bluish white to white; as the temperature increases, the smoke color gradually darkens, transitioning from a gray-white to a gray-black color, until burning. As shown by the color data analysis of these typical smoke images, it can be found that the colors of the smoke are mostly distributed in a relatively compact narrow band, and therefore the smoke color feature model generated by combustion is:
where N is the number of consecutive frames selected, typically 20<N<35，Ii,jIi，jFor each color component, mean, of a pixel point (i, j) in the RGB color modeli,jFor the mean value of the gray levels of the individual channels, Center is the Center of the gray level of the pixel of smoke, which is typically 160, with a parameter of 400,1＝135，2＝195。
the model of the colour characteristics of the smoke generated by the leaking chemicals is:
where N is the number of successive frames selected, typically 15<N<30，XkOne or two color components at pixel point (i, j) in the RGB color model, where k ∈ (R, G, B),
is XkThe complement of (c).
And ω is chosen according to the different colors of the gases, for example: chlorine is a yellow-green gas, then
And ω 50, nitrogen dioxide is a reddish brown gas, then
And ω 60.
2.2, as shown in fig. 3, the detection of the moving object is performed by using a dynamic background subtraction method, and the algorithm describes:
Dk(i，j)＝|fk(i，j)-Bk-1(i，j)|
wherein, (i, j) is the position coordinate of the discrete image pixel point, fk(i, j) is an input image, Bk-1(i, j) is a background model, Dk(i, j) detected motion region, MkThe (i, j) is the motion region binarization, and T15 is the binarization threshold. And obtaining a binary image after motion detection, then carrying out morphological processing and connectivity analysis on the binary image, if the area of a certain connected region is larger than a preset threshold value, indicating that a moving target is detected, considering that the connected region is an extracted target image, and simultaneously updating a background model in real time.
The image background removal based on movement adopts a high-efficiency background reconstruction and foreground extraction method, and the algorithm sequentially comprises the following steps: a. establishing a background sample set, b, judging background pixel points, c, updating the background sample set:
2.2.1, first, the establishment of a background sample set: the algorithm adopts the characteristic that pixel point neighborhoods have similar space-time distribution, randomly selects pixel neighborhood values to fill a model Sample set, and establishes a Sample set Sample (k, i, j) based on all pixel pointsM×NWhere M, N is the number of pixels in the horizontal and vertical directions, where 0 ≦ i ≦ M-1, 0 ≦ j ≦ N-1, and 1 ≦ K ≦ K, where K represents the number of samples in general, and K ≦ 2 in generalnN is 3, 4, 5, …, as formula
Sample(k,i,j)＝{υ1(i,j),υ2(i,j),...,υk(i,j),...,υK(i,j)}
Where the υ index k refers to the kth sample in the sample set.
2.2.2, secondly, distinguishing background pixel points: pixel point I in a new framei，jComparing size with elements in the sample set, provided that pixel Ii，jIf the number of the Euclidean distance from the element in the sample set to be less than R is greater than # minDISTANCE, the point is considered as a background point, wherein K is 25, # minDISTANCE is 2, R is 20, and the discrimination formula is
#{SR(υ(i,j))∩{υ1(i,j),υ2(i,j),...,υk(i,j),...,υK(i,j)}}≥#minDISTANCE
And if the number of the elements in the ith row and the jth column in the current frame, which are less than R away from the elements in the element Sample set, is more than # minDISTANCE, considering upsilon (i, j) as a background, and replacing Sample (k, i, j) with upsilon (i, j)M×NAs to which one is replaced, a random method is adopted for updating.
2.2.3, the updating of the sample set adopts a random strategy, namely, the probability of updating each sample is ensured to be equal, namely 1/K. Then the probability that the sample value P will not be updated at time t is (K-1)/K, and if the time is continuous, the probability that the sample value remains after dt times is:
the probability that the sample in the background model sample set is replaced is not related to the time t, that is, the random update strategy is applicable to the update of the background model sample set.
And 2.3, circulating the step 2.2 to realize the dynamic update of the background model, and then extracting the foreground by using a background subtraction method.
And 2.4, performing expansion corrosion by adopting a mathematical morphology method, filling the hollow, removing noise and extracting a complete foreground target.
The third step is specifically as follows:
3.1, smoke source tracking:
labeling of foreground connected regions:
3.1.1, creating a structure array containing the position, the gray average value and the area information of each connected region;
3.1.2, creating a linked list for storing information corresponding to the target section in the communication area;
and 3.1.3, horizontally and vertically scanning the image matrix, if a new unrecorded foreground target section is scanned, creating a node recording the position of the target section and flag bit information of whether the target section is marked, and inserting the node into a linked list.
And 3.1.4, scanning the upper and lower neighborhoods of the seed point by taking the head node of the linked list as a seed growing point, creating a node for recording the row where the target section is located, the column where the starting and ending positions are located and whether the marked flag bit exists or not when the unrecorded target section exists, and adding the node to the tail part of the linked list.
And 3.1.5, taking out a target section from the head of the linked list, judging and marking, and then deleting the head node until the linked list is empty, namely, a connected region is marked. And updating the information of the structure array every time one target section is marked, and completely recording the information of the connected region in the structure array when the whole connected region is marked.
3.2, positioning of a smoke source:
marking foreground targets by adopting a continuous multi-frame fitting method, recording the position and area information of each foreground, and establishing a tracking linked list for each appearing foreground target, wherein the length of the linked list is L (20)<＝L<30), the statistical information of the smoke characteristics of the continuous L frames is stored, the position (m, n) of smoke target particles is calculated, and the smoke target is moved to the central point (m) of the monitoring camera screen by the rotation angle theta of the cameracenter,ncenter) Calculating the smoke position by the rotation angle of the two cameras
Wherein(x, y) is the coordinate position of the smoke source, (x)1，y1) Is the coordinate position of the camera 1, theta1Is the rotation angle of the camera 1 for positioning the smoke source, (x)2，y2) Is the coordinate position of the camera 2, theta2Is the angle of rotation at which the camera 2 locates the smoke source.
According to the technical scheme, the smoke monitoring, positioning and alarming device disclosed by the invention adopts a single-frame dynamic background modeling algorithm which is high in operation speed and suitable for complex real-time changing environment according to the color characteristic of the fog substances generated by burning the substances and the color characteristic of the fog substances generated by leakage of the chemical substances and the motion characteristic of the smoke drift, so that the smoke monitoring, positioning and alarming on a large scene are realized. Experiments show that the algorithm can accurately distinguish floating clouds, flying flags, airplanes, kites, vehicles, pedestrians and smoke targets, and has good detection and alarm effects on large-area forests, factories, large-space chemical storage warehouses and the like. However, the detection rate of the experiment is background-sensitive, so that when a constructor installs a camera, a background which is relatively single, small in change and stable in light is constructed as much as possible, and the detection rate is improved.
The above embodiments are implemented on the premise of the technical solution of the present invention, and detailed embodiments and specific operation procedures are given, but the scope of the present invention is not limited to the above embodiments. The methods used in the above examples are conventional methods unless otherwise specified.
Claims (2)
1. A smoke monitoring method based on computer vision technology is characterized by comprising the following steps:
step one, video image preprocessing: the noise is filtered through a smoothing function, and then the smoothed image is enhanced, so that the contrast of the image is improved;
step two, smoke detection is carried out according to the color characteristics and the motion characteristics of the smoke: sequentially comprises three stages of 1) foreground extraction based on color, 2) image background removal based on motion, and 3) intersection operation of the foreground extracted by the two stages; performing intersection operation on the smoke target image extracted through the color information and the smoke target image extracted through the motion characteristic, wherein the operation result is the finally extracted smoke target;
step three, positioning calculation of the smoke source: tracking and positioning the smoke source of the smoke detected in the second step;
the smoke color characteristic model generated by the combustion of the substances in the step two is as follows:
where N is the number of consecutive frames selected, 20<N<35，meani,jFor the mean value of the gray levels of the individual channels, Center is the Center of the gray level of the pixel of smoke, and 160, with a parameter of 400,1＝135，2＝195；
the smoke color feature model generated by the leaked chemical substances in the step two is as follows:
where N is the number of successive frames selected, XkIs a certain color component at a pixel point (i, j) in the RGB color model, where k ∈ (R, G, B),
is XkThe complement of (a) is to be added,
the value of the sum omega is determined according to different colors of the gas;
in the third step, the smoke source tracking and positioning adopts a method based on the matching of a target segment of a connected region and continuous multiframes to mark foreground targets, the position and area information of each foreground is recorded, and a tracking linked list is established for each appearing foreground target, wherein the length of the linked list is L: 20<＝L<30, the statistical information of the smoke characteristics of the continuous L frames is stored, the position (m, n) of smoke target particles is calculated, and the smoke target is moved to a monitoring camera screen by a camera rotation angle thetaCenter point (m)center,ncenterFinally, calculating the position of the generated smoke according to the positions of more than two cameras;
wherein (x, y) is the coordinate position of the smoke source, (x)1，y1) Is the coordinate position of the camera 1, theta1Is the rotation angle of the camera 1 for positioning the smoke source, (x)2，y2) Is the coordinate position of the camera 2, theta2Is the angle of rotation at which the camera 2 locates the smoke source.
2. The computer vision technology-based smoke monitoring method as claimed in claim 1, wherein the motion-based image background removal in the second step is an efficient background reconstruction and foreground extraction method; sequentially comprises the following steps: a. establishing a background sample set, b, judging background pixel points, c, updating the background sample set:
a. establishing a background sample set: the establishment of the background Sample set is completed by one frame, the algorithm adopts the characteristic that the neighborhood of the pixel points has similar space-time distribution, the neighborhood value of the pixel points is randomly selected to fill the model Sample set, and the Sample set Sample (k, i, j) based on all the pixel points is establishedM×NWherein M, N is the number of pixel points in horizontal and vertical directions, i is more than or equal to 0 and less than or equal to M-1, j is more than or equal to 0 and less than or equal to N-1, K is more than or equal to 1 and less than or equal to K, and K is 2n，n＝3，4，5…；
b. And (3) distinguishing background pixel points: pixel point I in a new framei，jComparing size with elements in the sample set, provided that pixel Ii，jIf the number of the Euclidean distances from the elements in the sample set to be less than R is greater than # minDISTANCE, the point is considered as a background point; # minDISTANCE ═ 2, R ═ 20;
c. updating of background sample set: if any pixel in the current frame is considered to be a background point as long as the number of pixels which are less than R away from the element in the element Sample set is greater than # minDISTANCE, the pixel needs to be replaced by the pixel (k, i, j)M×NThe replacement of the pixels adopts a random method;
and (c) circularly performing the step (b) and the step (c) to realize the dynamic update of the background, then extracting the foreground by using a background subtraction method, performing expansion corrosion by adopting a mathematical morphology method, filling the cavity, removing the noise and extracting a complete foreground target.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710531954.9A CN107274374B (en) | 2017-07-03 | 2017-07-03 | Smoke monitoring method based on computer vision technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710531954.9A CN107274374B (en) | 2017-07-03 | 2017-07-03 | Smoke monitoring method based on computer vision technology |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107274374A CN107274374A (en) | 2017-10-20 |
| CN107274374B true CN107274374B (en) | 2020-09-22 |
Family
ID=60070733
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710531954.9A Active CN107274374B (en) | 2017-07-03 | 2017-07-03 | Smoke monitoring method based on computer vision technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107274374B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111353334A (en) * | 2018-12-21 | 2020-06-30 | 富士通株式会社 | Smoke detection method and device |
| CN112634151A (en) * | 2020-12-14 | 2021-04-09 | 深圳中兴网信科技有限公司 | Poisson fusion-based smoke data enhancement method, enhancement equipment and storage medium |
| CN113435393B (en) * | 2021-07-12 | 2023-05-09 | 北京林业大学 | Forest fire smoke root node detection method, device and equipment |
| CN114049737A (en) * | 2021-11-30 | 2022-02-15 | 东北林业大学 | Forest fire monitoring and fire extinguishing system based on FPGA and tracked robot |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8983180B2 (en) * | 2012-10-05 | 2015-03-17 | Industry Academic Cooperation Foundation Keimyung University | Method of detecting smoke of forest fire using spatiotemporal BoF of smoke and random forest |
| CN106897720A (en) * | 2017-01-11 | 2017-06-27 | 济南中维世纪科技有限公司 | A kind of firework detecting method and device based on video analysis |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103795935B (en) * | 2014-03-05 | 2017-12-12 | 吉林大学 | A kind of camera shooting type multi-target orientation method and device based on image rectification |
| CN105678295B (en) * | 2016-01-04 | 2019-03-19 | 武汉科技大学 | Gas heating flame method of real-time based on the analysis of ROI the average image |
| CN105701474B (en) * | 2016-01-15 | 2019-02-05 | 西安交通大学 | A kind of video smoke recognition methods of color combining and external physical characteristic |
| CN105788142B (en) * | 2016-05-11 | 2018-08-31 | 中国计量大学 | A kind of fire detection system and detection method based on Computer Vision |
| CN106650600A (en) * | 2016-10-17 | 2017-05-10 | 东南大学 | Forest smoke and fire detection method based on video image analysis |
-
2017
- 2017-07-03 CN CN201710531954.9A patent/CN107274374B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8983180B2 (en) * | 2012-10-05 | 2015-03-17 | Industry Academic Cooperation Foundation Keimyung University | Method of detecting smoke of forest fire using spatiotemporal BoF of smoke and random forest |
| CN106897720A (en) * | 2017-01-11 | 2017-06-27 | 济南中维世纪科技有限公司 | A kind of firework detecting method and device based on video analysis |
Non-Patent Citations (2)
| Title |
|---|
| Color Feature Selection for Smoke Detection in Videos;Gabriela Miranda等;《2014 12th IEEE International Conference on Industrial Informatics(INDIN)》;20141106;第31-36页 * |
| 基于视频的实时烟雾检测算法;张建明等;《计算机应用与软件》;20161231;第33 卷(第12 期);第248-252页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107274374A (en) | 2017-10-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107274374B (en) | Smoke monitoring method based on computer vision technology | |
| CN107705288B (en) | Infrared video detection method for dangerous gas leakage under strong interference of pseudo-target motion | |
| CN103069434B (en) | For the method and system of multi-mode video case index | |
| CN108665487B (en) | Transformer substation operation object and target positioning method based on infrared and visible light fusion | |
| CN106897681B (en) | Remote sensing image contrast analysis method and system | |
| CN110379020B (en) | Laser point cloud coloring method and device based on generation countermeasure network | |
| CN110188835B (en) | Data-enhanced pedestrian re-identification method based on generative confrontation network model | |
| CN109598794B (en) | Construction method of three-dimensional GIS dynamic model | |
| CN110210428B (en) | MSER-based smoke root node detection method in remote complex environment | |
| US10489674B2 (en) | Information extraction using image data | |
| CN104978567B (en) | Vehicle checking method based on scene classification | |
| CN109918971B (en) | Method and device for detecting number of people in monitoring video | |
| CN111383244B (en) | Target detection tracking method | |
| CN111985365A (en) | Straw burning monitoring method and system based on target detection technology | |
| Huerta et al. | Exploiting multiple cues in motion segmentation based on background subtraction | |
| Xiong et al. | Early smoke detection of forest fires based on SVM image segmentation | |
| CN112862150A (en) | Forest fire early warning method based on image and video multi-model | |
| Shi et al. | Video-based fire detection with saliency detection and convolutional neural networks | |
| CN112085778A (en) | Oblique photography illegal building detection method and system based on superpixels and morphology | |
| CN115311623A (en) | Equipment oil leakage detection method and system based on infrared thermal imaging | |
| JP2020030750A (en) | Shadow detection device, shadow detection method, shadow detection program, learning device, leaning method, and learning program | |
| CN113033386A (en) | High-resolution remote sensing image-based transmission line channel hidden danger identification method and system | |
| CN113628251B (en) | Smart hotel terminal monitoring method | |
| CN110910379B (en) | Incomplete detection method and device | |
| Xia et al. | A shadow detection method for remote sensing images using affinity propagation algorithm |
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 | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20220624 Address after: 130062 7th floor, 1153 Zhengyang Street, Lvyuan District, Changchun City, Jilin Province Patentee after: CHANGCHUN KUNLUN CONSTRUCTION CO.,LTD. Address before: 114044, No. 185, Qianshan Road, hi tech Zone, Liaoning, Anshan Patentee before: University of Science and Technology Liaoning |
|
| TR01 | Transfer of patent right |