[go: up one dir, main page]

CN110874592A - Forest fire smoke image detection method based on total bounded variation - Google Patents

Forest fire smoke image detection method based on total bounded variation Download PDF

Info

Publication number
CN110874592A
CN110874592A CN201911000036.9A CN201911000036A CN110874592A CN 110874592 A CN110874592 A CN 110874592A CN 201911000036 A CN201911000036 A CN 201911000036A CN 110874592 A CN110874592 A CN 110874592A
Authority
CN
China
Prior art keywords
image
smoke
suspected
value
tbv
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
Application number
CN201911000036.9A
Other languages
Chinese (zh)
Other versions
CN110874592B (en
Inventor
安明伟
姜敏敏
李洪昌
王雷
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing Vocational College Of Information Technology
Original Assignee
Nanjing Vocational College Of Information Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nanjing Vocational College Of Information Technology filed Critical Nanjing Vocational College Of Information Technology
Priority to CN201911000036.9A priority Critical patent/CN110874592B/en
Publication of CN110874592A publication Critical patent/CN110874592A/en
Application granted granted Critical
Publication of CN110874592B publication Critical patent/CN110874592B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/254—Analysis of motion involving subtraction of images
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/40—Analysis of texture
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20021—Dividing image into blocks, subimages or windows
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A40/00—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production
    • Y02A40/10—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production in agriculture
    • Y02A40/28—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production in agriculture specially adapted for farming

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Multimedia (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Computation (AREA)
  • Evolutionary Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Image Analysis (AREA)
  • Fire-Detection Mechanisms (AREA)

Abstract

The invention relates to the technical field of information processing, in particular to a forest fire smoke image detection method based on total bounded variation, which is characterized by comprising the following steps of: step 1: inputting images in a video sequence; step 2: adopting an image blocking method to obtain TBV values of blocked images, searching suspected smoke blocking areas from a blocking result image by comparing the TBV values twice, and finally extracting final suspected smoke areas by utilizing fusion clustering processing of feature data; and step 3: carrying out smoke motion characteristic analysis on the suspected smoke area by adopting an interframe difference method, if the suspected smoke area moves, judging as a fire smoke image, and giving a fire alarm; and if the suspected smoke area does not move, returning to the video input to perform the next image processing flow. The invention avoids errors caused by complex calculation, and has accurate and stable output result and good detection effect.

Description

Forest fire smoke image detection method based on total bounded variation
Technical Field
The invention relates to the technical field of information processing, in particular to a forest fire smoke image detection method based on total bounded variation.
Background
Forest fires attract attention of all countries in the world as disasters which cause the greatest harm to forests; the traditional detection technology based on the smoke sensor has a small monitoring range and high laying cost in a forest large area, and the sensor is easy to age and the sensitivity is reduced. At present, a video system is widely applied to forest monitoring, and partial fire incidents can be found through manual monitoring, but the forest area is wide, and video images are numerous, so that difficulty is brought to manual monitoring. In recent years, with the continuous development and progress of the exploration of image processing technology, video image processing and image recognition technology methods are becoming mature. The video smoke detection technology has become a novel forest fire detection and early warning method due to the advantages of short response time, high sensitivity, large coverage area and the like, has wide application prospect and is well paid attention by researchers at home and abroad.
The fire smoke has abundant image characteristics, and the current video smoke detection method mainly analyzes the fire smoke according to visual characteristics such as the movement, color, shape, transparency, texture and the like of the smoke; for example, Yuan et al propose a method of accumulating a motion model and rapidly estimating the motion direction of smoke using an integral map to detect smoke. Chen et al detect fire smoke based on changes in color and shape. Phillips and the like detect fire smoke according to the characteristics of video smoke shape irregularity and smoke area expansion; toreyin et al propose a smoke detection method based on contour and wavelet transforms. Aiming at the problem, the Tian and the like analyze a mixing mechanism of the smoke and the background, construct a set of smoke foreground extraction method, reduce background interference to a certain extent and improve the smoke identification accuracy. In the aspect of extracting the smoke texture features, related research methods are more, wherein the methods such as GLCM, LBP, Wavelet and the like are most widely applied; toreyin et al use texture features to detect smoke; yu et al realize a set of real-time flame and smoke detection system based on GLCM analysis of smoke texture; introducing LBP by Tian and the like to extract smoke texture characteristics; yuan et al propose a smoke detection method based on pyramid histogram sequence; however, the color features, the texture features and the like of smoke are comparatively dispersed, the variation range is large, more false detections exist in practical application, and some methods have no obvious advantages in forest fire smoke detection. And the feature research is limited to static low-level features such as color, contour, motion and the like, and is not enough to distinguish smoke from some suspected smoke objects, such as: clouds, fog and the like, and the research on high-level characteristics and space-time characteristics is less; the smoke of the forest fire presents various states in different environments, the difficulty of smoke visual feature extraction smoke detection is difficult, how to construct a stable and efficient feature extraction algorithm, and static and dynamic information in a video is fused, so that the key for reducing smoke false detection is realized.
Disclosure of Invention
The invention aims to provide a forest fire smoke image detection method based on total bounded variation, which avoids errors caused by complex calculation, and has accurate and stable output result and good detection effect.
In order to solve the technical problems, the technical scheme of the invention is as follows: a forest fire smoke image detection method based on total bounded variation includes the steps:
step 1: inputting images in a video sequence;
step 2: extracting a suspected smoke area of the forest fire in the image, specifically:
step 2.1: partitioning the image to form N image blocks, and calculating the TBV value of each image block by using a total bounded variation method;
step 2.2: solving the overall average value of the image according to the TBV value of each image block
Figure BDA0002241012990000021
And the total variance σ2If, if
Figure BDA0002241012990000022
Returning to the step 1 to process the next frame of image; if it is
Figure BDA0002241012990000023
Then the overall mean of the image is taken
Figure BDA0002241012990000024
As an initial dividing reference, comparing the TBV value of each image block with the initial dividing reference, determining the category of each image block in the image, and enabling the TBV not to exceed the overall average value
Figure BDA0002241012990000025
The image block is classified into a suspected fire smoke block;
step 2.3: fusing and clustering suspected fire smoke blocks to form characteristic data and extracting a final suspected smoke area;
and step 3: carrying out smoke motion characteristic analysis on the suspected smoke area by adopting an interframe difference method, if the suspected smoke area moves, judging as a fire smoke image, and giving a fire alarm; and if the suspected smoke area does not move, returning to the video input to perform the next image processing flow.
According to the above scheme, the specific method for calculating the TBV value of each image block in step 2.1 is as follows:
let f denote the original sharp image, f (x, y) denote the pixel value of the coordinate (x, y) pixel in the image; when the clear image is blurred due to smoke, the pixels of the blurred image are set as g (x, y), and the total bounded variation TBV of the clear image f is calculated respectivelyfAnd total bounded variation TBV of blurred image ggRespectively solving the following equations (1) and (2):
Figure BDA0002241012990000026
Figure BDA0002241012990000027
the expression of omega in the formulas (1) and (2) is shown as (3):
Ω＝{(x,y):0≤x≤a,0≤y≤b} (3)
where a is the image length and b is the image width.
According to the scheme, the step 2.2 specifically comprises the following steps:
step 2.21: calculating the overall average value of the image according to the TBV value of each image block
Figure BDA0002241012990000028
Global variance σ2(ii) a Global variance σ2Expressed as:
Figure BDA0002241012990000029
wherein ,TBVnThe TBV value of the nth image block in the image is obtained, and N is the total number of the image blocks in the image;
step 2.22: judging whether the image has smoke or not; the method specifically comprises the following steps:
if it is
Figure BDA0002241012990000031
Indicating that the image has no smoke or the smoke is distributed integrally, eliminating the smoke of the fire, and returning to the step 1 to process the next frame of image;
if it is
Figure BDA0002241012990000032
Not excluding fire smoke, executing step 2.23;
step 2.23: and (3) initially dividing the image block according to the TBV value: selecting an ensemble mean of images
Figure BDA0002241012990000033
As an initial dividing standard, comparing the TBV value of each image block with the initial dividing standard to determine the category of each image block in the image, and determining the category of each image block in the image when the TBV value of each image block is taken as the initial dividing standard
Figure BDA0002241012990000034
Namely when the TBV value of the nth image block is less than or equal to the overall average value, determining the type of the image block as suspected fire smokeBlock, form category data Mn(ii) a Grouping pixels corresponding to the division of the image block into a reference division to form an initial division M-M of the whole image1，M2，…，Mn。
According to the above scheme, in step 2.3, M is determined by using spectral clustering method1,M2,…MnCarrying out fusion clustering on the initially divided category data of the image block to cluster, forming feature data, and extracting a suspected smoke area, wherein the method specifically comprises the following steps:
according to
Figure BDA0002241012990000035
Can obtain the corresponding suspected fire smoke block M after the ith frame of image is dividedn(x, y) synthesizing the divided image blocks to obtain the suspected smog area M of the ith frame imagei 2(x, y) solved by equation (5):
Figure BDA0002241012990000036
wherein ,
Figure BDA0002241012990000037
representing the combination of the segmented image blocks into a complete video image; mN·M i(x, y) represents the Nth row and the Mth column of suspected fire smoke blocks in the ith frame of image;
finally, the suspected smoke area M of the ith frame imagei(x, y) is Mi 1(x, y) and Mi 2The intersection of (x, y). Solving by equation (6):
Figure BDA0002241012990000038
according to the formula (6), the t-th image can be sequentially extracted from the video fire smoke image1,t2,...,tiSuspected smoke region of image frame
Figure BDA0002241012990000039
Definition of
Figure BDA00022410129900000310
In turn is
Figure BDA00022410129900000311
Picture pixel number of smoke of medium fire; thereby, a suspected smoke region is extracted.
According to the scheme, the specific algorithm of the interframe difference method in the step 3 is as follows:
the image of the ith frame and the ith-1 frame in the video sequence is recorded as fi and fi-1The gray value of the corresponding pixel points of two frames is recorded as fi(x, y) and fi-1(x, y), subtracting the gray values of the corresponding pixel points of the two frames of images according to the formula (7), and taking the absolute value of the gray values to obtain a difference image Di：
Di＝|fi(x,y)-fi-1(x,y)| (7)
Setting a binary threshold value of the difference image as T, and carrying out binarization processing on the pixel points one by one according to a formula (8) to obtain a binary image R'i(x, y), wherein the point with the gray value of 255 is a foreground point, namely a motion target point, and the point with the gray value of 0 is a background point; to R'i(x, y) performing connectivity analysis to obtain an image R containing a complete moving targeti；
Figure BDA0002241012990000041
wherein ,fi(x, y) is the gray value of the suspected smoke area of the ith image frame, Di(x, y) is the smoke from tiImage frame to ti+1Growth of image frame, while t1<t2<...<ti；
The fire smoke growth rate of a suspected smoke area of a single frame and two continuous frames is not representative, and t with the same number of image frames or time interval is selected1,t2,...,tiThe taken value of i of the image frame depends on Di(x,y)>The number of times T is generated, described by equation (9):
Figure BDA0002241012990000042
in the formula (9), T is a threshold value, TvAnd if the suspected smoke area increases to be not more than the threshold value, the suspected smoke area does not move, the video is returned to carry out the processing flow of the next image.
The invention has the following beneficial effects:
an image quality evaluation method based on Total Bounded Variation (TBV) proposes that when an image becomes blurred for various reasons, the difference between image boundaries becomes small, and the Total Bounded Variation thereof gradually decreases; under the normal weather condition, the image boundary difference of the forest monitoring picture is large, after fire smoke exists, namely a large amount of noise is generated in a part of areas on a clear image, the thicker the smoke is, the more the noise is, and therefore the potential fire hazard can be judged; when the smoke concentration is increased due to a fire, the corresponding image fuzzy degree is increased, the difference of the texture boundary is reduced, the total bounded variation score is gradually reduced, the image fuzzy degree can be treated as a fuzzy image, the difference between the boundaries can be effectively represented based on the characteristic properties of the variation score, a relation model between the TBV and the image fuzzy degree is constructed on the basis of the TBV theory, and the TBV of an extreme value state can be used as a criterion for smoke concentration evaluation; the invention provides a forest fire smoke image detection method based on total bounded variation, which considers the actual condition of forest monitoring and solves an extreme value of a target function by the idea of block stable analysis so as to obtain a total bounded value, and then extracts a final suspected smoke area by utilizing the fusion clustering processing of characteristic data; in order to obtain a better smoke detection effect, after the smoke characteristic region is extracted, the smoke motion characteristic analysis is carried out, and then the fusion judgment is carried out to give out fire alarm.
The method shields the complex calculation of static characteristics of the smoke, such as color, texture, frequency characteristics and the like, and can extract a more complete suspected smoke area based on image blocking and TBV detection relative ratio according to the fuzzy degree of the smoke; after the smoke characteristic area is extracted, smoke motion characteristic analysis is carried out, and then fire alarm is given out through fusion judgment; when the smoke motion characteristics are analyzed, an interframe difference method is adopted, two adjacent frames are obtained in a video image sequence, difference operation is carried out, and finally the motion characteristics are extracted; the change of pixel values among pixel points in a moving object is not considered, the complete area of the object is not extracted, the smoke detection output can be accurately carried out only by obtaining the change of the outline of the object caused by the motion of smoke or not, the phenomenon of 'holes' in the interframe detection method is avoided, and the smoke detection effect is better. The method reduces the calculation amount, avoids errors caused by complex calculation, and has accurate and stable result output and wide application prospect.
Drawings
FIG. 1 is an overall flow chart of the smoke image detection method of the present invention;
FIG. 2 is a flowchart of a suspected smoke region extraction algorithm in this embodiment;
FIG. 3 is a block diagram of an image in the present embodiment;
FIG. 4 is a schematic diagram of three-dimensional division of data for initial division of image blocks;
FIG. 5 is a schematic diagram of data contour line division of initial division of image blocks;
FIG. 6 is a graph of image fusion clusters;
FIG. 7 is a flowchart illustrating the difference between frames according to the present embodiment;
fig. 8 is a schematic diagram illustrating a growth rate analysis in the present embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.
Referring to fig. 1 to 8, the invention is a method for detecting a forest fire smoke image based on total bounded variation, firstly extracting a suspected smoke area of a forest fire according to a detection algorithm of the total bounded variation fire smoke area, then performing motion characteristic analysis, giving a fire alarm if the suspected smoke area moves, and returning to a video input to perform a next image processing flow if the suspected smoke area does not move. The steps of the forest fire smoke detection method are shown in figure 1, and the steps comprise,
step 1: inputting images in a video sequence;
step 2: extracting a suspected smoke area of the forest fire in the image; the method specifically comprises the following steps: adopting an image blocking method to obtain TBV values of blocked images, searching suspected smoke blocking areas from a blocking result image by comparing the TBV values twice, and finally extracting final suspected smoke areas by utilizing fusion clustering processing of feature data;
and step 3: carrying out smoke motion characteristic analysis on the suspected smoke area by adopting an interframe difference method, if the suspected smoke area moves, judging as a fire smoke image, and giving a fire alarm; and if the suspected smoke area does not move, returning to the video input to perform the next image processing flow.
The extraction of suspected smoke areas in video-based smoke detection is an important step. The integrity of the suspected smoke region extraction will affect the extraction of smoke features, such as the area and perimeter of the smoke region and the diffusion characteristics of smoke, so that the overall smoke detection will have certain influence. For the analysis of the method of the present invention, a method based on image blocks and relative TBV detection is provided to extract a more complete suspected smoke area, in order to eliminate false detection caused by natural weather such as fog, variance calculation is performed on a block image TBV of a detection image, if the variance is smaller, the result is eliminated as the cause of fire smoke, and then block detection is performed on the image, wherein the flow chart of the suspected smoke area extraction method is shown in FIG. 2, and the method specifically comprises the following steps:
step 2.1: partitioning the image to form N image blocks, and calculating the TBV value of each image block by using a total bounded variation method; in this embodiment, the image is expanded and divided into 16 × 16 independent smaller images, as shown in fig. 3; calculating TBV values of all image blocks by using a total bounded variation method, aiming at solving the TBV values of a large-size image in a plurality of times so as to extract feature data; the specific method for calculating the TBV value of each image block comprises the following steps:
let f denote the original sharp image, f (x, y) denote the pixel value of the coordinate (x, y) pixel in the image; when the clear image is blurred due to smoke, the pixels of the blurred image are set as g (x, y), and the total bounded variation TBV of the clear image f is calculated respectivelyfAnd total bounded variation TBV of blurred image ggRespectively solving the following equations (1) and (2):
Figure BDA0002241012990000061
Figure BDA0002241012990000062
the expression of omega in the formulas (1) and (2) is shown as (3):
Ω＝{(x,y):0≤x≤a,0≤y≤b} (3)
where a is the image length and b is the image width.
Step 2.2: the suspected fire smoke patch was searched from the image by comparing TBV values twice: solving the overall average value of the image according to the TBV value of each image block
Figure BDA0002241012990000063
And the total variance σ2If, if
Figure BDA0002241012990000064
Returning to the step 1 to process the next frame of image; if it is
Figure BDA0002241012990000071
Then the overall mean of the image is taken
Figure BDA0002241012990000072
As an initial dividing reference, comparing the TBV value of each image block with the initial dividing reference, determining the category of each image block in the image, and enabling the TBV not to exceed the overall average value
Figure BDA0002241012990000073
The image block is classified into a suspected fire smoke block; the specific method comprises the following steps:
step 2.21: calculating the overall average value of the image according to the TBV value of each image block
Figure BDA0002241012990000074
Global variance σ2(ii) a Global variance σ2Expressed as:
Figure BDA0002241012990000075
wherein ,TBVnThe TBV value of the nth image block in the image is obtained, and N is the total number of the image blocks in the image;
step 2.22: judging whether the image has smoke or not; the method specifically comprises the following steps:
if it is
Figure BDA0002241012990000076
Indicating that the image has no smoke or the smoke is distributed integrally, eliminating the smoke of the fire, and returning to the step 1 to process the next frame of image;
if it is
Figure BDA0002241012990000077
Not excluding fire smoke, executing step 2.23;
step 2.23: and (3) initially dividing the image block according to the TBV value: to avoid loss of generality, the overall mean of the image is selected
Figure BDA0002241012990000078
As an initial dividing standard, comparing the TBV value of each image block with the initial dividing standard to determine the category of each image block in the image, and determining the category of each image block in the image when the TBV value of each image block is taken as the initial dividing standard
Figure BDA0002241012990000079
Namely when the TBV value of the nth image block is less than or equal to the overall average value, determining the type of the image block as a suspected fire smoke block to form type data Mn(ii) a By the method, the pixels corresponding to the division of the image block are classified into the reference division, and one initial division M of the whole image is formed1，M2，…，MnAs shown in fig. 4 and 5;
step 2.3: fusing and clustering suspected fire smoke blocks to form characteristic data and extracting a final suspected smoke area; the method specifically comprises the following steps: class data M by using spectral clustering methodnFor M ═ M1,M2,…MnCarrying out fusion clustering on the initially divided category data of the image block to cluster, forming feature data, and extracting a suspected smoke area, wherein the method specifically comprises the following steps:
according to
Figure BDA00022410129900000710
Can obtain the corresponding suspected fire smoke block M after the ith frame of image is dividedn(x, y) synthesizing the divided image blocks to obtain the suspected smog area M of the ith frame imagei 2(x, y) solved by equation (5):
Figure BDA00022410129900000711
wherein ,
Figure BDA00022410129900000712
representing the combination of the segmented image blocks into a complete video image; mN·M i(x, y) represents the Nth row and the Mth column of suspected fire smoke blocks in the ith frame of image;
finally, the suspected smoke area M of the ith frame imagei(x, y) is Mi 1(x, y) and Mi 2The intersection of (x, y). Solving by equation (6):
Figure BDA0002241012990000081
according to the formula (6), the t-th image can be sequentially extracted from the video fire smoke image1,t2,...,tiSuspected smoke region of image frame
Figure BDA0002241012990000082
Definition of
Figure BDA0002241012990000083
In turn is
Figure BDA0002241012990000084
Picture pixel number of smoke of medium fire; thereby, a suspected smoke region is extracted.
When fire smoke images are judged, smoke movement is irregular due to the influence of air flow on smoke particles, so that the smoke outline presents complex characteristics; for fire smoke detection, the method for judging and detecting whether smoke exists through single characteristics is inaccurate, in order to obtain better smoke detection effect, smoke characteristic analysis is carried out after smoke characteristic regions are extracted to carry out smoke fusion judgment, and smoke has unique characteristics in aspects of color space, irregular motion, main motion direction and the like due to the characteristics of visual blurring, translucency, particles and diffusion motion of the smoke, so that complex calculation of static characteristics of the smoke, such as color, texture and the like, is shielded. Therefore, when the characteristics of the smoke are analyzed, the smoke detection and output can be accurately carried out only by paying attention to the motion characteristics of the smoke; in early fire, smoke can continuously emerge from a smoke point and continuously spread to the periphery, the spreading mode is different from other moving objects, the edge contour of the area is always kept in an irregular form in the moving process of the moving object, and the dynamic characteristics of the video smoke are mainly described according to the characteristics that the smoke moves along with the air flow and shows the diffusibility and irregularity of the smoke movement.
The most easily realized method in all moving object detection algorithms is the interframe difference method (temporalldifference), which is based on the principle that two adjacent frames are obtained in a video image sequence, difference operation is carried out on the two adjacent frames, and finally, the motion characteristics are extracted. The algorithm has the greatest advantages that the complexity of calculation is avoided, the calculation is simple, the required time is short, the efficiency is greatly improved, and meanwhile, the algorithm is strong in environment adaptation. The invention realizes the extraction of the smoke region based on the TBV, can extract the smoke region according to the fuzzy degree of the smoke, does not need to consider the pixel value change between the pixel points in the moving object, does not need to extract the complete region of the object, and only needs to obtain the contour of the object without the change caused by the smoke motion, thereby the 'cavity' phenomenon which appears when the interframe detection method is adopted does not influence the detection result. In addition, the moving speed of the smoke target is low, and a good judgment result can be obtained only by reasonably selecting the threshold T, so that the moving characteristic of the smoke is obtained.
The video sequence collected by the camera has the characteristic of continuity, if no moving object exists in a scene, the change of continuous frames is weak, and if the moving object exists, the continuous frames can obviously change from frame to frame. The interframe difference method is based on the thought, because the target in the scene moves and the position of the image of the target in different image frames is different, the algorithm carries out difference operation on two or three continuous frames of images, the pixel points corresponding to different frames are subtracted, the absolute value of the gray difference is judged, when the absolute value exceeds a certain threshold value, the target can be judged to be the moving target, and therefore the target detection function is realized.
The operation process of the two-frame difference method is shown in FIG. 7, and the i-th frame and the i-1 th frame in the video sequence are recorded as fi and fi-1The gray value of the corresponding pixel points of two frames is recorded as fi(x, y) and fi-1(x, y), subtracting the gray values of the corresponding pixel points of the two frames of images according to the formula (7), and taking the absolute value of the gray values to obtain a difference image Di：
Di＝|fi(x,y)-fi-1(x,y)| (7)
Setting a binary threshold value of the difference image as T, and carrying out binarization processing on the pixel points one by one according to a formula (8) to obtain a binary image R'i(x, y), wherein the point with the gray value of 255 is a foreground point, i.e., a motion target point,the point with the gray value of 0 is the background point; to R'i(x, y) performing connectivity analysis to obtain an image R containing a complete moving targeti；
Figure BDA0002241012990000091
wherein ,fi(x, y) is the gray value of the suspected smoke area of the ith image frame, Di(x, y) is the smoke from tiImage frame to ti+1Growth of image frame, while t1<t2<...<tiAs shown in fig. 8;
the fire smoke growth rate of a suspected smoke area of a single frame and two continuous frames is not representative, and t with the same number of image frames or time interval is selected1,t2,...,tiThe taken value of i of the image frame depends on Di(x,y)>The number of times T is generated, described by equation (9):
Figure BDA0002241012990000092
in the formula (9), T is a threshold value, TvIf the suspected smoke area increases to exceed the threshold value, the suspected smoke area moves, and the existence of fire smoke is judged. In equation (9), a sampling termination frame t is definednStored as tnIf the video image has fire smoke, the sampling termination image frame is obtained, which represents that the fire smoke is accurately detected, and if the video image has no fire smoke, tnThe value of (2) will continue to the end of the video image and finally equals to the time length value of the video image, so the sampling end image frame is very critical for fire prediction alarm. Determining the presence of fire smoke does not mean that a fire smoke alarm can be given, as it may be a partial misjudgment and effect. The video images meeting the conditions effectively detect fire smoke and give fire alarm.
The foregoing is a more detailed description of the present invention that is presented in conjunction with specific embodiments, and the practice of the invention is not to be considered limited to those descriptions. For those skilled in the art to which the invention pertains, several simple deductions or substitutions can be made without departing from the spirit of the invention, and all shall be considered as belonging to the protection scope of the invention.

Claims (5)

1. A forest fire smoke image detection method based on total bounded variation is characterized by comprising the following steps:
step 1: inputting images in a video sequence;
step 2: extracting a suspected smoke area of the forest fire in the image, specifically:
step 2.1: partitioning the image to form N image blocks, and calculating the TBV value of each image block by using a total bounded variation method;
step 2.2: solving the overall average value of the image according to the TBV value of each image block
Figure FDA0002241012980000011
And the total variance σ2If, if
Figure FDA0002241012980000012
Returning to the step 1 to process the next frame of image; if it is
Figure FDA0002241012980000013
Then the overall mean of the image is taken
Figure FDA0002241012980000014
As an initial dividing reference, comparing the TBV value of each image block with the initial dividing reference, determining the category of each image block in the image, and enabling the TBV not to exceed the overall average value
Figure FDA0002241012980000015
The image block is classified into a suspected fire smoke block;
step 2.3: and fusing and clustering the suspected fire smoke blocks to form characteristic data and extracting a final suspected smoke area.
And step 3: carrying out smoke motion characteristic analysis on the suspected smoke area by adopting an interframe difference method, if the suspected smoke area moves, judging as a fire smoke image, and giving a fire alarm; and if the suspected smoke area does not move, returning to the video input to perform the next image processing flow.
2. The method for detecting forest fire smoke image based on total bounded variation as claimed in claim 1, wherein: the specific method for calculating the TBV value of each image block in step 2.1 is as follows:
let f denote the original sharp image, f (x, y) denote the pixel value of the coordinate (x, y) pixel in the image; when the clear image is blurred due to smoke, the pixels of the blurred image are set as g (x, y), and the total bounded variation TBV of the clear image f is calculated respectivelyfAnd total bounded variation TBV of blurred image ggRespectively solving the following equations (1) and (2):
Figure FDA0002241012980000016
Figure FDA0002241012980000017
the expression of omega in the formulas (1) and (2) is shown as (3):
Ω＝{(x,y):0≤x≤a,0≤y≤b} (3)
where a is the image length and b is the image width.
3. The forest fire smoke image detection method based on total bounded variation as claimed in claim 2, wherein: the step 2.2 specifically comprises the following steps:
step 2.21: calculating the overall average value of the image according to the TBV value of each image block
Figure FDA0002241012980000018
Global variance σ2(ii) a Global variance σ2Expressed as:
Figure FDA0002241012980000021
wherein ,TBVnThe TBV value of the nth image block in the image is obtained, and N is the total number of the image blocks in the image;
step 2.22: judging whether the image has smoke or not; the method specifically comprises the following steps:
if it is
Figure FDA0002241012980000022
Indicating that the image has no smoke or the smoke is distributed integrally, eliminating the smoke of the fire, and returning to the step 1 to process the next frame of image;
if it is
Figure FDA0002241012980000023
Not excluding fire smoke, executing step 2.23;
step 2.23: and (3) initially dividing the image block according to the TBV value: selecting an ensemble mean of images
Figure FDA0002241012980000024
As an initial dividing standard, comparing the TBV value of each image block with the initial dividing standard to determine the category of each image block in the image, and determining the category of each image block in the image when the TBV value of each image block is taken as the initial dividing standard
Figure FDA0002241012980000025
Namely when the TBV value of the nth image block is less than or equal to the overall average value, determining the type of the image block as a suspected fire smoke block to form type data Mn(ii) a Grouping pixels corresponding to the division of the image block into a reference division to form an initial division M-M of the whole image1，M2，…，Mn。
4. The method for detecting forest fire smoke image based on total bounded variation as claimed in claim 3, wherein: step 2.3 utilizes a spectral clustering method to pair M ═ M1,M2,…MnPerforming fusion clustering on category data of initial division of image blockClustering, forming characteristic data, and extracting suspected smoke areas, specifically:
according to
Figure FDA0002241012980000026
Can obtain the corresponding suspected fire smoke block M after the ith frame of image is dividedn(x, y) synthesizing the divided image blocks to obtain the suspected smog area M of the ith frame imagei 2(x, y) solved by equation (5):
Figure FDA0002241012980000027
wherein ,
Figure FDA0002241012980000028
representing the combination of the segmented image blocks into a complete video image; mN·M i(x, y) represents the Nth row and the Mth column of suspected fire smoke blocks in the ith frame of image;
finally, the suspected smoke area M of the ith frame imagei(x, y) is Mi 1(x, y) and Mi 2The intersection of (x, y). Solving by equation (6):
Mi(x,y)＝Mi 1(x,y)∩Mi 2(x,y) (6)
according to the formula (6), the t-th image can be sequentially extracted from the video fire smoke image1,t2,...,tiSuspected smoke region of image frame
Figure FDA0002241012980000029
Definition of
Figure FDA00022410129800000210
In turn is
Figure FDA00022410129800000211
Figure FDA00022410129800000212
Picture pixel number of smoke of medium fire; thereby, a suspected smoke region is extracted.
5. The forest fire smoke image detection method based on total bounded variation as claimed in claim 4, wherein: the specific algorithm of the interframe difference method in the step 3 is as follows:
the image of the ith frame and the ith-1 frame in the video sequence is recorded as fi and fi-1The gray value of the corresponding pixel points of two frames is recorded as fi(x, y) and fi-1(x, y), subtracting the gray values of the corresponding pixel points of the two frames of images according to the formula (7), and taking the absolute value of the gray values to obtain a difference image Di：
Di＝|fi(x,y)-fi-1(x,y)| (7)
Setting a difference image binarization threshold value as T, and carrying out binarization processing on pixel points one by one according to a formula (8) to obtain a binarization image Ri' (x, y), wherein, the point with the gray value of 255 is a foreground point, namely a motion target point, and the point with the gray value of 0 is a background point; to Ri' (x, y) performing connectivity analysis to obtain an image R containing a complete moving objecti；
Figure FDA0002241012980000031
wherein ,fi(x, y) is the gray value of the suspected smoke area of the ith image frame, Di(x, y) is the smoke from tiImage frame to ti+1Growth of image frame, while t1<t2<...<ti；
The fire smoke growth rate of a suspected smoke area of a single frame and two continuous frames is not representative, and t with the same number of image frames or time interval is selected1,t2,...,tiThe taken value of i of the image frame depends on Di(x,y)>The number of times T is generated, described by equation (9):
Figure FDA0002241012980000032
in the formula (9), T is a threshold value, TvAnd if the suspected smoke area increases to be not more than the threshold value, the suspected smoke area does not move, the video is returned to carry out the processing flow of the next image.
CN201911000036.9A 2019-10-21 2019-10-21 Forest fire smoke image detection method based on total bounded variation Active CN110874592B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911000036.9A CN110874592B (en) 2019-10-21 2019-10-21 Forest fire smoke image detection method based on total bounded variation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911000036.9A CN110874592B (en) 2019-10-21 2019-10-21 Forest fire smoke image detection method based on total bounded variation

Publications (2)

Publication Number Publication Date
CN110874592A true CN110874592A (en) 2020-03-10
CN110874592B CN110874592B (en) 2023-05-16

Family

ID=69717858

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911000036.9A Active CN110874592B (en) 2019-10-21 2019-10-21 Forest fire smoke image detection method based on total bounded variation

Country Status (1)

Country Link
CN (1) CN110874592B (en)

Cited By (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111476142A (en) * 2020-04-02 2020-07-31 深圳市泰和安科技有限公司 Video image detection method and device
CN111898549A (en) * 2020-07-31 2020-11-06 平安国际智慧城市科技股份有限公司 Fire monitoring method and device based on artificial intelligence, computer equipment and medium
CN112115878A (en) * 2020-09-21 2020-12-22 北京林业大学 Forest fire smoke root node detection method based on smoke area density
CN112560944A (en) * 2020-12-14 2021-03-26 广东电网有限责任公司珠海供电局 Charging pile fire detection method based on image recognition
CN112949453A (en) * 2021-02-26 2021-06-11 南京恩博科技有限公司 Training method of smoke and fire detection model, smoke and fire detection method and smoke and fire detection equipment
CN113989488A (en) * 2021-10-20 2022-01-28 青岛浩海网络科技股份有限公司 Firework point positioning method based on video monitoring
CN114155439A (en) * 2021-12-07 2022-03-08 河南鑫安利安全科技股份有限公司 Chemical enterprise safety risk early warning method and system based on visual perception
CN114332694A (en) * 2021-12-20 2022-04-12 苏州汇川控制技术有限公司 Elevator smoke identification method and system, terminal equipment and storage medium
CN114998218A (en) * 2022-05-11 2022-09-02 南京信息职业技术学院 Method and device for detecting defects of industrial products with uniform surfaces
CN115049955A (en) * 2022-05-19 2022-09-13 江苏南工科技集团有限公司 Fire detection analysis method and device based on video analysis technology
CN115223105A (en) * 2022-09-20 2022-10-21 万链指数(青岛)信息科技有限公司 Big data based risk information monitoring and analyzing method and system
CN115311658A (en) * 2022-10-12 2022-11-08 四川三思德科技有限公司 Forest fire prevention smoke alarm anti-interference processing method
CN115564710A (en) * 2022-08-31 2023-01-03 清远蓄能发电有限公司 Fire smoke detection method, device and storage medium based on LK optical flow method
CN116777707A (en) * 2023-06-20 2023-09-19 浙江宏远智能科技有限公司 Intelligent community management system and method
CN118799765A (en) * 2024-09-11 2024-10-18 浙江点创信息科技有限公司 A method and system for processing low-altitude images of unmanned aerial vehicles
CN119652937A (en) * 2024-11-29 2025-03-18 浙江久成消防设备有限公司 A fire information management system and method based on Internet of Things

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030117652A1 (en) * 1999-09-17 2003-06-26 Paul Lapstun Rotationally symmetric tags
CN101739551A (en) * 2009-02-11 2010-06-16 北京智安邦科技有限公司 Method and system for identifying moving objects
CN101739685A (en) * 2009-02-11 2010-06-16 北京智安邦科技有限公司 Moving object classification method and system thereof
CN109165577A (en) * 2018-08-07 2019-01-08 东北大学 A kind of early stage forest fire detection method based on video image

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030117652A1 (en) * 1999-09-17 2003-06-26 Paul Lapstun Rotationally symmetric tags
CN101739551A (en) * 2009-02-11 2010-06-16 北京智安邦科技有限公司 Method and system for identifying moving objects
CN101739685A (en) * 2009-02-11 2010-06-16 北京智安邦科技有限公司 Moving object classification method and system thereof
CN109165577A (en) * 2018-08-07 2019-01-08 东北大学 A kind of early stage forest fire detection method based on video image

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
周洁琼;邹北骥;朱岳;刘洁;: "一种改进的实时火焰模拟算法", 计算机工程与科学 *

Cited By (24)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111476142A (en) * 2020-04-02 2020-07-31 深圳市泰和安科技有限公司 Video image detection method and device
CN111898549A (en) * 2020-07-31 2020-11-06 平安国际智慧城市科技股份有限公司 Fire monitoring method and device based on artificial intelligence, computer equipment and medium
CN112115878A (en) * 2020-09-21 2020-12-22 北京林业大学 Forest fire smoke root node detection method based on smoke area density
CN112115878B (en) * 2020-09-21 2024-05-14 北京林业大学 A forest fire smoke root node detection method based on smoke area density
CN112560944B (en) * 2020-12-14 2023-08-08 广东电网有限责任公司珠海供电局 A fire detection method for charging piles based on image recognition
CN112560944A (en) * 2020-12-14 2021-03-26 广东电网有限责任公司珠海供电局 Charging pile fire detection method based on image recognition
CN112949453A (en) * 2021-02-26 2021-06-11 南京恩博科技有限公司 Training method of smoke and fire detection model, smoke and fire detection method and smoke and fire detection equipment
CN112949453B (en) * 2021-02-26 2023-12-26 南京恩博科技有限公司 Training method of smoke and fire detection model, smoke and fire detection method and equipment
CN113989488A (en) * 2021-10-20 2022-01-28 青岛浩海网络科技股份有限公司 Firework point positioning method based on video monitoring
CN114155439A (en) * 2021-12-07 2022-03-08 河南鑫安利安全科技股份有限公司 Chemical enterprise safety risk early warning method and system based on visual perception
CN114332694A (en) * 2021-12-20 2022-04-12 苏州汇川控制技术有限公司 Elevator smoke identification method and system, terminal equipment and storage medium
CN114998218A (en) * 2022-05-11 2022-09-02 南京信息职业技术学院 Method and device for detecting defects of industrial products with uniform surfaces
CN115049955A (en) * 2022-05-19 2022-09-13 江苏南工科技集团有限公司 Fire detection analysis method and device based on video analysis technology
CN115049955B (en) * 2022-05-19 2024-11-26 江苏南工科技集团有限公司 A fire detection analysis method and device based on video analysis technology
CN115564710A (en) * 2022-08-31 2023-01-03 清远蓄能发电有限公司 Fire smoke detection method, device and storage medium based on LK optical flow method
CN115223105A (en) * 2022-09-20 2022-10-21 万链指数(青岛)信息科技有限公司 Big data based risk information monitoring and analyzing method and system
CN115223105B (en) * 2022-09-20 2022-12-09 万链指数(青岛)信息科技有限公司 Big data based risk information monitoring and analyzing method and system
CN115311658B (en) * 2022-10-12 2023-01-10 四川三思德科技有限公司 Forest fire prevention smoke alarm anti-interference processing method
CN115311658A (en) * 2022-10-12 2022-11-08 四川三思德科技有限公司 Forest fire prevention smoke alarm anti-interference processing method
CN116777707A (en) * 2023-06-20 2023-09-19 浙江宏远智能科技有限公司 Intelligent community management system and method
CN116777707B (en) * 2023-06-20 2023-12-01 浙江宏远智能科技有限公司 Intelligent community management system and method
CN118799765A (en) * 2024-09-11 2024-10-18 浙江点创信息科技有限公司 A method and system for processing low-altitude images of unmanned aerial vehicles
CN118799765B (en) * 2024-09-11 2025-01-28 浙江点创信息科技有限公司 A method and system for processing low-altitude images of unmanned aerial vehicles
CN119652937A (en) * 2024-11-29 2025-03-18 浙江久成消防设备有限公司 A fire information management system and method based on Internet of Things

Also Published As

Publication number Publication date
CN110874592B (en) 2023-05-16

Similar Documents

Publication Publication Date Title
CN110874592B (en) Forest fire smoke image detection method based on total bounded variation
CN104392468B (en) Moving Object Detection Method Based on Improved Visual Background Extraction
CN103871029B (en) A kind of image enhaucament and dividing method
CN103077539B (en) Motion target tracking method under a kind of complex background and obstruction conditions
CN109409256A (en) A kind of forest rocket detection method based on 3D convolutional neural networks
CN102136059B (en) Video- analysis-base smoke detecting method
CN102982313B (en) The method of Smoke Detection
CN102201146B (en) Fire smoke recognition method in zero-illumination environment based on active infrared video
CN102298781B (en) Motion shadow detection method based on color and gradient characteristics
CN114299417A (en) Multi-target tracking method based on radar-vision fusion
CN104978567B (en) Vehicle checking method based on scene classification
CN110222644A (en) Forest fire smoke detection method based on image segmentation
CN103942557A (en) Coal-mine underground image preprocessing method
CN105046218B (en) A kind of multiple features traffic video smog detection method based on serial parallel processing
CN111027370A (en) Multi-target tracking and behavior analysis detection method
CN109919053A (en) A deep learning vehicle parking detection method based on surveillance video
CN106056139A (en) Forest fire smoke/fog detection method based on image segmentation
CN112258403A (en) Method for extracting suspected smoke area from dynamic smoke
CN114821441B (en) A deep learning-based method for identifying moving targets at airports combined with ADS-B information
CN104183142A (en) Traffic flow statistics method based on image visual processing technology
CN111046827A (en) Video smoke detection method based on convolutional neural network
CN115690564A (en) Outdoor fire smoke image detection method based on Recursive BIFPN network
CN110210428A (en) A kind of smog root node detection method under remote complex environment based on MSER
CN113052055A (en) Smoke detection method based on optical flow improvement and Yolov3
Abujayyab et al. Integrating object-based and pixel-based segmentation for building footprint extraction from satellite images

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