CN107590418A - A kind of video smoke recognition methods based on behavioral characteristics - Google Patents
A kind of video smoke recognition methods based on behavioral characteristics Download PDFInfo
- Publication number
- CN107590418A CN107590418A CN201610532520.6A CN201610532520A CN107590418A CN 107590418 A CN107590418 A CN 107590418A CN 201610532520 A CN201610532520 A CN 201610532520A CN 107590418 A CN107590418 A CN 107590418A
- Authority
- CN
- China
- Prior art keywords
- target
- moving
- smoke
- moving target
- area
- 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.)
- Pending
Links
- 239000000779 smoke Substances 0.000 title claims abstract description 85
- 238000000034 method Methods 0.000 title claims abstract description 40
- 230000003542 behavioural effect Effects 0.000 title abstract 3
- 230000008859 change Effects 0.000 claims description 30
- 230000003068 static effect Effects 0.000 claims description 19
- 239000011159 matrix material Substances 0.000 claims description 11
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 238000001914 filtration Methods 0.000 claims description 5
- 239000000284 extract Substances 0.000 claims description 3
- 230000010339 dilation Effects 0.000 claims 1
- 238000009472 formulation Methods 0.000 claims 1
- 239000000203 mixture Substances 0.000 claims 1
- 238000012545 processing Methods 0.000 abstract description 10
- 238000000605 extraction Methods 0.000 abstract description 9
- 238000013461 design Methods 0.000 abstract description 4
- 238000012360 testing method Methods 0.000 description 9
- 238000001514 detection method Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 230000008901 benefit Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 239000003086 colorant Substances 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000009792 diffusion process Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 229910052739 hydrogen Inorganic materials 0.000 description 1
- 239000001257 hydrogen Substances 0.000 description 1
- 125000004435 hydrogen atom Chemical class [H]* 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000007480 spreading Effects 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a kind of video smoke recognition methods based on behavioral characteristics, this method has real-time and accuracy concurrently.(1) the continuous key frame of input video is extracted first, noise is filtered by median filter, profile repairing improves definition；(2) modeled again by the prospect difference processing of successive frame for moving-target, behavioral characteristics are extracted with VR feature classifiers, moving-target classified with reference to exemplary dynamic feature, and identify wherein smoke target；(3) smoke target finally is carried out into similarity with exemplary smoke image to compare, improves recognition accuracy.It is in place of the characteristic of the present invention：Design VR feature classifiers, target dynamic feature under effective extraction complex scene, the interference such as people, car, trees, flag is removed by target dynamic feature difference, accurately identify wherein smoke target, provide a strong adaptability and have the smog recognition methods of real-time and accuracy rate, the smog early warning available for fire early period of origination under complex environment.
Description
Technical Field
The invention belongs to the technical field of computer image processing, and particularly relates to a video smoke identification method based on dynamic characteristics, which can be applied to a smoke and fire early warning and monitoring software system.
Background
The fire disaster is a multi-disaster and often causes a great amount of casualties and property loss, so that the symptoms of the fire disaster can be found in time, and the early warning can be realized to save huge losses of people and property, thereby generating huge economic and social benefits. However, the traditional manual monitoring has low efficiency, high cost and artificial false reporting and missing reporting; although the fire detection mode based on sensors such as smoke and temperature is improved, the fire detection mode is seriously influenced by factors such as physical environment, monitoring range, hardware cost and the like, so that the problems of high cost, high false alarm rate and the like are caused. For example, when the temperature sensor detects a temperature rise and gives an alarm, a fire usually occurs, the fire has spread, and the time reserved for people is short. In the multiple fire instances, the interval from the generation of the smoke to the occurrence of open fire is a period of time, if the smoke is in the early stage of smoke diffusion, the smoke can be accurately identified and early warned, the hidden fire danger can be put out by concentrating manpower and material resources in time, or valuable time can be won for escaping or fire extinguishing.
The smoke is generated at the initial stage of the fire or before the smoke is diffused, and the smoke diffusion range is far larger than that of the flame, so that the research on the smoke identification technology is more real-time and feasible, and the smoke early warning device is very suitable for early warning of the fire.
One of the core techniques for video-based smoke recognition is to extract key frames from the video stream in real time, and thereby perform deep feature extraction and pattern recognition. The quality of the smoke recognition method is determined according to the effectiveness and the accuracy of feature extraction and feature recognition of the target in the complex environment picture. The most common feature extraction adopted in smoke recognition comprises static feature extraction and dynamic feature extraction, wherein static features comprise contours, colors, gradient values and the like, and the smoke recognition based on the static features usually carries out feature recognition according to features extracted from a key frame. The dynamic characteristics comprise a motion trend, a contour transformation, an expansion trend and the like, and the smoke identification based on the dynamic characteristics is mainly characterized in that the statistical characteristics of continuous multiple frames need to be extracted. The current research considers that the static characteristic calculation amount is small, the efficiency is high, but the anti-interference capability is weak, and the accuracy is low; and the dynamic characteristic calculation amount is large, the prediction instantaneity is low, but the anti-interference capability is strong, and the accuracy is relatively high.
Although smoke has abundant characteristics, it is difficult to accurately identify smoke from a complex environment. The existing identification method and technology have the following main defects:
(1) Insufficient accuracy
The existing smoke detection is a static smoke identification method combined with appearance characteristics or a multi-characteristic fusion method screened according to the minimum value of each channel of a smoke area and a non-smoke area, and has high rate of missing report and false report. In the published "a video smoke recognition method combining color and appearance features" [ patent application No. 201610028532.5], a static smoke recognition method combining color and appearance features presents difficulties for fast expanding smoke recognition. Under a complex scene, the rate of false negative and false positive is higher.
(2) Insufficient real-time performance
The existing smoke identification method based on dynamic characteristics, for example, the issued smoke identification method based on fusion of multiple characteristics of videos [ patent application No. 201410245514.3] has no advantages in speed, and the issued smoke detection and fire early warning method based on track identification [ patent application No. 201510099921.2] is suitable for identifying smoke under the conditions of fluttering and spreading, but has no capability of generating initial expanded smoke; the issued smoke detection and early warning method adopts a method of reserving a cached video image and matching with picture cutting and blocking processing, so that more memory space and CPU processing time are occupied, and the real-time performance of smoke identification is seriously influenced.
(3) Is not highly adaptable
Although some algorithms have better effect under certain conditions, the detection performance of the algorithms is reduced under complex environments.
Disclosure of Invention
The invention aims to provide an effective smoke identification method based on dynamic characteristics aiming at the defects of low accuracy, insufficient real-time performance and low applicability of the existing smoke identification technology.
In order to solve the technical problem, the invention discloses a video smoke identification method based on dynamic characteristics, which comprises the following steps:
(1) Extracting continuous key frames of an input video through a callback function;
(2) Filtering tiny noise points in the key frame through a median filter, and improving the definition of a target through contour repairing;
(3) Method for pairing all moving targets D in complex scene through continuous beta key frame foreground difference i,j Modeling and forming a binary matrix V of the moving target area i,j 。
Dynamically distributing a three-dimensional array SAMPLES [ ] [ NUM _ SAMPLES ] to store the times of continuously detecting the foreground pixel points; if a certain pixel point is detected as a foreground for beta times continuously, a static area is considered to be misjudged as motion, the static area is updated to be a background point, and otherwise, the static area is a moving point:
in the formula: d i,j And the motion point mark is the motion point mark of a certain current pixel point, beta is the threshold value of the motion point mark, and i and j are the number of horizontal and vertical pixel points respectively.
Continuous D i,j Binary matrix V for forming moving target area i,j ；
In the formula: v describes all moving target areas in the picture, 0 is marked as no foreground pixel point, d i,j Are foreground pixels; wherein, V i,j Describing a certain moving target area, d i,j As the center point, wid is the width of the region and heg is the height of the region.
(4) According to V i,j The sequence of appearance numbers the moving object;
P k ＝(v i,j ) Occurrence of the kth (3)
In the formula: p k And identifying the kth occurring moving target, wherein the maximum value of k is controlled by a threshold value N, and N is the maximum number of moving targets which can be tracked simultaneously under the existing condition.
(5) Realizing moving target P according to moving target area coverage between continuous frames k The tracking identifier of (2);
here, P k Ensuring moving object P by tracking identification k The number k is constant during the whole movement, and this step is used for aligning the moving target P k And (4) motion tracking, and accurate moving object classification in step 7.
(6) Calculating the moving target P according to the change of the moving target area of the continuous M frames k The dynamic characteristics of (2):
average target area change:
mean area center position (Δ x, Δ y) change:
average zone width variation:
average zone height variation:
moving index:
expansion index:
Δexp＝Δwid*Δheg(10)
in the formula: r is a radical of hydrogen k Is a moving target P k The number of foreground pixel points (excluding background); m is the continuous frame number, and the change of the continuous M frames is a dynamic characteristic;
the area change (Δ r) and the expansion index (Δ exp) differ in that: the former is to measure the number of foreground pixel points in the moving target P, namely the target foreground area, and the latter is to measure the area of a rectangular area where the moving target is located, namely the target whole area.
(7) The VR feature classifier extracts dynamic features of the moving target and combines the moving target P in continuous M frames k Dynamic characteristics (delta r, delta move, delta exp) of people, vehicles, numbers, flags, smoke, flames and the like, and moving targets are classified by adopting threshold definition; when the gradual change target classification is satisfied, the suspected smoke is obtained, as shown in fig. 2.
In the formula:is the average target area change threshold;a mean region center position change threshold;is the average region width variation threshold;a region height variation threshold;is a motion index threshold;is an expansion index threshold;
the obtained gradual change target J is a suspected smoke target.
(8) Finally, the similarity of the gradual change target J and the typical smoke image is compared by adopting a K-means cooperative algorithm, and the similarity S is larger than a threshold valueThe identification of smoke is deemed complete.
According to the method, the VR dynamic feature classifier is designed, the dynamic features of the object in the complex scene are accurately extracted, the smoke target in the complex scene is accurately identified through the dynamic feature difference of the moving target, and the method has strong applicability, excellent real-time performance and accuracy in the complex scene, and can identify smoke with various different colors and morphological features.
Compared with the prior art, the invention has the following advantages:
(1) The method has stronger applicability and stronger anti-interference capability in a complex scene;
(2) The real-time performance is excellent;
(3) By combining with dynamic characteristics, missing reports and false reports caused by similar backgrounds are effectively removed, and the method has higher accuracy rate for complex backgrounds.
Drawings
FIG. 1 is a flow chart of a video smoke identification method based on dynamic characteristics according to the present invention;
FIG. 2 is a diagram illustrating an example of foreground detection in the present embodiment;
FIG. 3 is a design diagram of a VR feature classifier in accordance with the present invention;
FIG. 4 is an exemplary diagram of the dynamic feature "people-vehicle" of the present embodiment;
FIG. 5 is an exemplary diagram of the dynamic feature "Tree" of the present embodiment;
FIG. 6 is a diagram illustrating the detection effect of the dynamic feature "smoke" in the present embodiment;
Detailed Description
The technical solution of the present invention is further explained with reference to the accompanying drawings and the detailed description.
In order to effectively remove the false alarm and the missed report caused by similar backgrounds and shield the interference caused by the swinging of trees, the movement of people, the flutter of flags, balloons and other similar smoke objects, the invention designs a VR dynamic feature classifier, accurately extracts the dynamic features of the objects in a complex scene, and accurately identifies the smoke target in the complex scene through the dynamic feature difference of the target, and FIG. 1 is a flow chart of the method;
extracting all continuous key frames of the input video through a callback function;
filtering tiny noise points in the key frame through a median filter, wherein the median filter has a very obvious effect on processing common noise points, and the parameter is 0; the definition of the target is improved through contour repairing, the areas within a certain distance are mainly communicated and filled, and the communication distance is determined to be 3 pixels through 300 scene tests.
Modeling the moving target in the moving target area in the figure 2 (a) through continuous beta times of key frame foreground differences to form a binary matrix of the moving target area, wherein the figure 2 (b) is a human-vehicle binary matrix.
Dynamically distributing a three-dimensional array SAMPLES [ ] [ NUM _ SAMPLES ] to store the times of continuously detecting the foreground pixel points; and if a certain pixel point is detected as the foreground for beta times continuously, a static area is considered to be misjudged as motion, the static area is updated as a background point, and otherwise, the static area is a moving point.
Continuous D i,j Binary matrix V for forming moving target area i,j ；
FIG. 2 (b) shows foreground change of moving object in human-vehicle double region, and FIG. 2 (c) shows extracted moving object (human) 1 ) And (vehicle) 1 ) The foreground change and region binary matrix of (2) is used for extracting dynamic characteristics.
According to V i,j The sequence of occurrence numbers the moving object;
P k ＝(v i,j ) Occurrence of the kth (3)
Realizing moving target P according to moving target area coverage between continuous frames k The tracking identifier of (2);
here, P k Ensuring moving object P by tracking identification k The number k is constant during the whole movement, and this step is used for aligning the moving target P k And (7) accurate moving object classification in step 7.
Calculating the moving target P according to the change of the moving target area of the continuous M frames k The dynamic characteristics of (2):
average target area change:
mean area center position (Δ x, Δ y) variation:
average zone width variation:
average zone height variation:
mobility index:
expansion index:
Δexp＝Δwid*Δheg(10)
the area change (Δ r) and the expansion index (Δ exp) differ in that: the former is to measure the number of foreground pixel points in the moving target P, namely the foreground area of the target, and the latter is to measure the area of a rectangular region where the moving target is located, namely the whole region area of the target.
FIG. 3 is a diagram of the design of VR feature classifier, extracting dynamic features of moving targets, combining the moving targets P in continuous M frames k Dynamic characteristics (delta r, delta move, delta exp) of people, vehicles, numbers, flags, smoke, flames and the like, and moving objects are classified by adopting threshold definition.
Typical moving object dynamic characteristics are:
and isAnd isSuspected pedestrians, vehicles, etc.;
FIG. 4 (a), (b), (c) and (d) shows a group of pedestrian objects P 1 The continuous M = 4 uniformly selected key frame pictures, fig. 4 (e) (f) (g) (h) is a foreground change and region binary matrix V of the group of key frames, and the extracted dynamic features are Δ move =1.21 and Δ exp =0.02.
Typical shaking and flickering target dynamic characteristics are as follows:
and is provided withAnd isSuspected trees, flags, and the like;
FIG. 5 (a), (b), (c) and (d) show a group of shaking tree moving targets P 2 Is equal to 4 uniformly selected key frame pictures of the series of M =130 key frame pictures,
FIG. 5 (a), (b), (c) and (d) shows a test swaying tree at 1; 2, static branches for comparison test;
FIG. 5 (e) (f) (g) (h) is a foreground variation and region binary matrix V for the set of keyframes;
the extracted dynamic features Δ move =0.01, Δ exp =0.03. The static background interference is significantly removed.
Typical fade target dynamic characteristics are:
and isAnd isSuspected smoke, flame, etc.;
FIG. 6 (a) (b) (c) (d) is a set of smoke moving objectsMark P 3 The continuous M = 4 uniformly selected key frame pictures of 190;
FIG. 6 shows the test smoke 1 in the frames (a), (b), (c) and (d); 2, shaking trees for interference test;
FIG. 6 (e) (f) (g) (h) is the foreground variation and region binary matrix V for the set of keyframes;
the extracted dynamic features Δ move =0.13 and Δ exp =2.11. And obviously removing dynamic tree interference.
When the classification of the gradual change target is satisfied, the result is suspected smoke, and an obtained gradual change target J is a suspected smoke target, as shown in fig. 6.
After 300 scene tests and the combination of the dynamic characteristic analysis of 1000 moving targets, typical thresholds are as follows:
when the pedestrians and the vehicles continuously have M =130 key frames, the delta exp tends to be stable, and the delta exp is less than 0.05; Δ move is measured as the moving speed, Δ move >0.5;
when the trees and the flags are continuously M =130 key frames, the delta exp tends to be stable, and the delta exp is less than 0.05; Δ move tends to be stable, Δ move <0.05;
when smoke and flame continuously have M =190 key frames, the expansion speed is measured by delta exp, and the delta exp is more than 1.35; Δ move tends to be stable, Δ move <0.05;
performing similarity comparison on the gradual change target J and the typical smoke image by adopting a K-means cooperative algorithm, wherein the similarity S is greater than a threshold valueThe identification of smoke is deemed complete.
The method of this embodiment is used to detect people, cars, trees, and smoke in fig. 2 and fig. 4 to fig. 6 (because of the particularity prescribed by the invention and the patent laws, the figures only show the existence effect of the main target), wherein the image in fig. 2 (a) includes grey cars and black pedestrians, the images in fig. 4 (a to d) include black pedestrians and grey electric cars, the images in fig. 4 (a to d) include grey cars and black pedestrians, the images in fig. 5 (a to d) include green dynamic trees and green static trees, and the images in fig. 6 (a to d) include green dynamic trees, pedestrians and yellow smoke.
According to the processing of the present embodiment, the processing results for fig. 4 (a to d) are obtained, where fig. 4 (e to h) are the results of the dynamic test of pedestrians.
According to the processing of this embodiment, the processing results of fig. 5 (a-d) are obtained, wherein fig. 5 (e-h) are dynamic test results of trees, which can well remove green static trees and backgrounds and keep dynamic areas; therefore, the dynamic feature extraction method has the advantages of accurate dynamic feature extraction, strong anti-interference capability and strong applicability in a complex scene;
according to the process of this embodiment, the results of the process of fig. 6 (a-d) can be obtained, wherein fig. 6 (e-h) are the results of the dynamic test of smoke. Dynamic trees, pedestrians and backgrounds can be well removed, and a dynamic smoke area is reserved; therefore, the invention can be said to combine with the dynamic characteristics, can effectively remove the false negative and false positive brought by the similar background, the dynamic characteristics are extracted accurately, and the smoke recognition precision is higher.
In the real-time video processing, dynamic characteristics of objects such as pedestrians, vehicles, trees, flags and the like tend to be stable when M =130 key frames are continuous, and dynamic characteristics of flame and smoke objects tend to be stable when M =190 key frames are continuous, so that the smoke identification method adopting dynamic characteristic extraction in a complex scene has excellent real-time performance.
While the present invention provides a method for identifying smoke in a video with dynamic characteristics, the above description is only one embodiment of the present invention, and it should be noted that several modifications and embellishments without departing from the principle of the present invention should be regarded as the protection scope of the present invention.
Claims (8)
1. A video smoke identification method based on dynamic characteristics is characterized by comprising the following steps:
(1) Extracting continuous key frames of the input video through a callback function;
(2) Filtering the key frame obtained in the step 1 by a median filter to filter noise points; performing contour repairing to improve the definition of the target and obtain a working key frame;
(3) Carrying out continuous beta times of foreground difference on the working key frames obtained in the step 2 to obtain all moving targets D in the complex scene i,j Identifying;
dynamically allocating the times of continuously detecting the foreground pixel points of the three-dimensional array samples [ ] [ ] ]; if a certain pixel point is detected as a foreground for beta times continuously, a static area is considered to be misjudged as motion, the static area is updated to be a background point, and otherwise, the static area is a moving point:
in the formula: d i,j The number of the horizontal and vertical pixels is the moving point mark of a certain current pixel, beta is the threshold value of the moving point mark, and i and j are the number of the horizontal and vertical pixels respectively.
(4) For the moving target D obtained in the step 3 i,j Identification for adjacent discrimination, adjacent D i,j Binary matrix V for forming moving target area i,j ；
(5) For V obtained in step 4 i,j According to V i,j The sequence of occurrence numbers the moving target, produce the set P of moving target, its formulation is;
P k ＝(v i,j ) the k occurrence (3)
In the formula: p k Identifying the kth occurring moving target, the maximum value of k being controlled by a threshold value N, N being the maximum number of moving targets that can be tracked simultaneously under existing conditions,
(6) Carrying out moving target area coverage between adjacent frames on the moving target P obtained in the step 5 to realize the kth moving target P k The tracking identifier of (2);
(7) For the moving target P obtained in the step 5 k Detecting moving target region of continuous M frames, and calculating P k The dynamic features of (a), comprising: average target area change (Δ r), average region center position (Δ x, Δ y), average region width change (Δ wid), average region height change (Δ heg), mobility index (Δ move), dilation index (Δ exp);
(8) For the moving target obtained in the step 7P k And (3) extracting feature changes by using a VR feature classifier, and classifying the moving target by adopting threshold definition to obtain a gradual change target J as a suspected smoke target.
(9) Comparing the similarity of the suspected smoke target J obtained in the step 8 and the typical smoke image by adopting a K-means collaborative filtering algorithm optimized by a clustering algorithm, wherein the similarity S is greater than a threshold valueThe identification of smoke is deemed complete.
2. The method for video smoke recognition based on dynamic features of claim 1, wherein in step (3), the threshold β for the number of consecutive frames defining the moving foreground and background is chosen to be 50.
3. The method for identifying smoke in a video based on dynamic features as claimed in claim 1, wherein in the step (4), the moving objects D adjacent to each other are selected i,j Formed moving target region binary matrix V i,j Is as follows;
in the formula: v describes the whole moving target area in the picture, 0 is marked as no foreground pixel point, d i,j Are foreground pixels; wherein, V i,j Describing a certain moving target area, d i,j As the center point, wid is the width of the region and heg is the height of the region.
4. The video smoke recognition method based on dynamic features of claim 1, wherein in step (5), the threshold of the maximum number N of moving objects that can be tracked simultaneously under the existing conditions is selected to be 50.
5. The video smoke recognition method based on dynamic features of claim 1, wherein in step (6), the video smoke recognition method based on dynamic featuresRealizing a moving object P k The formula of (2) is as follows:
6. the video smoke identification method based on dynamic characteristics as claimed in claim 1, wherein in the step (7), the moving object P is calculated according to the change of the moving object region of the continuous M frames k The formula is as follows:
average target area change:
mean area center position (Δ x, Δ y) variation:
average zone width variation:
average zone height variation:
moving index:
expansion index:
Δexp＝Δwid*Δheg (10)
in the formula: r is k Is a moving target P k The number of foreground pixel points (excluding background); m is the continuous frame number, and the change of the continuous M frames is a dynamic characteristic; the area change (Δ r) measures the number of foreground pixel points in the moving target P, i.e. the foreground area of the moving target, and the expansion index (Δ exp) measures the area of a rectangular region where the moving target is located, i.e. the whole rectangular region area of the target.
7. The method according to claim 1, wherein in step (8), the VR feature classifier extracts dynamic features of the moving object, and combines the moving object P in consecutive M frames k Dynamic characteristics (delta r, delta move, delta exp) of people, vehicles, numbers, flags, smoke, flames and the like, a threshold value is adopted to classify moving objects, and the formula is as follows:
the VR feature classifier is designed as shown in FIG. 2:
in the formula:is an average target area change threshold;a mean region center position change threshold;is the average region width variation threshold;is a region height variation threshold;is a motion index threshold;is an expansion index threshold; the obtained gradual change target J is suspected smoke.
8. The video smoke identification method based on dynamic characteristics as claimed in claim 1, wherein in the step (9), a K-means collaborative filtering algorithm optimized by a clustering algorithm is adopted for similarity comparison.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610532520.6A CN107590418A (en) | 2016-07-08 | 2016-07-08 | A kind of video smoke recognition methods based on behavioral characteristics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610532520.6A CN107590418A (en) | 2016-07-08 | 2016-07-08 | A kind of video smoke recognition methods based on behavioral characteristics |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107590418A true CN107590418A (en) | 2018-01-16 |
Family
ID=61046379
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610532520.6A Pending CN107590418A (en) | 2016-07-08 | 2016-07-08 | A kind of video smoke recognition methods based on behavioral characteristics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107590418A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109357977A (en) * | 2018-09-29 | 2019-02-19 | 佛山市云米电器科技有限公司 | Smog sub-district domain space based on time connection rechecks method |
| CN113378629A (en) * | 2021-04-27 | 2021-09-10 | 阿里云计算有限公司 | Method and device for detecting abnormal vehicle in smoke discharge |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1852428A (en) * | 2006-05-25 | 2006-10-25 | 浙江工业大学 | Intelligent tunnel safety monitoring apparatus based on omnibearing computer vision |
| CN101441771A (en) * | 2008-12-19 | 2009-05-27 | 中国科学技术大学 | Video fire hazard smoke detecting method based on color saturation degree and movement mode |
| US20090315722A1 (en) * | 2008-06-20 | 2009-12-24 | Billy Hou | Multi-wavelength video image fire detecting system |
| 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 |
| US20110058706A1 (en) * | 2008-05-08 | 2011-03-10 | Utc Fire & Secunity | System and method for video detection of smoke and flame |
| CN102609727A (en) * | 2012-03-06 | 2012-07-25 | 中国人民解放军理工大学工程兵工程学院 | Fire flame detection method based on dimensionless feature extraction |
| CN103020628A (en) * | 2012-11-30 | 2013-04-03 | 北京理工大学 | Smoke detection method based on red, green and blue (RGB) contrast image and target shape |
| CN103971114A (en) * | 2014-04-23 | 2014-08-06 | 天津航天中为数据系统科技有限公司 | Forest fire detection method based on aerial remote sensing |
| CN104794486A (en) * | 2015-04-10 | 2015-07-22 | 电子科技大学 | Video smoke detecting method based on multi-feature fusion |
| US20160063832A1 (en) * | 2014-08-29 | 2016-03-03 | Robert Bosch Gmbh | Fire detection apparatus, method for evaluating at least one fire detection feature and computer program |
| CN105741480A (en) * | 2016-03-17 | 2016-07-06 | 福州大学 | Fire and smoke detection method based on image identification |
-
2016
- 2016-07-08 CN CN201610532520.6A patent/CN107590418A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1852428A (en) * | 2006-05-25 | 2006-10-25 | 浙江工业大学 | Intelligent tunnel safety monitoring apparatus based on omnibearing computer vision |
| US20110058706A1 (en) * | 2008-05-08 | 2011-03-10 | Utc Fire & Secunity | System and method for video detection of smoke and flame |
| US20090315722A1 (en) * | 2008-06-20 | 2009-12-24 | Billy Hou | Multi-wavelength video image fire detecting system |
| CN101441771A (en) * | 2008-12-19 | 2009-05-27 | 中国科学技术大学 | Video fire hazard smoke detecting method based on color saturation degree and movement mode |
| 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 |
| CN102609727A (en) * | 2012-03-06 | 2012-07-25 | 中国人民解放军理工大学工程兵工程学院 | Fire flame detection method based on dimensionless feature extraction |
| CN103020628A (en) * | 2012-11-30 | 2013-04-03 | 北京理工大学 | Smoke detection method based on red, green and blue (RGB) contrast image and target shape |
| CN103971114A (en) * | 2014-04-23 | 2014-08-06 | 天津航天中为数据系统科技有限公司 | Forest fire detection method based on aerial remote sensing |
| US20160063832A1 (en) * | 2014-08-29 | 2016-03-03 | Robert Bosch Gmbh | Fire detection apparatus, method for evaluating at least one fire detection feature and computer program |
| CN104794486A (en) * | 2015-04-10 | 2015-07-22 | 电子科技大学 | Video smoke detecting method based on multi-feature fusion |
| CN105741480A (en) * | 2016-03-17 | 2016-07-06 | 福州大学 | Fire and smoke detection method based on image identification |
Non-Patent Citations (4)
| Title |
|---|
| WEI YE等: "Dynamic texture based smoke detection using Surfacelet transform and HMT model", 《FIRE SAFETY JOURNAL》, vol. 73, pages 91 - 101, XP029211362, DOI: 10.1016/j.firesaf.2015.03.001 * |
| YASHAR DELDJOO等: "A novel fuzzy-based smoke detection system using dynamic and static smoke features", 《2015 23RD IRANIAN CONFERENCE ON ELECTRICAL ENGINEERING》, pages 729 - 733 * |
| 史玉坤等: "利用多特征判别的烟雾检测方法研究", 《信号处理》, vol. 31, no. 10, pages 1336 - 1341 * |
| 郭炜强: "基于运动特征的林火烟雾图像检测技术研究", 《中国优秀硕士学位论文全文数据库 (信息科技辑)》, vol. 2012, no. 09, pages 138 - 473 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109357977A (en) * | 2018-09-29 | 2019-02-19 | 佛山市云米电器科技有限公司 | Smog sub-district domain space based on time connection rechecks method |
| CN109357977B (en) * | 2018-09-29 | 2021-04-02 | 佛山市云米电器科技有限公司 | Smoke subregion space rechecking method based on time contact |
| CN113378629A (en) * | 2021-04-27 | 2021-09-10 | 阿里云计算有限公司 | Method and device for detecting abnormal vehicle in smoke discharge |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | 3D parallel fully convolutional networks for real-time video wildfire smoke detection | |
| CN106845443B (en) | Video flame detection method based on multi-feature fusion | |
| CN103824070B (en) | A kind of rapid pedestrian detection method based on computer vision | |
| Zhao et al. | SVM based forest fire detection using static and dynamic features | |
| CN109145689A (en) | A kind of robot fire detection method | |
| Li et al. | Abandoned objects detection using double illumination invariant foreground masks | |
| CN109918971B (en) | Method and device for detecting number of people in monitoring video | |
| CN103400111A (en) | Method for detecting fire accident on expressway or in tunnel based on video detection technology | |
| CN110490043A (en) | A kind of forest rocket detection method based on region division and feature extraction | |
| Wang et al. | A new fire detection method using a multi-expert system based on color dispersion, similarity and centroid motion in indoor environment | |
| CN110874592A (en) | Forest fire smoke image detection method based on total bounded variation | |
| CN114202646A (en) | Infrared image smoking detection method and system based on deep learning | |
| Dimitropoulos et al. | Flame detection for video-based early fire warning systems and 3D visualization of fire propagation | |
| CN104463869A (en) | Video flame image composite recognition method | |
| CN103400113A (en) | Method for detecting pedestrian on expressway or in tunnel based on image processing | |
| Chen et al. | Fire detection using spatial-temporal analysis | |
| CN106096523A (en) | The detection of a kind of Moving Object in Video Sequences and Deviant Behavior parser | |
| Khondaker et al. | Computer vision-based early fire detection using enhanced chromatic segmentation and optical flow analysis technique. | |
| CN114885119A (en) | Intelligent monitoring alarm system and method based on computer vision | |
| Su et al. | A new local-main-gradient-orientation HOG and contour differences based algorithm for object classification | |
| CN108090397A (en) | Pedestrian detecting system based on infrared image | |
| CN106815567B (en) | Flame detection method and device based on video | |
| CN110991245A (en) | Real-time smoke detection method based on deep learning and optical flow method | |
| CN107590418A (en) | A kind of video smoke recognition methods based on behavioral characteristics | |
| CN107729811B (en) | Night flame detection method based on scene modeling |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20190129 Address after: 110 000 No. 64 North Street of Yellow River, Huanggu District, Shenyang City, Liaoning Province Applicant after: Yin Hang Address before: 110000 Zhengyi Road 27-3, Deyi Development Zone, Shenbei New District, Shenyang City, Liaoning Province Applicant before: Yin Hang Applicant before: Cao Guoqiang |
|
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180116 |