CN111639610A - Fire recognition method and system based on deep learning - Google Patents
Fire recognition method and system based on deep learning Download PDFInfo
- Publication number
- CN111639610A CN111639610A CN202010491922.2A CN202010491922A CN111639610A CN 111639610 A CN111639610 A CN 111639610A CN 202010491922 A CN202010491922 A CN 202010491922A CN 111639610 A CN111639610 A CN 111639610A
- Authority
- CN
- China
- Prior art keywords
- fire
- flame
- image
- module
- neural network
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention discloses a fire recognition method and a fire recognition system based on deep learning, which are characterized in that based on deep learning technology, the existing fire images are analyzed and trained to obtain a neural network model, the images acquired by visible light real-time video are subjected to fire recognition, the characteristics of a single-frame image and the characteristics of a multi-frame image time sequence are combined, weather and geographic data information are combined, fire judgment, influence analysis and prevention and control assistance are comprehensively carried out, the influence of environmental factors on recognition results is avoided, and the recognition accuracy is greatly improved.
Description
Technical Field
The invention relates to the field of fire recognition, in particular to a fire recognition method and system based on deep learning.
Background
The fire fighting safety system usually adopts an image recognition technology to recognize fire, and the principle of recognizing fire by using the image recognition technology generally comprises the following steps: 1. target segmentation, mainly used for finding the position where there may be fire in the digital image, through means such as edge detection, feature space clustering, etc. to identify, 2, feature extraction, mainly used for extracting visual features of the firework target, such as color, shape, texture, spatial relationship, etc.; 3. and comprehensively judging and extracting the characteristics of the region obtained by target segmentation, and judging the similarity between the characteristics and the prior characteristics of the target object.
The existing fire recognition system, such as the infrared light video fire recognition system disclosed in patent document CN201120213456.8, adopts infrared thermal imaging technology to recognize fire, but the infrared thermal imaging technology has great application limitation under different meteorological conditions, and the deployment cost in application is high, which is not beneficial to large-scale popularization; at present, some technologies based on a neural network are also available in fire recognition systems, for example, a forest fire early warning method and system based on deep learning in patent document CN201810419169.9, which can automatically recognize fire through learning of the neural network, but are prone to high false recognition under certain specific conditions, such as: images of sunlight, canula aurea, fire clouds and the like are mistakenly identified as fire conditions; the cloud and fog are mistakenly identified as the situation of fire dense smoke, and the reliability of fire identification cannot be ensured.
Therefore, a fire recognition method and a fire recognition system based on deep learning are needed, and the problems that the existing fire recognition method and system are large in use limitation, high in arrangement cost and inaccurate in recognition result can be solved.
Disclosure of Invention
The invention aims to provide a fire recognition method and a fire recognition system based on deep learning, and aims to solve the problems that the existing fire recognition method and the existing fire recognition system are high in use limitation, high in arrangement cost and inaccurate in recognition result.
In order to achieve the purpose, the invention provides the following scheme:
the invention provides a fire recognition method based on deep learning, which comprises the following steps:
s1, collecting a fire image sample by using collection equipment, labeling information of dense smoke and flame in the image sample to obtain a training sample, training the training sample by using a computer to establish a neural network model, carrying out inference and identification on a real-time visible light video, and identifying whether the dense smoke and/or the flame in the visible light video meet a fire standard or not;
s2, establishing a convolutional neural network by using a computer to extract image features of the information of the smoke and the flame in the single picture, establishing a cyclic neural network by using the computer to extract the image features of the information of the smoke and the flame in the video multi-picture time sequence, and finally performing reasoning identification on the image features of the single picture and the image features of the video multi-picture time sequence by using a full-connection neural network to output fire information and correct the fire information;
and S3, collecting weather and geographic data through collection equipment, and comprehensively distinguishing, analyzing influence and assisting in prevention and control of the fire by combining the weather and the geographic data through a computer.
Preferably, S1 specifically includes the following steps:
(1) acquiring a fire image sample through network picture, video crawling and field video shooting, preprocessing the image sample in a uniform format and size and sample distribution and arrangement, and enhancing by adopting a general data enhancement method to obtain a sample image set;
(2) marking the images in the sample image set in a manual mode, wherein the marked contents are whether the smoke and the flame exist or not and the specific positions of the smoke and the flame to form a training sample set;
(3) disorganizing images in the training sample set, taking a part of images as a training data set, taking the rest of images as a test data set, and carrying out neural network model training in a random mode, wherein an optimization function adopts an Adam algorithm;
(4) the neural network model is converged through multiple times of training, high-dimensional characteristics of the smoke and flame patterns are learned to be recognized, the position of a target object in the patterns is detected, training completion parameters and the model are stored, and real-time video reasoning recognition is carried out;
(5) and (4) bringing the image which generates the false recognition into the next round of training, and performing repeated iteration to enhance the capability of the model for resisting the false recognition.
More preferably, the method for inferential identification in step (4) is:
a. judging whether the fire standard is achieved according to the intensity, duration and position information of the dense smoke, and reporting the fire if the fire standard is achieved;
b. judging whether the flame independent judgment fire standard is met or not according to the intensity, duration and position information of the flame, and reporting the fire if the flame independent judgment fire standard is met;
c. comprehensively judging whether the fire standard is met according to the intensity, duration and position information of the dense smoke and the flame, and reporting the fire if the fire standard is met.
Preferably, S2 specifically includes the following steps:
(1) extracting the image characteristics of the dense smoke and the flame in the single image through the established deep convolution neural network, wherein the extracted information comprises the color, the state, the texture, the size, the position, the illumination and the density of the dense smoke and the flame, and is used for the next-level network identification;
(2) extracting image characteristics of the dense smoke and the flame in the video multi-picture time sequence through the established deep circulation neural network, wherein the extracted information comprises colors, states, textures, sizes, positions, illumination and density of the dense smoke and the flame, and is used for the next-stage network identification;
(3) and based on the extracted single-picture image characteristics and the extracted video multi-picture time sequence image characteristics, adopting a full-connection neural network to carry out reasoning identification and outputting fire information.
More preferably, in the step (1), the convolution layer adopts a VGG-like mode, convolution kernels adopt 1 × 1 and 3 × 3 more, the depth of the neural network is increased by utilizing small convolution, and a 3 × 1 and 1 × 3 convolution kernel optimization mode is used at the same time; obtaining new characteristics by adopting a full patch mode under 2 x 2; adopting a Relu activation function, wherein the expression is as follows: y isi＝Max(0,xi) Adopting a LeakyRelu activation function, wherein the expression is as follows:
and a universal L2 regularization penalty mode is adopted to avoid overfitting, the probability of 0.5 is adopted for discarding, and a BatchNormal layer is introduced into a part of convolution layers to further generate a bloom model.
More preferably, in step (2), a recurrent neural network is established based on the output features of the convolutional neural network, and the recurrent neural network comprises a transverse multistage time sequence and a longitudinal plurality of hidden layers; the RNN mode of the long-short term memory network is adopted to solve the gradient disappearance problem, and the expression is as follows:
ft＝σ(Wfxt+Ufht-1+bf)ft＝σ(Wfxt+Ufht-1+bf)
it＝σ(Wixt+Uiht-1+bi)it＝σ(Wixt+Uiht-1+bi)
ot＝σ(Woxt+Uoht-1+
bo)ot＝σ(Woxt+Uoht-1+bo)
ct～＝tanh(Wcxt+Ucht-1+bc)
the activation function adopts sigmoid and tanh functions, and is respectively applied to threshold output and data activation output.
More preferably, in step (3), the identification inference process is: judging whether an object concerned by people exists, carrying out Classication Classification judgment on the position with the object, identifying whether the region is dense smoke or flame, reasoning confidence coefficient of identifying the dense smoke or the flame output by each region, representing the credibility of the identification result, reasoning and outputting position and size correction information, and finally correcting the position and the size of the identified dense smoke and flame region.
Corresponding to the identification method, the invention also provides a fire identification system based on deep learning, which comprises an image acquisition module, wherein the image acquisition module is connected with an identification module and is used for identifying dense smoke and flame of the image acquired by the image acquisition module;
the identification module is connected with a discrimination module and is used for discriminating dense smoke and/or flame of the identification result;
the identification module is connected with an alarm module and used for giving out fire alarm according to the judgment result;
the extraction module is connected with the image acquisition module and is used for extracting the characteristic information of the image;
the comprehensive reasoning system comprises a comprehensive reasoning module and is characterized by further comprising a climate collecting module and a geography collecting module, wherein the extracting module, the climate collecting module and the geography collecting module are all connected with the comprehensive reasoning module, and the comprehensive reasoning module is used for comprehensively judging to obtain a fire recognition result according to image characteristics, climate data information and geography data information.
Compared with the prior art, the invention has the following beneficial technical effects:
1. according to the fire identification method and system based on deep learning, the visible light video is adopted for image acquisition, compared with an infrared thermal imaging technology, the method and system are lower in arrangement cost, do not relate to the problem that weather influences thermal imaging, and are wider in application range.
2. According to the fire condition identification method and system based on deep learning, the single-picture image characteristics and the video multi-picture time sequence image characteristics are combined, and comprehensive judgment of dense smoke and flame is performed, so that the problem of false identification caused by environmental factors is effectively solved, and the identification precision is greatly improved.
3. The fire identification method and the fire identification system based on deep learning provided by the invention are combined with comprehensive judgment of climate information and geographic information, so that the influence of natural environment on identification is avoided, and the identification accuracy is further improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a flowchart of S1 in the method and system for fire recognition based on deep learning according to the present invention
Fig. 2 is a flow chart of S2 in the method and system for fire recognition based on deep learning according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a fire recognition method and a fire recognition system based on deep learning, and aims to solve the problems that the existing fire recognition method and system are high in use limitation, high in arrangement cost and inaccurate in recognition result.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Example 1:
the embodiment provides a fire recognition method based on deep learning, which comprises the following steps:
s1, collecting a fire image sample by using collection equipment, labeling information of dense smoke and flame in the image sample to obtain a training sample, training the training sample by using a computer to establish a neural network model, carrying out inference and identification on a real-time visible light video, and identifying whether the dense smoke and/or the flame in the visible light video meet a fire standard or not;
s2, establishing a convolutional neural network by using a computer to extract image features of the information of the smoke and the flame in the single picture, establishing a cyclic neural network by using the computer to extract the image features of the information of the smoke and the flame in the video multi-picture time sequence, and finally performing reasoning identification on the image features of the single picture and the image features of the video multi-picture time sequence by using a full-connection neural network to output fire information and correct the fire information;
and S3, collecting weather and geographic data through collection equipment, and comprehensively distinguishing, analyzing influence and assisting in prevention and control of the fire by combining the weather and the geographic data through a computer.
Specifically, as shown in fig. 1, S1 specifically includes the following steps:
(1) acquiring a fire image sample through network picture, video crawling and field video shooting, preprocessing the image sample in a uniform format and size and sample distribution and arrangement, and enhancing by adopting a general data enhancement method to obtain a sample image set;
(2) marking the images in the sample image set in a manual mode, wherein the marked contents are whether the smoke and the flame exist or not and the specific positions of the smoke and the flame to form a training sample set;
(3) disorganizing images in the training sample set, taking a part of images as a training data set, taking the rest of images as a test data set, and carrying out neural network model training in a random mode, wherein an optimization function adopts an Adam algorithm;
(4) the neural network model is converged through multiple times of training, high-dimensional characteristics of the smoke and flame patterns are learned to be recognized, the position of a target object in the patterns is detected, training completion parameters and the model are stored, and real-time video reasoning recognition is carried out;
(5) and (4) bringing the image which generates the false recognition into the next round of training, and performing repeated iteration to enhance the capability of the model for resisting the false recognition.
Further, the method for reasoning and identifying in the step (4) comprises the following steps:
a. judging whether the fire standard is achieved according to the intensity, duration and position information of the dense smoke, and reporting the fire if the fire standard is achieved;
b. judging whether the flame independent judgment fire standard is met or not according to the intensity, duration and position information of the flame, and reporting the fire if the flame independent judgment fire standard is met;
c. comprehensively judging whether the fire standard is met according to the intensity, duration and position information of the dense smoke and the flame, and reporting the fire if the fire standard is met.
Further, as shown in fig. 2, S2 specifically includes the following steps:
(1) extracting the image characteristics of the dense smoke and the flame in the single image through the established deep convolution neural network, wherein the extracted information comprises the color, the state, the texture, the size, the position, the illumination and the density of the dense smoke and the flame, and is used for the next-level network identification;
(2) extracting image characteristics of the dense smoke and the flame in the video multi-picture time sequence through the established deep circulation neural network, wherein the extracted information comprises colors, states, textures, sizes, positions, illumination and density of the dense smoke and the flame, and is used for the next-stage network identification;
(3) and based on the extracted single-picture image characteristics and the extracted video multi-picture time sequence image characteristics, adopting a full-connection neural network to carry out reasoning identification and outputting fire information.
Go toStep (1), the convolution layer adopts a VGG-like mode, convolution kernels mostly adopt 1 × 1 and 3 × 3, the depth of a neural network is increased by utilizing small convolution, and a 3 × 1 and 1 × 3 convolution kernel optimization mode is used at the same time; obtaining new characteristics by adopting a full patch mode under 2 x 2; adopting a Relu activation function, wherein the expression is as follows: y isi＝Max(0,xi) Adopting a LeakyRelu activation function, wherein the expression is as follows:
and a universal L2 regularization penalty mode is adopted to avoid overfitting, the probability of 0.5 is adopted for discarding, and a BatchNormal layer is introduced into a part of convolution layers to further generate a bloom model.
Further, in the step (2), a cyclic neural network is established based on the output characteristics of the convolutional neural network, and comprises a transverse multistage time sequence and a longitudinal plurality of hidden layers; the RNN mode of the long-short term memory network is adopted to solve the gradient disappearance problem, and the expression is as follows:
ft＝σ(Wfxt+Ufht-1+bf)ft＝σ(Wfxt+Ufht-1+bf)
it＝σ(Wixt+Uiht-1+bi)it＝σ(Wixt+Uiht-1+bi)
ot＝σ(Woxt+Uoht-1+
bo)ot＝σ(Woxt+Uoht-1+bo)
ct～＝tanh(Wcxt+Ucht-1+bc)
the activation function adopts sigmoid and tanh functions, and is respectively applied to threshold output and data activation output.
Further, in step (3), the identification inference process is as follows: judging whether an object concerned by people exists, carrying out Classication Classification judgment on the position with the object, identifying whether the region is dense smoke or flame, reasoning confidence coefficient of identifying the dense smoke or the flame output by each region, representing the credibility of the identification result, reasoning and outputting position and size correction information, and finally correcting the position and the size of the identified dense smoke and flame region.
As an application of the identification method, the embodiment further provides a fire identification system based on deep learning, which comprises an image acquisition module, wherein the image acquisition module is connected with an identification module and is used for identifying dense smoke and flame of an image acquired by the image acquisition module; the identification module is connected with a discrimination module and is used for discriminating dense smoke and/or flame of the identification result; the identification module is connected with an alarm module and used for giving out fire alarm according to the judgment result; the extraction module is connected with the image acquisition module and is used for extracting the characteristic information of the image; the comprehensive reasoning module is used for comprehensively distinguishing according to the image characteristics, the climate data information and the geographic data information to obtain a fire recognition result.
The fire recognition system based on deep learning provided by the embodiment has the working principle that: the image acquisition module sends acquired image information to the recognition module, the recognition module automatically recognizes the image information according to the neural network model and sends a recognition result to the judgment module, the judgment module judges according to a preset fire standard, and the alarm module is controlled to give an alarm if the fire standard is met; meanwhile, the extraction module extracts the characteristics of the image acquired by the image acquisition module and sends the image to the comprehensive reasoning module, and the comprehensive reasoning module combines the weather information and the geographic data acquired by the climate acquisition module and the geographic acquisition module to comprehensively study and judge the fire, analyze the influence and assist in prevention and control so as to obtain a processing scheme.
The principle and the implementation mode of the invention are explained by applying specific examples, and the description of the above examples is only used for helping understanding the method and the core idea of the invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In summary, this summary should not be construed to limit the present invention.
Claims (8)
1. A fire behavior recognition method based on deep learning is characterized in that: the method comprises the following steps:
s1, collecting a fire image sample by using collection equipment, labeling information of dense smoke and flame in the image sample to obtain a training sample, training the training sample by using a computer to establish a neural network model, carrying out inference and identification on a real-time visible light video, and identifying whether the dense smoke and/or the flame in the visible light video meet a fire standard or not;
s2, establishing a convolutional neural network by using a computer to extract image features of the information of the smoke and the flame in the single picture, establishing a cyclic neural network by using the computer to extract the image features of the information of the smoke and the flame in the video multi-picture time sequence, and finally performing reasoning identification on the image features of the single picture and the image features of the video multi-picture time sequence by using a full-connection neural network to output fire information and correct the fire information;
and S3, collecting weather and geographic data through collection equipment, and comprehensively distinguishing, analyzing influence and assisting in prevention and control of the fire by combining the weather and the geographic data through a computer.
2. The fire recognition method based on deep learning of claim 1, wherein: s1 specifically includes the following steps:
(1) acquiring a fire image sample through network picture, video crawling and field video shooting, preprocessing the image sample in a uniform format and size and sample distribution and arrangement, and enhancing by adopting a general data enhancement method to obtain a sample image set;
(2) marking the images in the sample image set in a manual mode, wherein the marked contents are whether the smoke and the flame exist or not and the specific positions of the smoke and the flame to form a training sample set;
(3) disorganizing images in the training sample set, taking a part of images as a training data set, taking the rest of images as a test data set, and carrying out neural network model training in a random mode, wherein an optimization function adopts an Adam algorithm;
(4) the neural network model is converged through multiple times of training, high-dimensional characteristics of the smoke and flame patterns are learned to be recognized, the position of a target object in the patterns is detected, training completion parameters and the model are stored, and real-time video reasoning recognition is carried out;
(5) and (4) bringing the image which generates the false recognition into the next round of training, and performing repeated iteration to enhance the capability of the model for resisting the false recognition.
3. The fire recognition method based on deep learning of claim 2, wherein: the reasoning and identifying method in the step (4) comprises the following steps:
a. judging whether the fire standard is achieved according to the intensity, duration and position information of the dense smoke, and reporting the fire if the fire standard is achieved;
b. judging whether the flame independent judgment fire standard is met or not according to the intensity, duration and position information of the flame, and reporting the fire if the flame independent judgment fire standard is met;
c. comprehensively judging whether the fire standard is met according to the intensity, duration and position information of the dense smoke and the flame, and reporting the fire if the fire standard is met.
4. The fire recognition method based on deep learning of claim 1, wherein: s2 specifically includes the following steps:
(1) extracting the image characteristics of the dense smoke and the flame in the single image through the established deep convolution neural network, wherein the extracted information comprises the color, the state, the texture, the size, the position, the illumination and the density of the dense smoke and the flame, and is used for the next-level network identification;
(2) extracting image characteristics of the dense smoke and the flame in the video multi-picture time sequence through the established deep circulation neural network, wherein the extracted information comprises colors, states, textures, sizes, positions, illumination and density of the dense smoke and the flame, and is used for the next-stage network identification;
(3) and based on the extracted single-picture image characteristics and the extracted video multi-picture time sequence image characteristics, adopting a full-connection neural network to carry out reasoning identification and outputting fire information.
5. The fire recognition method based on deep learning of claim 4, wherein: in the step (1), the convolution layer adopts a mode similar to VGG, and the convolution layer is coiledThe most of the kernels adopt 1 × 1 and 3 × 3, the depth of the neural network is increased by utilizing small convolution, and a 3 × 1, 1 × 3 convolution kernel optimization mode is used; obtaining new characteristics by adopting a full patch mode under 2 x 2; adopting a Relu activation function, wherein the expression is as follows: y isi＝Max(0,xi) Adopting a LeakyRelu activation function, wherein the expression is as follows:
and a universal L2 regularization penalty mode is adopted to avoid overfitting, the probability of 0.5 is adopted for discarding, and a BatchNormal layer is introduced into a part of convolution layers to further generate a bloom model.
6. The fire recognition method based on deep learning of claim 4, wherein: in the step (2), a cyclic neural network is established based on the output characteristics of the convolutional neural network, and comprises a transverse multistage time sequence and a plurality of longitudinal hidden layers; the RNN mode of the long-short term memory network is adopted to solve the gradient disappearance problem, and the expression is as follows:
ft＝σ(Wfxt+Ufht-1+bf)ft＝σ(Wfxt+Ufht-1+bf)
it＝σ(Wixt+Uiht-1+bi)it＝σ(Wixt+Uiht-1+bi)
ot＝σ(Woxt+Uoht-1+
bo)ot＝σ(Woxt+Uoht-1+bo)
ct～＝tanh(Wcxt+Ucht-1+bc)
the activation function adopts sigmoid and tanh functions, and is respectively applied to threshold output and data activation output.
7. The fire recognition method based on deep learning of claim 4, wherein: in the step (3), the identification reasoning process is as follows: judging whether an object concerned by people exists, carrying out Classication Classification judgment on the position with the object, identifying whether the region is dense smoke or flame, reasoning confidence coefficient of identifying the dense smoke or the flame output by each region, representing the credibility of the identification result, reasoning and outputting position and size correction information, and finally correcting the position and the size of the identified dense smoke and flame region.
8. The utility model provides a condition of a fire identification system based on deep learning which characterized in that: the device comprises an image acquisition module, wherein the image acquisition module is connected with an identification module and is used for identifying dense smoke and flame of an image acquired by the image acquisition module;
the identification module is connected with a discrimination module and is used for discriminating dense smoke and/or flame of the identification result;
the identification module is connected with an alarm module and used for giving out fire alarm according to the judgment result;
the extraction module is connected with the image acquisition module and is used for extracting the characteristic information of the image;
the comprehensive reasoning system comprises a comprehensive reasoning module and is characterized by further comprising a climate collecting module and a geography collecting module, wherein the extracting module, the climate collecting module and the geography collecting module are all connected with the comprehensive reasoning module, and the comprehensive reasoning module is used for comprehensively judging to obtain a fire recognition result according to image characteristics, climate data information and geography data information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010491922.2A CN111639610A (en) | 2020-06-03 | 2020-06-03 | Fire recognition method and system based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010491922.2A CN111639610A (en) | 2020-06-03 | 2020-06-03 | Fire recognition method and system based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111639610A true CN111639610A (en) | 2020-09-08 |
Family
ID=72330314
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010491922.2A Pending CN111639610A (en) | 2020-06-03 | 2020-06-03 | Fire recognition method and system based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111639610A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112216055A (en) * | 2020-09-24 | 2021-01-12 | 佛山市天然气高压管网有限公司 | Fire early warning detection method and system for gas field station |
| CN112801148A (en) * | 2021-01-14 | 2021-05-14 | 西安电子科技大学 | Fire recognition and positioning system and method based on deep learning |
| CN113743328A (en) * | 2021-09-08 | 2021-12-03 | 无锡格林通安全装备有限公司 | Flame detection method and device based on long-term and short-term memory model |
| CN113971861A (en) * | 2021-11-26 | 2022-01-25 | 安徽理工大学 | Intelligent fire monitoring system based on Internet of things |
| CN117558116A (en) * | 2024-01-11 | 2024-02-13 | 山东奥深智能工程有限公司 | Fire control early warning system based on big data |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090315722A1 (en) * | 2008-06-20 | 2009-12-24 | Billy Hou | Multi-wavelength video image fire detecting system |
| CN108985376A (en) * | 2018-07-17 | 2018-12-11 | 东北大学 | It is a kind of based on convolution-Recognition with Recurrent Neural Network rotary kiln sequence operating mode's switch method |
| CN109920191A (en) * | 2019-01-16 | 2019-06-21 | 深圳壹账通智能科技有限公司 | Fire alarm method, device, computer equipment and storage medium |
| CN110443969A (en) * | 2018-05-03 | 2019-11-12 | 中移(苏州)软件技术有限公司 | A kind of fire point detecting method, device, electronic equipment and storage medium |
| CN110503021A (en) * | 2019-08-19 | 2019-11-26 | 温州大学 | Fire hazard smoke detecting method based on time compression track characteristic identification |
-
2020
- 2020-06-03 CN CN202010491922.2A patent/CN111639610A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090315722A1 (en) * | 2008-06-20 | 2009-12-24 | Billy Hou | Multi-wavelength video image fire detecting system |
| CN110443969A (en) * | 2018-05-03 | 2019-11-12 | 中移(苏州)软件技术有限公司 | A kind of fire point detecting method, device, electronic equipment and storage medium |
| CN108985376A (en) * | 2018-07-17 | 2018-12-11 | 东北大学 | It is a kind of based on convolution-Recognition with Recurrent Neural Network rotary kiln sequence operating mode's switch method |
| CN109920191A (en) * | 2019-01-16 | 2019-06-21 | 深圳壹账通智能科技有限公司 | Fire alarm method, device, computer equipment and storage medium |
| CN110503021A (en) * | 2019-08-19 | 2019-11-26 | 温州大学 | Fire hazard smoke detecting method based on time compression track characteristic identification |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112216055A (en) * | 2020-09-24 | 2021-01-12 | 佛山市天然气高压管网有限公司 | Fire early warning detection method and system for gas field station |
| CN112216055B (en) * | 2020-09-24 | 2021-05-11 | 佛山市天然气高压管网有限公司 | Fire early warning detection method and system for gas field station |
| CN112801148A (en) * | 2021-01-14 | 2021-05-14 | 西安电子科技大学 | Fire recognition and positioning system and method based on deep learning |
| CN113743328A (en) * | 2021-09-08 | 2021-12-03 | 无锡格林通安全装备有限公司 | Flame detection method and device based on long-term and short-term memory model |
| CN113971861A (en) * | 2021-11-26 | 2022-01-25 | 安徽理工大学 | Intelligent fire monitoring system based on Internet of things |
| CN117558116A (en) * | 2024-01-11 | 2024-02-13 | 山东奥深智能工程有限公司 | Fire control early warning system based on big data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111639610A (en) | Fire recognition method and system based on deep learning | |
| Tao et al. | Smoke detection based on deep convolutional neural networks | |
| CN108510467B (en) | SAR image target identification method based on depth deformable convolution neural network | |
| Han et al. | Object detection in optical remote sensing images based on weakly supervised learning and high-level feature learning | |
| CN107016357B (en) | Video pedestrian detection method based on time domain convolutional neural network | |
| CN107480620B (en) | Remote sensing image automatic target identification method based on heterogeneous feature fusion | |
| CN105069481B (en) | Natural scene multiple labeling sorting technique based on spatial pyramid sparse coding | |
| CN109740639B (en) | Wind cloud satellite remote sensing image cloud detection method and system and electronic equipment | |
| CN107247930A (en) | SAR image object detection method based on CNN and Selective Attention Mechanism | |
| CN111814638B (en) | Security scene flame detection method based on deep learning | |
| CN111680632A (en) | Smoke and fire detection method and system based on deep learning convolutional neural network | |
| CN110490043A (en) | A kind of forest rocket detection method based on region division and feature extraction | |
| Ahmad et al. | Overhead view person detection using YOLO | |
| CN113011357A (en) | Depth fake face video positioning method based on space-time fusion | |
| CN111814690B (en) | Target re-identification method, device and computer readable storage medium | |
| Fang et al. | Camouflage people detection via strong semantic dilation network | |
| CN111046827A (en) | Video smoke detection method based on convolutional neural network | |
| CN113128481A (en) | Face living body detection method, device, equipment and storage medium | |
| CN113469050A (en) | Flame detection method based on image subdivision classification | |
| CN114821014A (en) | Multi-mode and counterstudy-based multi-task target detection and identification method and device | |
| Shu et al. | Center-point-guided proposal generation for detection of small and dense buildings in aerial imagery | |
| Premarathna et al. | An elephant detection system to prevent human-elephant conflict and tracking of elephant using deep learning | |
| CN113221655B (en) | Face spoofing detection method based on feature space constraint | |
| CN111291712B (en) | Forest fire recognition method and device based on interpolation CN and capsule network | |
| CN110334703B (en) | Ship detection and identification method in day and night image |
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 |