CN114298167A - Tunnel fire detection method based on YOLO neural network - Google Patents
Tunnel fire detection method based on YOLO neural network Download PDFInfo
- Publication number
- CN114298167A CN114298167A CN202111509268.4A CN202111509268A CN114298167A CN 114298167 A CN114298167 A CN 114298167A CN 202111509268 A CN202111509268 A CN 202111509268A CN 114298167 A CN114298167 A CN 114298167A
- Authority
- CN
- China
- Prior art keywords
- training
- model
- fire
- data set
- 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
Landscapes
- Alarm Systems (AREA)
- Fire-Detection Mechanisms (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the field of fire detection, in particular to a tunnel fire detection method based on a YOLO neural network, which comprises the following steps: in the invention, a fire picture data set is produced; establishing a training model based on the data set; marking the fire picture in the data set by using roboflow; inputting the training model into a training model to train the model; the trained model is connected to a monitoring camera of the expressway for real-time monitoring and early warning, a training result can be simulated, the result shows that the mAP reaches more than 0.9, the model has a good identification effect on the tunnel fire incident, and the monitoring of the fire is effectively improved.
Description
Technical Field
The invention belongs to the field of fire detection, and particularly relates to a tunnel fire detection method based on a YOLO neural network.
Background
Among the safety accidents occurring in the road tunnel, fire is the most harmful one. Once a fire disaster occurs in the tunnel, the fire disaster is very difficult to extinguish, and serious economic loss and casualties are easily caused. In order to prevent fire accidents in road tunnels, various early fire detection technologies have been proposed. With the great popularization of video monitoring, the automatic fire detection and early warning based on image video has great research significance and use value. The traditional flame detection is generally to design a feature extraction algorithm according to prior knowledge, manually extract flame dynamic or static features, and then perform flame identification.
According to the traditional flame detection method based on computer vision, it is difficult and time-consuming to design an artificial feature extraction algorithm according to priori knowledge, and when facing different complex environments and variable flame types, the generalization capability of the traditional flame detection method is often insufficient, so that the problem of inaccurate fire video detection in the complex environments is solved.
Disclosure of Invention
Aiming at the problems, the invention provides a tunnel fire detection method based on a YOLO neural network, which comprises the following steps:
making a fire picture training data set;
marking the fire picture in the data set by using roboflow;
establishing a training model by using a YOLOv5 algorithm based on the data set;
inputting the marked fire picture into a training model to train the model;
and accessing the trained model into a monitoring camera of the expressway for real-time monitoring and early warning.
Preferably, the data set comprises a training set, a validation set and a test set;
the training model comprises a YOLO neural network.
Preferably, the number ratio of the pictures in the training set, the verification set and the test set is: 6:2:2.
Preferably, the training model includes a training algorithm, and the training algorithm includes the following steps:
pre-training: pre-training the fire picture on ImageNet, wherein the pre-trained classification model adopts the first 20 convolutional layers, and then an average-pool layer and a full-connection layer are added;
after pre-training, adding 4 convolution layers and 2 full-connection layers which are initialized randomly on 20 convolution layers obtained by pre-training;
network prediction: using the NMS algorithm: firstly, finding the box with the highest confidence coefficient from all the detection boxes, then calculating the IOU of the box and the rest boxes one by one, and if the value of the IOU is greater than a certain threshold value, removing the box; then repeating the above process for the rest detection frames until all the detection frames are processed;
analyzing the prediction result of the network: and judging the category of the prediction frame, and detecting the accuracy of the prediction result through an NMS algorithm.
Preferably, the convolutional neural network analyzes the feature vector in the target portion, the feature vector is encoded by a low-level visual descriptor, and the low-level visual descriptor includes SIFT, Haar, HOG and SURF.
Preferably, the system further comprises a detection system and a network platform, wherein the detection system is connected with the model, the detection system is connected with the network platform, and the network platform is connected with an alarm system.
The invention has the beneficial effects that:
in the invention, a data set is made based on a YOLOv5 algorithm; establishing a training model based on the data set; marking the fire picture in the data set by using roboflow; inputting the training model into a training model to train the model; the trained model is connected to a monitoring camera of the expressway for real-time monitoring and early warning, a training result can be simulated, the result shows that the mAP reaches more than 0.9, the model has a good identification effect on the tunnel fire incident, and the monitoring of the fire is effectively improved.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a flowchart of a tunnel fire detection method based on the YOLO neural network according to the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, 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 some, but not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, are within the scope of the present invention.
The embodiment of the invention provides a tunnel fire detection method based on a YOLO neural network, which comprises the following steps:
making a fire picture training data set;
marking the fire picture in the data set by using roboflow;
establishing a training model by using a YOLOv5 algorithm based on the data set;
inputting the marked fire picture into a training model to train the model;
and accessing the trained model into a monitoring camera of the expressway for real-time monitoring and early warning.
Preferably, the data set includes a training set, a verification set and a test set, and the sets are divided into several sets mainly for improving model accuracy;
the training model comprises a YOLO neural network.
Preferably, the number ratio of the pictures in the training set, the verification set and the test set is: 6:2:2.
The training model comprises a training algorithm, and the training algorithm comprises the following steps:
pre-training: pre-training the fire picture on ImageNet, wherein the pre-trained classification model adopts the first 20 convolutional layers, and then an average-pool layer and a full-connection layer are added;
after pre-training, adding 4 convolution layers and 2 full-connection layers which are initialized randomly on 20 convolution layers obtained by pre-training;
network prediction: using the NMS algorithm: firstly, finding the box with the highest confidence coefficient from all the detection boxes, then calculating the IOU of the box and the rest boxes one by one, and if the value of the IOU is greater than a certain threshold value, removing the box; then repeating the above process for the rest detection frames until all the detection frames are processed;
analyzing the prediction result of the network: and judging the category of the prediction frame, and detecting the accuracy of the prediction result through an NMS algorithm.
The convolutional neural network analyzes the feature vectors in the target part, the feature vectors are coded by low-level visual descriptors, and the low-level visual descriptors comprise SIFT, Haar, HOG and SURF.
The system comprises a model, a detection system and a network platform, wherein the detection system is connected with the model, the detection system is connected with the network platform, and the network platform is connected with an alarm system.
In the invention, a data set is made based on a YOLOv5 algorithm; establishing a training model based on the data set; marking the fire picture in the data set by using roboflow; inputting the training model into a training model to train the model; the trained model is connected to a monitoring camera of the expressway for real-time monitoring and early warning, a training result can be simulated, the result shows that the mAP reaches more than 0.9, the model has a good identification effect on tunnel fire accidents, and the monitoring of fire is effectively improved.
Although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (6)
1. A tunnel fire detection method based on a YOLO neural network is characterized by comprising the following steps:
making a fire picture training data set;
marking the fire picture in the data set by using roboflow;
establishing a training model by using a YOLOv5 algorithm based on the data set;
inputting the marked fire picture into a training model to train the model;
and accessing the trained model into a monitoring camera of the expressway for real-time monitoring and early warning.
2. The method of claim 1, wherein the data set comprises a training set, a validation set, and a test set;
the training model comprises a YOLO neural network.
3. The method as claimed in claim 2, wherein the ratio of the number of pictures in the training set, the verification set and the test set is: 6:2:2.
4. The method of claim 1, wherein the training model comprises a training algorithm, and the training algorithm comprises the following steps:
pre-training: pre-training the fire picture on ImageNet, wherein the pre-trained classification model adopts the first 20 convolutional layers, and then an average-pool layer and a full-connection layer are added;
after pre-training, adding 4 convolutional layers and 2 full-connection layers which are initialized randomly on the 20 convolutional layers obtained by pre-training;
network prediction: using the NMS algorithm: firstly, finding out the box with the maximum confidence coefficient from all the detection boxes, then calculating the IOU of the box and the rest boxes one by one, and if the IOU value is greater than a certain threshold value, removing the box; then repeating the above process for the rest detection frames until all the detection frames are processed;
analyzing the prediction result of the network: and judging the category of the prediction frame, and detecting the accuracy of the prediction result through an NMS algorithm.
5. The method of claim 4, wherein the convolutional neural network analyzes the feature vector in the target portion, the feature vector is encoded by a low-level visual descriptor, and the low-level visual descriptor comprises SIFT, Haar, HOG and SURF.
6. The method for detecting tunnel fire based on the YOLO neural network as claimed in claim 1, further comprising a detection system and a network platform, wherein the detection system is connected with the model, the detection system is connected with the network platform, and the network platform is connected with an alarm system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111509268.4A CN114298167A (en) | 2021-12-10 | 2021-12-10 | Tunnel fire detection method based on YOLO neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111509268.4A CN114298167A (en) | 2021-12-10 | 2021-12-10 | Tunnel fire detection method based on YOLO neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114298167A true CN114298167A (en) | 2022-04-08 |
Family
ID=80966891
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111509268.4A Pending CN114298167A (en) | 2021-12-10 | 2021-12-10 | Tunnel fire detection method based on YOLO neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114298167A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115147715A (en) * | 2022-04-14 | 2022-10-04 | 山东浪潮科学研究院有限公司 | Fire detection method and device based on TinyML |
| CN115753848A (en) * | 2022-12-02 | 2023-03-07 | 中山大学孙逸仙纪念医院深汕中心医院 | Remote online monitoring system and method for radiographic inspection |
-
2021
- 2021-12-10 CN CN202111509268.4A patent/CN114298167A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115147715A (en) * | 2022-04-14 | 2022-10-04 | 山东浪潮科学研究院有限公司 | Fire detection method and device based on TinyML |
| CN115753848A (en) * | 2022-12-02 | 2023-03-07 | 中山大学孙逸仙纪念医院深汕中心医院 | Remote online monitoring system and method for radiographic inspection |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Huang et al. | Detection algorithm of safety helmet wearing based on deep learning | |
| CN114298167A (en) | Tunnel fire detection method based on YOLO neural network | |
| CN109815797B (en) | Living body detection method and apparatus | |
| AU2017389535A1 (en) | Image tampering detection method and system, electronic apparatus and storage medium | |
| CN110827505A (en) | Smoke segmentation method based on deep learning | |
| CN110674790B (en) | Abnormal scene processing method and system in video monitoring | |
| CN111738054B (en) | Behavior anomaly detection method based on space-time self-encoder network and space-time CNN | |
| CN111626169B (en) | Image-based railway dangerous falling rock size judgment method | |
| CN112200081A (en) | Abnormal behavior identification method and device, electronic equipment and storage medium | |
| CN111813997A (en) | Intrusion analysis method, device, equipment and storage medium | |
| CN113221890A (en) | OCR-based cloud mobile phone text content supervision method, system and system | |
| CN111008574A (en) | Key person track analysis method based on body shape recognition technology | |
| CN112785578A (en) | Road crack detection method and system based on U-shaped codec neural network | |
| CN113298077A (en) | Transformer substation foreign matter identification and positioning method and device based on deep learning | |
| Tao et al. | Smoke vehicle detection based on spatiotemporal bag-of-features and professional convolutional neural network | |
| CN115937237A (en) | Local feature extraction method based on edge transform domain | |
| CN115410119A (en) | Violent movement detection method and system based on adaptive generation of training samples | |
| CN112329858B (en) | Image recognition method for breakage fault of anti-loosening iron wire of railway motor car | |
| CN111898440B (en) | Mountain fire detection method based on three-dimensional convolutional neural network | |
| CN113255519A (en) | Crane lifting arm identification system and multi-target tracking method for power transmission line dangerous vehicle | |
| CN117475353A (en) | Video-based abnormal smoke identification method and system | |
| CN115690584A (en) | SSD (solid State drive) -based improved power distribution room foreign matter detection method | |
| CN115311458A (en) | Real-time expressway pedestrian intrusion event detection method based on multi-task learning | |
| CN115205747A (en) | Automatic road detection system based on image recognition | |
| CN114898137A (en) | Face recognition-oriented black box sample attack resisting method, device, equipment and medium |
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 |