CN111767907B - Method of multi-source data fire detection system based on GA and VGG network - Google Patents
Method of multi-source data fire detection system based on GA and VGG network Download PDFInfo
- Publication number
- CN111767907B CN111767907B CN202010913174.2A CN202010913174A CN111767907B CN 111767907 B CN111767907 B CN 111767907B CN 202010913174 A CN202010913174 A CN 202010913174A CN 111767907 B CN111767907 B CN 111767907B
- Authority
- CN
- China
- Prior art keywords
- vgg
- data sources
- network
- different data
- source data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Biophysics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Remote Sensing (AREA)
- Multimedia (AREA)
- Astronomy & Astrophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (AREA)
- Genetics & Genomics (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a method of a multi-source data fire detection system based on GA and VGG networks. The present invention uses multi-source data as a source of fire determination data. Firstly, converting images into proper space according to real-time images acquired by a satellite and an unmanned aerial vehicle, respectively carrying out image recognition on different data sources by using a VGG network, and then optimizing the weights of the different data sources by using a Genetic Algorithm (GA) to obtain the optimal weight. And finally, evaluating the identification results of different data sources. The fire disaster identification rate can be improved through multi-source data fusion and a VGG network, the environment information can be monitored accurately and efficiently, natural disasters can be found in time, damage can be stopped in time, and the life and property safety of the nation and people can be guaranteed.
Description
Technical Field
The invention relates to the field of deep learning, in particular to a method for designing a multi-source data fire detection system based on a GA (genetic algorithm) and a VGG (virtual ground gateway) network.
Background
With the rapid development of economy and the continuous progress of science and technology, the utilization rate of electronic equipment is continuously improved, and a fire disaster becomes a main disaster which harms the property and the health of the nation. For example, the Australian jungle fire in 2019 burns for 7 months, which greatly endangers the personal and property safety of human beings.
In recent years, with the continuous progress of deep learning technology, the deep learning technology based on deep learning is greatly developed, wherein the vision group of oxford university proposes a VGG network, and the VGG network is widely applied to various tasks in the vision field. The VGG Network is called Visual Geometry Group Network in English, adopts a small-sized convolution kernel, and is widely applied to face recognition, target detection, image segmentation and the like. The general Algorithm of GA English is designed and proposed according to the evolution law of organisms in nature, and when a complex combination optimization problem is solved, better optimization results can be obtained faster compared with some conventional optimization algorithms.
Disclosure of Invention
To solve the above existing problems. The invention provides a method of a multi-source data fire detection system based on GA and VGG networks, which is used for carrying out weighted identification processing on various data sources and carrying out efficient and accurate fire evaluation. To achieve this object:
the invention provides a method of a multi-source data fire detection system based on GA and VGG networks, which comprises the following steps,
step 1: converting the images into RGB image space according to the real-time images acquired by the satellite and the unmanned aerial vehicle;
step 2: image preprocessing, adjusting the image to 224 × 224, calculating the average value of three channels, and subtracting the average value from each pixel;
and step 3: sending the preprocessed output image into a VGG network for network training;
and 4, step 4: carrying out identification weighting processing on different data sources;
and 5: optimizing the weights of different data sources by using a genetic algorithm GA fitness function to obtain an optimal weight;
step 6: performing performance test on the trained VGG model and the optimal weight values of different data sources by using the test sample;
and 7: fine-tuning the optimal weight of the VGG network and the genetic algorithm GA to obtain an optimal evaluation model;
and 8: and identifying multi-source data by using the optimal evaluation model for fire monitoring.
As a further improvement of the present invention, the formula of the image pixel preprocessing in step 2 is:
wherein the content of the first and second substances,
which are the three color channels of the RGB image color space, respectively.
As a further improvement of the present invention, in step 3, the preprocessed output image is sent to the VGG network for network training, the input is deformed, and then nonlinear processing is performed through the ReLU, where a formula of a ReLU activation function is:
wherein the content of the first and second substances,
is the ReLU input variable.
As a further improvement of the present invention, the identification weighting processing formula of the different data sources in step 4 is:
wherein n is the number of samples,
outputs the recognition result weight for VGG, and
，
the result of the recognition is output for the VGG,
is the decision threshold.
As a further improvement of the present invention, the genetic algorithm GA fitness function in step 5 is:
wherein the content of the first and second substances,
in order to train the sample labels,
and identifying a result for training the model.
The method of the multisource data fire detection system based on the GA and the VGG network has the advantages that:
1. according to the invention, a multi-source data source is utilized, so that the fire disaster judgment is more accurate;
2. according to the method, the optimal value of the multi-source data weight can be found by utilizing the genetic algorithm GA, so that the accuracy of the model is improved;
3. the invention utilizes the VGG network, and the generalization capability is very good;
4. the algorithm of the invention is simple to realize and has low cost.
Drawings
FIG. 1 is a system architecture diagram;
FIG. 2 is a flow chart of system execution;
fig. 3 is a VGG network model parameter diagram.
Detailed Description
The invention is described in further detail below with reference to the following detailed description and accompanying drawings:
the invention provides a method of a multi-source data fire detection system based on GA and VGG networks, which is used for carrying out weighted identification processing on various data sources and carrying out efficient and accurate fire evaluation. Fig. 1 is a system architecture diagram.
The working flow chart of the invention is shown in figure 2.
Firstly, converting an image into an RGB image space according to a real-time image acquired by a satellite and an unmanned aerial vehicle; the image is adjusted to 224 x 224 size, the average of the three channels is calculated and subtracted at each pixel, and the formula for image pixel pre-processing is:
wherein the content of the first and second substances,
three colors of the RGB image color space, respectivelyA color channel.
Secondly, sending the preprocessed output images into a VGG network for network training, and carrying out recognition weighting processing on different data sources;
the input is deformed, and nonlinear processing is carried out through a ReLU, wherein the formula of a ReLU activation function is as follows:
wherein the content of the first and second substances,
is the ReLU input variable.
The identification weighting processing formula of different data sources is as follows:
wherein n is the number of samples,
outputs the recognition result weight for VGG, and
，
the result of the recognition is output for the VGG,
is the decision threshold. Where the parameters of the G network are shown in figure 3.
Finally, optimizing the weights of different data sources by using a genetic algorithm GA to obtain an optimal weight;
performing performance test on the trained VGG model and the optimal weight values of different data sources by using the test sample, and fine-tuning the optimal weight values of the VGG network and the genetic algorithm GA to obtain an optimal evaluation model;
the genetic algorithm GA fitness function is as follows:
wherein the content of the first and second substances,
in order to train the sample labels,
and identifying a result for training the model.
The Genetic Algorithm (GA) algorithm flow is as follows:
1) randomly generating a population as a primary solution to the problem (typically, the primary solution may be significantly different from the optimal solution, which is tolerable as long as it is ensured that the primary solution is randomly generated to ensure the diversity of individual genes);
2) searching a proper coding scheme to code individuals in the population, wherein common coding schemes such as floating point number coding or binary coding can be selected (note that different coding schemes directly influence implementation details of subsequent genetic operators);
3) calculating the fitness of each individual in the population by taking the function value of the multimodal function as the fitness of the individual (the calculated fitness provides a basis for subsequent individual selection);
4) selecting parents and parents participating in reproduction according to the fitness, wherein the selection principle is that individuals with higher fitness are more likely to be selected (so that individuals with low fitness are continuously eliminated);
5) performing genetic operation on the selected parents and the parents, namely copying genes of the parents and the parents, and generating offspring by adopting operators such as crossover, mutation and the like (on the basis of reserving excellent genes to a greater extent, the mutation increases the diversity of the genes, thereby improving the probability of finding the optimal solution);
6) and judging whether to continue executing the algorithm or to find out the individual with the highest fitness in all the descendants as a solution to return and end the program according to a certain criterion (the judgment criterion can be a set threshold value of the solution, a specified iteration number and the like).
The above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention in any way, but any modifications or equivalent variations made according to the technical spirit of the present invention are within the scope of the present invention as claimed.
Claims (5)
1. A method of a multi-source data fire detection system based on GA and VGG network comprises the following steps,
step 1: converting the images into RGB image space according to the real-time images acquired by the satellite and the unmanned aerial vehicle;
step 2: image preprocessing, adjusting the image to 224 × 224, calculating the average value of three channels, and subtracting the average value from each pixel;
and step 3: sending the preprocessed output image into a VGG network for network training;
and 4, step 4: carrying out identification weighting processing on different data sources;
and 5: optimizing the weights of different data sources by using a genetic algorithm GA fitness function to obtain an optimal weight;
step 6: performing performance test on the trained VGG model and the optimal weight values of different data sources by using the test sample;
and 7: fine-tuning the optimal weight of the VGG network and the genetic algorithm GA to obtain an optimal evaluation model;
and 8: and identifying multi-source data by using the optimal evaluation model for fire monitoring.
2. The method of claim 1, wherein the method comprises the steps of;
the formula of the image pixel preprocessing in the step 2 is as follows:
wherein the content of the first and second substances,
which are the three color channels of the RGB image color space, respectively.
3. The method of claim 1, wherein the method comprises the steps of;
in the step 3, the preprocessed output image is sent to a VGG network for network training, the input is deformed, and nonlinear processing is performed through a ReLU, wherein the formula of a ReLU activation function is as follows:
wherein the content of the first and second substances,
is the ReLU input variable.
4. The method of claim 1, wherein the method comprises the steps of;
the identification weighting processing formula of different data sources in the step 4 is as follows:
wherein n is the number of samples,
outputs the recognition result weight for VGG, and
，
the result of the recognition is output for the VGG,
is the decision threshold.
5. The method of claim 1, wherein the method comprises the steps of;
the genetic algorithm GA fitness function in the step 5 is as follows:
wherein the content of the first and second substances,
in order to train the sample labels,
and identifying a result for training the model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010913174.2A CN111767907B (en) | 2020-09-03 | 2020-09-03 | Method of multi-source data fire detection system based on GA and VGG network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010913174.2A CN111767907B (en) | 2020-09-03 | 2020-09-03 | Method of multi-source data fire detection system based on GA and VGG network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111767907A CN111767907A (en) | 2020-10-13 |
| CN111767907B true CN111767907B (en) | 2020-12-15 |
Family
ID=72729235
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010913174.2A Active CN111767907B (en) | 2020-09-03 | 2020-09-03 | Method of multi-source data fire detection system based on GA and VGG network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111767907B (en) |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100383805C (en) * | 2005-11-03 | 2008-04-23 | 复旦大学 | Method for sorting characters of ground object through interfusion of satellite carried microwave and infrared remote sensing |
| CN105096333A (en) * | 2015-09-06 | 2015-11-25 | 河海大学常州校区 | Segmentation method for infrared thermal imaging image of forest fire |
| CN108510057A (en) * | 2017-02-27 | 2018-09-07 | 顾泽苍 | A kind of constructive method of the neural network model of ultra-deep confrontation study |
| CN111582345A (en) * | 2020-04-29 | 2020-08-25 | 中国科学院重庆绿色智能技术研究院 | Target identification method for complex environment under small sample |
-
2020
- 2020-09-03 CN CN202010913174.2A patent/CN111767907B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN111767907A (en) | 2020-10-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112101426B (en) | Unsupervised learning image anomaly detection method based on self-encoder | |
| CN109711463B (en) | Attention-based important object detection method | |
| EP3690740B1 (en) | Method for optimizing hyperparameters of auto-labeling device which auto-labels training images for use in deep learning network to analyze images with high precision, and optimizing device using the same | |
| US11308714B1 (en) | Artificial intelligence system for identifying and assessing attributes of a property shown in aerial imagery | |
| JP2020123330A (en) | Method for acquiring sample image for label acceptance inspection from among auto-labeled images utilized for neural network learning, and sample image acquisition device utilizing the same | |
| CN113806746B (en) | Malicious code detection method based on improved CNN (CNN) network | |
| CN112163628A (en) | Method for improving target real-time identification network structure suitable for embedded equipment | |
| CN114842267A (en) | Image classification method and system based on label noise domain self-adaption | |
| CN116994069B (en) | Image analysis method and system based on multi-mode information | |
| CN114255403A (en) | Optical remote sensing image data processing method and system based on deep learning | |
| CN112149526B (en) | Lane line detection method and system based on long-distance information fusion | |
| CN116910752B (en) | Malicious code detection method based on big data | |
| CN112926661A (en) | Method for enhancing image classification robustness | |
| CN113127857A (en) | Deep learning model defense method for adversarial attack and deep learning model | |
| CN115546196A (en) | Knowledge distillation-based lightweight remote sensing image change detection method | |
| CN115035599A (en) | Armed personnel identification method and armed personnel identification system integrating equipment and behavior characteristics | |
| CN114662605A (en) | Flame detection method based on improved YOLOv5 model | |
| CN114358249A (en) | Target recognition model training method, target recognition method and device | |
| CN114119532A (en) | Building change detection method based on remote sensing image and twin neural network | |
| CN111767907B (en) | Method of multi-source data fire detection system based on GA and VGG network | |
| CN115965836A (en) | Human behavior posture video data amplification system and method with controllable semantics | |
| CN115019096A (en) | Armed personnel equipment detection method and armed personnel equipment detection system based on dynamic neural network | |
| CN111858999B (en) | Retrieval method and device based on segmentation difficult sample generation | |
| CN111144492B (en) | Scene map generation method for mobile terminal virtual reality and augmented reality | |
| CN113887634A (en) | Improved two-step detection-based electric safety belt detection and early warning method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20230302 Address after: Room 1801, 18th floor, building 1, yard 1, No. 81, Beiqing Road, Haidian District, Beijing 100094 Patentee after: Zhongkexing Tuwei Tianxin (Beijing) Technology Co.,Ltd. Address before: Room 905 and 906, 9 / F, building 4, No. 18, Jiangdong Street, Jialing, Jianye District, Nanjing City, Jiangsu Province, 210000 Patentee before: JIANGSU QUAN QUAN INFORMATION TECHNOLOGY CO.,LTD. |
|
| TR01 | Transfer of patent right | ||
| CP01 | Change in the name or title of a patent holder |
Address after: Room 1801, 18th floor, building 1, yard 1, No. 81, Beiqing Road, Haidian District, Beijing 100094 Patentee after: Zhongkexing Tuwei Tianxin Technology Co.,Ltd. Address before: Room 1801, 18th floor, building 1, yard 1, No. 81, Beiqing Road, Haidian District, Beijing 100094 Patentee before: Zhongkexing Tuwei Tianxin (Beijing) Technology Co.,Ltd. |
|
| CP01 | Change in the name or title of a patent holder |