CN109460744A - A kind of video monitoring system based on deep learning - Google Patents
A kind of video monitoring system based on deep learning Download PDFInfo
- Publication number
- CN109460744A CN109460744A CN201811418016.9A CN201811418016A CN109460744A CN 109460744 A CN109460744 A CN 109460744A CN 201811418016 A CN201811418016 A CN 201811418016A CN 109460744 A CN109460744 A CN 109460744A
- Authority
- CN
- China
- Prior art keywords
- module
- classification results
- image
- video monitoring
- classification
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Multimedia (AREA)
- Alarm Systems (AREA)
Abstract
The invention discloses a kind of video monitoring systems based on deep learning, including image capture module, image preprocessing and screening module, packet forward module, two-stage classification module and classification results statistical module；Described image acquisition module connects each monitoring source, acquires the image data in each monitoring source；Described image pretreatment and screening module are handled and are screened to acquired image；The packet forward module carries out Unified coding to processed image data and sends it to two-stage classification module；The security risk that the two-stage classification module carries out suspicious abnormality detection, the Anomaly target detection of the second level and production area to image data detects；Classification results that the classification results statistical module counts obtain simultaneously are made classification results by the consecutive hours intersexuality judgement to classification results and are determined.The present invention can be effectively combined existing video monitoring system, can be realized the extraction of real-time production area security information, achieve the purpose that prevention and control security risk.
Description
Technical field
The present invention relates to field of video monitoring, more particularly to a kind of video monitoring system based on deep learning.
Background technique
Safety in production is one of the important policy of social development and construction, and production must be safe, and safety is in order to produce.Implementing
It must be noted that the safe and healthy and indispensable welfare project of worker while increasing production and practising economy.
In terms of enterprise security monitoring, there are complete video monitoring system, and energy in existing major part Utilities Electric Co.
The video monitoring and video recording of main security node are enough covered, but its actual conditions goes actively to patrol there is still a need for very more manpowers
It is day and night checked depending on and to monitor.Moreover, in current video monitoring being pacified respectively in each area to be monitored
If a camera, and in monitoring center setting, technically display window is checked for owner, and when discovery has abnormal, by
Monitor notifies its corresponding staff that scene is gone to check.It is possible that occurring directly notifying relevant work personnel
Ground situation, on the other hand, different emergencies be need differently people check sub-categoryly, therefore the processing of monitor video and
Display is not targeted, also not prompt enough, can greatly reduce the availability of monitor video in this way and for abnormal thing
The treatment effeciency of part.
Summary of the invention
Goal of the invention: according to above-mentioned deficiency, the invention proposes the video monitoring systems based on deep learning, can be effective
Be integrated to existing enterprise's video monitoring system, can be realized the extraction of real-time production area security information, reach prevention
With the purpose of control security risk.
Technical solution:
A kind of video monitoring system based on deep learning, including image capture module, image preprocessing and screening module,
Packet forward module, two-stage classification module and classification results statistical module；
Described image acquisition module connects each monitoring source, acquires the image data in each monitoring source；
Described image pretreatment and screening module go described image acquisition module acquired image using median filtering
It makes an uproar, and by second order Sobel filtering to the corresponding edge extracting of area-of-interest progress and by being split to described image
It calculates and extracts engineering threshold value, abnormal picture frame is filtered out by engineering threshold value；
The packet forward module pre-processes described image and the processed image data of screening module carries out unified volume
Code simultaneously sends it to two-stage classification module；The coding will include affiliated group information, current time information and pretreatment knot
The ID of fruit information setting uniqueness；Affiliated group information is grouped according to Exception Type or processing scheme；
The two-stage classification module carries out suspicious abnormality detection, second to the image data that the packet forward module is sent
The Anomaly target detection of grade and the security risk detection of production area；The suspicious abnormality detection carries out specific objective image
Initial screening obtains the picture frame of specific objective；The Anomaly target detection of the second level and the security risk inspection of production area
The master drawing training by convolutional neural networks to actual scene gathered in advance is surveyed, to the pipeline gas or liquid under no target
Or mist particles leakage is monitored to obtain abnormal alarm information, to the picture frame for the specific objective that suspicious abnormality detection obtains
The target intensity of anomaly carried out under target is distinguished；
Classification results that two-stage classification module described in the classification results statistical module counts obtains and by being tied to classification
Classification results judgement is made in the consecutive hours intersexuality judgement of fruit, and corresponding result data is saved in local；The classification results
Determine to use Time Continuous sex determination, i.e., by judging whether that identical result occur in set period of time obtains finally judging knot
Fruit, and the storage of obtained result is optimized to locally for line drag.
Described image acquisition module is using the monitoring device of addition output function or the place of executable image data extraction
Manage unit.
Described image pretreatment and screening module using the monitoring device for having image real time transfer function or can be performed
The processing unit of image real time transfer.
The two-stage classification module includes first order neural network and second level neural network；The first order neural network
Be made of a neural network model, the neural network model select RCNN, Fast R-CNN, Faster R-CNN, FPN,
YOLO v1, YOLO v2, SSD or RetinaNet；The second level neural network uses convolutional neural networks model, according to reality
The number for applying scene determines the neural network model quantity.
Setting time described in the classification results judgement of the classification results statistical module is according to different requirement and difference
Scene determine.
It further include Master Control Center, the Master Control Center includes user management module, message managing module and early warning and shows
Show module；
The user management module is for user information registration, certification and creation and deletes；The user management module
Including user information update module and authority management module；The user information update module is used for registering new user and modification
Family information, the user information include user name, name, work number, working group；The authority management module is used for the institute to user
There is grade to carry out delineation of power, the Permission Levels of different operating group are divided；
The message managing module is for receiving and transmitting each intermodule event message；
The early warning and display module, which are used to receive classification results and are sent to terminal to corresponding result, makes warning simultaneously
It is output to display equipment.
The terminal uses the warning light with information warning or the prompting bell with information warning, the display equipment
For cell phone client or page end.
The utility model has the advantages that the present invention is based on the video monitoring systems of deep learning can be effectively combined existing enterprise's view
Frequency monitoring system can be realized the extraction of real-time production area security information, achieve the purpose that prevention and control security risk.
Detailed description of the invention
Fig. 1 is that the present invention is based on structural schematic diagrams total in the video monitoring system preferred embodiment of deep learning.
Fig. 2 is that the present invention is based on partial structure diagrams in the video monitoring system preferred embodiment of deep learning.
Specific embodiment
In the following with reference to the drawings and specific embodiments, the present invention is furture elucidated.
Fig. 1 is that the present invention is based on structural schematic diagrams total in the video monitoring system preferred embodiment of deep learning.Such as Fig. 1
It is shown, video monitoring system of the invention include image capture module, image preprocessing and screening module, packet forward module,
Two-stage classification module, classification results statistical module and Master Control Center；
Described image acquisition module connects the monitoring source of video monitoring system of the present invention, for acquiring the image in each monitoring source
Data can be used the monitoring device for being added to output function or can execute the processing unit of the extraction of image data；
Described image pretreatment and screening module go described image acquisition module acquired image using median filtering
It makes an uproar, and by second order Sobel filtering to the corresponding edge extracting of area-of-interest (ROI) progress and by being carried out to described image
Separation calculation extracts engineering threshold value, filters out abnormal picture frame by engineering threshold value, reduces most of interference image, can be with
Using the monitoring device for having image real time transfer function or the processing unit of image real time transfer can be executed；
The packet forward module is used to unite to described image pretreatment and the processed image data of screening module
One encodes and sends it to two-stage classification module, the monitoring device for having encoding function can be used or can execute coding
The processing unit of processing；The coding includes that affiliated group information, current time information and the setting of pre-processed results information are unique
The ID of property；Affiliated group information is grouped according to Exception Type or processing scheme；
The suspicious abnormality detection of image data progress that the two-stage classification module is used to send the packet forward module,
The Anomaly target detection of the second level and the security risk detection of production area；At the beginning of suspicious abnormality detection is used for specific objective image
Screening, obtains the picture frame of specific objective, is carried out in the environment of TensorFlow using Fast RCNN identification network；At this
In invention, by being screened to the individual features under different scenes using different image detection algorithms；The exception of the second level
The detection of the security risk of target detection and production area is by convolutional neural networks to the master drawing of actual scene gathered in advance
Training, under no target pipeline gas or liquid or mist particles leakage be monitored to obtain abnormal alarm information, it is right
The picture frame for the specific objective that suspicious abnormality detection obtains carries out the target intensity of anomaly under target and is distinguished；
The two-stage classification module includes first order neural network and second level neural network；First order neural network is main
Be made of a neural network model, available neural network have RCNN, Fast R-CNN, Faster R-CNN, FPN,
YOLO v1,YOLO v2,SSD,RetinaNet.Second level neural network uses convolutional neural networks model, with specific reference to implementation
The number of scene is related, practical selection scheme can be small-sized traditional convolutional neural networks such as AlexNet, GoogleNet,
VGGNet。
The classification results statistical module be used to count classification results that the two-stage classification module obtains and by point
The consecutive hours intersexuality judgement of class result is made more accurate classification results and is determined, and corresponding result data is saved in this
Ground；Classification results determine to use Time Continuous sex determination, i.e., by judging whether that identical result occur in set period of time obtains
Final judging result, and the storage of obtained result is optimized to locally for line drag；In the present invention, when the setting
Between determined according to different requirement and different scenes.
The Master Control Center manages the interaction of all modules, including user management module, message managing module and early warning
And display module；The user management module is for user information registration, certification and creation and deletes；The user management mould
Block includes user information update module and authority management module；The user information update module is for registering new user and modification
User information, the user information include user name, name, work number, working group；The authority management module is used for enterprise
All position hierarchies carry out delineation of power, divide to the Permission Levels of different operating group；The message managing module is used for
Receive the event message between transfer module；The early warning and display module are for receiving classification results and sending to corresponding result
Made to terminal and warn and be output to display equipment, actual terminal source of early warning can be warning light with information warning or
Person is the prompting bell with information warning, and display equipment is mainly present on cell phone client or page end.
The preferred embodiment of the present invention has been described above in detail, but during present invention is not limited to the embodiments described above
Detail can carry out a variety of equivalents to technical solution of the present invention (in full within the scope of the technical concept of the present invention
Amount, shape, position etc.), these equivalents belong to protection of the invention.
Claims (7)
1. a kind of video monitoring system based on deep learning, it is characterised in that: including image capture module, image preprocessing and
Screening module, packet forward module, two-stage classification module and classification results statistical module；
Described image acquisition module connects each monitoring source, acquires the image data in each monitoring source；
Described image pretreatment and screening module denoise described image acquisition module acquired image using median filtering, and
Corresponding edge extracting is carried out to area-of-interest by second order Sobel filtering and is mentioned by being split calculating to described image
Engineering threshold value is taken, abnormal picture frame is filtered out by engineering threshold value；
The packet forward module pre-processes described image and the processed image data of screening module carries out Unified coding simultaneously
Send it to two-stage classification module；The coding will include affiliated group information, current time information and pre-processed results letter
The ID of breath setting uniqueness；Affiliated group information is grouped according to Exception Type or processing scheme；
The two-stage classification module carries out suspicious abnormality detection, the second level to image data that the packet forward module is sent
The detection of the security risk of Anomaly target detection and production area；The suspicious abnormality detection carries out primary dcreening operation to specific objective image
Choosing, obtains the picture frame of specific objective；The Anomaly target detection of the second level and the security risk detection of production area are logical
Convolutional neural networks are crossed to the training of the master drawing of actual scene gathered in advance, under no target pipeline gas or liquid or
Mist particles leakage is monitored to obtain abnormal alarm information, carries out to the picture frame for the specific objective that suspicious abnormality detection obtains
There is the target intensity of anomaly under target to be distinguished；
Classification results that two-stage classification module described in the classification results statistical module counts obtains and by classification results
Classification results judgement is made in the judgement of consecutive hours intersexuality, and corresponding result data is saved in local；The classification results determine
Using Time Continuous sex determination, i.e., by judging whether that identical result occur in set period of time obtains final judging result, and
And the storage of obtained result is optimized to locally for line drag.
2. video monitoring system according to claim 1, it is characterised in that: described image acquisition module is using addition output
The processing unit of the monitoring device of function or executable image data extraction.
3. video monitoring system according to claim 1, it is characterised in that: described image pretreatment and screening module use
Have the monitoring device of image real time transfer function or the processing unit of executable image real time transfer.
4. video monitoring system according to claim 1, it is characterised in that: the two-stage classification module includes first order mind
Through network and second level neural network；The first order neural network is made of a neural network model, the neural network
Model selection RCNN, Fast R-CNN, Faster R-CNN, FPN, YOLOv1, YOLO v2, SSD or RetinaNet；Described
Secondary Neural Networks use convolutional neural networks model, determine the neural network model quantity according to the number of implement scene.
5. video monitoring system according to claim 1, it is characterised in that: the classification knot of the classification results statistical module
Setting time described in fruit judgement is determined according to different requirements and different scenes.
6. video monitoring system according to claim 1, it is characterised in that: it further include Master Control Center, the Master Control Center
Including user management module, message managing module and early warning and display module；
The user management module is for user information registration, certification and creation and deletes；The user management module includes
User information update module and authority management module；The user information update module is for registering new user and modification user's letter
Breath, the user information includes user name, name, work number, working group；The authority management module is used for all etc. of user
Grade carries out delineation of power, divides to the Permission Levels of different operating group；
The message managing module is for receiving and transmitting each intermodule event message；
The early warning and display module are made warning and are exported for receiving classification results and being sent to terminal to corresponding result
To display equipment.
7. video monitoring system according to claim 6, it is characterised in that: the terminal uses mentioning with information warning
Show that perhaps the display equipment of the prompting bell with information warning is cell phone client or page end to lamp.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811418016.9A CN109460744B (en) | 2018-11-26 | 2018-11-26 | Video monitoring system based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811418016.9A CN109460744B (en) | 2018-11-26 | 2018-11-26 | Video monitoring system based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109460744A true CN109460744A (en) | 2019-03-12 |
| CN109460744B CN109460744B (en) | 2021-08-27 |
Family
ID=65611632
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811418016.9A Active CN109460744B (en) | 2018-11-26 | 2018-11-26 | Video monitoring system based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109460744B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110553151A (en) * | 2019-07-17 | 2019-12-10 | 石化盈科信息技术有限责任公司 | pipeline leakage monitoring method and system |
| CN110687132A (en) * | 2019-10-08 | 2020-01-14 | 嘉兴凡视智能科技有限公司 | Intelligent visual detection system for foreign matters and bubbles in liquid based on deep learning algorithm |
| CN112004053A (en) * | 2020-07-20 | 2020-11-27 | 浙江大华技术股份有限公司 | Method and device for electronically amplifying monitoring image and computer equipment |
| CN112307821A (en) * | 2019-07-29 | 2021-02-02 | 顺丰科技有限公司 | Video stream processing method, device, equipment and storage medium |
| CN112861681A (en) * | 2021-01-29 | 2021-05-28 | 长兴云尚科技有限公司 | Pipe gallery video intelligent analysis method and system based on cloud processing |
| CN113625603A (en) * | 2021-07-27 | 2021-11-09 | 金鹏电子信息机器有限公司 | Security monitoring management system and management method based on big data |
| CN115081957A (en) * | 2022-08-18 | 2022-09-20 | 山东超华环保智能装备有限公司 | Useless management platform of danger of keeping in and monitoring useless |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106845890A (en) * | 2016-12-23 | 2017-06-13 | 济南博图信息技术有限公司 | A kind of storage monitoring method and device based on video monitoring |
| WO2017136137A1 (en) * | 2016-02-04 | 2017-08-10 | Nec Laboratories America, Inc. | Semantic segmentation based on global optimization |
| CN108805002A (en) * | 2018-04-11 | 2018-11-13 | 杭州电子科技大学 | Monitor video accident detection method based on deep learning and dynamic clustering |
| CN108830457A (en) * | 2018-05-14 | 2018-11-16 | 国网辽宁省电力有限公司 | Operation of power networks monitoring and control panorama management system based on artificial intelligence |
-
2018
- 2018-11-26 CN CN201811418016.9A patent/CN109460744B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017136137A1 (en) * | 2016-02-04 | 2017-08-10 | Nec Laboratories America, Inc. | Semantic segmentation based on global optimization |
| CN106845890A (en) * | 2016-12-23 | 2017-06-13 | 济南博图信息技术有限公司 | A kind of storage monitoring method and device based on video monitoring |
| CN108805002A (en) * | 2018-04-11 | 2018-11-13 | 杭州电子科技大学 | Monitor video accident detection method based on deep learning and dynamic clustering |
| CN108830457A (en) * | 2018-05-14 | 2018-11-16 | 国网辽宁省电力有限公司 | Operation of power networks monitoring and control panorama management system based on artificial intelligence |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110553151A (en) * | 2019-07-17 | 2019-12-10 | 石化盈科信息技术有限责任公司 | pipeline leakage monitoring method and system |
| CN110553151B (en) * | 2019-07-17 | 2021-05-07 | 石化盈科信息技术有限责任公司 | Pipeline leakage monitoring method and system |
| CN112307821A (en) * | 2019-07-29 | 2021-02-02 | 顺丰科技有限公司 | Video stream processing method, device, equipment and storage medium |
| CN110687132A (en) * | 2019-10-08 | 2020-01-14 | 嘉兴凡视智能科技有限公司 | Intelligent visual detection system for foreign matters and bubbles in liquid based on deep learning algorithm |
| CN112004053A (en) * | 2020-07-20 | 2020-11-27 | 浙江大华技术股份有限公司 | Method and device for electronically amplifying monitoring image and computer equipment |
| CN112861681A (en) * | 2021-01-29 | 2021-05-28 | 长兴云尚科技有限公司 | Pipe gallery video intelligent analysis method and system based on cloud processing |
| CN112861681B (en) * | 2021-01-29 | 2023-04-07 | 长兴云尚科技有限公司 | Pipe gallery video intelligent analysis method and system based on cloud processing |
| CN113625603A (en) * | 2021-07-27 | 2021-11-09 | 金鹏电子信息机器有限公司 | Security monitoring management system and management method based on big data |
| CN115081957A (en) * | 2022-08-18 | 2022-09-20 | 山东超华环保智能装备有限公司 | Useless management platform of danger of keeping in and monitoring useless |
| CN115081957B (en) * | 2022-08-18 | 2022-11-15 | 山东超华环保智能装备有限公司 | Useless management platform of danger of keeping in and monitoring useless |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109460744B (en) | 2021-08-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109460744A (en) | A kind of video monitoring system based on deep learning | |
| CN107911653B (en) | Intelligent video monitoring module, system, method and storage medium for residence | |
| CN103108159B (en) | Electric power intelligent video analyzing and monitoring system and method | |
| CN106454250A (en) | Intelligent recognition and early warning processing information platform | |
| CN107452104A (en) | A kind of control method for vehicle and system of the vehicle bayonet socket based on intelligent monitoring | |
| CN104581081B (en) | Passenger flow analysing method based on video information | |
| KR100956978B1 (en) | The system which integrated manages a traffic accident notice area | |
| CN110674790B (en) | Abnormal scene processing method and system in video monitoring | |
| CN103514694A (en) | Intrusion detection monitoring system | |
| CN111564224A (en) | Intelligent monitoring system with health monitoring function and implementation method thereof | |
| CN104821060A (en) | Robbery early warning method and device | |
| CN106529373A (en) | Human face recognition monitoring alarm method and system | |
| CN202771441U (en) | Face recognition monitoring system | |
| CN110047272A (en) | A kind of intelligent transportation pedestrian behavior monitoring and alarming system based on big data | |
| CN101127143A (en) | Automatic tracking display system for warn | |
| CN109635631A (en) | A kind of fire watch room personnel recognition methods on duty based on artificial intelligence | |
| CN202058203U (en) | Radioactive source dynamic monitoring and management system | |
| CN115880631A (en) | Power distribution station fault identification system, method and medium | |
| CN109741574A (en) | A kind of comprehensive monitor system based on intelligent safety helmet | |
| CN109447878A (en) | City big data security administration platform | |
| CN204929052U (en) | Internet security protection control intelligent management system | |
| CN110188617A (en) | A kind of machine room intelligent monitoring method and system | |
| CN113887310A (en) | Worker standard dressing identification method, system and medium | |
| CN113628172A (en) | Intelligent detection algorithm for personnel handheld weapons and smart city security system | |
| CN105148399B (en) | A kind of monitoring device and monitoring method of et al. Ke product |
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 |