[go: up one dir, main page]

CN107480578A - A kind of video detection system and method using crowd behaviour analysis - Google Patents

A kind of video detection system and method using crowd behaviour analysis Download PDF

Info

Publication number
CN107480578A
CN107480578A CN201610405681.9A CN201610405681A CN107480578A CN 107480578 A CN107480578 A CN 107480578A CN 201610405681 A CN201610405681 A CN 201610405681A CN 107480578 A CN107480578 A CN 107480578A
Authority
CN
China
Prior art keywords
video
crowd
safety index
module
people
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
Application number
CN201610405681.9A
Other languages
Chinese (zh)
Inventor
徐杰
王博
陈训逊
王东安
包秀国
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
National Computer Network and Information Security Management Center
Original Assignee
National Computer Network and Information Security Management Center
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by National Computer Network and Information Security Management Center filed Critical National Computer Network and Information Security Management Center
Priority to CN201610405681.9A priority Critical patent/CN107480578A/en
Publication of CN107480578A publication Critical patent/CN107480578A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10016—Video; Image sequence
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20212—Image combination
    • G06T2207/20224—Image subtraction
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30232—Surveillance
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • G06V20/44—Event detection

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Image Analysis (AREA)

Abstract

The present invention is a kind of video detecting method and system using crowd behaviour analysis, including video acquisition module, for gathering video data；Pretreatment module, for being pre-processed to the video collected；Crowd density determination module, for pretreated video, calculating the characteristic information with " people " significantly identified；Safety index generation module, for output safety index；Crowd's unusual checking module, for the probably behavior cruelly for differentiating whether there is robbery in the region, fighting；The beneficial effects of the present invention are：This method is based on deep learning and optical flow method and analyzed, and is analyzed by crowd density, safety index calculates and the analysis of crowd's abnormal behaviour, effectively detects probably behavior cruelly present in video, can be applied to photographic technique field.

Description

A kind of video detection system and method using crowd behaviour analysis
Technical field
It is more particularly to a kind of to be analyzed using crowd behaviour the invention belongs to the network information security, technical field of video monitoring Video detection system and method.
Background technology
Terrorist propagates video without restraint using network, has turned into the sudden and violent probably multiple root sex chromosome mosaicism to take place frequently of activity, has had Serious social harm.The dissemination of severe strike video, it is the important effective means of the sudden and violent probably crime of containment.In order to detect Video obtains the semantic description to scene, establishes high-rise Scene Semantics, it is necessary to understood exactly the content of video scene Contacting between description and low-level visual feature, it could identify and sudden and violent probably scene whether is included in positioning video.Traditional intelligence regards The object of frequency monitoring research is mainly target detection and classification, target following, object matching and target identification etc., it is difficult to " sudden and violent Probably " this scene makes correct description and identification.
The content of the invention
The technical problems to be solved by the invention are the defects of overcoming prior art, there is provided one kind is analyzed using crowd behaviour Video detection system and method, using the intelligent video scene understanding technology analyzed towards crowd behaviour, carry out real-time high-precision The crowd's quantity semantic information and crowd's abnormal behaviour voice messaging in complex scene are obtained, so as to identify video.
The technical scheme is that a kind of video detecting method analyzed using crowd behaviour, is comprised the following steps：
(1) video is gathered；
(2) video of collection is pre-processed；
(3) the population analysis technology based on deep learning, by mass data sample analysis, is learnt by computer self The difference of people and other targets, calculate the characteristic information with " people " significantly identified；
(4) characteristic information based on the people indicated, judge the unusual condition of various crowds, and combine a variety of crowd's rows For output safety index；
(5) by safety index compared with the threshold value set, the video higher than the threshold value is considered as security video；Less than this The video of threshold value, the region optical flow method that is combined using background subtraction and pyramid L-K algorithms extract moving region, and adopt With speed maximum come differentiate whether there is robbery in the region, fight cruelly probably behavior.
A kind of video detection system analyzed using crowd behaviour, the system are included：
Video acquisition module, for gathering video data；
Pretreatment module, for being pre-processed to the video collected；
Crowd density determination module, for pretreated video, the population analysis technology based on deep learning, passing through Mass data sample analysis, people and the difference of other targets are learnt by computer self, calculating has what is significantly identified The characteristic information of " people ", give this feature information transmission to safety index generation module；
Safety index generation module, for the characteristic information sent according to crowd density determination module, judge various crowds Unusual condition, and combine a variety of crowd behaviours, output safety index, and the safety index is sent to one behavior of crowd inspection Survey module；
Crowd's unusual checking module, for the height according to safety index, for the relatively low video of safety index, profit The region optical flow method being combined with background subtraction and pyramid L-K algorithms extracts moving region, and uses speed maximum Come the probably behavior cruelly for differentiating whether there is robbery in the region, fighting.
The beneficial effects of the present invention are：This method is based on deep learning and optical flow method is analyzed, analyzed by crowd density, Safety index calculates and the analysis of crowd's abnormal behaviour, effectively detects probably behavior cruelly present in video, can be applied to video Real-time monitoring field.
Brief description of the drawings
Fig. 1 is the video detection calculation flow chart proposed by the present invention analyzed using crowd behaviour
Embodiment
Below, carried out as described in detail below for the present invention with reference to accompanying drawing：
The system of the present invention includes following several parts：
Video acquisition module, for gathering video data；
Pretreatment module, for being pre-processed to the video collected；
Crowd density determination module：Whether conventional method is judged crowded in scene using the statistics of number.But because of monitoring Scene area is different, and simple demographics can not judge the intensive situation of personnel in scene, and user is more paid close attention in scene Crowd density, demographics provide only as assistance data.Population analysis technology of the invention based on deep learning, is no longer adopted The mode combined with artificial defined feature goes to judge whether target is " people ".But by mass data sample analysis, allow calculating Machine voluntarily learns the difference of people and other targets, calculates the characteristic information from level to level that can significantly identify " people ".Because sample covers Lid rate is larger, and during machine learning, the technology can effectively break through illuminance abrupt variation, and background is complicated, and human body parts block, should With the single difficult point for waiting conventional art of scene.
Safety index generation module：Population density is anticipated as one of attribute of crowd to actual combat work with important guidance Justice, but only can not just be made a policy with this index.Because when i.e. convenient crowd density is relatively low, various emergency situations equally have can Can fear event cruelly.Such as number have accumulated more people suddenly than sparse scene；Certain passage, there are more people to be detained；Have in scene Individual strenuous exercise；Unordered state of crowd etc..Above various states, when crowd density is smaller, it is also possible to send out Raw harmfulness event.Based on various crowd's unusual conditions mentioned above, invention introduces the concept of " confusion ", and combine more Kind crowd behaviour, output safety index.
Crowd's unusual checking module：For the relatively low video of safety index, the present invention passes through background subtraction and gold The region optical flow method that word tower L-K algorithms combine differentiates in video whether have to extract moving region using speed maximum The probably behavior cruelly such as plunder, fight.This method has certain robustness, has preferable differentiation to normal behaviour and abnormal behaviour Degree, can be applied to monitor in real time.
Described is only the instantiation of the present invention, any equivalent transformation based on the inventive method basis, belongs to this hair Within bright protection domain.

Claims (2)

1. a kind of video detecting method analyzed using crowd behaviour, it is characterised in that comprise the following steps：
(1) video is gathered；
(2) video of collection is pre-processed；
(3) the population analysis technology based on deep learning, by mass data sample analysis, by computer self learn people and The difference of other targets, calculate the characteristic information with " people " significantly identified；
(4) characteristic information based on the people indicated, judge the unusual condition of various crowds, and combine a variety of crowd behaviours, it is defeated Go out safety index；
(5) by safety index compared with the threshold value set, the video higher than the threshold value is considered as security video；Less than the threshold value Video, the region optical flow method being combined using background subtraction and pyramid L-K algorithms extracts moving region, and using speed The probably behavior cruelly for spending maximum to differentiate whether there is robbery in the region, fight.
2. a kind of video detection system analyzed using crowd behaviour, it is characterised in that the system includes：
Video acquisition module, for gathering video data；
Pretreatment module, for being pre-processed to the video collected；
Crowd density determination module, for pretreated video, the population analysis technology based on deep learning, by a large amount of Data sample is analyzed, and is learnt people and the difference of other targets by computer self, is calculated with " people's " significantly identified Characteristic information, give this feature information transmission to safety index generation module；
Safety index generation module, for the characteristic information sent according to crowd density determination module, judge that various crowds' is different Normal situation, and a variety of crowd behaviours, output safety index are combined, and the safety index is sent to one behavioral value mould of crowd Block；
Crowd's unusual checking module, for the height according to safety index, for the relatively low video of safety index, utilize the back of the body The region optical flow method that scape calculus of finite differences and pyramid L-K algorithms are combined is sentenced to extract moving region using speed maximum The probably behavior cruelly for whether do not have robbery in the region, fighting.
CN201610405681.9A 2016-06-08 2016-06-08 A kind of video detection system and method using crowd behaviour analysis Pending CN107480578A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610405681.9A CN107480578A (en) 2016-06-08 2016-06-08 A kind of video detection system and method using crowd behaviour analysis

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610405681.9A CN107480578A (en) 2016-06-08 2016-06-08 A kind of video detection system and method using crowd behaviour analysis

Publications (1)

Publication Number Publication Date
CN107480578A true CN107480578A (en) 2017-12-15

Family

ID=60594349

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610405681.9A Pending CN107480578A (en) 2016-06-08 2016-06-08 A kind of video detection system and method using crowd behaviour analysis

Country Status (1)

Country Link
CN (1) CN107480578A (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108470154A (en) * 2018-02-27 2018-08-31 燕山大学 A kind of large-scale crowd salient region detection method
CN109299657A (en) * 2018-08-14 2019-02-01 清华大学 Method and device for group behavior recognition based on semantic attention retention mechanism
CN110390226A (en) * 2018-04-16 2019-10-29 杭州海康威视数字技术股份有限公司 Crowd's event recognition method, device, electronic equipment and system
CN111160150A (en) * 2019-12-16 2020-05-15 盐城吉大智能终端产业研究院有限公司 Video monitoring crowd behavior identification method based on depth residual error neural network convolution
CN113743184A (en) * 2021-06-08 2021-12-03 中国人民公安大学 Abnormal behavior crowd detection method and device based on element mining and video analysis
CN115311591A (en) * 2021-12-09 2022-11-08 北京市基础设施投资有限公司 Early warning method and device for abnormal behaviors and intelligent camera

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090222388A1 (en) * 2007-11-16 2009-09-03 Wei Hua Method of and system for hierarchical human/crowd behavior detection
CN102682303A (en) * 2012-03-13 2012-09-19 上海交通大学 Crowd exceptional event detection method based on LBP (Local Binary Pattern) weighted social force model
CN103693532A (en) * 2013-12-26 2014-04-02 江南大学 Method of detecting violence in elevator car

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090222388A1 (en) * 2007-11-16 2009-09-03 Wei Hua Method of and system for hierarchical human/crowd behavior detection
CN102682303A (en) * 2012-03-13 2012-09-19 上海交通大学 Crowd exceptional event detection method based on LBP (Local Binary Pattern) weighted social force model
CN103693532A (en) * 2013-12-26 2014-04-02 江南大学 Method of detecting violence in elevator car

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
廖飞 等: "深度学习的人群分析系统进行大流量人群监控决策", 《安全&自动化》 *

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108470154A (en) * 2018-02-27 2018-08-31 燕山大学 A kind of large-scale crowd salient region detection method
CN108470154B (en) * 2018-02-27 2021-08-24 燕山大学 A large-scale crowd saliency region detection method
CN110390226A (en) * 2018-04-16 2019-10-29 杭州海康威视数字技术股份有限公司 Crowd's event recognition method, device, electronic equipment and system
CN110390226B (en) * 2018-04-16 2021-09-21 杭州海康威视数字技术股份有限公司 Crowd event identification method and device, electronic equipment and system
CN109299657A (en) * 2018-08-14 2019-02-01 清华大学 Method and device for group behavior recognition based on semantic attention retention mechanism
CN109299657B (en) * 2018-08-14 2020-07-03 清华大学 Group behavior identification method and device based on semantic attention retention mechanism
CN111160150A (en) * 2019-12-16 2020-05-15 盐城吉大智能终端产业研究院有限公司 Video monitoring crowd behavior identification method based on depth residual error neural network convolution
CN113743184A (en) * 2021-06-08 2021-12-03 中国人民公安大学 Abnormal behavior crowd detection method and device based on element mining and video analysis
CN113743184B (en) * 2021-06-08 2023-08-29 中国人民公安大学 Abnormal Behavior Crowd Detection Method and Device Based on Element Mining and Video Analysis
CN115311591A (en) * 2021-12-09 2022-11-08 北京市基础设施投资有限公司 Early warning method and device for abnormal behaviors and intelligent camera

Similar Documents

Publication Publication Date Title
CN107480578A (en) A kind of video detection system and method using crowd behaviour analysis
Chelli et al. A machine learning approach for fall detection and daily living activity recognition
CN110458101B (en) Method and equipment for monitoring the physical signs of prisoners based on the combination of video and equipment
CN109376639B (en) Accompanying personnel early warning system and method based on portrait recognition
CN109145789A (en) Power supply system safety work support method and system
CN105787472B (en) A method for abnormal behavior detection based on spatiotemporal Laplacian feature map learning
CN106571014A (en) Method for identifying abnormal motion in video and system thereof
CN111091060B (en) Fall and violence detection method based on deep learning
CN105404895A (en) Abnormal state identification method and identification system
CN115865988B (en) Passenger ship passenger tread event monitoring system and method by utilizing mobile phone sensor network
CN110930632B (en) Early warning system based on artificial intelligence
CN110909672A (en) Smoking action recognition method based on double-current convolutional neural network and SVM
Wang et al. Accelerometer-based human fall detection using sparrow search algorithm and back propagation neural network
CN105844269A (en) Information processing method used for fall-down detection and information processing system thereof
CN118053261A (en) Anti-spoofing early warning method, device, equipment and medium for smart campus
CN116778657A (en) A method and system for intelligently identifying intrusion behavior
CN116563758A (en) Lion head goose monitoring method, device, equipment and storage medium
Long et al. An image-based fall detection system using you only look once (yolo) algorithm to monitor elders’ fall events
CN119694062A (en) A smart fire safety integrated big data platform system
CN113239772B (en) Personnel gathering early warning method and system in self-service bank or ATM environment
Al-Tamimi et al. Face mask detection based on algorithm YOLOv5s
Arpa et al. A machine learning and deep learning integrated model to detect criminal activities
CN117333980A (en) Be used for wisdom power plant access control management system
Philip et al. Elderly fall detection using deep learning techniques
Luna-Perejon et al. Sampling Frequency Evaluation on Recurrent Neural Networks Architectures for IoT Real-time Fall Detection Devices.

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
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20171215

WD01 Invention patent application deemed withdrawn after publication