CN109117721A - A kind of pedestrian hovers detection method - Google Patents

A kind of pedestrian hovers detection method Download PDF

Info

Publication number
CN109117721A
CN109117721A CN201810734131.0A CN201810734131A CN109117721A CN 109117721 A CN109117721 A CN 109117721A CN 201810734131 A CN201810734131 A CN 201810734131A CN 109117721 A CN109117721 A CN 109117721A
Authority
CN
China
Prior art keywords
pedestrian
detection
algorithm
duration
hovers
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
CN201810734131.0A
Other languages
Chinese (zh)
Inventor
应艳丽
贠周会
王欣欣
谢吉朋
吴斌
叶超
王旭
黄江林
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.)
Jiangxi Hongdu Aviation Industry Group Co Ltd
Original Assignee
Jiangxi Hongdu Aviation Industry Group Co Ltd
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 Jiangxi Hongdu Aviation Industry Group Co Ltd filed Critical Jiangxi Hongdu Aviation Industry Group Co Ltd
Priority to CN201810734131.0A priority Critical patent/CN109117721A/en
Publication of CN109117721A publication Critical patent/CN109117721A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/103—Static body considered as a whole, e.g. static pedestrian or occupant recognition
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
    • G06T7/251—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments involving models
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/277—Analysis of motion involving stochastic approaches, e.g. using Kalman filters
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • 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; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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
    • G06T2207/30196—Human being; Person
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Human Computer Interaction (AREA)
  • Image Analysis (AREA)

Abstract

The invention belongs to Digital Image Processing and mode identification technology, hover detection method more particularly to a kind of pedestrian.It still can trace into pedestrian after the of short duration disappearance of pedestrian or partial occlusion, target will not be lost, continue to record pedestrian movement track and tracks duration, and judge whether Wander behavior to have occurred in monitoring area according to pedestrian movement's path length and tracking duration, it realizes intelligent pedestrian and hovers detection.

Description

A kind of pedestrian hovers detection method
Technical field
The invention belongs to Digital Image Processing and mode identification technology, hover the side of detection more particularly to a kind of pedestrian Method.
Background technique
With social development and the development of smart city, attention of the intelligent video monitoring system by more and more people. The people is also increasingly stronger to the security requirement of living environment, and intelligent video monitoring system has reformed into safety guarantee and prevented Defend one kind of system important tool and means.
Pedestrian hovers one kind as pedestrian's unusual checking of detection, is the important component in intelligent monitor system One of.It is monitored specific region using video analysis the relevant technologies, when detecting that someone walks up and down in the area Overlong time when, judge whether pedestrian has Wander behavior, to execute corresponding preset program.Pedestrian, which hovers, to be detected at present mainly Applied to the safety monitoring of public place and the behavioural analysis field of business place.With the propulsion of intelligent monitor system, row The application scenarios of people's Wander behavior detection will gradually increase, and importance also can be improved increasingly.
But the complexity of real monitoring scene makes the requirement of the Processing Algorithm to intelligent monitoring video higher and higher, needs It wants algorithm that can overcome a variety of different interference, possesses stronger robustness.Pedestrian hovers to detect and usually be transported according to target at present Dynamic path length and tracking duration are to determine whether for Wander behavior.But such methods still can not accurately obtain mesh at present Motion profile when being at least partially obscured after perhaps of short duration disappearance and occurring is marked because when target is at least partially obscured or of short duration disappearance When occurring afterwards again, it is more likely that target can be lost, the motion profile and duration before causing all are lost, and fresh target is re-used as It is calculated, so that testing result is not accurate enough.
Summary of the invention
It hovers detection method the purpose of the present invention is to provide a kind of pedestrian, it can be in the of short duration disappearance of pedestrian or part Still pedestrian can be traced into after blocking, will not lose target, continue to record pedestrian movement track and tracks duration, and according to row People's motion profile length judges whether Wander behavior has occurred in monitoring area with tracking duration, realizes intelligent pedestrian and hovers Detection.
The present invention to achieve the goals above, adopts the following technical scheme that
A kind of pedestrian hovers detection method, and the pedestrian hovers, and detection method includes the following steps:
The first step, in terms of the pedestrian detection in monitoring area, using target pedestrian detection algorithm to collected each Frame image carries out pedestrian detection；
When there is the pedestrian being at least partially obscured, using based on the pedestrian detection algorithm blocked；
Pedestrian if it exists can obtain the location information of pedestrian by pedestrian detection algorithm；
Second step when constant testing is to the pedestrian, obtains pedestrian position information in pedestrian detection in pedestrian track side face On the basis of, track following is carried out to the pedestrian detected using target tracking algorism；When of short duration disappear occurs in the pedestrian detected After mistake again the phenomenon that occurring when, when disappearance of short duration because of pedestrian, can not detect the location information of the pedestrian, therefore use based on prediction Track algorithm the location information of the pedestrian is predicted, to obtain the motion profile of pedestrian；And then count the track of pedestrian Length and tracking duration；
Third step, in terms of abnormal behaviour judgement, by analyzing Wander behavior feature, according to the motion profile length of pedestrian It with tracking duration, is compared with preset threshold value, determines whether the pedestrian is target of hovering in the monitoring area.
Further, in the first step, pedestrian detection is using CNN algorithm or CNN derivative based on deep learning Algorithm；
Based on the pedestrian detection algorithm blocked in the first step, using positive and negative with pedestrian's unshielding and partial occlusion Sample database carries out network model training using CNN algorithm under deep learning frame, obtains with unshielding and partial occlusion row The network model of people's detectability, then CNN pedestrian detection algorithm is using the network model of the pre-training to collected each Frame image carries out pedestrian target detection, obtains the location information of monitoring area one skilled in the art.
Further, the second step includes following sub-step:
A. when continuing to detect the pedestrian tracked in the first step, using the target tracking algorism based on detection to pedestrian Track following is carried out, and records the coordinate information of the pedestrian；
B. when pedestrian the case where of short duration disappearance occurs without detecting pedestrian in the first step, using the mesh based on prediction Mark track algorithm predicts the current location information of pedestrian, realizes pedestrian track tracking, and records the coordinate letter of the pedestrian Breath；
C. judge whether target pedestrian is to trace into for the first time, if so, recording the time T0 that the pedestrian traces into for the first time；
D. path length calculating is carried out according to the pedestrian movement's trajectory coordinates information traced into；
E. according to the time T1 currently traced into, to count the duration T (T1-T0) that pedestrian hovers in monitoring area.
Further, determine that pedestrian whether there is Wander behavior in the third step, first hovered duration T and pre- according to tracking If time threshold of hovering be compared, if be less than the threshold value, there is no Wander behaviors by the pedestrian；If being greater than or waiting In the threshold value, then continue to be judged according to pursuit path length；I.e. by pursuit path length and preset path length threshold value It is compared, if path length is less than the threshold value, the pedestrian is still without generation Wander behavior；If path length is also greater than pre- If threshold value, then determine that Wander behavior has occurred in the pedestrian in monitoring area.
Further, by obtaining pedestrian to there is the facture of phenomenon again after pedestrian's partial occlusion or of short duration disappearance Be at least partially obscured or of short duration disappearance after again occur in the case where pedestrian motion profile and tracking duration；When in monitoring area When the motion profile length and tracking duration of pedestrian are more than or equal to preset threshold value, determine that the pedestrian sends out in monitoring area Wander behavior is given birth to.
Beneficial effects of the present invention:
Hover detection method the present invention provides a kind of pedestrian, the human hair that can be expert at first portion block or of short duration disappearance after Pedestrian still can be traced into occur again in the case where, target will not be lost, continue to record pedestrian movement track and tracks duration, And judge whether Wander behavior has occurred in monitoring area according to pedestrian movement's path length and tracking duration, realize intelligence Pedestrian hovers the purpose of detection, and the present invention is of great significance and application value.
Detailed description of the invention
Fig. 1 is that pedestrian in the embodiment of the present invention for video monitoring hovers the frame-type flow chart of detection method；
Fig. 2 is the pedestrian tracking process flow diagram based on detection when detecting pedestrian in the embodiment of the present invention；
Fig. 3 is the pedestrian tracking process flow based on prediction when not detecting pedestrian in the embodiment of the present invention.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention more comprehensible, relevant drawings and implementation will be passed through below Example, invention is further described in detail.
A kind of pedestrian hovers detection method, and the pedestrian hovers, and detection method includes the following steps:
The first step, in terms of the pedestrian detection in monitoring area, using target pedestrian detection algorithm to collected each Frame image carries out pedestrian detection；
When there is the pedestrian being at least partially obscured, using based on the pedestrian detection algorithm blocked；
Pedestrian if it exists can obtain the location information of pedestrian by pedestrian detection algorithm；
Second step when constant testing is to the pedestrian, obtains pedestrian position information in pedestrian detection in pedestrian track side face On the basis of, track following is carried out to the pedestrian detected using target tracking algorism；When of short duration disappear occurs in the pedestrian detected After mistake again the phenomenon that occurring when, when disappearance of short duration because of pedestrian, can not detect the location information of the pedestrian, therefore use based on prediction Track algorithm the location information of the pedestrian is predicted, to obtain the motion profile of pedestrian；And then count the track of pedestrian Length and tracking duration；
Third step, in terms of abnormal behaviour judgement, by analyzing Wander behavior feature, according to the motion profile length of pedestrian With tracking duration, be compared with preset threshold value, determine the pedestrian whether when the monitoring area in target of hovering.
Further, in the first step, pedestrian detection is using CNN algorithm or CNN derivative based on deep learning Algorithm；
Based on the pedestrian detection algorithm blocked in the first step, using positive and negative with pedestrian's unshielding and partial occlusion Sample database carries out network model training using CNN algorithm under deep learning frame, obtains with unshielding and partial occlusion row The network model of people's detectability, then CNN pedestrian detection algorithm is using the network model of the pre-training to collected each Frame image carries out pedestrian target detection, obtains the location information of monitoring area one skilled in the art.
Further, the second step includes following sub-step:
A. when continuing to detect the pedestrian tracked in the first step, using the target tracking algorism based on detection to pedestrian Track following is carried out, and records the coordinate information of the pedestrian；
B. when pedestrian the case where of short duration disappearance occurs without detecting pedestrian in the first step, using the mesh based on prediction Mark track algorithm predicts the current location information of pedestrian, realizes pedestrian track tracking, and records the coordinate letter of the pedestrian Breath；
C. judge whether target pedestrian is to trace into for the first time, if so, recording the time T0 that the pedestrian traces into for the first time；
D. path length calculating is carried out according to the pedestrian movement's trajectory coordinates information traced into；
E. according to the time T1 currently traced into, to count the duration T (T1-T0) that pedestrian hovers in monitoring area.
Further, determine that pedestrian whether there is Wander behavior in the third step, first hovered duration T and pre- according to tracking If time threshold of hovering be compared, if be less than the threshold value, there is no Wander behaviors by the pedestrian；If being greater than or waiting In the threshold value, then continue to be judged according to pursuit path length；I.e. by pursuit path length and preset path length threshold value It is compared, if path length is less than the threshold value, the pedestrian is still without generation Wander behavior；If path length is also greater than pre- If threshold value, then determine that Wander behavior has occurred in the pedestrian in monitoring area.
Further, by obtaining pedestrian to there is the facture of phenomenon again after pedestrian's partial occlusion or of short duration disappearance Be at least partially obscured or of short duration disappearance after again occur in the case where pedestrian motion profile and tracking duration；When in monitoring area When the motion profile length and tracking duration of pedestrian are more than or equal to preset threshold value, determine that the pedestrian sends out in monitoring area Wander behavior is given birth to.
Fig. 1 is that pedestrian in the embodiment of the present invention for video monitoring hovers the frame-type flow chart of detection method.It is such as attached Shown in figure Fig. 1, this method includes 5 steps: image preprocessing 110, target pedestrian tracking 130, is hesitated at target pedestrian detection 120 Behavior of wandering determines 140 and target designation 150 of hovering.
In step 110, video image frame is obtained from monitor video, some pretreatments is carried out to image, such as image is gone It makes an uproar, the processing such as image enhancement, change of scale, improves the quality of image, prepare for next step target detection.
Target pedestrian detecting step 120 detects the target pedestrian in monitoring scene according to pretreated image information. The algorithm of target detection used in this step is CNN algorithm or CNN derivative algorithm based on deep learning.The detection of the algorithm Process is broadly divided into three steps, first the positive and negative sample set of preparation pedestrian, network model training, pedestrian detection.
Environment is complicated in the monitoring scene of reality, often because various extraneous factors cause pedestrian to be at least partially obscured.So with In the positive negative example base of pedestrian of training network model, pedestrian's positive sample concentrate the image not to be blocked comprising pedestrian and The image that pedestrian is at least partially obscured；Pedestrian's negative sample concentrates the image comprising not pedestrian.
Network model training uses CNN algorithm or CNN derivative algorithm based on deep learning frame to be trained, such as The deep learning Open Framework of the current main-streams such as tensorflow frame, caffe frame, be arranged training process in learning rate, The parameters such as convergent iterations number, convolution nucleus number, these parameters influence whether network model for pedestrian detection accuracy, therefore It needs constantly to carry out arameter optimization in training process.
Target pedestrian detection finally is carried out using the network model and CNN algorithm or CNN derivative algorithm of training, because model is examined Consider the case where pedestrian is at least partially obscured, so be not blocked or partial occlusion in the case where is able to detect that row in pedestrian People, while the pedestrian's coordinate location information that record detects.
In step 130, target pedestrian tracking needs to be divided into two kinds of situations and carries out different processing, i.e., detects in step 120 The case where to the pedestrian tracked facture and facture the case where pedestrian is not detected.
In the present embodiment, as shown in Fig. 2, being pedestrian tracking flow chart when detecting tracking pedestrians, mainly use Target tracking algorism based on detection tracks pedestrian, and the location information that will test is believed as the position of tracking pedestrians Breath.And the motion profile of the pedestrian is recorded, while judging whether pedestrian is tracked to for the first time, if so, record target pedestrian is first The secondary time T0 being tracked to.
In the present embodiment, as shown in figure 3, being pedestrian tracking flow chart when pedestrian is not detected, mainly uses and be based on The target tracking algorism of prediction, such as the target following of mean shift algorithm, Kalman filtering algorithm, particle filter algorithm mainstream Algorithm.It further says, particle filter algorithm is specifically used in the present embodiment, if in n-th frame, the movement rail of tracking pedestrians Mark location information is xn, just with x in the (n+1)th framenCentered on according to state transition function predict generate m sampling particle {xi}I=0 ..., m, posterior probability density is approached using the weighted sum of these particles, obtains the location information of tentative predictionIt recyclesEstimated location information truth value x is corrected with state Posterior distrbutionpn+1.The true value is For the pedestrian position information traced into present frame.Meanwhile record the motion profile of the pedestrian, and judge pedestrian whether for the first time by It traces into, if so, the time T0 that record target pedestrian is tracked to for the first time.
In step 140, the judgement of pedestrian's Wander behavior.First according to tracking hover duration T and it is preset hover time threshold into Row compares, if being less than the threshold value, there is no Wander behaviors by the pedestrian；If being greater than or equal to the threshold value, continue root Judged according to pursuit path length.Pursuit path length is compared with preset path length threshold value, if track is long Degree is less than the threshold value, then the pedestrian is still without generation Wander behavior；If path length also greater than preset threshold, then determines the row Wander behavior has occurred in people in monitoring area.
In step 150, the pedestrian for being judged as hovering in video monitoring is demarcated, determines mesh using red rectangle circle Pedestrian is marked, is warned in manner shown.
It is merely exemplary in above-described embodiment to illustrate the principle of the present invention and process rather than limit, without departing from essence of the invention Under the premise of mind and principle, the present invention in embodiment can there are many variations.Therefore, all that those skilled in the art are come Say it is obvious change (equivalent replacement, improvement etc.), should all be included within the scope of the claims cover.This hair Bright range claimed is only defined by described claims.

Claims (5)

  1. The detection method 1. a kind of pedestrian hovers, it is characterised in that: the pedestrian hovers, and detection method includes the following steps:
    The first step, in terms of the pedestrian detection in monitoring area, using target pedestrian detection algorithm to collected each frame figure As carrying out pedestrian detection；
    When there is the pedestrian being at least partially obscured, using based on the pedestrian detection algorithm blocked；
    Pedestrian if it exists can obtain the location information of pedestrian by pedestrian detection algorithm；
    Second step when constant testing is to the pedestrian, obtains the base of pedestrian position information in pedestrian detection in pedestrian track side face On plinth, track following is carried out to the pedestrian detected using target tracking algorism；After there is of short duration disappearance in the pedestrian detected When the phenomenon that occurring again, when disappearance of short duration because of pedestrian, can not detect the location information of the pedestrian, thus use based on prediction with Track algorithm predicts the location information of the pedestrian, to obtain the motion profile of pedestrian；And then count the path length of pedestrian With tracking duration；
    Third step, in terms of abnormal behaviour judgement, by analyze Wander behavior feature, according to the motion profile length of pedestrian and with Track duration is compared with preset threshold value, determines whether the pedestrian is target of hovering in the monitoring area.
  2. The detection method 2. pedestrian according to claim 1 hovers, it is characterised in that:
    In the first step, pedestrian detection is using CNN algorithm or CNN derivative algorithm based on deep learning；
    Based on the pedestrian detection algorithm blocked in the first step, the positive negative sample with pedestrian's unshielding and partial occlusion is utilized Library carries out network model training using CNN algorithm under deep learning frame, obtains examining with unshielding and partial occlusion pedestrian The network model of survey ability, then CNN pedestrian detection algorithm is using the network model of the pre-training to collected each frame figure As carrying out pedestrian target detection, the location information of monitoring area one skilled in the art is obtained.
  3. The detection method 3. pedestrian according to claim 1 hovers, it is characterised in that:
    The second step includes following sub-step:
    A. when continuing to detect the pedestrian tracked in the first step, pedestrian is carried out using the target tracking algorism based on detection Track following, and record the coordinate information of the pedestrian；
    B. when in the first step pedestrian occur of short duration disappearance the case where without detecting pedestrian when, using the target based on prediction with Track algorithm predicts the current location information of pedestrian, realizes pedestrian track tracking, and record the coordinate information of the pedestrian；
    C. judge whether target pedestrian is to trace into for the first time, if so, recording the time T0 that the pedestrian traces into for the first time；
    D. path length calculating is carried out according to the pedestrian movement's trajectory coordinates information traced into；
    E. according to the time T1 currently traced into, to count the duration T(T1-T0 that pedestrian hovers in monitoring area).
  4. The detection method 4. pedestrian according to claim 1 hovers, it is characterised in that:
    Determine that pedestrian whether there is Wander behavior in the third step, is first hovered and duration T and preset hovered the time according to tracking Threshold value is compared, if being less than the threshold value, there is no Wander behaviors by the pedestrian；If being greater than or equal to the threshold value, Continuation is judged according to pursuit path length；Pursuit path length is compared with preset path length threshold value, if Path length is less than the threshold value, then the pedestrian is still without generation Wander behavior；If path length is then sentenced also greater than preset threshold Wander behavior has occurred in the fixed pedestrian in monitoring area.
  5. The detection method 5. pedestrian according to claim 1 hovers, it is characterised in that:
    By to there is the facture of phenomenon again after pedestrian's partial occlusion or of short duration disappearance, obtain pedestrian being at least partially obscured or The motion profile of pedestrian and tracking duration occur after the of short duration disappearance of person again in the case where；When the motion profile of monitoring area one skilled in the art When length and tracking duration are more than or equal to preset threshold value, determine that Wander behavior has occurred in the pedestrian in monitoring area.
CN201810734131.0A 2018-07-06 2018-07-06 A kind of pedestrian hovers detection method Pending CN109117721A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810734131.0A CN109117721A (en) 2018-07-06 2018-07-06 A kind of pedestrian hovers detection method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810734131.0A CN109117721A (en) 2018-07-06 2018-07-06 A kind of pedestrian hovers detection method

Publications (1)

Publication Number Publication Date
CN109117721A true CN109117721A (en) 2019-01-01

Family

ID=64823063

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810734131.0A Pending CN109117721A (en) 2018-07-06 2018-07-06 A kind of pedestrian hovers detection method

Country Status (1)

Country Link
CN (1) CN109117721A (en)

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110264497A (en) * 2019-06-11 2019-09-20 浙江大华技术股份有限公司 Track determination method and device, the storage medium, electronic device of duration
CN110688896A (en) * 2019-08-23 2020-01-14 北京正安维视科技股份有限公司 Pedestrian loitering detection method
CN110969115A (en) * 2019-11-28 2020-04-07 深圳市商汤科技有限公司 Pedestrian event detection method and device, electronic equipment and storage medium
CN111160203A (en) * 2019-12-23 2020-05-15 中电科新型智慧城市研究院有限公司 Loitering and lingering behavior analysis method based on head and shoulder model and IOU tracking
CN111461041A (en) * 2020-04-07 2020-07-28 西安交通大学 Multi-factor joint abnormal pedestrian distinguishing method based on generation of countermeasure network
CN111860318A (en) * 2020-07-20 2020-10-30 杭州品茗安控信息技术股份有限公司 Construction site pedestrian loitering detection method, device, equipment and storage medium
CN112640419A (en) * 2020-02-28 2021-04-09 深圳市大疆创新科技有限公司 Following method, movable platform, device and storage medium
CN112633150A (en) * 2020-12-22 2021-04-09 中国华戎科技集团有限公司 Target trajectory analysis-based retention loitering behavior identification method and system
CN112733814A (en) * 2021-03-30 2021-04-30 上海闪马智能科技有限公司 Deep learning-based pedestrian loitering retention detection method, system and medium
CN112990058A (en) * 2021-03-30 2021-06-18 北京邮电大学 Multi-target pedestrian loitering detection method based on trajectory analysis
CN113408333A (en) * 2021-04-27 2021-09-17 上海工程技术大学 Method for distinguishing pedestrian traffic behaviors in subway station based on video data
CN113495270A (en) * 2020-04-07 2021-10-12 富士通株式会社 Monitoring device and method based on microwave radar
CN113591904A (en) * 2021-06-17 2021-11-02 浙江大华技术股份有限公司 Sojourn time statistical method, goods adjusting method and related device
CN114066944A (en) * 2022-01-17 2022-02-18 天津聚芯光禾科技有限公司 Optical module production workshop worker post behavior analysis method based on pedestrian tracking

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101577006A (en) * 2009-06-15 2009-11-11 北京中星微电子有限公司 Loitering detecting method and loitering detecting system in video monitoring
CN101751549A (en) * 2008-12-03 2010-06-23 财团法人工业技术研究院 Method for tracking moving object
CN101770648A (en) * 2009-01-06 2010-07-07 北京中星微电子有限公司 Video monitoring based loitering system and method thereof
CN101853511A (en) * 2010-05-17 2010-10-06 哈尔滨工程大学 Anti-shelter target trajectory predicting and tracking method
WO2012074366A2 (en) * 2010-12-02 2012-06-07 Mimos Bhd. A system and a method for detecting a loitering event
CN103116959A (en) * 2013-01-25 2013-05-22 上海博超科技有限公司 Analyzing and recognizing method for abnormal behaviors in intelligent videos
CN105184812A (en) * 2015-07-21 2015-12-23 复旦大学 Target tracking-based pedestrian loitering detection algorithm
CN106128053A (en) * 2016-07-18 2016-11-16 四川君逸数码科技股份有限公司 A kind of wisdom gold eyeball identification personnel stay hover alarm method and device
CN107564034A (en) * 2017-07-27 2018-01-09 华南理工大学 The pedestrian detection and tracking of multiple target in a kind of monitor video

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101751549A (en) * 2008-12-03 2010-06-23 财团法人工业技术研究院 Method for tracking moving object
CN101770648A (en) * 2009-01-06 2010-07-07 北京中星微电子有限公司 Video monitoring based loitering system and method thereof
CN101577006A (en) * 2009-06-15 2009-11-11 北京中星微电子有限公司 Loitering detecting method and loitering detecting system in video monitoring
CN101853511A (en) * 2010-05-17 2010-10-06 哈尔滨工程大学 Anti-shelter target trajectory predicting and tracking method
WO2012074366A2 (en) * 2010-12-02 2012-06-07 Mimos Bhd. A system and a method for detecting a loitering event
CN103116959A (en) * 2013-01-25 2013-05-22 上海博超科技有限公司 Analyzing and recognizing method for abnormal behaviors in intelligent videos
CN105184812A (en) * 2015-07-21 2015-12-23 复旦大学 Target tracking-based pedestrian loitering detection algorithm
CN106128053A (en) * 2016-07-18 2016-11-16 四川君逸数码科技股份有限公司 A kind of wisdom gold eyeball identification personnel stay hover alarm method and device
CN107564034A (en) * 2017-07-27 2018-01-09 华南理工大学 The pedestrian detection and tracking of multiple target in a kind of monitor video

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
WANLI OUYANG ET AL.: "Joint Deep Learning for Pedestrian Detection", 《2013 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION》 *
王斌: "基于深度学习的行人检测", 《中国优秀硕士学位论文全文数据库 信息科技辑》 *
赵卫峰 等: "基于运动轨迹分量的行人徘徊行为检测研究", 《视频应用与工程》 *

Cited By (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110264497B (en) * 2019-06-11 2021-09-17 浙江大华技术股份有限公司 Method and device for determining tracking duration, storage medium and electronic device
CN110264497A (en) * 2019-06-11 2019-09-20 浙江大华技术股份有限公司 Track determination method and device, the storage medium, electronic device of duration
CN110688896A (en) * 2019-08-23 2020-01-14 北京正安维视科技股份有限公司 Pedestrian loitering detection method
WO2021103423A1 (en) * 2019-11-28 2021-06-03 深圳市商汤科技有限公司 Method and apparatus for detecting pedestrian events, electronic device and storage medium
CN110969115A (en) * 2019-11-28 2020-04-07 深圳市商汤科技有限公司 Pedestrian event detection method and device, electronic equipment and storage medium
TWI757756B (en) * 2019-11-28 2022-03-11 大陸商深圳市商湯科技有限公司 Pedestrian event detection method and detection device, electronic device and computer-readable storage medium
CN111160203A (en) * 2019-12-23 2020-05-15 中电科新型智慧城市研究院有限公司 Loitering and lingering behavior analysis method based on head and shoulder model and IOU tracking
CN111160203B (en) * 2019-12-23 2023-05-16 中电科新型智慧城市研究院有限公司 Loitering and stay behavior analysis method based on head-shoulder model and IOU tracking
CN112640419A (en) * 2020-02-28 2021-04-09 深圳市大疆创新科技有限公司 Following method, movable platform, device and storage medium
CN112640419B (en) * 2020-02-28 2022-07-05 深圳市大疆创新科技有限公司 Following method, movable platform, device and storage medium
CN111461041A (en) * 2020-04-07 2020-07-28 西安交通大学 Multi-factor joint abnormal pedestrian distinguishing method based on generation of countermeasure network
CN113495270A (en) * 2020-04-07 2021-10-12 富士通株式会社 Monitoring device and method based on microwave radar
CN111860318A (en) * 2020-07-20 2020-10-30 杭州品茗安控信息技术股份有限公司 Construction site pedestrian loitering detection method, device, equipment and storage medium
CN112633150A (en) * 2020-12-22 2021-04-09 中国华戎科技集团有限公司 Target trajectory analysis-based retention loitering behavior identification method and system
CN112990058A (en) * 2021-03-30 2021-06-18 北京邮电大学 Multi-target pedestrian loitering detection method based on trajectory analysis
CN112733814B (en) * 2021-03-30 2021-06-22 上海闪马智能科技有限公司 Deep learning-based pedestrian loitering retention detection method, system and medium
CN112733814A (en) * 2021-03-30 2021-04-30 上海闪马智能科技有限公司 Deep learning-based pedestrian loitering retention detection method, system and medium
CN113408333A (en) * 2021-04-27 2021-09-17 上海工程技术大学 Method for distinguishing pedestrian traffic behaviors in subway station based on video data
CN113591904A (en) * 2021-06-17 2021-11-02 浙江大华技术股份有限公司 Sojourn time statistical method, goods adjusting method and related device
CN114066944A (en) * 2022-01-17 2022-02-18 天津聚芯光禾科技有限公司 Optical module production workshop worker post behavior analysis method based on pedestrian tracking
CN114066944B (en) * 2022-01-17 2022-04-12 天津聚芯光禾科技有限公司 Optical module production workshop worker post behavior analysis method based on pedestrian tracking

Similar Documents

Publication Publication Date Title
CN109117721A (en) A kind of pedestrian hovers detection method
Chaudhary et al. Multiple anomalous activity detection in videos
CN105574501B (en) A kind of stream of people's video detecting analysis system
CN110443109A (en) Abnormal behaviour monitor processing method, device, computer equipment and storage medium
CN111726586A (en) Production system operation standard monitoring and reminding system
EP3035235B1 (en) Method for setting a tridimensional shape detection classifier and method for tridimensional shape detection using said shape detection classifier
CN105184258A (en) Target tracking method and system and staff behavior analyzing method and system
CN108197575A (en) A kind of abnormal behaviour recognition methods detected based on target detection and bone point and device
CN108229407A (en) A kind of behavioral value method and system in video analysis
CN103854027A (en) Crowd behavior identification method
CN111126153B (en) Safety monitoring method, system, server and storage medium based on deep learning
CN105227907A (en) Based on the nothing supervision anomalous event real-time detection method of video
CN106792883A (en) Sensor network abnormal deviation data examination method and system
Liu et al. Exemplar-AMMs: Recognizing crowd movements from pedestrian trajectories
Senaratne et al. Comparing two video-based techniques for driver fatigue detection: classification versus optical flow approach
Yang et al. Bird's-eye view social distancing analysis system
Rodriguez et al. Deep learning applied to capacity control in commercial establishments in times of COVID-19
CN104077571B (en) A kind of crowd's anomaly detection method that model is serialized using single class
CN103092925B (en) A kind of video abstraction generating method and device
CN108364303A (en) A kind of video camera intelligent-tracking method with secret protection
CN109615641A (en) Multiple target pedestrian tracking system and tracking based on KCF algorithm
CN108171187A (en) A kind of abnormal behaviour automatic identifying method and device based on the extraction of bone point
Kaltsa et al. Timely, robust crowd event characterization
Liang et al. Blink detection and duration estimation by using adaptive threshold with considering individual difference
Ding et al. Identifying unsafe behaviors of construction workers through an unsupervised multi-anomaly GAN approach

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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20190101