CN108830204B - Method for detecting abnormality in target-oriented surveillance video - Google Patents

Method for detecting abnormality in target-oriented surveillance video Download PDF

Info

Publication number
CN108830204B
CN108830204B CN201810558987.7A CN201810558987A CN108830204B CN 108830204 B CN108830204 B CN 108830204B CN 201810558987 A CN201810558987 A CN 201810558987A CN 108830204 B CN108830204 B CN 108830204B
Authority
CN
China
Prior art keywords
target
abnormal
abnormality
foreground
appearance
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201810558987.7A
Other languages
Chinese (zh)
Other versions
CN108830204A (en
Inventor
李小丹
李卫海
刘乾坤
刘斌
俞能海
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.)
University of Science and Technology of China USTC
Original Assignee
University of Science and Technology of China USTC
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 University of Science and Technology of China USTC filed Critical University of Science and Technology of China USTC
Priority to CN201810558987.7A priority Critical patent/CN108830204B/en
Publication of CN108830204A publication Critical patent/CN108830204A/en
Application granted granted Critical
Publication of CN108830204B publication Critical patent/CN108830204B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

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
    • 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
    • G06V20/42—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items of sport video content
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/215—Motion-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras
    • G06T7/77—Determining position or orientation of objects or cameras using statistical methods
    • 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
    • G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
    • 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
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/07—Target detection

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computational Linguistics (AREA)
  • Image Analysis (AREA)
  • Closed-Circuit Television Systems (AREA)

Abstract

The invention discloses a method for detecting abnormity in a target-oriented surveillance video, which comprises the following steps: for each video frame, extracting three-channel data fused with appearance and motion information, and sending the three-channel data into a target detection network to extract the position, size and category of a foreground target in the video frame; judging whether the foreground target belongs to position abnormality and appearance abnormality according to the position, the size and the category of each foreground target, and obtaining a corresponding target abnormality score; when the two types of abnormal objects are not included, extracting the manual characteristics of the foreground object to judge whether the abnormal objects belong to the abnormal movement, and obtaining a corresponding object abnormal score; tracking abnormal targets which are not abnormal in position to obtain a final abnormal target set and corresponding target abnormal scores; and determining whether the corresponding abnormal target is abnormal or not by judging whether the target abnormality score exceeds an abnormality threshold or not. The method can improve the accuracy of target anomaly detection.

Description

Method for detecting abnormality in target-oriented surveillance video
Technical Field
The invention relates to the technical field of anomaly detection, in particular to a method for detecting anomalies in a target-oriented surveillance video.
Background
Detection of abnormal behavior in video is an important problem in the field of computer vision, and aims to find abnormal events in video in time given a video or when the video is processed in real time in an online system. Anomaly detection is also a difficult problem because anomalies cannot be listed, so that anomaly detection cannot be a supervised problem. The algorithm needs to automatically detect and locate behaviors (violating traffic rules, fighting, illegal theft, etc.) in the video which do not conform to the conventional mode. The existing methods are mainly divided into two types:
(1) a method of block-based and analyzing features thereof. The core idea of the algorithm is to divide a video into blocks, extract features of the blocks and analyze whether the blocks are abnormal or not. Such methods lack semantic understanding of the scene, are easy to split a single target, are not easy for individual analysis, and can limit the detection effect.
(2) A method based on trajectory analysis. The method comprises the steps of obtaining a track graph of a moving target in a video by using a target tracking algorithm, and then analyzing whether a certain track belongs to an abnormal track or not by using a specific track analysis method, wherein the method can only analyze the abnormity of speed, direction and the like on a track level, neglects scene information, and sharply reduces the tracking effect due to shielding.
Furthermore, current algorithms rarely take into account location anomalies.
Disclosure of Invention
The invention aims to provide a method for detecting the abnormality in a target-oriented surveillance video, which can improve the accuracy of target abnormality detection.
The purpose of the invention is realized by the following technical scheme:
a method for detecting abnormality in a target-oriented surveillance video includes:
for each video frame, extracting three-channel data fused with appearance and motion information, and sending the three-channel data into a target detection network to extract the position, size and category of a foreground target in the video frame;
judging whether the foreground target belongs to position abnormality and appearance abnormality according to the position, the size and the category of each foreground target, and obtaining a corresponding target abnormality score; when the two types of abnormal objects are not included, extracting the manual characteristics of the foreground object to judge whether the abnormal objects belong to the abnormal movement, and obtaining a corresponding object abnormal score;
tracking abnormal targets which are not abnormal in position to obtain a final abnormal target set and corresponding target abnormal scores; and determining whether the corresponding abnormal target is abnormal or not by judging whether the target abnormality score exceeds an abnormality threshold or not.
According to the technical scheme provided by the invention, on one hand, the target in the video is extracted by using target detection, so that the target can be effectively prevented from being split, and accurate analysis on a single target is facilitated; in addition, the existing target detection has a poor detection effect on dim and fuzzy scenes, so that the target detection based on the dynamic graph is provided, and the target close to the background color can be effectively detected. On the other hand, target detection and abnormal tracking are effectively combined, the condition of missing detection possibly caused by target detection can be effectively avoided, meanwhile, the target detection can provide a plurality of groups for tracking, and the tracking is not influenced by shielding and is more accurate.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on the drawings without creative efforts.
Fig. 1 is a flowchart of an anomaly detection method in a target-oriented surveillance video according to an embodiment of the present invention;
FIG. 2 is a diagram of a dynamic graph and a result of target detection based on the dynamic graph according to an embodiment of the present invention;
FIG. 3 is a diagram illustrating the effect of anomaly detection on a data set UCSD according to an embodiment of the present invention;
fig. 4 is a schematic diagram illustrating a comparison result between the scheme of the present invention and the existing algorithm provided in the embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention are clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present invention without making any creative effort, shall fall within the protection scope of the present invention.
The embodiment of the invention provides a method for detecting abnormality in a target-oriented surveillance video, which mainly comprises the following steps as shown in figure 1:
1. for each video frame, three-channel data fused with appearance and motion information is extracted and sent to a target detection network to extract the position, size and category of a foreground target in the video frame.
In the embodiment of the invention, for each video frame, the optical flow information and the corresponding gray-scale map are calculated, and the optical flow direction and size in the gray-scale map and the optical flow information form three-channel data similar to an HSI image; the three-channel data fuses appearance and motion information, namely, dynamic images.
And after a dynamic graph corresponding to each video frame is obtained, the dynamic graph is sent into a target detection network (RFCN) for identification, and the position, the size and the category of a foreground target are extracted.
Fig. 2 shows an RGB image and a target detection effect image corresponding to the proposed dynamic image, and it can be seen from the image that the left dynamic image can make foreground targets in the image clearer, and some targets in the dynamic image cannot be detected by using the original image for target detection, which proves the effectiveness of the target detection algorithm based on the dynamic image, because the colors of the target and the background are too close to each other in appearance and are difficult to detect only by appearance information.
2. Judging whether the foreground target belongs to position abnormality and appearance abnormality according to the position, the size and the category of each foreground target, and obtaining a corresponding target abnormality score; when the two types of abnormal objects are not included, extracting the manual characteristics of the foreground object to judge whether the abnormal objects belong to the abnormal movement, and obtaining a corresponding abnormal object score.
In the embodiment of the invention, the anomaly detection is mainly divided into: position anomaly detection, appearance anomaly detection and motion anomaly detection; mainly as follows:
1) and detecting position abnormality.
Using a principal component analysis method to respectively take the background and the foreground of an input video frame as a low-rank matrix and an error sparse matrix, and combining the position and the size of a foreground target in continuous video frames to obtain an active region in an image; judging whether the lower body of the current foreground target is completely positioned in the inactive area or not according to the position and the size of the current foreground target; if yes, the current foreground target belongs to position abnormality, and the corresponding target abnormality score is recorded as 1; otherwise, the position is not abnormal, and whether the appearance is abnormal is judged.
For example, after obtaining an active region in an image, an inactive region, e.g., a lawn; when a certain pedestrian tramples the lawn, the pedestrian is regarded as a foreground object, and the detection process can be used for judging that the lower half of the pedestrian is located in the inactive area, so that the foreground object belongs to the position abnormality.
2) And detecting appearance abnormity.
If the category of the current foreground target does not appear in the training set and the probability is greater than a set value (for example, 0.9), the current foreground target belongs to appearance abnormality, and the probability is the corresponding target abnormality score.
3) Motion anomaly detection
If the target does not belong to the two types of abnormalities, whether the abnormal motion mode exists or not is continuously judged according to the manual characteristics corresponding to the foreground target.
In the embodiment of the invention, two manual features are provided, namely optical flow angle histogram variance HVOFA and average kinetic energy En;
setting total N pixel points of a current foreground target, equally dividing an optical flow angle interval into B intervals, and counting the frequency of each interval angle in the target by an optical flow angle Histogram (HOFA), namely:
Figure BDA0001682657940000041
in the above formula, fbRepresenting the frequency of occurrence of the corresponding optical flow angle in the b-th interval;
then, the optical flow angle histogram variance HVOFA is calculated by:
Figure BDA0001682657940000042
in the above formula, the first and second carbon atoms are,
Figure BDA0001682657940000043
represents the average frequency; the above formula is converted from the Cauchy inequality. The inequality is taken when all angles are in the same interval. In a crowd scene, for example, a car, a skateboard and the like are all abnormal, the motion directions of pixel points in a foreground target are basically consistent, so that a large HOFA exists, pedestrians have a lot of limb motions, the motion directions of the pixel points are distributed discretely, the HOFA is small, and the HOFA can be used for effectively discovering the abnormality.
However, if there are some abnormalities such as skating or other objects with a fast speed, the average kinetic energy En can be used for judgment. Suppose that the magnitude of the optical flow velocity of the ith pixel point is viThen, the average kinetic energy En of the current foreground target is:
Figure BDA0001682657940000044
if the difference between the HOFA and the En of the current foreground target and the average value of the HOFA and the En of all foreground targets with the same pixel point number exceeds a preset standard, the current foreground target is abnormal in motion, and the corresponding target abnormal score is the sum of HVOFA and En.
3. Tracking abnormal targets which are not abnormal in position to obtain a final abnormal target set and corresponding target abnormal scores; and determining whether the corresponding abnormal target is abnormal or not by judging whether the target abnormality score exceeds an abnormality threshold or not.
In the embodiment of the invention, the abnormal target obtained in the last step is tracked to reduce the missing detection and the false detection of the target. Specifically, for each anomalous target, it may not be captured in some frames due to illumination, occlusion, etc., resulting in some missed detections.
In the embodiment of the invention, when a certain foreground target is detected to be abnormal in position, the corresponding target abnormal score is 1; generally, the target anomaly maximum is defined as 1, so a positional anomaly can be directly identified as an anomaly. That is, the present step mainly tracks an abnormal target other than a positional abnormality (i.e., an appearance abnormality or a motion abnormality).
Placing abnormal targets with non-position abnormality into an abnormal target set, taking each abnormal target in the abnormal target set as an initial abnormal target, tracking by adopting a target tracking method to complement the position of each video frame, wherein the tracked abnormal target has the same target abnormality score as that of the corresponding initial abnormal target; if a certain abnormal target appears in the j-th frame and the k-th frame, tracking according to the target information of the j-th frame, when the k-th frame is tracked, if the overlap area between the tracked bounding box (target frame) and the bounding box obtained by target abnormal detection in the step 2 is more than P percent (for example, 60 percent) of the union of the two, the two are the same target, and the subsequent video frame continues subsequent tracking by taking the bounding box obtained by target abnormal detection as an initial target until the next new bounding box is encountered;
for some abnormal targets which are detected by mistake, clustering the displacement of the centers of the abnormal targets in five frames according to the extracted motion tracks; assume that the center position of the target at the j-th frame is (x)j,yj) Then its displacement within five frames is:
Figure BDA0001682657940000051
if the calculated displacement offset is smaller than the set threshold value, the displacement offset is regarded as false detection, the abnormal target which is detected by the false detection is marked as a normal target, and the abnormal target is removed from the abnormal target set;
after the processing, all foreground targets with abnormal appearance and abnormal motion can be determined, and the foreground targets determined before are combined to be detected as foreground targets with abnormal positions, so that a final abnormal target set is formed; judging whether the target abnormality score of each abnormal target exceeds an abnormality threshold value, if so, determining that all pixels in the abnormal target are abnormal; otherwise, the corresponding abnormal target is normal. After the process is finished, the system gives a prompt according to the finally detected abnormal target and displays the position of the abnormal target.
Based on the scheme of the embodiment of the invention, abnormal segments can be quickly found out from a large number of monitoring videos, and personnel can be assisted to quickly find out required segments from a large number of videos.
As shown in fig. 3, in order to perform an anomaly detection effect map on the data set UCSD based on the above scheme, the boxed portions in the left and right images are the finally detected anomaly targets.
In addition, in order to illustrate the effect of the above-mentioned scheme of the present invention, the comparison is performed with the conventional algorithm, and the comparison result is shown in fig. 4, and in fig. 4(a) to 4(d), the larger the curve coverage area is, the higher the algorithm accuracy is. It can be seen that the effect of the scheme of the invention is superior to that of the majority of the schemes at present; the calculation performance and calculation efficiency of some existing algorithms similar to the effect of the scheme of the invention are far lower than those of the scheme of the invention.
In summary, the above solution of the embodiment of the present invention has the following advantages compared with the prior art:
firstly, the method uses target detection to extract the target in the video, can effectively avoid splitting the target, and is more beneficial to accurately analyzing a single target. Secondly, the existing target detection has poor detection effect on dim and fuzzy scenes, so the invention provides the target detection based on the dynamic graph, and the target which is close to the background color can be effectively detected.
Secondly, the invention provides a new optical flow angle histogram variance of manual features, which has scale invariance because optical flow assignment information is not used, so that the features can still have good capability of distinguishing anomalies when relating to a scene with depth of field.
Thirdly, the target detection and the abnormal tracking are effectively combined, the missed detection of the target possibly caused by the target detection can be effectively avoided, and meanwhile, the target detection can provide a plurality of target reference frames for the tracking, so that the tracking is not influenced by shielding and is more accurate.
Through the above description of the embodiments, it is clear to those skilled in the art that the above embodiments can be implemented by software, and can also be implemented by software plus a necessary general hardware platform. With this understanding, the technical solutions of the embodiments can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a usb disk, a removable hard disk, etc.), and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods according to the embodiments of the present invention.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims (4)

1. A method for detecting abnormality in a target-oriented surveillance video, comprising:
for each video frame, extracting three-channel data fused with appearance and motion information, and sending the three-channel data into a target detection network to extract the position, size and category of a foreground target in the video frame;
judging whether the foreground target belongs to position abnormality and appearance abnormality according to the position, the size and the category of each foreground target, and obtaining a corresponding target abnormality score; when the two types of abnormal objects are not included, extracting the manual characteristics of the foreground object to judge whether the abnormal objects belong to the abnormal movement, and obtaining a corresponding object abnormal score;
tracking abnormal targets which are not abnormal in position to obtain a final abnormal target set and corresponding target abnormal scores; determining whether the corresponding abnormal target is abnormal by judging whether the target abnormality score exceeds an abnormality threshold;
extracting two manual features, namely an optical flow angle histogram variance HVOFA and an average kinetic energy En, of a foreground object which does not belong to position abnormality and appearance abnormality;
setting total N pixel points of a current foreground target, equally dividing an optical flow angle interval into B intervals, and counting the frequency of each interval angle in the target by an optical flow angle Histogram (HOFA), namely:
Figure FDA0003225805420000011
in the above formula, fbRepresenting the frequency of occurrence of the corresponding optical flow angle in the b-th interval;
then, the optical flow angle histogram variance HVOFA is calculated by:
Figure FDA0003225805420000012
in the above formula, the first and second carbon atoms are,
Figure FDA0003225805420000013
represents the average frequency;
suppose that the magnitude of the optical flow velocity of the ith pixel point is viThen, the average kinetic energy En of the current foreground target is:
Figure FDA0003225805420000014
if the difference between the HOFA and the En of the current foreground target and the average value of the HOFA and the En of all foreground targets with the same pixel point number exceeds a preset standard, the current foreground target is abnormal in motion, and the corresponding target abnormal score is the sum of HVOFA and En.
2. The method of detecting abnormalities in object-facing surveillance video according to claim 1,
calculating optical flow information and a corresponding gray-scale image of each video frame, wherein the optical flow direction and size in the gray-scale image and the optical flow information form three-channel data; the three-channel data fuses appearance and motion information, namely, dynamic images.
3. The method as claimed in claim 1, wherein the determining whether the foreground object belongs to position abnormality and appearance abnormality according to the position, size and category of each foreground object, and obtaining the corresponding object abnormality score comprises:
using a principal component analysis method to respectively take the background and the foreground of an input video frame as a low-rank matrix and an error sparse matrix, and combining the position and the size of a foreground target in continuous video frames to obtain an active region in an image; judging whether the lower body of the current foreground target is completely positioned in the inactive area or not according to the position and the size of the current foreground target; if yes, the current foreground target belongs to position abnormality, and the corresponding target abnormality score is recorded as 1; otherwise, the position is not abnormal, and whether the position belongs to the appearance abnormity is judged;
and if the category of the current foreground target does not appear in the training set and the probability is greater than a set value, determining that the current foreground target belongs to abnormal appearance, wherein the probability is the corresponding target abnormal score.
4. The method of detecting abnormalities in object-facing surveillance video according to claim 1,
placing abnormal targets with non-position abnormality into an abnormal target set, taking each abnormal target in the abnormal target set as an initial abnormal target, tracking by adopting a target tracking method to complement the position of each video frame, wherein the tracked abnormal target has the same target abnormality score as that of the corresponding initial abnormal target; if a certain abnormal target appears in the jth frame and the kth frame, tracking according to the target information of the jth frame, when the kth frame is tracked, if the overlap area between the tracked bounding box and the bounding box obtained by target detection is greater than P% of the union of the two, the two are the same target, and continuing subsequent tracking by taking the bounding box obtained by target detection as an initial target in subsequent video frames until the next bounding box is encountered;
for some abnormal targets which are detected by mistake, clustering the displacement of the centers of the abnormal targets in five frames according to the extracted motion tracks; assume that the center position of the target at the j-th frame is (x)j,yj) Then its displacement within five frames is:
Figure FDA0003225805420000021
if the calculated displacement offset is smaller than the set threshold value, the displacement offset is regarded as false detection, the abnormal target which is detected by the false detection is marked as a normal target, and the abnormal target is removed from the abnormal target set;
after the processing, determining all foreground targets with abnormal appearance and abnormal motion, and detecting the foreground targets with abnormal positions by combining the previously determined foreground targets to form a final abnormal target set; judging whether the target abnormality score of each abnormal target exceeds an abnormality threshold value, if so, determining that all pixels in the abnormal target are abnormal; otherwise, the corresponding abnormal target is normal.
CN201810558987.7A 2018-06-01 2018-06-01 Method for detecting abnormality in target-oriented surveillance video Active CN108830204B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810558987.7A CN108830204B (en) 2018-06-01 2018-06-01 Method for detecting abnormality in target-oriented surveillance video

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810558987.7A CN108830204B (en) 2018-06-01 2018-06-01 Method for detecting abnormality in target-oriented surveillance video

Publications (2)

Publication Number Publication Date
CN108830204A CN108830204A (en) 2018-11-16
CN108830204B true CN108830204B (en) 2021-10-19

Family

ID=64146799

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810558987.7A Active CN108830204B (en) 2018-06-01 2018-06-01 Method for detecting abnormality in target-oriented surveillance video

Country Status (1)

Country Link
CN (1) CN108830204B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111325073B (en) * 2018-12-17 2024-02-20 上海交通大学 Monitoring video abnormal behavior detection method based on motion information clustering
CN110390262B (en) * 2019-06-14 2023-06-30 平安科技（深圳）有限公司 Video analysis method, device, server and storage medium
CN113705274B (en) * 2020-05-20 2023-09-05 杭州海康威视数字技术股份有限公司 Climbing behavior detection method and device, electronic equipment and storage medium
CN112364716A (en) * 2020-10-23 2021-02-12 岭东核电有限公司 Nuclear power equipment abnormal information detection method and device and computer equipment

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1976445A (en) * 2006-12-15 2007-06-06 上海鼎业电子科技有限公司 Multi-camera monitoring system based on three-dimensional video frequency dynamic tracking and tracking method thereof
CN101719216A (en) * 2009-12-21 2010-06-02 西安电子科技大学 Movement human abnormal behavior identification method based on template matching
CN102547244A (en) * 2012-01-17 2012-07-04 深圳辉锐天眼科技有限公司 Video monitoring method and system
CN103310466A (en) * 2013-06-28 2013-09-18 安科智慧城市技术（中国）有限公司 Single target tracking method and achievement device thereof
CN103473791A (en) * 2013-09-10 2013-12-25 惠州学院 Method for automatically recognizing abnormal velocity event in surveillance video
CN105741324A (en) * 2016-03-11 2016-07-06 江苏物联网研究发展中心 Moving object detection identification and tracking method on moving platform
CN106407903A (en) * 2016-08-31 2017-02-15 四川瞳知科技有限公司 Multiple dimensioned convolution neural network-based real time human body abnormal behavior identification method
CN106682619A (en) * 2016-12-28 2017-05-17 上海木爷机器人技术有限公司 Object tracking method and device

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10325169B2 (en) * 2016-10-09 2019-06-18 Airspace Systems, Inc. Spatio-temporal awareness engine for priority tree based region selection across multiple input cameras and multimodal sensor empowered awareness engine for target recovery and object path prediction

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1976445A (en) * 2006-12-15 2007-06-06 上海鼎业电子科技有限公司 Multi-camera monitoring system based on three-dimensional video frequency dynamic tracking and tracking method thereof
CN101719216A (en) * 2009-12-21 2010-06-02 西安电子科技大学 Movement human abnormal behavior identification method based on template matching
CN102547244A (en) * 2012-01-17 2012-07-04 深圳辉锐天眼科技有限公司 Video monitoring method and system
CN103310466A (en) * 2013-06-28 2013-09-18 安科智慧城市技术（中国）有限公司 Single target tracking method and achievement device thereof
CN103473791A (en) * 2013-09-10 2013-12-25 惠州学院 Method for automatically recognizing abnormal velocity event in surveillance video
CN105741324A (en) * 2016-03-11 2016-07-06 江苏物联网研究发展中心 Moving object detection identification and tracking method on moving platform
CN106407903A (en) * 2016-08-31 2017-02-15 四川瞳知科技有限公司 Multiple dimensioned convolution neural network-based real time human body abnormal behavior identification method
CN106682619A (en) * 2016-12-28 2017-05-17 上海木爷机器人技术有限公司 Object tracking method and device

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
Histograms of Optical Flow Orientation and Magnitude to Detect Anomalous Events in Videos;Rensso Victor Hugo Mora Colque,et al.;《IEEE》;20161208;第673-682页 *
一种融合光流的分通道帧差目标检测方法;夏瑜 等;《弹箭与制导学报》;20120228;第32卷(第1期);第175-178、184页 *
监控视频异常行为检测算法研究;李强;《中国优秀硕士学位论文全文数据库 信息科技辑》;20180115;第8、21、43-44、47-50页 *

Also Published As

Publication number Publication date
CN108830204A (en) 2018-11-16

Similar Documents

Publication Publication Date Title
CN108053427B (en) Improved multi-target tracking method, system and device based on KCF and Kalman
CN108062349B (en) Video monitoring method and system based on video structured data and deep learning
CN108830204B (en) Method for detecting abnormality in target-oriented surveillance video
CN108009473B (en) Video structuralization processing method, system and storage device based on target behavior attribute
CN108052859B (en) Abnormal behavior detection method, system and device based on clustering optical flow characteristics
Bertini et al. Multi-scale and real-time non-parametric approach for anomaly detection and localization
Bondi et al. Real-time people counting from depth imagery of crowded environments
CN109145708B (en) Pedestrian flow statistical method based on RGB and D information fusion
CN108230364B (en) Foreground object motion state analysis method based on neural network
KR101764845B1 (en) A video surveillance apparatus for removing overlap and tracking multiple moving objects and method thereof
EP2891990A1 (en) Method and device for monitoring video digest
US10271018B2 (en) Method of detecting critical objects from CCTV video using metadata filtering
CN104134222A (en) Traffic flow monitoring image detecting and tracking system and method based on multi-feature fusion
CN104966304A (en) Kalman filtering and nonparametric background model-based multi-target detection tracking method
Salehi et al. An automatic video-based drowning detection system for swimming pools using active contours
CN111145223A (en) Multi-camera personnel behavior track identification analysis method
KR101472674B1 (en) Method and apparatus for video surveillance based on detecting abnormal behavior using extraction of trajectories from crowd in images
WO2013075295A1 (en) Clothing identification method and system for low-resolution video
CN110598570A (en) Pedestrian abnormal behavior detection method and system, storage medium and computer equipment
Yamamoto et al. Multiple players tracking and identification using group detection and player number recognition in sports video
CN114783037B (en) Object re-recognition method, object re-recognition apparatus, and computer-readable storage medium
Hossain et al. Fast-D: When non-smoothing color feature meets moving object detection in real-time
CN115661698A (en) Escalator passenger abnormal behavior detection method, system, electronic device and storage medium
US20200394802A1 (en) Real-time object detection method for multiple camera images using frame segmentation and intelligent detection pool
CN111091583B (en) Long-term target tracking method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant