CN106780547A - Monitor video velocity anomaly mesh object detection method is directed to based on kinergety model - Google Patents
Monitor video velocity anomaly mesh object detection method is directed to based on kinergety model Download PDFInfo
- Publication number
- CN106780547A CN106780547A CN201611213597.3A CN201611213597A CN106780547A CN 106780547 A CN106780547 A CN 106780547A CN 201611213597 A CN201611213597 A CN 201611213597A CN 106780547 A CN106780547 A CN 106780547A
- Authority
- CN
- China
- Prior art keywords
- kinergety
- value
- model
- space block
- monitor video
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
Landscapes
- Image Analysis (AREA)
Abstract
Monitor video velocity anomaly mesh object detection method is directed to based on kinergety model the invention discloses one kind, including：Consider the motion vector information of 8 neighborhoods of abnormal object and its surrounding, set up kinergety model, using data item and differences plus and to represent the kinergety value of each space block；Each space block to frame of video enters row boundary value extraction, and each boundary value represents its correspondence space block, distinguishes the boundary value of proper motion and abnormal motion；Two graders based on sigmoid functions are classified to the kinergety value in test set, using the difference of the kinergety value of each piece of test set and the boundary value for extracting as input, and export the probable value of [0,1]；The corresponding probable value of space block is bigger, represents that the possibility that abnormal motion, i.e. velocity anomaly occur is higher.Effectively expressive movement intensity mode of the invention, complexity is relatively low, balances detection accuracy and detection efficiency, is capable of the abnormal target of real-time detection monitor video medium velocity.
Description
Technical field
Supervised the present invention relates to video monitoring abnormality detection technical field, more particularly to a kind of being directed to based on kinergety model
Control the detection method of video speed abnormal object.
Background technology
In recent years, taking place frequently with some attacks of terrorism and colony's incident of violence, people to City Life Quality and
The requirement that living safety is taken precautions against is growing day by day.Although various video monitoring equipments have been placed in various crowded public affairs extensively
Place altogether, such as railway station, subway station, hospital, square and cell, but current video monitoring system be merely able to it is right
Certain monitoring scene carries out simple video record, by transmission of video to Control Room, artificial scene is carried out by monitoring personnel
Monitoring；Or store monitor video, it is used for searching evidence obtaining afterwards.This traditional monitoring system lack real-time and
It is intelligent, very dependent on the experience and subjective judgement of monitoring personnel.Therefore, in order to make monitoring system automatically divide in real time
Analysis video content, judges the abnormal object and anomalous event in monitor video, and auxiliary monitoring personnel is judged, it is necessary to carry significantly
The degree of accuracy and the quality of public place safety precaution that monitor video high judges extremely.
Anomalous event under common scene be it is diversified, for example：Plunder, fall, invading, crowd gathers, the stream of people is inverse
OK, object hypervelocity etc..Wherein, widely, practicality is very high for the range of application of detection speed abnormal object.In recent years
Research in, many scientific research personnel propose different object speed method for detecting abnormality based on different technologies.
Helbing et al. calculates the reciprocal force of crowd based on movable information proposition social force model, and direction is judged with this
And the exception of speed, but the method is for lacking the sparse scene of the crowd of enough movable informations and not applying to[1]；Adam et al.
The multiple monitoring unit fixed using position is counted to the low-level features of scene, the hypervelocity thing of detection hypervelocity diverse location
Body, but the method is limited to the quantity and density of monitoring unit[2]；Weixin et al. extracts the dynamic unity and coherence in writing of object in scene
Feature describes the object of different conditions, and so that this is to exception and is normally carried out classification, but the characteristic extraction procedure time of the method answers
Miscellaneous degree is high, and live effect is not good[3]；The Vikas synthesis speed of object, size and textural characteristics, by a multistage classifier pair
Exception is judged, but the method is limited to the selection of template and the setting of hierarchical detection mechanism[4]。
The content of the invention
Monitor video velocity anomaly mesh object detection method is directed to based on kinergety model the invention provides one kind, this
Invention considers the movable information of target object and its approaching object simultaneously, has preferable robustness to the detection of exercise intensity,
The velocity anomaly target under different scenes can be effectively detected, it is described below：
One kind is directed to monitor video velocity anomaly mesh object detection method, the detection method bag based on kinergety model
Include following steps：
Consider the motion vector information of 8 neighborhoods of abnormal object and its surrounding, set up kinergety model, the model bag
Include：Data item and differences, using data item and differences plus and to represent the kinergety value of each space block；
Each space block to frame of video enters row boundary value extraction, and each boundary value represents its correspondence space block,
Distinguish the boundary value of proper motion and abnormal motion；
Two graders based on sigmoid functions are classified to the kinergety value in test set, each with test set
The difference of the kinergety value of block and the boundary value for extracting exports the probable value of [0,1] as input；
The corresponding probable value of space block is bigger, represents that the possibility that abnormal motion, i.e. velocity anomaly occur is higher.
The detection method also includes：
With reference to image pyramid, using Lucas-Kanade optical flow methods on different scale to monitor video in each frame
Light stream vector is extracted, optical flow field is set up, the optical flow field of each frame is divided into nonoverlapping space block of M × N, obtain each space
The motion vector of block.
The motion vector is specially：
Bi.j=(fx,y(h, v) | x=1,2 ..., M, y=1,2 ..., N)
I=1,2 ..., P, j=1,2 ..., Q
Wherein, Bi,jRepresent that position is the space block of (i, j)；fx,y(h, v) is represented and is contained in block Bi,jIn position for (x,
Y) motion vector, comprising a horizontal component h and vertical component v；M is the horizontal length of each space block；N is each
The vertical length of space block；P is the space number of blocks of each frame horizontal direction；Q is the space number of blocks of each frame vertical direction.
Methods described also includes：Motion vector is quantified.
The boundary value is extracted and is specially：
If data growth rate is more than default certain threshold value, the kinergety value of previous data point is considered as border
Value.
The beneficial effect of technical scheme that the present invention is provided is：Kinergety model proposed by the present invention can effective earth's surface
Up to exercise intensity pattern, algorithm complex is relatively low, balances detection accuracy and detection efficiency, being capable of real-time detection monitor video
The abnormal target of medium velocity.
Brief description of the drawings
Fig. 1 gives the flow chart for monitor video velocity anomaly mesh object detection method based on kinergety model；
Fig. 2 gives has the abnormal monitoring scene exemplary plot of local velocity；
A () is the schematic diagram of the motor vehicle of fast running in pavement；B () is the signal of bicycle walked in pavement
Figure.
Fig. 3 gives the exemplary plot of local velocity's abnormal object and its kinergety value；
A (), (c) are the schematic diagram of any two frame comprising abnormal motion in test set；B (), (d) are respectively (a), (c)
The schematic diagram of the kinergety value of respective frame.
Fig. 4 gives the exemplary plot that kinergety boundary value is estimated using the method for data growth rate；
A () is all frames in training set, the example counts the kinergety value of square marked region；B () transports for the block
Kinetic energy value distribution situation from small to large, it can be seen that most of value goes to zero, and there is few very big noise figure；
C () is the data growth rate of kinergety value, when data growth rate is excessive, then it is assumed that the value is beyond proper motion energy
The scope of value；D () is the final kinergety boundary value (two dotted line intersection points) chosen, overstriking dotted line represents proper motion energy
The scope of value.
Fig. 5 gives final kinergety boundary value figure；
Fig. 6 gives the present invention ratio with the performance of Social Force, MDT and MPPCA methods on UCSD data sets
Compared with result figure；
A () is the schematic diagram of the frame level abnormality detection performance of UCSD Ped1 test sets；B () is UCSD Ped2 test sets
The schematic diagram of frame level abnormality detection performance；C () is the schematic diagram of the Pixel-level abnormality detection performance of UCSD Ped2 test sets.
Fig. 7 gives the present invention visualization result figure (mark abnormal object) on UCSD data sets.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, further is made to embodiment of the present invention below
Ground is described in detail.
Embodiment 1
The embodiment of the present invention, based on scene motion information, is proposed for the detection of the velocity anomaly target under monitoring scene
A kind of kinergety model describes object of which movement pattern, and by the classification to different zones kinergety value, realizes speed
The real-time detection of abnormal object is spent, referring to Fig. 1, the method is comprised the following steps：
101：The movable information of target area is extracted, exercise intensity value is obtained；
102：Diversity factor calculating is carried out to target area and its 8 neighborhoods using SSD measurements, a movement differential value is obtained；
103：Added using two and represent the kinergety value of central area object, and then divided by kinergety value
Class, carrys out the abnormal target of detection speed.
In sum, the kinergety model that the embodiment of the present invention is proposed considers target object and its approaching object simultaneously
Movable information, have preferable robustness to the detection of exercise intensity, can effectively detect the velocity anomaly under different scenes
Target.
Embodiment 2
The scheme in embodiment 1 is further introduced with reference to specific computing formula, it is described below：
201：Extract movable information；
Motion vector is the key character that can express object movable information, and optical flow method is to calculate motion vector the most
One of effective method.Embodiment of the present invention combination image pyramid, using Lucas-Kanade optical flow methods on different scale
Each frame in monitor video extracts light stream vector, sets up optical flow field.Afterwards, the optical flow field of each frame is divided into size for M
Nonoverlapping space block of × N, common P × Q.For each piece, motion vector is expressed as follows：
Bi.j=(fx,y(h, v) | x=1,2 ..., M, y=1,2 ..., N)
I=1,2 ..., P, j=1,2 ..., Q
Wherein, Bi,jRepresent that position is the space block of (i, j)；fx,y(h, v) is represented and is contained in block Bi,jIn position for (x,
Y) motion vector, it includes a horizontal component h and vertical component v.Intensity positioned at the motion vector of (x, y) is expressed as：
magx,y=| | fx,y(h,v)||2
For the ease of statistics, the motion vector of each pixel is carried out appropriate quantization by the embodiment of the present invention, and experiment shows
Preferably, specific quantification manner is as follows for Detection results when quantification gradation is 16：The motion vector for finding each space block first is strong
The maximum max mag of angle valuex,y, by interval [0, max magx,y] 16 deciles, fall the motion arrow on each by stages such as small
Amount intensity level is quantified as the size of the interval midpoint value, finally gives magx,y, i.e., the motion after quantified on (x, y) position
The intensity of vector.
202：Kinergety model；
For local anomaly, abnormal object must be surrounded by normal target.Therefore, the embodiment of the present invention is simultaneously
Consider the motion vector information of 8 neighborhoods of abnormal object and its surrounding, set up a kinergety model.The model includes two：
Data item and differences, it is as follows：
Wherein, DataTerm (i, j) represents data item；DifferenceTerm (i, j) represents differences；Space block Bi,j
In main motion intensity be expressed as Dom magi,j x,y, it is the intensity level of the most motion vector of quantity in the block.In differences
In, Si,jRepresent central block Bi,j8 neighborhood regions, Bm,nIt is belonging to a space block in 8 neighborhood.Central block is adjacent
The exercise intensity diversity factor in region is as follows with SSD measurement representations：
Wherein, SSD (Bi,j,Bm,n) it is two quadratic sums of the error of image block corresponding pixel points, measurement two spaces are fast
Exercise intensity difference；To be located at the motion vector intensity on (x, y) position in the space block of serial number (i, j)；To be located at the motion vector intensity on (x, y) position in the space block of serial number (m, n)；NfIt is Bm,nIn motion vector
Quantity.It is bigger with associating for central block due to the region that distance center block is nearer.Therefore, it is pointed to the sky under different distance
Between block assign different weighted value ωm,nTo adjust the weight of different zones.
Using data item and differences plus and to represent the kinergety value of each space block, adjusted by α and β
The proportion of whole two, it is as follows：
E (i, j)=α DataTerm (i, j)+β DifferenceTerm (i, j)
203：Boundary value is extracted；
Each space block to all frame of video of training set calculates kinergety value, and counts each space block fortune
The distribution of kinetic energy value.In order to reduce the influence of the noise of optical flow field, it should weed out excessive value in kinergety value, and look for
To proper motion and the boundary value of abnormal motion.The embodiment of the present invention is by calculating the number of the kinergety value of each space block
Boundary value is extracted according to growth rate.If data growth rate is more than default certain threshold value, by the motion energy of previous data point
Value is considered as boundary value.According to experiment, the threshold value is set as that 0.5 effect is best.
204：Space block sort.
Due to position of the monitoring camera in monitoring scene so that moving object can be detected difference in different regions
Exercise intensity value and kinergety value.Generally, the object nearer apart from camera possesses bigger exercise intensity.Therefore, it is right
Each space block of frame of video enters row boundary value extraction, and each boundary value represents its correspondence space block, distinguishes normal fortune
The boundary value of dynamic and abnormal motion.
The embodiment of the present invention is entered by two graders based on sigmoid functions to the kinergety value in test set
Row classification, two grader is using the difference of the kinergety value of each piece of test set and the boundary value for extracting as input and defeated
Go out the probable value of [0,1].The corresponding probable value of space block is bigger, represent occur abnormal motion, i.e. velocity anomaly can
Energy property is higher.That is, the kinergety value of the space block is excessive beyond normal range (NR).Two graders are as follows：
Wherein, E ' (i, j) and E (i, j) represent boundary energy value and kinergety value to be tested respectively.
In sum, the kinergety model that the embodiment of the present invention is proposed considers target object and its approaching object simultaneously
Movable information, have preferable robustness to the detection of exercise intensity, can effectively detect the velocity anomaly under different scenes
Target.
Embodiment 3
Feasibility checking is carried out to the scheme in embodiment 1 and 2 with reference to specific accompanying drawing, experimental data, is referred to down
Text description：
Experiment carries out emulation testing on MATLAB platforms to UCSD data sets.UCSD data sets are widely used in part
In Anomaly target detection, it includes two subset --- Ped1 and Ped2 of different scenes.Two training sets of subset are only wrapped
It is the moving target comprising some velocity anomalies is such as walked in test set bicycle, quick containing proper motion (pedestrian of walking)
Motor vehicle, slide plate of traveling etc..For frame level detection, as long as there is a pixel to be detected as exception in test process,
Then the frame is to be designated as exception；For Pixel-level abnormality detection, as long as in test process, being designated as abnormal pixel and accounting for real different
More than the 40% of normal target pixel points, then can be considered that detection is correct, on the contrary detection mistake.
The embodiment of the present invention is entered with Social Force, MPPCA, the MDT methods for also belonging to low-level features detection method
Row compares, and comparative result is as follows：
In frame level Detection task, this method can be seen that by the ROC curve result of Fig. 6 (a), (b) and put up the best performance：
Comparable with MDT methods in Ped1, gap is small；MDT methods are performed more than in Ped2.As shown in table 1, it is of the invention average
Error rate is close with MDT-temporal.In Pixel-level Detection task, this can be seen that by the ROC curve result of Fig. 6 (c)
The verification and measurement ratio of invention is small with MDT method gaps.But as shown in Table 2, the efficiency of algorithm of this method is much higher by other method, warp
Real-time Anomaly target detection can be realized substantially after crossing optimization.
The other vision response test of the frame level of table 1 (PED1 and PED2)
The verification and measurement ratio of table 2 and efficiency of algorithm (PED2)
Fig. 7 (a) and (b) illustrate visualization result：The region that white is labeled as in figure is the region that velocity anomaly occurs,
Should be apparent that the present invention effectively detects motor vehicle, bicycle, slide plate uniform velocity abnormal object.
In sum, the kinergety model that the embodiment of the present invention is proposed considers target object and its approaching object simultaneously
Movable information, have preferable robustness to the detection of exercise intensity, can effectively detect the velocity anomaly under different scenes
Target.
Bibliography
[1]Helbing,Dirk,and Peter Molnar."Social force model for pedestrian
dynamics."Physical review E 51.5(1995):4282.
[2]Adam,Amit,et al."Robust real-time unusual event detection using
multiple fixed-location monitors."IEEE Transactions on Pattern Analysis and
Machine Intelligence 30.3(2008):555-560.
[3]Li,Weixin,Vijay Mahadevan,and Nuno Vasconcelos."Anomaly detection
and localization in crowdedscenes."IEEE transactions on pattern analysis and
machine intelligence 36.1(2014):18-32.
[4]Reddy,Vikas,Conrad Sanderson,and Brian C.Lovell."Improved anomaly
detection in crowded scenesvia cell-based analysis of foreground speed,size
and texture."CVPR 2011WORKSHOPS.IEEE,2011.
It will be appreciated by those skilled in the art that accompanying drawing is a schematic diagram for preferred embodiment, the embodiments of the present invention
Sequence number is for illustration only, and the quality of embodiment is not represented.
The foregoing is only presently preferred embodiments of the present invention, be not intended to limit the invention, it is all it is of the invention spirit and
Within principle, any modification, equivalent substitution and improvements made etc. should be included within the scope of the present invention.
Claims (5)
1. it is a kind of that monitor video velocity anomaly mesh object detection method is directed to based on kinergety model, it is characterised in that described
Detection method is comprised the following steps：
Consider the motion vector information of 8 neighborhoods of abnormal object and its surrounding, set up kinergety model, the model includes：Number
According to item and differences, using data item and differences plus and to represent the kinergety value of each space block；
Each space block to frame of video enters row boundary value extraction, and each boundary value represents its correspondence space block, distinguishes
The boundary value of proper motion and abnormal motion；
Two graders based on sigmoid functions are classified to the kinergety value in test set, with each piece of test set
The difference of kinergety value and the boundary value for extracting exports the probable value of [0,1] as input；
The corresponding probable value of space block is bigger, represents that the possibility that abnormal motion, i.e. velocity anomaly occur is higher.
2. it is according to claim 1 it is a kind of based on kinergety model for monitor video velocity anomaly target detection side
Method, it is characterised in that the detection method also includes：
With reference to image pyramid, using Lucas-Kanade optical flow methods on different scale to monitor video in each frame extract
Light stream vector, sets up optical flow field, and the optical flow field of each frame is divided into nonoverlapping space block of M × N, obtains each space block
Motion vector.
3. a kind of inspection that monitor video velocity anomaly target is directed to based on kinergety model according to claim 1 and 2
Survey method, it is characterised in that the motion vector is specially：
Bi.j=(fx,y(h, v) | x=1,2 ..., M, y=1,2 ..., N)
I=1,2 ..., P, j=1,2 ..., Q
Wherein, Bi,jRepresent that position is the space block of (i, j)；fx,y(h, v) is represented and is contained in block Bi,jIn position be (x, y)
Motion vector, comprising a horizontal component h and vertical component v；M is the horizontal length of each space block；N is each space
The vertical length of block；P is the space number of blocks of each frame horizontal direction；Q is the space number of blocks of each frame vertical direction.
4. a kind of inspection that monitor video velocity anomaly target is directed to based on kinergety model according to claim 1 and 2
Survey method, it is characterised in that methods described also includes：Motion vector is quantified.
5. it is according to claim 1 it is a kind of based on kinergety model for monitor video velocity anomaly target detection side
Method, it is characterised in that the boundary value is extracted and is specially：
If data growth rate is more than default certain threshold value, the kinergety value of previous data point is considered as boundary value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611213597.3A CN106780547B (en) | 2016-12-24 | 2016-12-24 | Monitor video velocity anomaly mesh object detection method is directed to based on kinergety model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611213597.3A CN106780547B (en) | 2016-12-24 | 2016-12-24 | Monitor video velocity anomaly mesh object detection method is directed to based on kinergety model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106780547A true CN106780547A (en) | 2017-05-31 |
| CN106780547B CN106780547B (en) | 2019-06-18 |
Family
ID=58920688
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611213597.3A Active CN106780547B (en) | 2016-12-24 | 2016-12-24 | Monitor video velocity anomaly mesh object detection method is directed to based on kinergety model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106780547B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110392302A (en) * | 2018-04-16 | 2019-10-29 | 北京陌陌信息技术有限公司 | Video is dubbed in background music method, apparatus, equipment and storage medium |
| CN110580708A (en) * | 2018-06-11 | 2019-12-17 | 杭州海康威视数字技术股份有限公司 | Rapid movement detection method and device and electronic equipment |
| CN110580504A (en) * | 2019-08-27 | 2019-12-17 | 天津大学 | Video abnormal event detection method based on self-feedback mutual exclusion subclass mining |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101325691A (en) * | 2007-06-14 | 2008-12-17 | 清华大学 | Method and apparatus for tracing a plurality of observation model with fusion of differ durations |
| CN102156961A (en) * | 2009-12-22 | 2011-08-17 | 索尼公司 | Image processing apparatus, image processing method, and program |
-
2016
- 2016-12-24 CN CN201611213597.3A patent/CN106780547B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101325691A (en) * | 2007-06-14 | 2008-12-17 | 清华大学 | Method and apparatus for tracing a plurality of observation model with fusion of differ durations |
| CN102156961A (en) * | 2009-12-22 | 2011-08-17 | 索尼公司 | Image processing apparatus, image processing method, and program |
Non-Patent Citations (4)
| Title |
|---|
| ALEXANDER SHEKHOVTSOV ER.AL: "Efficient MRF Deformation Model for Non-Rigid Image Matching", 《2007 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
| TOBIAS SENST ET.AL: "II-LK – A Real-Time Implementation for Sparse Optical Flow", 《INTERNATIONAL CONFERENCE IMAGE ANALYSIS AND RECOGNITION》 * |
| YANG CONG ET.AL: "Video Anomaly Search in Crowded Scenes via Spatio-Temporal Motion Context", 《IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY》 * |
| 王乔 等: "基于整体能量模型的异常行为检测", 《计算机应用研究》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110392302A (en) * | 2018-04-16 | 2019-10-29 | 北京陌陌信息技术有限公司 | Video is dubbed in background music method, apparatus, equipment and storage medium |
| CN110580708A (en) * | 2018-06-11 | 2019-12-17 | 杭州海康威视数字技术股份有限公司 | Rapid movement detection method and device and electronic equipment |
| CN110580708B (en) * | 2018-06-11 | 2022-05-31 | 杭州海康威视数字技术股份有限公司 | Rapid movement detection method and device and electronic equipment |
| CN110580504A (en) * | 2019-08-27 | 2019-12-17 | 天津大学 | Video abnormal event detection method based on self-feedback mutual exclusion subclass mining |
| CN110580504B (en) * | 2019-08-27 | 2023-07-25 | 天津大学 | Video abnormal event detection method based on self-feedback mutual exclusion subclass mining |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106780547B (en) | 2019-06-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106910203B (en) | The quick determination method of moving target in a kind of video surveillance | |
| CN110135269B (en) | Fire image detection method based on mixed color model and neural network | |
| CN102324016B (en) | Statistical method for high-density crowd flow | |
| CN105894701B (en) | The identification alarm method of transmission line of electricity external force damage prevention Large Construction vehicle | |
| CN101339655B (en) | Visual sense tracking method based on target characteristic and bayesian filtering | |
| CN105761261B (en) | A method of detection camera suffers artificial malicious sabotage | |
| CN106204640A (en) | A kind of moving object detection system and method | |
| CN105426820B (en) | More people's anomaly detection methods based on safety monitoring video data | |
| CN106128053A (en) | A kind of wisdom gold eyeball identification personnel stay hover alarm method and device | |
| CN107622258A (en) | A kind of rapid pedestrian detection method of combination static state low-level image feature and movable information | |
| CN106203274A (en) | Pedestrian's real-time detecting system and method in a kind of video monitoring | |
| Nallaivarothayan et al. | An MRF based abnormal event detection approach using motion and appearance features | |
| CN105678803A (en) | Video monitoring target detection method based on W4 algorithm and frame difference | |
| CN104123544A (en) | Video analysis based abnormal behavior detection method and system | |
| CN105354563A (en) | Depth and color image combined human face shielding detection early-warning device and implementation method | |
| Tian et al. | A novel video-based smoke detection method using image separation | |
| CN105574488A (en) | Low-altitude aerial infrared image based pedestrian detection method | |
| CN103745230A (en) | Adaptive abnormal crowd behavior analysis method | |
| Karpagavalli et al. | Estimating the density of the people and counting the number of people in a crowd environment for human safety | |
| CN102930248A (en) | Crowd abnormal behavior detection method based on machine learning | |
| CN103489012B (en) | Crowd density detecting method and system based on support vector machine | |
| CN106127814A (en) | A kind of wisdom gold eyeball identification gathering of people is fought alarm method and device | |
| CN104680554B (en) | Compression tracking and system based on SURF | |
| CN105868734A (en) | Power transmission line large-scale construction vehicle recognition method based on BOW image representation model | |
| CN106780547B (en) | Monitor video velocity anomaly mesh object detection method is directed to based on kinergety model |
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 |