CN106339690A - Video object flow detecting method and system based on noise elimination and auxiliary determination line - Google Patents
Video object flow detecting method and system based on noise elimination and auxiliary determination line Download PDFInfo
- Publication number
- CN106339690A CN106339690A CN201610771414.3A CN201610771414A CN106339690A CN 106339690 A CN106339690 A CN 106339690A CN 201610771414 A CN201610771414 A CN 201610771414A CN 106339690 A CN106339690 A CN 106339690A
- Authority
- CN
- China
- Prior art keywords
- target
- module
- tracking
- area
- interest
- 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
Links
Classifications
-
- 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
- G06V20/53—Recognition of crowd images, e.g. recognition of crowd congestion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
Abstract
The invention relates to a video object flow detecting method and system based on noise elimination and an auxiliary determination line. Object detection is carried out on a region of interest in a video image cyclically, and objects are tracked and refreshed; a safe region is arranged outside the region of interest to count the people flow; and whether all present tracked objects are mis-detection objects is determined cyclically, and the determined mis-detection objects are deleted. According to the method and system, the instantaneity is high, and certain accuracy can be ensured in a complex scene.
Description
Technical field
The present invention relates to a kind of technology of Video processing analysis technical field, specifically a kind of eliminated based on noise and
The video object flow rate testing methods of auxiliary judgment line and system.
Background technology
Target flow statistics is the problem very in video analysis with researching value, in monitoring system, resource management, agriculture
In the protection of industry production, the research of Animal behaviour and species, all there is using value.And in complicated scene, target flow
Statistics is also more challenging.
In terms of target flow statistics, the regression and statistical method development based on area-of-interest (roi) is relatively early, is primarily upon
Pedestrian's quantity in a region.It generally extracts the feature of each section after each crowd segmentation in an image, and
Obtain the regression function figure in the space of feature and the crowd's quantity in crowd's block.Such as a.b.chan et al. exists
" conference on computer vision and pattern recognition 2008 " (international computer in 2008
Vision and pattern recognition meeting) on deliver " privacy preserving crowd monitoring:counting
In people without people models or tracking " (model of not employment and the people's stream statistics followed the tracks of) paper
Propose, divided area is the typical characteristic that can show the pedestrian's quantity in partitioning portion, as long as the extraction of feature can root
Give each pixel suitable weight according to the visual angle of scene, then divided area and pedestrian's quantity there is the pass of approximately linear
System.Commonly use some nonlinear regression algo in practice and include Gaussian process recurrence or Bayesian regression.This algorithm is broadly divided into
Training and two stages of prediction, it is characterized in that accuracy is higher, but is required for substantial amounts of staking-out work in each scene,
It is difficult in practical application.
Another method Ji Yu lines of interest (loi) is then based on one section of sequence of pictures, counts in a period of time window, wears
Get over the statistical method of the target of certain lines of interest, kim b et al. delivered in 2008 " a method counting
A kind of method of the people flow rate statistical proposing in pedestrians in crowded scenes " paper, to obtain in video sequence
The prediction of pedestrian stream quantity based on the light stream arrived, is entered by integrating the pixel number obtaining.The feature of the method is to be not required to
To carry out demarcation and the training process of complexity as the regression and statistical method based on area-of-interest, so in new scene bottom
Administration is relatively easy to.But it is very sensitive to crowding phenomenon.Larger statistical error can be produced in complex scene.It is thus desirable to
The algorithm that a kind of complexity is relatively low and has certain accuracy in complex scene.
The existing people flow rate statistical method based on Video Analysis Technology, has by training number of people characteristic model, frame-to-frame differences
Obtain foreground point, number of people feature extraction and the identification of moving object, body local feature identification and number of people characteristic area are followed the tracks of
Count to realize flow of the people.But this kind of technology cannot eliminate the interference to foreground information for the conditions such as illumination, and for static or slow
The pedestrian of speed cannot detect, therefore easily cause and repeatedly pass through detection zone border in same person, cause repeat count.
Content of the invention
The present invention is directed to deficiencies of the prior art, proposes a kind of to eliminate and the regarding of auxiliary judgment line based on noise
Frequency target flow detection method and system, the particular state letter of each target being obtained according to target detection and target following
Breath, and eliminate to obtain overall traffic statistics result with auxiliary judgment line with reference to noise;Complexity of the present invention is relatively low, real-time
Good, and can guarantee that certain accuracy rate under complex scene.
The present invention is achieved by the following technical solutions:
The present invention relates to a kind of video object flow rate testing methods based on noise elimination and auxiliary judgment line, cyclically right
Carry out target detection, as in area-of-interest in the image of start frame in detection video in area-of-interest in video image
All targets, and it is tracked;Then outside area-of-interest, setting safety zone carries out people flow rate statistical;Finally
Cycle criterion currently all tracked targets be whether error detection target, the target of then deletion error detection is to be made an uproar in this way
Sound eliminates.
The present invention relates to a kind of system realizing said method, comprising: module of target detection, target tracking module, artificial abortion
Amount statistical module and noise cancellation module, wherein: module of target detection be connected with target tracking module and transmission region in inspection
The information of all targets measuring, target tracking module is connected with people flow rate statistical module and transmits all targets tracing into
Information, people flow rate statistical module is connected with output module and transmits real-time demographics result, noise cancellation module and target
Tracking module is connected and transmits the information of the target that needs are deleted.
Technique effect
Compared with prior art, the present invention is by periodically being detected to the target in area-of-interest and being followed the tracks of,
Reduce certain operand, improve the real-time of algorithm;The strategy of safety zone and noise remove then can be for traveling
The pedestrian stopping in way can correctly follow the tracks of, and can eliminate repeat count and row based on auxiliary judgment line and noise remove
The statistical number of person error that people's flase drop brings, has been obviously improved the accuracy rate of target flow statistics, has allowed the present invention under complex scene
Target flow statistics have and preferably applied.
Brief description
Fig. 1 is the flow chart of the inventive method;
Fig. 2 is area-of-interest schematic diagram;
Fig. 3 is safety zone schematic diagram;
Fig. 4 eliminates schematic diagram for noise;
Fig. 5 is embodiment detection process schematic diagram.
Specific embodiment
As shown in figure 1, the present embodiment comprises the following steps:
The first step, delimitation area-of-interest.In the present embodiment, detection target is taking pedestrian as a example.As Fig. 1, with two bases
This to delimit a region perpendicular to the line of pedestrian's direct of travel, and the detect and track of pedestrian is then based on this region.
As all targets in area-of-interest in the image of start frame in second step, detection video, and it is carried out
Follow the tracks of.
In the present embodiment, described target detection refers to using k.huang et al. in " asia pacific signal
And processing association conference 2014 " (Asia-Pacific signal with process society conference) opinion of delivering
Civilian " improved human head and shoulder detection with local main gradient and
Tracklets based feature, " the detection algorithm of a kind of middle head based on people proposing and shoulder.
In the present embodiment, described tracking refers to: searches near the position of the previous frame target in next two field picture
Similar color lump, searches nearest with previous frame target location Euclidean distance then it is assumed that being same target;Then periodically
The tracking that ground carries out target refreshes.
Described tracking refreshes and refers to: carry out target detection, by the target in testing result with existing carried out with
All targets of track calculate overlapping fraction s respectively two-by-two, if all overlap fractions, all not less than given threshold, illustrate that this target is
Newly occur in the target in region, new tracking is set up to it.
Described periodicity, preferably carries out a secondary tracking refreshing every 10 frames.
Described overlapping fractionWherein: rtIt is some target area that tracking result is outlined, rdIt is
The area of some target that testing result is outlined.
3rd step as shown in figure 3, delimit safety zone, that is, region determined by the two lines of area-of-interest both sides with
Carry out artificial abortion's statistics of variables.This two lines is parallel with the boundary line of area-of-interest, the direction advanced perpendicular to pedestrian, and away from
The boundary line of area-of-interest is nearer, preferably by 10 pixels of boundary line interested horizontal translation.Specifically, for each frame
Each target tracking result, when tracked target passes safety zone border in area-of-interest, then flow of the people system
Meter Jia one, and two borders characterize two direction of travel of pedestrian, you can to count the flow of the people of two direction of travel respectively.
4th step, noise remove.Because aforesaid operations may detect target in the place originally not having pedestrian, lead to
The generation of noise, so need to remove noise to lift accuracy rate.As shown in figure 4, in the present embodiment often through 3 frames, deleting
Displacement is zero target.
According to above step, the test video collected for us is analyzed, as shown in figure 5, being video interception, mainly
There are three longer videos, be from left to right followed successively by video 1, video 2, video 3.
Video 1 | Video 2 | Video 3 | |
Effective strength | 169 | 63 | 120 |
Algorithm statistical number of person | 171 | 57 | 114 |
Error | 0.012 | 0.095 | 0.05 |
Result shows, even in the case of the so highdensity artificial abortion of video 3, Detection accuracy also has 95%.And
In the so relatively simple scene of video 1, rate of accuracy reached to more than 98%.This experiment shows that algorithm can carry out essence to flow of the people
True statistics.
Above-mentioned be embodied as can by those skilled in the art on the premise of without departing substantially from the principle of the invention and objective with difference
Mode local directed complete set is carried out to it, protection scope of the present invention is defined by claims and is not embodied as institute by above-mentioned
Limit, each implementation in the range of it is all by the constraint of the present invention.
Claims (10)
1. a kind of eliminated based on noise and auxiliary judgment line video object flow rate testing methods it is characterised in that cyclically right
Carry out target detection, as in area-of-interest in the image of start frame in detection video in area-of-interest in video image
All targets, and it is tracked;Then outside area-of-interest, setting safety zone carries out people flow rate statistical;Finally
Cycle criterion currently all tracked targets be whether error detection target, the target of then deletion error detection is to be made an uproar in this way
Sound eliminates.
2. method according to claim 1, is characterized in that, described tracking refers to: the previous frame in next two field picture
Searching for similar color lump near the position of target, searching nearest with previous frame target location Euclidean distance then it is assumed that being same
One target;Then the tracking periodically carrying out target refreshes.
3. method according to claim 2, is characterized in that, described tracking refreshing refers to: by the target in testing result
Calculate overlapping fraction with the existing all targets carrying out following the tracks of respectively two-by-two, when all overlap fractions are all not less than setting
Threshold value, illustrates that this target is the target newly occurring in region, sets up new tracking to it.
4. method according to claim 3, is characterized in that, described periodicity, is to carry out a secondary tracking brush every 10 frames
Newly.
5. method according to claim 3, is characterized in that, described overlapping fractionWherein: rtIt is to follow the tracks of knot
Some target area that fruit is outlined, rdIt is the area of some target that testing result is outlined.
6. method according to claim 1, is characterized in that, described people flow rate statistical refers to: each for each frame
The tracking result of individual target, when tracked target passes safety zone border in area-of-interest, then people flow rate statistical adds one,
Article two, border characterizes two direction of travel of pedestrian, you can to count the flow of the people of two direction of travel respectively.
7. method according to claim 1, is characterized in that, currently whether all tracked targets are for described cycle criterion
The target of error detection, that is, often through 3 frames, deletes the target that displacement is zero.
8. method according to claim 1, is characterized in that, the border of described boundary of safe region and area-of-interest
Line is parallel, the direction advanced perpendicular to pedestrian.
9. the method according to claim 1 or 9, is characterized in that, described boundary of safe region is by border interested
10 pixels of line horizontal translation.
10. a kind of system realizing any of the above-described claim methods described is it is characterised in that include: module of target detection, mesh
Mark tracking module, people flow rate statistical module and noise cancellation module, wherein: module of target detection is connected with target tracking module
And the information of all targets detecting in transmission region, target tracking module is connected with people flow rate statistical module and transmits all
The information of the target tracing into, people flow rate statistical module is connected with output module and transmits real-time demographics result, noise
Cancellation module is connected with target tracking module and transmits the information of the target that needs are deleted.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610771414.3A CN106339690A (en) | 2016-08-30 | 2016-08-30 | Video object flow detecting method and system based on noise elimination and auxiliary determination line |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610771414.3A CN106339690A (en) | 2016-08-30 | 2016-08-30 | Video object flow detecting method and system based on noise elimination and auxiliary determination line |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106339690A true CN106339690A (en) | 2017-01-18 |
Family
ID=57824035
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610771414.3A Pending CN106339690A (en) | 2016-08-30 | 2016-08-30 | Video object flow detecting method and system based on noise elimination and auxiliary determination line |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106339690A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107330386A (en) * | 2017-06-21 | 2017-11-07 | 厦门中控智慧信息技术有限公司 | A kind of people flow rate statistical method and terminal device |
CN108537089A (en) * | 2017-03-01 | 2018-09-14 | 开利公司 | Flow of the people estimating system and flow of the people estimating and measuring method |
CN109697392A (en) * | 2017-10-23 | 2019-04-30 | 北京京东尚科信息技术有限公司 | Draw the method and device of target object thermodynamic chart |
CN110443097A (en) * | 2018-05-03 | 2019-11-12 | 北京中科晶上超媒体信息技术有限公司 | A kind of video object extract real-time optimization method and system |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101777188A (en) * | 2010-03-12 | 2010-07-14 | 华中科技大学 | Real-time bus passenger flow volume statistical method |
CN101814149A (en) * | 2010-05-10 | 2010-08-25 | 华中科技大学 | Self-adaptive cascade classifier training method based on online learning |
CN101847206A (en) * | 2010-04-21 | 2010-09-29 | 北京交通大学 | Pedestrian traffic statistical method and system based on traffic monitoring facilities |
CN103986910A (en) * | 2014-05-20 | 2014-08-13 | 中国科学院自动化研究所 | Method and system for passenger flow statistics based on cameras with intelligent analysis function |
CN105554451A (en) * | 2015-12-10 | 2016-05-04 | 天津艾思科尔科技有限公司 | Image sensor with flow statistical function |
-
2016
- 2016-08-30 CN CN201610771414.3A patent/CN106339690A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101777188A (en) * | 2010-03-12 | 2010-07-14 | 华中科技大学 | Real-time bus passenger flow volume statistical method |
CN101847206A (en) * | 2010-04-21 | 2010-09-29 | 北京交通大学 | Pedestrian traffic statistical method and system based on traffic monitoring facilities |
CN101814149A (en) * | 2010-05-10 | 2010-08-25 | 华中科技大学 | Self-adaptive cascade classifier training method based on online learning |
CN103986910A (en) * | 2014-05-20 | 2014-08-13 | 中国科学院自动化研究所 | Method and system for passenger flow statistics based on cameras with intelligent analysis function |
CN105554451A (en) * | 2015-12-10 | 2016-05-04 | 天津艾思科尔科技有限公司 | Image sensor with flow statistical function |
Non-Patent Citations (2)
Title |
---|
LUCA BERTINETTO 等: "Staple:Complementary Learners for Real-Time Tracking", 《ARXIV:1512.01355V1》 * |
王瑞: "基于机器视觉的人流量双向统计技术研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108537089A (en) * | 2017-03-01 | 2018-09-14 | 开利公司 | Flow of the people estimating system and flow of the people estimating and measuring method |
CN108537089B (en) * | 2017-03-01 | 2024-03-15 | 开利公司 | People flow estimation system and people flow estimation method |
CN107330386A (en) * | 2017-06-21 | 2017-11-07 | 厦门中控智慧信息技术有限公司 | A kind of people flow rate statistical method and terminal device |
CN109697392A (en) * | 2017-10-23 | 2019-04-30 | 北京京东尚科信息技术有限公司 | Draw the method and device of target object thermodynamic chart |
CN110443097A (en) * | 2018-05-03 | 2019-11-12 | 北京中科晶上超媒体信息技术有限公司 | A kind of video object extract real-time optimization method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106096577B (en) | A kind of target tracking method in camera distribution map | |
CN104933710B (en) | Based on the shop stream of people track intelligent analysis method under monitor video | |
US10445567B2 (en) | Pedestrian head identification method and system | |
CN103530874B (en) | People stream counting method based on Kinect | |
Mukherjee et al. | Anovel framework for automatic passenger counting | |
CN103310444B (en) | A kind of method of the monitoring people counting based on overhead camera head | |
CN101794385B (en) | Multi-angle multi-target fast human face tracking method used in video sequence | |
CN106339690A (en) | Video object flow detecting method and system based on noise elimination and auxiliary determination line | |
CN106203513B (en) | A kind of statistical method based on pedestrian's head and shoulder multi-target detection and tracking | |
CN106097391A (en) | A kind of multi-object tracking method identifying auxiliary based on deep neural network | |
CN105303191A (en) | Method and apparatus for counting pedestrians in foresight monitoring scene | |
CN104751491A (en) | Method and device for tracking crowds and counting pedestrian flow | |
CN104408406B (en) | Personnel based on frame difference method and background subtraction leave the post detection method | |
CN108229407A (en) | A kind of behavioral value method and system in video analysis | |
CN103164858A (en) | Adhered crowd segmenting and tracking methods based on superpixel and graph model | |
CN105160297A (en) | Masked man event automatic detection method based on skin color characteristics | |
CN104835147A (en) | Method for detecting crowded people flow in real time based on three-dimensional depth map data | |
CN110874592A (en) | Forest fire smoke image detection method based on total bounded variation | |
CN109145708A (en) | A kind of people flow rate statistical method based on the fusion of RGB and D information | |
CN103646257A (en) | Video monitoring image-based pedestrian detecting and counting method | |
CN106056624A (en) | Unmanned aerial vehicle high-definition image small target detecting and tracking system and detecting and tracking method thereof | |
CN105957108A (en) | Passenger flow volume statistical system based on face detection and tracking | |
CN111160203A (en) | Loitering and lingering behavior analysis method based on head and shoulder model and IOU tracking | |
CN103646254A (en) | High-density pedestrian detection method | |
CN103325115A (en) | Pedestrian counting monitoring method based on head top camera |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170118 |
|
RJ01 | Rejection of invention patent application after publication |