CN109064686A - A kind of ATM trailing detection method based on human body segmentation - Google Patents
A kind of ATM trailing detection method based on human body segmentation Download PDFInfo
- Publication number
- CN109064686A CN109064686A CN201810940588.7A CN201810940588A CN109064686A CN 109064686 A CN109064686 A CN 109064686A CN 201810940588 A CN201810940588 A CN 201810940588A CN 109064686 A CN109064686 A CN 109064686A
- Authority
- CN
- China
- Prior art keywords
- human body
- segmentation
- atm
- detection method
- target
- 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
- G07—CHECKING-DEVICES
- G07F—COIN-FREED OR LIKE APPARATUS
- G07F19/00—Complete banking systems; Coded card-freed arrangements adapted for dispensing or receiving monies or the like and posting such transactions to existing accounts, e.g. automatic teller machines
- G07F19/20—Automatic teller machines [ATMs]
- G07F19/207—Surveillance aspects at ATMs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/269—Analysis of motion using gradient-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Human Computer Interaction (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of, and the ATM based on human body segmentation trails detection method.The present invention detected from image using the technology of deep learning using human body as foreground target in ATM protective cabin, then expert is carried out to the size of dimensions of human figure under different cameral, different angle to estimate, divide human body using partitioning algorithm, to calculate the number in ATM protective cabin, if being judged to detecting trailing more than 2 people, alarm.If statistical number of person is 1 people, trailing is not detected.The present invention uses the technology of deep learning, carrys out the human body in segmented image as target prospect, avoids the problem of human body that long-time is stopped in traditional background modeling method is as background, improves the accuracy rate of human body target foreground detection.
Description
Technical field
The invention belongs to technical field of computer vision, are related to a kind of ATM trailing detection method based on human body segmentation.
Background technique
ATM protective cabin based on demographics, which trails detection, has very important realistic meaning, can protect bank and
The generation of ATM crime dramas is lowered in the property safety of the common people, prevention.
The people counting algorithm for trailing detection currently used for ATM protective cabin has:
1) it human body segmentation's algorithm based on light stream: is obtained by optic flow technique tired perpendicular through the space-time light stream for mixing informant's body
Then product figure obtains the number by mixing line using human body segmentation's algorithm analysis light stream figure.The algorithm is in camera angle, illumination etc.
When changing, accuracy rate meeting sharp fall needs to manually adjust system parameter, increases cost of labor.
2) detection human body the human body target detection algorithm of target scale estimation: is estimated by target scale using deep learning
Target, to count human body number.The algorithm is a kind of algorithm end to end, and place one's entire reliance upon sample data, flat in sample data
Weighing apparatus property is bad, and in scene type situation not abundant enough, training gained algorithm is not high in certain outstanding scene accuracys rate.
3) it the deep learning algorithm based on light stream: is obtained by optic flow technique and is accumulated by the space-time light stream for mixing informant's body
Then figure obtains the number by mixing line using the technology analysis light stream figure of deep learning.The algorithm accuracy rate and robustness compared with
The first algorithm improves, but is still limited by optic flow technique, and when acute variation occurs for light, calculating will appear
Very big error, and calculate complex.
Summary of the invention
Deficiency existing for detection method is trailed based on demographics for existing, the purpose of the present invention is to provide a kind of bases
Detection method is trailed in the ATM of human body segmentation.
The present invention detected from image using the technology of deep learning using human body as foreground target, then to difference
The size progress expert of dimensions of human figure estimates under camera, different angle, divides human body using partitioning algorithm, to accurately unite
Meter number realizes that the ATM personnel of high-accuracy trail detection.
The method of the present invention the following steps are included:
Step 1 carries out image preprocessing to the image of ATM protective cabin vertical camera acquisition.
Step 2 trains deep learning model using a large amount of rich and varied sample datas, makes the deep learning mould obtained
Type can detected from image using human body as foreground target.
Step 3 estimates the dimensions of human figure progress expert under different cameral, different angle.
Step 4, the human body target size estimated according to expert, select maximum a posteriori using Markov Monte Carlo Method
The segmentation result of probability value, to complete human body segmentation's task in foreground target.
Step 5, the human body number of statistics segmentation are then considered as and trail if it is greater than 2 people, trigger alarm.
Beneficial effects of the present invention:
1. carrying out the human body in segmented image as target prospect using the technology of deep learning, avoiding traditional background and build
The human body that long-time is stopped in mould method regards the problem of background, improves the accuracy rate of human body target foreground detection.
2. overlapping is substantially absent in target body, therefore can be big in the image of the vertical camera acquisition of ATM protective cabin
The big calculation amount for reducing MCMC and seeking maximum a posteriori probability reduces and calculates cost and improve demographics and trail the effect of detection
Rate.
Specific embodiment
It in order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below will be in the embodiment of the present invention
Technical solution carry out clear, complete description, it is clear that described embodiments are only a part of the embodiments of the present invention, and
The embodiment being not all of.Based on the embodiment of the present invention, those of ordinary skill in the art are not before making creative work
Every other embodiment obtained is put, shall fall within the protection scope of the present invention.
The technical scheme adopted by the invention is that: before human body is used as by the technology in ATM protective cabin using deep learning
Scape target detected from image, then carries out expert to the size of dimensions of human figure under different cameral, different angle and estimates, benefit
Divide human body with partitioning algorithm, to calculate the number in ATM protective cabin, if being judged to detecting trailing more than 2 people, carries out
Alarm.If statistical number of person is 1 people, trailing is not detected, specifically comprises the following steps:
1, image preprocessing is carried out to the image of ATM protective cabin vertical camera acquisition.
2, deep learning model is trained using a large amount of rich and varied sample datas, makes the deep learning model obtained can
Human body detected from image as foreground target using high-accuracy.
3, expert is carried out to the dimensions of human figure under different cameral, different angle to estimate.
4, the human body target size estimated according to expert, selects maximum a posteriori using Markov Monte Carlo MCMC
The segmentation result of probability value, to complete human body segmentation's task in foreground target.
5, the human body number of statistics segmentation is then considered as and trails if it is greater than 2 people, triggers alarm.
Embodiment:
1. the image of the vertical camera acquisition of pair ATM protective cabin carries out image preprocessing.
2. the people of different building shape posture, the people of difference dressing train as sample data using under different illumination conditions
Deep learning model goes out human body foreground target in image using model inspection.
3. estimating using expert system to the dimensions of human figure under different cameral, different angle, an interval value is obtained.
4. seeking maximum a posteriori probability using MCMC according to the dimensions of human figure interval value that previous step obtains.
In finding process, two factors: overlapping, the expert of foreground target pixel and segmentation prediction human body pixel are only considered
System estimates the size of dimensions of human figure.
By MCMC methodology come the number of random change human body or position or size, and utilize formula:
P (I | θ)=Πi∈IP(i|θ)
Calculate maximum posteriori probability value, wherein I indicates image.
When the maximum a posteriori probability of acquisition than last time change it is big when, record current state, current state carry out again with
Machine changes, until iteration 500 times.
Record optimal value of the final value as maximum a posteriori probability.
5, human body segmentation is carried out using optimal maximum posteriori probability value, carries out demographics.When number is greater than 1, in advance
There is personnel's trailing in police.
The foregoing is only a preferred embodiment of the present invention, is not intended to limit the scope of the present invention, should
Understand, the present invention is not limited to implementation as described herein, the purpose of these implementations description is to help this field
In technical staff practice the present invention.
Claims (5)
1. a kind of ATM based on human body segmentation trails detection method, it is characterised in that method includes the following steps:
Step 1 carries out image preprocessing to the image of ATM protective cabin vertical camera acquisition;
Step 2 trains deep learning model using a large amount of rich and varied sample datas, makes the deep learning model obtained can
Human body detected from image as foreground target;
Step 3 estimates the dimensions of human figure progress expert under different cameral, different angle;
Step 4, the human body target size estimated according to expert, select maximum a posteriori probability using Markov Monte Carlo Method
The segmentation result of value, to complete human body segmentation's task in foreground target;
Step 5, the human body number of statistics segmentation are then considered as and trail if it is greater than 2 people, trigger alarm.
2. a kind of ATM based on human body segmentation according to claim 1 trails detection method, it is characterised in that: in step 2
The sample data uses under different illumination conditions, the people of different building shape posture, different dressings people as sample data.
3. a kind of ATM based on human body segmentation according to claim 1 or 2 trails detection method, it is characterised in that: most
Greatly in posterior probability values finding process, only consider two factors: foreground target pixel is overlapping, special with segmentation prediction human body pixel
Family estimates the size of human body target.
4. a kind of ATM based on human body segmentation according to claim 3 trails detection method, it is characterised in that: Ma Erke
Number, position or the size of random change human body are needed in husband's Monte Carlo Method.
5. a kind of ATM based on human body segmentation according to claim 4 trails detection method, it is characterised in that: work as acquisition
Maximum posteriori probability value when changing big than the last time, record current state, the random change human body again under current state
Number, position or size simultaneously calculate maximum posteriori probability value, until iteration 500 times.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810940588.7A CN109064686A (en) | 2018-08-17 | 2018-08-17 | A kind of ATM trailing detection method based on human body segmentation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810940588.7A CN109064686A (en) | 2018-08-17 | 2018-08-17 | A kind of ATM trailing detection method based on human body segmentation |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109064686A true CN109064686A (en) | 2018-12-21 |
Family
ID=64687274
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810940588.7A Pending CN109064686A (en) | 2018-08-17 | 2018-08-17 | A kind of ATM trailing detection method based on human body segmentation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109064686A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109711334A (en) * | 2018-12-26 | 2019-05-03 | 浙江捷尚视觉科技股份有限公司 | A kind of ATM trailing event detecting method based on space-time optical flow field |
Citations (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
AU2002301177A1 (en) * | 2001-09-26 | 2003-06-12 | Rockwell Electronic Commerce Technologies, Llc | System and method for predictive customer contacts |
CN101814183A (en) * | 2010-01-08 | 2010-08-25 | 清华大学 | Image segmentation method and system |
CN102034318A (en) * | 2010-11-01 | 2011-04-27 | 秦小枢 | Safety monitoring device matched with ATM (automatic teller machine) machine |
CN102855475A (en) * | 2012-09-17 | 2013-01-02 | 广州杰赛科技股份有限公司 | School bus monitoring method and school bus monitoring system |
CN105138979A (en) * | 2015-08-19 | 2015-12-09 | 南京理工大学 | Method for detecting the head of moving human body based on stereo visual sense |
CN205153609U (en) * | 2015-11-28 | 2016-04-13 | 段继丹 | Finance pavilion of withdrawing money by oneself |
CN105513221A (en) * | 2015-12-30 | 2016-04-20 | 四川川大智胜软件股份有限公司 | ATM (Automatic Teller Machine) cheat-proof device and system based on three-dimensional human face identification |
CN105718857A (en) * | 2016-01-13 | 2016-06-29 | 兴唐通信科技有限公司 | Human body abnormal behavior detection method and system |
WO2016154298A1 (en) * | 2015-03-23 | 2016-09-29 | Temple University-Of The Commonwealth System Of Higher Education | System and method for automatic interpretation of eeg signals using a deep learning statistical model |
CN106228709A (en) * | 2016-07-18 | 2016-12-14 | 四川君逸数码科技股份有限公司 | It is a kind of that wisdom gold eyeball identification is single adds paper money alarm method and device |
CN106373240A (en) * | 2016-09-14 | 2017-02-01 | 浙江维融电子科技股份有限公司 | Intelligent monitoring system of self-service bank and monitoring method thereof |
CN107424153A (en) * | 2017-04-18 | 2017-12-01 | 辽宁科技大学 | Face cutting techniques based on deep learning and Level Set Method |
CN207110556U (en) * | 2017-08-10 | 2018-03-16 | 四川五洲隆科技有限公司 | A kind of big visual angle safety protection cabin |
CN107911663A (en) * | 2017-11-27 | 2018-04-13 | 江苏理工学院 | A kind of elevator passenger hazardous act intelligent recognition early warning system based on Computer Vision Detection |
US9984579B1 (en) * | 2016-06-28 | 2018-05-29 | Amazon Technologies, Inc. | Unmanned aerial vehicle approach notification |
-
2018
- 2018-08-17 CN CN201810940588.7A patent/CN109064686A/en active Pending
Patent Citations (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
AU2002301177A1 (en) * | 2001-09-26 | 2003-06-12 | Rockwell Electronic Commerce Technologies, Llc | System and method for predictive customer contacts |
CN101814183A (en) * | 2010-01-08 | 2010-08-25 | 清华大学 | Image segmentation method and system |
CN102034318A (en) * | 2010-11-01 | 2011-04-27 | 秦小枢 | Safety monitoring device matched with ATM (automatic teller machine) machine |
CN102855475A (en) * | 2012-09-17 | 2013-01-02 | 广州杰赛科技股份有限公司 | School bus monitoring method and school bus monitoring system |
WO2016154298A1 (en) * | 2015-03-23 | 2016-09-29 | Temple University-Of The Commonwealth System Of Higher Education | System and method for automatic interpretation of eeg signals using a deep learning statistical model |
CN105138979A (en) * | 2015-08-19 | 2015-12-09 | 南京理工大学 | Method for detecting the head of moving human body based on stereo visual sense |
CN205153609U (en) * | 2015-11-28 | 2016-04-13 | 段继丹 | Finance pavilion of withdrawing money by oneself |
CN105513221A (en) * | 2015-12-30 | 2016-04-20 | 四川川大智胜软件股份有限公司 | ATM (Automatic Teller Machine) cheat-proof device and system based on three-dimensional human face identification |
CN105718857A (en) * | 2016-01-13 | 2016-06-29 | 兴唐通信科技有限公司 | Human body abnormal behavior detection method and system |
US9984579B1 (en) * | 2016-06-28 | 2018-05-29 | Amazon Technologies, Inc. | Unmanned aerial vehicle approach notification |
CN106228709A (en) * | 2016-07-18 | 2016-12-14 | 四川君逸数码科技股份有限公司 | It is a kind of that wisdom gold eyeball identification is single adds paper money alarm method and device |
CN106373240A (en) * | 2016-09-14 | 2017-02-01 | 浙江维融电子科技股份有限公司 | Intelligent monitoring system of self-service bank and monitoring method thereof |
CN107424153A (en) * | 2017-04-18 | 2017-12-01 | 辽宁科技大学 | Face cutting techniques based on deep learning and Level Set Method |
CN207110556U (en) * | 2017-08-10 | 2018-03-16 | 四川五洲隆科技有限公司 | A kind of big visual angle safety protection cabin |
CN107911663A (en) * | 2017-11-27 | 2018-04-13 | 江苏理工学院 | A kind of elevator passenger hazardous act intelligent recognition early warning system based on Computer Vision Detection |
Non-Patent Citations (2)
Title |
---|
易娴: "密集群体分割与异常行为检测的研究与实现", 《信息科技辑》 * |
朱志磊: "关于ATM机隔间内尾随检测算法的研究", 《信息科技辑》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109711334A (en) * | 2018-12-26 | 2019-05-03 | 浙江捷尚视觉科技股份有限公司 | A kind of ATM trailing event detecting method based on space-time optical flow field |
CN109711334B (en) * | 2018-12-26 | 2021-02-05 | 浙江捷尚视觉科技股份有限公司 | ATM trailing event detection method based on space-time optical flow field |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103489199B (en) | video image target tracking processing method and system | |
CN102799883B (en) | Method and device for extracting movement target from video image | |
CN110852283A (en) | Helmet wearing detection and tracking method based on improved YOLOv3 | |
CN104268851A (en) | ATM self-service business hall behavior analysis method based on depth information | |
CN110379168B (en) | Traffic vehicle information acquisition method based on Mask R-CNN | |
CN109344690B (en) | People counting method based on depth camera | |
CN108776974B (en) | A kind of real-time modeling method method suitable for public transport scene | |
CN103164711A (en) | Regional people stream density estimation method based on pixels and support vector machine (SVM) | |
CN106845325B (en) | A kind of information detecting method and device | |
CN108764167A (en) | A kind of target of space time correlation recognition methods and system again | |
CN105261039B (en) | A kind of self-adaptative adjustment target tracking algorism based on depth image | |
CN110633643A (en) | Abnormal behavior detection method and system for smart community | |
CN103488993A (en) | Crowd abnormal behavior identification method based on FAST | |
CN103456009B (en) | Object detection method and device, supervisory system | |
CN104820995A (en) | Large public place-oriented people stream density monitoring and early warning method | |
CN112541938A (en) | Pedestrian speed measuring method, system, medium and computing device | |
CN102609945A (en) | Automatic registration method of visible light and thermal infrared image sequences | |
CN104933735A (en) | A real time human face tracking method and a system based on spatio-temporal context learning | |
CN104463121A (en) | Crowd density information obtaining method | |
CN104217442B (en) | Aerial video moving object detection method based on multiple model estimation | |
CN108471497A (en) | A kind of ship target real-time detection method based on monopod video camera | |
CN103984937A (en) | Pedestrian counting method based on optical flow method | |
CN108764338A (en) | A kind of pedestrian tracking algorithm applied to video analysis | |
CN108230351A (en) | Sales counter evaluation method and system based on binocular stereo vision pedestrian detection | |
CN109064686A (en) | A kind of ATM trailing detection method based on human body segmentation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181221 |