CN103400148B - Video analysis-based bank self-service area tailgating behavior detection method - Google Patents
Video analysis-based bank self-service area tailgating behavior detection method Download PDFInfo
- Publication number
- CN103400148B CN103400148B CN201310332227.1A CN201310332227A CN103400148B CN 103400148 B CN103400148 B CN 103400148B CN 201310332227 A CN201310332227 A CN 201310332227A CN 103400148 B CN103400148 B CN 103400148B
- Authority
- CN
- China
- Prior art keywords
- bank self
- video image
- help coverage
- help
- 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.)
- Expired - Fee Related
Links
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a video analysis-based bank self-service area tailgating behavior detection method, which relates to the technical field of safety monitoring and aims to improve the safety protection effect at a bank self-service area. The method comprises the following steps: picking up real-time video images of the bank self-service area by using a camera at an overlook view angle; analyzing the picked video images to detect the number of people in the self-service area and the time for each person to enter the bank self-service area from the images; then judging whether the tailgating behaviors exist at the bank self-service area according to the detection result. The method provided by the invention is suitable for safety monitoring of the bank self-service area.
Description
Technical field
The present invention relates to safety monitoring technology, particularly relate to the technology that behavioral value method is trailed in a kind of bank self-help coverage based on video analysis.
Background technology
In existing bank safe precaution system, the safety defense monitoring system adopted lacks the real-time detection analysis and the effectively process that bank self-help coverage are existed to the behavior of threat public safety.And the behavior of this threat public safety often some major crimes event early stage performance, such as, trailing the behavior entering bank self-help coverage must be the prelude trailing robbery event.Detect and find the generation of this behavior for crime prevention event, preventing the upgrading of crime dramas in time, and process detection crime dramas all plays an important role afterwards.
Bank self-help coverage refers to that the client that bank sets up can complete deposit voluntarily, withdraws the money, transfer accounts and the operation of financial services equipment and the isolated area of wait such as inquiry.
The behavior of trailing of bank self-help coverage refer to personnel enter Self-Service region operation equipment or etc. after a period of time to be serviced, other staff enter the behavior in this region again.
At present, behavior of trailing for bank self-help coverage is taken precautions against only to take and is installed the simple measures such as anti-tailing linkage door and safeguard, these measures are all the passive type precautionary measures, its strick precaution effect is poor, the behavior of trailing cannot be found in real time, with prevention, prevent and process the crime dramas in self-help bank, improving the security and stability of society.
Summary of the invention
For the defect existed in above-mentioned prior art, technical matters to be solved by this invention be to provide a kind of can find bank self-help coverage in real time trail behavior, behavioral value method is trailed in the bank self-help coverage based on video analysis of improving safety precaution effect.
In order to solve the problems of the technologies described above, behavioral value method is trailed in a kind of bank self-help coverage based on video analysis provided by the present invention, and it is characterized in that, concrete steps are as follows:
Step S1: set one and trail discrimination threshold, described in trail discrimination threshold be a time length value;
Step S2: utilize video camera, with the real time video image of the viewing angles bank self-help coverage overlooked, and the video image of shooting covers whole bank self-help coverage;
Step S3: utilize the real time video image that video processing equipment analytical procedure S2 takes, detect in video image the destination number being positioned at bank self-help coverage, and each target enters the time of bank self-help coverage, then judge according to testing result;
If testing result judges to be positioned at the destination number of bank self-help coverage more than 1 people in video image, and the entry time difference entered between the first two target in bank self-help coverage is greater than and trails discrimination threshold, then judge that bank self-help coverage exists and trail behavior, otherwise then judge that bank self-help coverage does not exist the behavior of trailing.
Further, in described step S2, the cloth of video camera sets up an office and is positioned at the top of bank self-help coverage.
Further, in described step S3, first set one and trail timer, the timing according to trailing timer judges the behavior of trailing;
If video images detection result judges to be positioned at the destination number of bank self-help coverage in video image as zero, and trails the clocking value non-zero of timer, then will trail timer zero setting;
If video images detection result judges to be positioned at the destination number of bank self-help coverage in video image as 1, and the clocking value trailing timer is zero, then timer initiation timing is trailed in order;
If video images detection result judges to be positioned at the destination number of bank self-help coverage more than 1 in video image, and the clocking value trailing timer is greater than and trails discrimination threshold, then judge that bank self-help coverage exists and trail behavior, otherwise then judge that bank self-help coverage does not exist the behavior of trailing.
Further, in described step S3, the concrete steps detecting personnel in bank self-help coverage from video image are as follows:
Step S31: adopt algorithm of target detection to detect and obtain the pixel set that all targets entered in bank self-help coverage occupy in video image;
Step S32: the pixel set obtained according to step S31, adopts object extraction algorithm to obtain position in video image of each target of entering in bank self-help coverage and size;
Step S33: extract each target according to step S32 from video image, adopts effective target feature recognition algorithms, identifies quantity and the color textural characteristics of the effective target in video image.
Further, in described step S31, the algorithm of target detection adopted is Gaussian Mixture Background Algorithm.
Further, in described step S32, the object extraction algorithm adopted has two, respectively is region-growing method, K characteristics of mean clustering procedure by execution sequence.
Behavioral value method is trailed in bank self-help coverage based on video analysis provided by the invention, utilize video camera with the real time video image of the viewing angles bank self-help coverage overlooked, the video image of recycling associated picture analytical algorithm to shooting is analyzed, therefrom detect the personnel amount in bank self-help coverage, and everyone enters the time of bank self-help coverage, and then detect and whether there are personnel and trail the behavior entering Self-Service region, what can find bank self-help coverage in real time trails behavior, improve safety precaution effect, contribute to prevention, prevent and process the crime dramas in self-help bank, improve the security and stability of society.
Embodiment
Be described in further detail the present invention below in conjunction with specific embodiment, but the present embodiment is not limited to the present invention, every employing analog structure of the present invention and similar change thereof, all should list protection scope of the present invention in.
Behavioral value method is trailed in a kind of bank self-help coverage based on video analysis that the embodiment of the present invention provides, and it is characterized in that, concrete steps are as follows:
Step S1: set one and trail discrimination threshold, described in trail discrimination threshold be a time length value;
Step S2: utilize video camera, with the real time video image of the viewing angles bank self-help coverage overlooked, and the video image of shooting covers whole bank self-help coverage;
Wherein, the cloth of described video camera sets up an office and is positioned at the top of bank self-help coverage;
Step S3: utilize the real time video image that video processing equipment analytical procedure S2 takes, detect in video image target (i.e. personnel) quantity being positioned at bank self-help coverage, and each target enters the time of bank self-help coverage, then judge according to testing result;
If testing result judges to be positioned at the destination number of bank self-help coverage more than 1 people in video image, and the entry time difference entered between the first two target in bank self-help coverage is greater than and trails discrimination threshold, then judge that bank self-help coverage exists and trail behavior, otherwise then judge that bank self-help coverage does not exist the behavior of trailing.
In the step S3 of the embodiment of the present invention, first set one and trail timer, timing according to trailing timer judges the behavior of trailing, described timer of trailing can be mounted in software timer in video processing equipment, also can be mounted in the electronic time unit of porch, bank self-help coverage;
If video images detection result judges to be positioned at the destination number of bank self-help coverage in video image as zero, and trails the clocking value non-zero of timer, then will trail timer zero setting;
If video images detection result judges to be positioned at the destination number of bank self-help coverage in video image as 1, and the clocking value trailing timer is zero, then timer initiation timing is trailed in order;
If video images detection result judges to be positioned at the destination number of bank self-help coverage more than 1 in video image, and the clocking value trailing timer is greater than and trails discrimination threshold, then judge that bank self-help coverage exists and trail behavior, otherwise then judge that bank self-help coverage does not exist the behavior of trailing.
In the step S3 of the embodiment of the present invention, after judging that behavior is trailed in the existence of bank self-help coverage, namely send warning message.
In the step S3 of the embodiment of the present invention, detect the concrete steps being positioned at bank self-help coverage internal object in video image as follows:
Step S31: adopt algorithm of target detection to detect and obtain the pixel set (namely entering the personnel's spatial information in bank self-help coverage) that all targets entered in bank self-help coverage occupy in video image;
Step S32: the pixel set obtained according to step S31, adopts object extraction algorithm to obtain position in video image of each target of entering in bank self-help coverage and size (namely entering position and the size of each personnel in bank self-help coverage);
Step S33: extract each target according to step S32 from video image, adopt effective target feature recognition algorithms, identify the quantity of the effective target in video image and color textural characteristics (namely entering the color textural characteristics of personnel amount in bank self-help coverage and each personnel).
In existing video image analysis method, the algorithm of target detection for detecting and obtain the pixel set that all targets occupy in video image mainly contains background subtraction class algorithm, mistiming sorting algorithm, light stream class algorithm.
The ultimate principle of background subtraction class algorithm is the pixel value utilizing the parameter model of background to carry out approximate background image, present frame and background image are carried out the detection that differential comparison realizes moving region, wherein distinguish larger pixel region and be considered to moving region, and the less pixel region of difference is considered to background area.
Mistiming sorting algorithm is one of moving target detecting method conventional in video image analysis method, and ultimate principle is exactly adopt the time difference based on pixel to extract moving region in image by closing value at adjacent two frames of image sequence or three interframe.
The ultimate principle of light stream class algorithm calculates optical flow field, namely under suitable smoothness constraint condition, according to the spatio-temporal gradient estimating motion field of image sequence, detected moving target and scene by the change analyzing sports ground.
In the step S31 of the embodiment of the present invention, the algorithm of target detection adopted is the Gaussian Mixture Background Algorithm in background subtraction class algorithm, this algorithm is prior art, the ultimate principle of this algorithm is: in video image, gray difference is there is between object and background, the grey level histogram of video image can present and background, target multimodal one to one, the grey level histogram multimodal characteristic of video image is considered as the superposition of multiple Gaussian distribution, the segmentation of background in video image and target can be realized.
In the step S31 of the embodiment of the present invention, the Gaussian Mixture Background Algorithm adopted adopts multiple (being generally 3 to 5) gauss hybrid models to characterize the feature of each pixel in video image, mate with gauss hybrid models by each pixel in this image after often obtaining a new frame video image, the pixel that the match is successful is judged to be background dot, otherwise be then judged to be foreground point, the pixel that all targets namely entering bank self-help coverage occupy in video image;
In other embodiment of the present invention, also can realize the Gaussian Mixture Background Algorithm that adopts in the existing algorithm of target detection alternative steps S31 of identical function, such as mistiming sorting algorithm, light stream class algorithm, or other background subtraction class algorithm with other.
In existing video image analysis method, the object extraction algorithm for obtaining the position of each target in video image and size mainly contains the object extraction algorithm of the object extraction algorithm based on Threshold segmentation, the object extraction algorithm based on edge segmentation, the object extraction algorithm based on region segmentation, the object extraction algorithm based on model, feature based cluster and neural network.
Object extraction algorithm based on Threshold segmentation adopts, by one or several threshold value, the grey level histogram of image is divided into several class, and the pixel of gray-scale value in image in same class is attributed to same object, and then the extraction of realize target.
Based on the object extraction algorithm of edge segmentation based on the violent principle of the edge pixel gray-value variation between zones of different, the single order of spatial gradation or Second Order Differential Operator is adopted to carry out rim detection, carry out the extraction of realize target by detecting the edge comprising zones of different, conventional edge detection operator has: Roberts operator, Laplace operator, Prewitt operator, Sobel operator, Robinson operator, Kirsch operator and Canny operator etc.
Based on the object extraction algorithm of region segmentation based on the space local feature of the images such as gradation of image, texture, color and image pixel statistical uniformity, pixel in image is incorporated in each object or region, and then Iamge Segmentation is become several zoness of different, the extraction of realize target, the existing algorithm based on region segmentation mainly contains: region-growing method, split degree method, watershed segmentation method.
Based on the object extraction algorithm of model based on certain model, image segmentation problem is converted to the Solve problems of objective function, and then the extraction of realize target, the object extraction algorithm based on model mainly contains: Markov random field model, active contour model.
The parameter that the object extraction algorithm of feature based cluster and neural network mainly utilizes artificial intelligence approach to obtain for Iamge Segmentation, then splits image based on this parameter;
Wherein, pixel in image space characteristic of correspondence spatial point represents by the object extraction algorithm of feature based cluster, according to their gatherings at feature space, feature space is split, then they are mapped go back to original image space, obtain segmentation result, and then the extraction of realize target, existing feature clustering method has: K characteristics of mean clustering procedure, Fuzzy C-Means Cluster Algorithm;
Wherein, the object extraction algorithm based on neural network obtains linear decision function by training multi-layer perception（MLP）, and then classifying to pixel with decision function reaches the object of segmentation, and then the extraction of realize target.
In the step S32 of the embodiment of the present invention, the object extraction algorithm adopted is prior art, is specially the object extraction algorithm of region-growing method in conjunction with K characteristics of mean clustering procedure, and the concrete implementation step of this algorithm is as follows:
Step S321: adopt region-growing method, each pixel in the pixel set obtained with step S31 for sub pixel point, and sets up growth district Gaussian distribution using the gray-scale value of these pixels as mathematical expectation;
Step S322: each pixel meeting growth district Gaussian distribution in each sub pixel point surrounding neighbors is merged in the region at each sub pixel point place as growing point respectively, again using each growing point as new sub pixel point, repeat this step to occur to there is no new growing point, can the growth district of pixel set that obtains of obtaining step S31, and then obtain the position of each target in video image that enter in bank self-help coverage;
Step S323: adopt K characteristics of mean clustering procedure, choose the average point of each growth district generated by step S322 as cluster centre;
Step S324: calculate the distance of each sample to cluster centre, each sample is grouped into the class from its that nearest cluster centre place, and according to calculating the data object mean value of each cluster formed, obtain new cluster centre, repeat this step to the adjacent cluster centre obtained for twice not change, then show that sample adjustment terminates, clustering criteria function is restrained, and can obtain the size of each target in video image entered in bank self-help coverage.
In other embodiment of the present invention, also can realize the region-growing method that adopts in the existing object extraction algorithm alternative steps S32 of the identical function object extraction algorithm in conjunction with K characteristics of mean clustering procedure with other.
In the step S33 of the embodiment of the present invention, adopt effective target feature recognition algorithms to be prior art, this algorithm first from the effective pixel point each target extracted for sample, set up the color histogram of target corresponding to each sample point again, adopt the formation of bayes decision criterion based on the object statistics feature classifiers of color probability density function, and then feature extraction is carried out to each target, obtain the spatial color Nogata distribution characteristics data of each target, and then according to the spatial color Nogata distribution characteristics data of each target, adaptive template matching algorithm is utilized to carry out target differentiation and identification,
When adaptive template matching algorithm carries out target differentiation and identifies, first according to the statistics size of target each in video image, corresponding template is estimated for each target respectively sets one, with respectively estimating template, compartition is carried out to each corresponding order target area again, acquisition is estimated template with each and is estimated target one to one, eachly estimate the size of target and the corresponding consistent size estimating template, then respectively target is estimated to each according to the spatial color Nogata distribution characteristics data of each target and carry out dimension correction, merge the cut zone with same characteristic features data, and adjust each combined region estimate template size, what it can be used as subsequent video frame estimates template, cut zone after adjustment merges is effective target, its quantity is the quantity of effective target, its spatial color Nogata distribution characteristics data are the color textural characteristics of effective target.
In other embodiment of the present invention, also can realize with other the effective target feature recognition algorithms that adopts in existing effective target feature recognition algorithms alternative steps S33 of identical function.
Claims (3)
1. a behavioral value method is trailed in the bank self-help coverage based on video analysis, and it is characterized in that, concrete steps are as follows:
Step S1: set one and trail discrimination threshold, described in trail discrimination threshold be a time length value;
Step S2: utilize video camera, with the real time video image of the viewing angles bank self-help coverage overlooked, and the video image of shooting covers whole bank self-help coverage;
Step S3: utilize the real time video image that video processing equipment analytical procedure S2 takes, detect in video image the destination number being positioned at bank self-help coverage, and each target enters the time of bank self-help coverage, then judge according to testing result;
If testing result judges to be positioned at the destination number of bank self-help coverage more than 1 people in video image, and the entry time difference entered between the first two target in bank self-help coverage is greater than and trails discrimination threshold, then judge that bank self-help coverage exists and trail behavior, otherwise then judge that bank self-help coverage does not exist the behavior of trailing;
In described step S3, the concrete steps detecting personnel in bank self-help coverage from video image are as follows:
Step S31: adopt algorithm of target detection to detect and obtain the pixel set that all targets entered in bank self-help coverage occupy in video image;
Step S32: the pixel set obtained according to step S31, adopts object extraction algorithm to obtain position in video image of each target of entering in bank self-help coverage and size;
Step S33: extract each target according to step S32 from video image, adopts effective target feature recognition algorithms, identifies quantity and the color textural characteristics of the effective target in video image;
In described step S31, the algorithm of target detection adopted is Gaussian Mixture Background Algorithm;
In described step S32, the object extraction algorithm adopted is the object extraction algorithm of region-growing method in conjunction with K characteristics of mean clustering procedure, the growth district of the pixel set that this algorithm first adopts region-growing method obtaining step S31 to obtain, then adopt K characteristics of mean clustering procedure to obtain the size of each target in video image entered in bank self-help coverage.
2. behavioral value method is trailed in the bank self-help coverage based on video analysis according to claim 1, it is characterized in that: in described step S2, and the cloth of video camera sets up an office and is positioned at the top of bank self-help coverage.
3. behavioral value method is trailed in the bank self-help coverage based on video analysis according to claim 1, it is characterized in that: in described step S3, and first set one and trail timer, the timing according to trailing timer judges the behavior of trailing;
If video images detection result judges to be positioned at the destination number of bank self-help coverage in video image as zero, and trails the clocking value non-zero of timer, then will trail timer zero setting;
If video images detection result judges to be positioned at the destination number of bank self-help coverage in video image as 1, and the clocking value trailing timer is zero, then timer initiation timing is trailed in order;
If video images detection result judges to be positioned at the destination number of bank self-help coverage more than 1 in video image, and the clocking value trailing timer is greater than and trails discrimination threshold, then judge that bank self-help coverage exists and trail behavior, otherwise then judge that bank self-help coverage does not exist the behavior of trailing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310332227.1A CN103400148B (en) | 2013-08-02 | 2013-08-02 | Video analysis-based bank self-service area tailgating behavior detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310332227.1A CN103400148B (en) | 2013-08-02 | 2013-08-02 | Video analysis-based bank self-service area tailgating behavior detection method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103400148A CN103400148A (en) | 2013-11-20 |
| CN103400148B true CN103400148B (en) | 2015-04-01 |
Family
ID=49563766
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310332227.1A Expired - Fee Related CN103400148B (en) | 2013-08-02 | 2013-08-02 | Video analysis-based bank self-service area tailgating behavior detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103400148B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106210635A (en) * | 2016-07-18 | 2016-12-07 | 四川君逸数码科技股份有限公司 | A kind of wisdom gold eyeball identification is moved through method and apparatus of reporting to the police |
| CN106961576A (en) * | 2017-03-10 | 2017-07-18 | 盛视科技股份有限公司 | video anti-tailing method, device and system |
| CN109376639B (en) * | 2018-10-16 | 2021-12-17 | 上海弘目智能科技有限公司 | Accompanying personnel early warning system and method based on portrait recognition |
| CN109784274B (en) * | 2018-12-29 | 2021-09-14 | 杭州励飞软件技术有限公司 | Method for identifying trailing and related product |
| CN111144231B (en) * | 2019-12-09 | 2022-04-15 | 深圳市鸿逸达科技有限公司 | Self-service channel anti-trailing detection method and system based on depth image |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1658670A (en) * | 2004-02-20 | 2005-08-24 | 上海银晨智能识别科技有限公司 | Intelligent tracking monitoring system with multi-camera |
| CN101079171A (en) * | 2007-06-01 | 2007-11-28 | 北京汇大通业科技有限公司 | Intelligent video monitoring system of bank self-aid apparatus |
| CN103116959A (en) * | 2013-01-25 | 2013-05-22 | 上海博超科技有限公司 | Analyzing and recognizing method for abnormal behaviors in intelligent videos |
-
2013
- 2013-08-02 CN CN201310332227.1A patent/CN103400148B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1658670A (en) * | 2004-02-20 | 2005-08-24 | 上海银晨智能识别科技有限公司 | Intelligent tracking monitoring system with multi-camera |
| CN101079171A (en) * | 2007-06-01 | 2007-11-28 | 北京汇大通业科技有限公司 | Intelligent video monitoring system of bank self-aid apparatus |
| CN103116959A (en) * | 2013-01-25 | 2013-05-22 | 上海博超科技有限公司 | Analyzing and recognizing method for abnormal behaviors in intelligent videos |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103400148A (en) | 2013-11-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103400120B (en) | Video analysis-based bank self-service area push behavior detection method | |
| CN105979210B (en) | A kind of pedestrian's identifying system based on the more ball video camera arrays of multiple gun | |
| Yang et al. | Real-time and accurate segmentation of moving objects in dynamic scene | |
| CN107085714B (en) | Forest fire detection method based on video | |
| CN103400148B (en) | Video analysis-based bank self-service area tailgating behavior detection method | |
| Zhao et al. | A people counting system based on face detection and tracking in a video | |
| Zang et al. | Robust background subtraction and maintenance | |
| CN103927520A (en) | Method for detecting human face under backlighting environment | |
| Vosters et al. | Background subtraction under sudden illumination changes | |
| CN109918971A (en) | Number detection method and device in monitor video | |
| Huang et al. | Motion detection with pyramid structure of background model for intelligent surveillance systems | |
| Lin et al. | Collaborative pedestrian tracking and data fusion with multiple cameras | |
| Karpagavalli et al. | Estimating the density of the people and counting the number of people in a crowd environment for human safety | |
| CN101483763A (en) | Digital video processing method oriented to social security monitoring and apparatus thereof | |
| CN109410222B (en) | Flame detection method and device | |
| CN105354540A (en) | Video analysis based method for implementing person fall-down behavior detection | |
| CN109948474A (en) | AI thermal imaging all-weather intelligent monitoring method | |
| CN106228709A (en) | It is a kind of that wisdom gold eyeball identification is single adds paper money alarm method and device | |
| Ullah et al. | Gaussian mixtures for anomaly detection in crowded scenes | |
| CN109919068B (en) | Real-time monitoring method for adapting to crowd flow in dense scene based on video analysis | |
| KR20200060868A (en) | multi-view monitoring system using object-oriented auto-tracking function | |
| Seidenari et al. | Dense spatio-temporal features for non-parametric anomaly detection and localization | |
| CN103971100A (en) | Video-based camouflage and peeping behavior detection method for automated teller machine | |
| Lai et al. | Robust little flame detection on real-time video surveillance system | |
| CN106650639A (en) | Monitoring system pedestrian re-identification method suitable for security and protection |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20150401 Termination date: 20200802 |