CN101609580A - Method for intelligently protecting bank self-service equipment - Google Patents
Method for intelligently protecting bank self-service equipment Download PDFInfo
- Publication number
- CN101609580A CN101609580A CNA2008100392464A CN200810039246A CN101609580A CN 101609580 A CN101609580 A CN 101609580A CN A2008100392464 A CNA2008100392464 A CN A2008100392464A CN 200810039246 A CN200810039246 A CN 200810039246A CN 101609580 A CN101609580 A CN 101609580A
- Authority
- CN
- China
- Prior art keywords
- video
- frame
- text
- mode
- information
- 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
Images
Landscapes
- Alarm Systems (AREA)
Abstract
The present invention relates to a kind of method for intelligently protecting bank self-service equipment, realize: (1) system initialization by following steps; (2) video obtains; (3) algorithm detects; (4) message management and alarm; (5) testing result is inserted database; The invention has the beneficial effects as follows: the original video two field picture that obtains by camera chain and video frame images handled, can accurately judge the various unusual conditions that occur in the current video, and reported to the police at any time, help bank clerk in the shortest time, to make the precautionary measures.
Description
Technical field
The present invention relates to a kind of bank self-aid apparatus, relate in particular to the guard method of this bank self-aid apparatus.
Background technology
The monitoring rank of bank ATM machine system is low excessively in the prior art; just common video monitoring; without any the intelligent alarm measure; utilize ATM swindle or obtain client's password and reach and implement aspect the crime at the delinquency prevention molecule; can not effectively protect client's interests; so the monitoring rank of raising ATM system is so that better protection bank and client's interests just become a urgent demand.
Summary of the invention
The technical issues that need to address of the present invention have provided a kind of method for intelligently protecting bank self-service equipment, are intended to solve the above problems.
In order to solve the problems of the technologies described above, the present invention realizes by following steps:
System initialization;
Video obtains;
Algorithm detects;
Message management and alarm;
Testing result is inserted database;
Compared with prior art, the invention has the beneficial effects as follows: the original video two field picture that obtains by camera chain and video frame images handled, can accurately judge the various unusual conditions that occur in the current video, and reported to the police at any time, help bank clerk in the shortest time, to make the precautionary measures.
Description of drawings
Fig. 1 is a process flow diagram of the present invention;
Fig. 2 is anomalies health check-up flow gauge figure;
Embodiment
Below in conjunction with accompanying drawing and embodiment the present invention is described in further detail:
The present invention realizes by following steps:
1, system initialization;
2, video obtains;
3, algorithm detects;
4, message management and alarm;
5, testing result is inserted database;
In step 1: described initialization comprises: the judgement of unusual object detection, unusual condition is selected as acquiescence; The loading of image size, formatting; The loading of various algorithms and initialization;
In step 2: comprise two kinds of video capture pattern: A, natural light+video camera or B, infrared LED light filling lamp+video camera; Wherein: Mode B is additional to the application of Mode A: under the unfavorable state of light, can enable Mode B automatically and obtain desirable video image, be convenient to the diagnosis of video algorithm and the normal operation of system;
In step 3: comprise the diagnosis of unusual object detection and abnormal conditions;
Described unusual object detection realizes by following steps:
Video image is decomposed, select foreground image, analyze foreground picture;
Carry out the connected domain analysis of text constituent element, and according to the regularity of distribution of text, it is text filed to be with " connected domain of the regular arrangement of a plurality of mutual vicinities " that criterion is discerned;
The how text filed zone of carrying out is made up, reach video image Chinese version location.
Below above-mentioned step is done following analysis:
Coloured image and frame of video
Coloured image or frame of video are one 24 images, and this means in the sub-picture can have 2 at most
24Plant color, because the employed color of text accounts for only a few wherein after all, therefore be necessary that the position of carrying out earlier the RGB color gamut reduces and quantification work, 24 coloured images comprise 8 R, G, B, rgb image can be converted into 256 color images, or get high 2 in 24 very color per 8 simply, like this, 24 coloured images can be reduced to 6 coloured images quickly, and number of colours also die-offs to 64 (2
6) plant.
The extraction of color reduction and foreground picture
In fact, the color reduction does not need the obvious of ten minutes with the division of these two processes of extraction of foreground picture, and their something in common is and noise spot will be removed, and candidate's text pixel extraction is come out.Determine noise spot, can adopt the instrument of color histogram as a judgement, if that is: in the sub-picture, total number of certain color pixel is just several altogether, can conclude that these several pixels are noise spots.
For the image after the color reduction, though its number of color is a lot, to compare with the image of not reduction, number significantly reduces, and this figure is referred to as many-valued figure.
A multivalue image can be broken down into U elemental map image set I={I in theory
i, each I
iThe pixel value that image is all arranged, that is,
Generally speaking, think that text in the multivalue image is that color with unanimity presents basically, also has one or several color value.If text is made up of multiple color or texture, then the background at general its place is made up of consistent basically color.Therefore, image always can be divided into foreground image I
FWith background image I
B, and know I
F+ I
B=I, I
FAnd I
BDo not have and occur simultaneously.
(a) if text is a monodrome, I then
F=∪ I
eAnd
(b) if text is many-valued, I then
F=I-I
BAnd I
B=∪ I
bAnd
According to above-mentioned theory, can carry out color sub-clustering (Color Clustering) to coloured image, make color gamut in the text quantize to the color class of some, make pixel be assigned to the color class close, indicate the connection constituent element of candidate characters for each color class again with they primitive colors.
Connected domain (Connected Component) is calculated
Image is carried out Run-Length Coding (run length), make up connected domain (line adjacency graph) according to the distance of swimming.Algorithm is as follows:
Each distance of swimming of the initial row of input picture is considered and is connected domain independently
To the subsequent rows do in the image
{
To each distance of swimming Rc circulation do in the current line
{
If (certain distance of swimming of Rc and lastrow is communicated with, and these two distances of swimming have identical color)
Then (connected domain of Rc being incorporated into the distance of swimming place of lastrow)
else
Then (with Rc as a new connected domain)
}
While (distance of swimming of image remaining row not circulation finishes)
Differentiate the text pixel
Connected domain according to the distance of swimming makes up comprises too much redundant information, must leach these information, to extract the information of text, reaches the purpose of final localization of text.
The connected domain of text pixel has following discrimination method:
1) connected domain is too little, deletion;
2) connected domain is too big, deletion;
3) the serious disproportion of connected domain (farsighted greater than wide, or wide much larger than height), deletion;
4) if current connected domain is isolated, and the connected domain scope is very little, is then directly deleted;
5) if current connected domain and other connected domains are overlapping, or the adjacent and close then merging of color value;
6) each connected domain is done variance, variance yields is too little, can consider deletion;
7) each connected domain is x, the y direction projection is found out crest, trough;
Merge the text connected domain
In the text pixel column, the height of each connected domain, roomy cause approaching; The number of connected domain and the length of line of text are proportional; The top margin of connected domain, base differ in certain scope; Spacing between the connected domain satisfies certain requirement, and promptly the connected domain of text satisfies certain regularly arranged.
The text connected domain that identifies, some overlapping staggered phenomenon can occur, or dislocation is interrupted each other, and this just needs adjacent connected domain merging is become a big connected domain piece.During concrete the merging, according to differentiating that rule is judged.
Text filed for after merging utilizes the geometric properties of text, further optimized.Such as, for horizontal text, the width of final connected domain piece generally all can be greater than text filed height.
The identification text obtains text filed
Text connected domain to obtaining is merged, and can judge text filed according to adjacent a plurality of connected domain rules.
The diagnosis of described abnormal conditions comprises: illumination effect, effect of jitter and block influence:
Described illumination effect: if Video Detection information is from the Mode A in the step 2, but light is bad, video is fuzzy, anomalies health check-up method of determining and calculating can't normally start so, at this moment, will send assistance information, the automatic start-up mode B of system's meeting, this moment, the video council of Mode A got clogged, and system obtains the video information from Mode B, and system exception situation evaluation algorithm will start;
Described effect of jitter: if the shake of video camera is bigger, then will inevitably influence the quality of particular video frequency frame, make that the result of text detection is not accurate enough.To this, the method of signature analysis is done in employing to each frame of video, extract the feature on the frame of video, adopt the zonule to do the method for mixed Gaussian background modeling (Gaussian Mixture Model), extract the background information in the frame of video, compare with the multiframe background information, if shake, then the background information in the present frame can have greatly changed, by comparing with the background information of the preceding frame picture of number, also the background information in the script frame so just can be eliminated to a great extent because the influence that DE Camera Shake caused finally obtains regional aim;
Describedly block influence: if video camera is blocked, will lose relevant important video information, the quality of influence monitoring greatly.Therefore, adopt the method for multiframe mixed Gaussian background modeling (GaussianMixture Model), draw the target context in the video information earlier, again current each frame is done analysis, whether be blocked to judge present frame, if be blocked, then the background information in the present frame will most ofly disappear, so just can be judged accurately, be reached the purpose of timely warning.
The mixed Gauss model of prior art (GMM) is one can handle multi-modal situation and adaptive model.It describes the distribution of the value of a pixel with the mixing of a plurality of Gaussian distribution.Each surface under the multi-modal situation is described by a Gaussian distribution, and the Gaussian distribution of describing background surface is wherein arranged, and the distribution on description prospect surface is also arranged.In the GMM model, the brightness value of each pixel comes modeling with the mixing of K Gaussian distribution, and each Gaussian distribution has different weights ω respectively
jAnd priority, they are always according to priority order ordering from high to low.Get suitable surely background weights part and threshold value, only preceding several within this threshold value distribute and just are considered to background distributions, and other then are that prospect distributes.Mixed Gauss model can be described multi-modal background, and has good adaptivity.
In step 4: the message signale that other modules are sent in the main coherent system of this module, and handle these message signales, result is returned to corresponding module; If the system of drawing needs to report to the police, then send warning message immediately and give system, system can in time start flashlamp, report to the police (warning is exactly to start flashlamp, and the screen that do not stop by flashlamp dodges, to reach the effect of timely warning).
Message processing module can make other resume module reach in the communication synchronously, make the communication of total system integrate and reach consistent; This module can individual processing from the information of a module, also can handle information simultaneously from a plurality of modules.
Message processing module makes the cooperation between each module of system become simple, flexibly, configurable. make the total system framed structure have extremely strong extensibility, each intermodule of system both can be worked alone, cooperation that again can be continuously, but also can add new functional module as required.
In step 5: testing result is inserted database, unusual object that record this time detects or the unusual condition that is diagnosed: Word message, unusual condition, time, key frame picture.
Come the present invention is illustrated below by an example:
(1) system starts camara module earlier, default mode A starts, obtain video flowing by capture card, then the video streaming image that obtains is sent to the algorithm detection module, this module is carried out quality testing to the video streaming image of sending into immediately, if video quality reaches requirement, then subsequent algorithm starts immediately, if video quality can not meet the demands, message is through behind the message processing module, then can send corresponding message immediately and give system, after system obtains message, can start video capture mode B (adding the LED light filling) immediately, the image/video stream that collects by Mode B will be sent to the algorithm detection module again, the algorithm detection module will detect video quality again, if quality satisfies, then subsequent algorithm starts, if video quality can not meet the demands, system will continue repeated acquisition, meet the demands up to video quality;
(2) video flowing that satisfies of the quality that collects by (1) will be fed to other parts of algorithm detection module, comprises the diagnosis of text detection and abnormal conditions.Detect by the core respective algorithms, this algoritic module will initiate a message to message module, after message module is handled this message, can immediately result be issued system, and system starts corresponding module according to the message result who obtains, and comprises the module of warning;
(3) if in (2), system sends the indication of reporting to the police, and then alarm module can be reported to the police immediately, and the detected corresponding warning key frame picture of algorithm is saved in the database of system immediately, checks constantly in order to follow-up;
(4) along with the continuous input of video flowing, what all modules of whole system were not stopped analyzes video flowing, and preserves analysis result, reaches the purpose of monitoring constantly.
Claims (4)
1, a kind of method for intelligently protecting bank self-service equipment, realize by following steps:
(1), system initialization;
(2), video obtains;
(3), algorithm detects;
(4), message management and alarm;
(5), testing result is inserted database.
2, method for intelligently protecting bank self-service equipment according to claim 1, wherein in step (1): described initialization comprises: the judgement of unusual object detection, unusual condition is selected as acquiescence; The loading of image size, formatting; The loading of various algorithms and initialization;
In step (2): comprise two kinds of video capture pattern: A, natural light+video camera or B, infrared LED light filling lamp+video camera; Wherein: Mode B is additional to the application of Mode A: under the unfavorable state of light, can enable Mode B automatically and obtain desirable video image, be convenient to the diagnosis of video algorithm and the normal operation of system;
In step (3): comprise the diagnosis of unusual object detection and abnormal conditions;
In step (4): the message signale that other modules are sent in the coherent system, and handle these message signales, result is returned to corresponding module; If the system of drawing needs to report to the police, then send warning message immediately and give system, system can in time start flashlamp, reports to the police;
In step (5): testing result is inserted database, unusual object that record this time detects or the unusual condition that is diagnosed: Word message, unusual condition, time, key frame picture.
3, method for intelligently protecting bank self-service equipment according to claim 2, wherein unusual object detection realizes by following steps:
Video image is decomposed, select foreground image, analyze foreground picture;
Carry out the connected domain analysis of text constituent element, and according to the regularity of distribution of text, it is text filed to be with " connected domain of the regular arrangement of a plurality of mutual vicinities " that criterion is discerned;
The how text filed zone of carrying out is made up, reach video image Chinese version location.
4, method for intelligently protecting bank self-service equipment according to claim 2, wherein abnormal conditions diagnosis comprises: illumination effect, effect of jitter and block influence;
Described illumination effect: if Video Detection information is from the Mode A in the step (2), but light is bad, video is fuzzy, anomalies health check-up method of determining and calculating can't normally start so, at this moment, will send assistance information, the automatic start-up mode B of system's meeting, this moment, the video council of Mode A got clogged, and system obtains the video information from Mode B, and system exception situation evaluation algorithm will start;
Described effect of jitter: if the shake of video camera is bigger, then will inevitably influence the quality of particular video frequency frame, make that the result of text detection is not accurate enough; Each frame of video is done the method for signature analysis, extract the feature on the frame of video, adopt the zonule to do the method for mixed Gaussian background modeling, extract the background information in the frame of video, compare with the multiframe background information; By comparing with the background information of the preceding frame picture of number, also the background information in the script frame finally obtains regional aim;
Describedly block influence: if video camera is blocked, will lose relevant important video information, the quality of influence monitoring greatly; Adopt multiframe mixed Gaussian background modeling (method, draw the target context in the video information earlier, again current each frame is done analysis, whether be blocked to judge present frame, if be blocked, then the background information in the present frame will most ofly disappear, and so just can be judged accurately, reaches timely warning.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA2008100392464A CN101609580A (en) | 2008-06-20 | 2008-06-20 | Method for intelligently protecting bank self-service equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA2008100392464A CN101609580A (en) | 2008-06-20 | 2008-06-20 | Method for intelligently protecting bank self-service equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101609580A true CN101609580A (en) | 2009-12-23 |
Family
ID=41483325
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNA2008100392464A Pending CN101609580A (en) | 2008-06-20 | 2008-06-20 | Method for intelligently protecting bank self-service equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101609580A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101807248A (en) * | 2010-03-30 | 2010-08-18 | 上海银晨智能识别科技有限公司 | System for detecting the existence of people in ATM machine video scene and method thereof |
| CN102859565A (en) * | 2010-04-26 | 2013-01-02 | 传感电子有限责任公司 | Method and system for security system tampering detection |
| CN107040759A (en) * | 2017-04-12 | 2017-08-11 | 合肥才来科技有限公司 | Intelligent monitor system applied to bank |
| WO2018127153A1 (en) * | 2017-01-06 | 2018-07-12 | 中兴通讯股份有限公司 | Alarm method, alarm device and terminal |
| CN110647858A (en) * | 2019-09-29 | 2020-01-03 | 上海依图网络科技有限公司 | Video occlusion judgment method and device and computer storage medium |
-
2008
- 2008-06-20 CN CNA2008100392464A patent/CN101609580A/en active Pending
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101807248A (en) * | 2010-03-30 | 2010-08-18 | 上海银晨智能识别科技有限公司 | System for detecting the existence of people in ATM machine video scene and method thereof |
| CN102859565A (en) * | 2010-04-26 | 2013-01-02 | 传感电子有限责任公司 | Method and system for security system tampering detection |
| CN102859565B (en) * | 2010-04-26 | 2015-06-03 | 传感电子有限责任公司 | Method and system for security system tampering detection |
| US9286778B2 (en) | 2010-04-26 | 2016-03-15 | Sensormatic Electronics, LLC | Method and system for security system tampering detection |
| WO2018127153A1 (en) * | 2017-01-06 | 2018-07-12 | 中兴通讯股份有限公司 | Alarm method, alarm device and terminal |
| CN107040759A (en) * | 2017-04-12 | 2017-08-11 | 合肥才来科技有限公司 | Intelligent monitor system applied to bank |
| CN110647858A (en) * | 2019-09-29 | 2020-01-03 | 上海依图网络科技有限公司 | Video occlusion judgment method and device and computer storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105675623B (en) | It is a kind of based on the sewage color of sewage mouth video and the real-time analysis method of flow detection | |
| CN102348128B (en) | Surveillance camera system having camera malfunction detection function | |
| CN109215020A (en) | Ultra-high-tension power transmission line fault recognition method based on computer vision | |
| CN104809887B (en) | A kind of retrograde detection method of vehicle on expressway and autoalarm | |
| US8301577B2 (en) | Intelligent monitoring system for establishing reliable background information in a complex image environment | |
| CN101908142A (en) | Feature analysis-based video flame detecting method | |
| CN109636777A (en) | A kind of fault detection method of traffic lights, system and storage medium | |
| CN101609580A (en) | Method for intelligently protecting bank self-service equipment | |
| CN103761516B (en) | ATM abnormal human face detection based on video monitoring | |
| CN109785615A (en) | A kind of early warning shunting subsystem for public security checking station | |
| CN103761529A (en) | Open fire detection method and system based on multicolor models and rectangular features | |
| CN103260049A (en) | Intelligent skynet video quality diagnostic system | |
| CN103324910A (en) | Fire alarming method based on video detection | |
| CN107993470A (en) | Count down traffic signal lamp condition detection method and the monitoring system based on the method | |
| CN104240239A (en) | Method for detecting local road segment hazy weather based on road image | |
| CN112784821A (en) | Building site behavior safety detection and identification method and system based on YOLOv5 | |
| CN106872488A (en) | A kind of double surface defect visible detection methods of rapid large-area transparent substrate and device | |
| CN106454330A (en) | Fuzziness anomaly detection method for video signals | |
| WO2006028571B1 (en) | Method for visual inspection of printed matter on moving lids | |
| CN101222618A (en) | Methods and systems for identifying events for a vehicle | |
| CN102542673A (en) | Automatic teller machine (ATM) pre-warning method and system based on computer vision | |
| CN110363109B (en) | Computer vision-based coverage type oil leakage detection system and method | |
| CN105282542B (en) | The detection method and system of abnormal striped in a kind of video image | |
| CN103959309A (en) | Region selection for counterfeit determinations | |
| CN117935452A (en) | Workshop security system and method based on visual multi-mode analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Open date: 20091223 |