A kind of Intellectualized Video Monitoring method and systemTechnical field
The present invention relates to the technical field of network monitoring, especially a kind of Intellectualized Video Monitoring method and system.
Background technique
With information technology, network technology, the rapid development of image processing techniques: digitlization, the intelligent video of networkingAnalysis system then teacher's monitoring field forefront one of application development direction.Intelligent monitoring is by camera supervised, to screenIt is acquired through row, intellectual analysis, automatic early-warning；More automatically captured information, more intelligently analyzes information, more accurate ground through rowJudgement, more the offer service of active.
In railway as national basis facility and the popular vehicles, railway passenger station general staff is intensive, so respectivelyKind burst accident emerges one after another such as: falling down, passes through white safety line, elevator is hurted sb.'s feelings, and gathering of people is run.And currentVideo monitoring system can only play the role of post-mordem forensics.
The patent of invention of Publication No. CN106982334A disclose a kind of passenger flow monitor device for subway station andMethod, passenger flow monitor device include several video cameras, video analysis host and monitoring alarm host, each video camera differenceIt is set on a platform of subway station, several video cameras are electrically connected with video analysis host respectively, video analysis hostIt is electrically connected with monitoring alarm host；Each video camera acquires the image information of corresponding platform in real time respectively, and picture is believedBreath is sent to video analysis host, and the image information of each platform that video analysis host process receives generates intensity of passenger flowDistributed data, and intensity of passenger flow distributed data is sent to monitoring alarm host, the visitor that monitoring alarm host analysis receivesCurrent density distributed data generates passenger flow induction information, and exports passenger flow induction information.This method is merely by passenger flowDensity is monitored, and when there is crowded situation, is dredged the stream of people, and other burst things can not be coped withPart.
The patent of invention of Publication No. CN106154223A discloses a kind of intelligence based on environmental monitoring and video monitoringChange monitor operating system, including environmental parameter monitoring device, video capture device, main control unit, operational and controlled key button, emergency unit,The output end of user terminal and storage equipment, environmental parameter monitoring apparatus and video capturing device is connected to main control unit, operatesKey is connected to main control unit, and the output end of main control unit is connected to emergency unit, user terminal and storage equipment, master controlUnit can be such that user terminal selecting display environment parameter detection equipment and/or video catches based on the order that operational and controlled key button inputsThe data of equipment acquisition are obtained, monitor operating system of the present invention is provided existing while using video monitoring live viewField environmental parameter, and can be manipulated according to the variation of environmental parameter.The system excessively carried out only for the stream of people it is arduous andIt dredges, there is no the processing methods of more emergency events, and are all real-time perfoming monitoring, can not carry out afterwards to videoAnalysis.
Summary of the invention
In order to solve the above technical problems, the present invention proposes a kind of Intellectualized Video Monitoring method and system, to monitoringVideo uses secondary analysis method, and normal video, doubtful problem video and problem video is separated, and different reservations is arrangedTime facilitates the later period to search, and the various emergency events occurred simultaneously for station can using the type of alarm of different stageThe various problems that effective solution station is easy to appear.
There is provided a kind of Intellectualized Video Monitoring methods for the first object of the present invention, including IP Camera is used to carry outMonitoring, further comprising the steps of:
Step 1: the region at station is divided；
Step 2: setting the monitoring rules in the different regions；
Step 3: obtaining monitoring video and analyzed；
Step 4: after there are abnormal conditions, proposing alarm.
Preferably, waiting area, corridor area, entrance area, exit region and the exclusion area that the region division isAt least one of domain.
In any of the above-described scheme preferably, the monitoring rules include whether occurring abnormal feelings in monitoring areaCondition.
In any of the above-described scheme preferably, the abnormal conditions includes the abnormal aggregation, abnormal stagnant of crowdStay, theft personnel occur, online wanted circular personnel occur, more people simultaneously run, twisting is run, exception is hovered, people's case separate and rushEnter at least one of abstinence region.
In any of the above-described scheme preferably, the step 3 includes that the IP Camera passes through network transmission moduleVideo is sent to background data center.
In any of the above-described scheme preferably, the analysis includes initial screening and postsearch screening.
In any of the above-described scheme preferably, the initial screening, which refers to, carries out first time screening for video, and filtering out canVideo clip is doubted, the suspicious video clip is deposited into suspicious video storage modules, remaining video clip is stored in normal viewFrequency memory module.
In any of the above-described scheme preferably, the suspicious video clip refers at least one abnormal feelings of appearanceThe video clip of condition.
In any of the above-described scheme preferably, the postsearch screening refers to using biopsy method and/or face knowledgeOther method analyzes the suspicious video clip, and video is sent to problem video storage modules the problem of will obtain.
Be preferably in any of the above-described scheme ten, the biopsy method the following steps are included:
Step 01: identifying face from video ten, record the human face data detected；
Step 02: m important parameter being chosen as living body by the deep learning of big data and judges parameter；
Step 03: calculating living body change degree D, and be compared with the threshold value of setting.
In any of the above-described scheme preferably, the calculation formula of the living body change degree D are as follows:
Wherein, n-th of parameter variation value is Δ Xn, parameter coefficient is αn。
In any of the above-described scheme preferably, the threshold value includes model threshold D1, photo threshold value D2With living body threshold valueD3。
In any of the above-described scheme preferably, the step 03 includes when the living body change degree D is greater than living body threshold valueD3When, then the face is living body.
In any of the above-described scheme preferably, the face identification method the following steps are included:
Step 11: by the face pixelation in video, obtaining face pixel；
Step 12: by carrying out intellectual analysis to the face pixel, finding the face pixel, each is not advisedThe point then changed, is set as face key point；
Step 13: each face key point is compared using a large amount of human face datas, extrapolate T key point andIts location information, wherein T is natural number；
Step 14: a certain key point of setting is coordinate origin, according to the relative coordinate of the key point, by machineStudy is compared with big data, establishes the transverse and longitudinal coordinate coefficient of each key point, and calculate recognition of face according to following equationThe threshold value d of middle face comparison, formula are as follows:
Wherein, x represents key point difference value, and a represents the coefficient of each key point, which key point k represents, and n represents totalHow many a key points altogether
In any of the above-described scheme preferably, the step 3 further includes extracting from described problem video storage modulesDescribed problem video, and described problem video is classified.
It is preferably in any of the above-described scheme ten, the step 4 includes being alarmed according to the grade setting of described problem videoRank.
In any of the above-described scheme preferably, the alarm is mentioned including rear end Client-Prompt, by station broadcastAt least one of show, notify station patrol police and alarm to public security organ.
There is provided a kind of intelligent video monitoring systems for the second object of the present invention, including the network for being monitoredCamera and storage device further include with lower module:
Region division module: it is divided for the region to station；
Rule settings module: for setting the monitoring rules in the different regions；
Analysis module: for obtaining monitoring video and being analyzed；
Alarm module: after there are abnormal conditions, alarm is proposed；
The storage device includes in suspicious video storage modules, normal video memory module and problem video storage modulesIt is at least one.
Preferably, waiting area, corridor area, entrance area, exit region and the exclusion area that the region division isAt least one of domain.
In any of the above-described scheme preferably, the monitoring rules include whether occurring abnormal feelings in monitoring areaCondition.
In any of the above-described scheme preferably, the abnormal conditions includes the abnormal aggregation, abnormal stagnant of crowdStay, theft personnel occur, online wanted circular personnel occur, more people simultaneously run, twisting is run, exception is hovered, people's case separate and rushEnter at least one of abstinence region.
In any of the above-described scheme preferably, the analysis module, which has, passes through net using the IP CameraVideo is sent to the function of background data center by network transmission module.
In any of the above-described scheme preferably, the analysis includes initial screening and postsearch screening.
In any of the above-described scheme preferably, the initial screening, which refers to, carries out first time screening for video, and filtering out canVideo clip is doubted, the suspicious video clip is deposited into the suspicious video storage modules, remaining video clip is stored in instituteState normal video memory module.
In any of the above-described scheme preferably, the suspicious video clip refers at least one abnormal feelings of appearanceThe video clip of condition.
It is preferably in any of the above-described scheme ten, the postsearch screening refers to using biopsy method and/or face knowledgeOther method analyzes the suspicious video clip, and video is sent to described problem video storage mould the problem of will obtainBlock.
In any of the above-described scheme preferably, the biopsy method the following steps are included:
Step 01: identifying face from video, record the human face data detected；
Step 02: m important parameter being chosen as living body by the deep learning of big data and judges parameter；
Step 03: calculating living body change degree D, and be compared with the threshold value of setting.
In any of the above-described scheme preferably: the calculation formula of the living body change degree D are as follows:
Wherein, n-th of parameter variation value is Δ Xn, parameter coefficient is αn。
In any of the above-described scheme preferably, the threshold value includes model threshold D1, photo threshold value D2With living body threshold valueD3。
In any of the above-described scheme preferably, the step 03 includes when the living body change degree D is greater than living body threshold valueD3When, then the face is living body.
In any of the above-described scheme preferably, the face identification method the following steps are included:
Step 11: by the face pixelation in video, obtaining face pixel；
Step 12: by carrying out intellectual analysis to the face pixel, finding the face pixel, each is not advisedThe point then changed, is set as face key point；
Step 13: each face key point is compared using a large amount of human face datas, extrapolate T key point andIts location information, wherein T is natural number；
Step 14: a certain key point of setting is coordinate origin, according to the relative coordinate of the key point, by machineStudy is compared with big data, establishes the transverse and longitudinal coordinate coefficient of each key point, and calculate recognition of face according to following equationThe threshold value d of middle face comparison, formula are as follows:
Wherein, x represents key point difference value, and a represents the coefficient of each key point, which key point k represents, and n represents totalHow many a key points altogether
It is preferably in any of the above-described scheme ten, the analysis module has from described problem video storage modulesDescribed problem video is extracted, and to the function that described problem video is classified.
In any of the above-described scheme preferably, the alarm module has the grade setting report according to described problem videoThe function of alert rank.
In any of the above-described scheme preferably, the alarm is mentioned including rear end Client-Prompt, by station broadcastAt least one of show, notify station patrol police and alarm to public security organ.
The invention proposes a kind of Intellectualized Video Monitoring method and systems, are known using biopsy method and/or faceOther method analyzes the suspicious video clip, can rapidly judge whether station occurs emergency event or exceptionEvent, and make timely and effectively alert process.
Detailed description of the invention
Fig. 1 is the flow chart of a preferred embodiment of Intellectualized Video Monitoring method according to the invention.
Fig. 2 is the module map of a preferred embodiment of intelligent video monitoring system according to the invention.
Fig. 3 is the flow chart of an embodiment of the biopsy method of Intellectualized Video Monitoring method according to the invention.
Fig. 4 is the flow chart of an embodiment of the face identification method of Intellectualized Video Monitoring method according to the invention.
Fig. 5 is the flow chart of another preferred embodiment of Intellectualized Video Monitoring method according to the invention.
Specific embodiment
The present invention is further elaborated with specific embodiment with reference to the accompanying drawing.
Embodiment one
As shown in Fig. 2, intelligent video monitoring system includes: that IP Camera 200, region division module 210, rule are setCover half block 220, analysis module 230, background data center 240, storage device 250 and alarm module 260, tenth, storageDevice 250 includes normal video memory module 251, suspicious video storage modules 252 and problem video storage modules 253.
IP Camera 200: for carrying out shooting monitoring to monitoring area, and video is sent to analysis module230。
Region division module 210: dividing for the region to station, and region division result is sent to network shootingFirst 200 and background data center 240.
Rule settings module 220: for obtaining region division as a result, and according to region division from background data center 240As a result the monitoring rules in the different regions are set.
Analysis module 230: for obtaining monitoring video and being analyzed；By the no problem video hair after primary screeningNormal video memory module 251 is given, problematic video is sent to suspicious video storage modules 252；After postsearch screeningNo problem video be sent to normal video memory module 251, problematic video is sent to problem video storage modules253, problematic video is classified, and analysis result and classification results are sent to background data center 240.
Background data center 240: for receiving the information of analysis module 230, from problem video storage modules 253Extraction problem video, and warning message is sent to alarm module 260.
Normal video memory module 251: for storing normal video, retention cycle is 90 days.
Suspicious video storage modules 252: for storing suspicious video, retention cycle is 180 big.
Problem video storage modules 253.For storage problem video, retention cycle is 500 days.
Alarm module 260: after there are abnormal conditions, different grades of report is issued according to the classification situation of problem videoAlert, type of alarm includes rear end Client-Prompt, passes through station broadcast, notice station patrol police and alarm to public security organDeng.
As shown in Figure 1, executing step 100, monitoring area is monitored using IP Camera 200.Execute step110, region division module 210 divides the region at station, waiting area that region division is, corridor area, inlet regionDomain, exit region and abstinence region.Step 120 is executed, rule settings module 220 sets the monitoring rule in the different regionsThen, monitoring rules include whether occurring abnormal conditions in monitoring area, abnormal conditions include crowd abnormal aggregation,Abnormal delay, theft personnel occur, online wanted circular personnel occur, more people are run simultaneously, twisting is run, exception is hovered, people's caseSeparate and swarm into abstinence region etc..Step 130 is executed, analysis module 230 obtains IP Camera 200 and passes through network transmissionMonitoring video that module sends over simultaneously is analyzed.The analysis work of analysis module 230 includes initial screening and secondary sieveChoosing.Initial screening, which refers to, carries out first time screening for video, filters out suspicious video clip, the suspicious video clip (is referred toThere is the video clip of at least one abnormal conditions) suspicious video storage modules 252 are deposited into, by remaining video clipIt is stored in normal video memory module 251.Postsearch screening refer to using biopsy method and/or face identification method to it is described canDoubtful video clip is analyzed, and video is sent to problem video storage modules 253 the problem of will obtain.
11. biopsy method is the following steps are included: step 300: identifying face from video, record the face detectedData；Step 310: m important parameter being chosen as living body by the deep learning of big data and judges parameter；Step 320: calculatingLiving body change degree D, and be compared with the threshold value of setting.The calculation formula of body change degree D are as follows:
Wherein, n-th of parameter variation value is Δ Xn, parameter coefficient is αn。
Threshold value includes model threshold D1, photo threshold value D2With living body threshold value D3,
Model threshold D1< photo threshold value D2< living body threshold value D3,
1) when the living body change degree D is greater than model threshold D1Less than photo threshold value D2When, determine the face for model；
2) when the living body change degree D is greater than photo threshold value D2Less than living body threshold value D3When, determine the face for photo；
3) when the living body change degree D is greater than living body threshold value D3When, determine the face for living body.
Face identification method is the following steps are included: step 400: by the face pixelation in video, obtaining face pixel；Step 410: by carrying out intellectual analysis to the face pixel, finding the face pixel, each is irregularPoint is set as face key point；Step 420: each face key point being compared using a large amount of human face datas, is extrapolatedT key point and its location information (wherein T is natural number, in the present embodiment, T=269)；Step 430: setting is a certain describedKey point is coordinate origin, according to the relative coordinate of the key point, is compared by machine learning and big data, and each institute is establishedThe transverse and longitudinal coordinate coefficient of key point is stated, and calculates the threshold value d of face comparison in recognition of face according to following equation, formula is as follows:
Wherein, x represents key point difference value, and a represents the coefficient of each key point, which key point k represents, and n represents totalHow many a key points altogether.
Step 140 is executed, after there are abnormal conditions, alarm module 260 is alarmed according to the grade setting of described problem videoRank, and propose to alarm, alarm include rear end Client-Prompt, by station broadcast, notice station patrol police and to public affairsPacify organ's alarm etc..
Embodiment two
As shown in figure 5, the present invention provides a kind of station intelligent monitor systems, including network hard disk video recorder, netNetwork camera, client access terminal, and video information manages platform, and video intelligent, broadcast interface machine, passenger facilities are broadcasted flatPlatform.
The present embodiment provides a kind of data methods, which comprises area is divided AT STATION.Monitoring area installationUpper network camera be monitored to region.IP Camera acquisition module: using high definition IPC, to realize that high definition regardsScreen acquisition using independent IP address network segment, forms setting for front end and more monitoring by video information by network transmission moduleStandby link.Video resource is passed in NVR by IP transmission network.Video store function stores mould using NVR in this programmeFormula carries out distributed storage to real-time video.Meanwhile video intelligent system analysis module, according to the practical application at station, notOccur violating predefined therefore rule behavior in the scene of same monitor camera, intelligent analysis module can be to sending prompt letterBreath.Information management module receives the prompt information from intelligent analysis module, will be prompted to information record.It is mentioned in client current bound faceShow information.Abnormality alarming module simultaneously, the alarm mode that different exception rules are answered according to set by station connect station advertisementSystem carries out station commercial announcements.
A point picture is carried out to regional scope based on above-mentioned, described network monitoring camera head, is waited according to required be divided intoRegion, corridor area, monitoring area.Monitoring camera carries out the acquisition of high definition screen using high definition IPC, while can satisfy front endDifferent type, the IPC of different function can be used in the different demands of plurality of application scenes.
Based on above-mentioned, the intelligent analysis module in real time analyzes monitoring region, which, which can be, does not adviseArbitrary shape then.Intelligent video analysis is judged whether for number within the scope of statistic mixed-state based on the rule that station is fixedGathering of people occurs, personnel are run, and personnel cross the border, and twisting is run, and personnel cross the border, and Face datection, personnel hover, the separation of people's caseDeng.
The behavior wherein to break the rules defines numerical value and can be configured by station employees:
Testing staff is detained and assembles for a long time in video, while can exclude to be lined up ticket checking etc. in waiting hall normallyBehavior can set the threshold value of gathering of people quantity, reach number of thresholds once discovery gathering of people answers automatic prompt or alarmAggregation can just be can be regarded as；Can be set gathering of people lag time, time end then not counting assemble；System can be with automatic fitration tripVisitor is lined up ticket checking, goes to war and waits normal behaviours, but occurs being detained for a long time and be also considered aggregation.Classification report equally also can be setAlert: level-one, second level, three-level corresponds to different number personnel, and there is a situation where assemble；
In the area people behavior of detection, emphasis detects more people runs simultaneously, once the personnel of discovery run prompt orAlarm, set people's movement speed, or run number one or more, more than set valve system automatic prompt and reportIt is alert；
The present invention can also be in key area abstinence region, for the intruder of abstinence region or forbidden line automatically through rowPrompt, alarm.Setting is put on railway platform, it can be determined that whether train reaches.Wherein abstinence region can be set polygon orIrregular area.
The embodiment of the present invention is also connected with the broadcast system at station, and the broadcasting speech of station employees' color setting works as system promptSomeone makes infringement, on the one hand system one side rear end Client-Prompt is broadcasted corresponding by station broadcastPrompt.
Embodiment three
In the present embodiment ten, will monitoring region hair be divided into waiting area, corridor area, entrance area, exit region andAbstinence region etc., for different regions, the rule of monitoring is different.
For waiting area, it is monitored for following situation, there are following situations in any, then it alarms:
1) the abnormal aggregation of crowd:
2) the abnormal delay of crowd；
3) more than two people abnormal limbs behavior or contacts；
4) more than two people armed to stand facing each other or have a fist fight；
5) special personnel (thief, illegally attracts customers, wanted criminal at ox) occurs；
6) more people run (in the same direction or incorgruous) simultaneously；
7) enter and be prohibited from entering region.
For corridor area, it is monitored for following situation, there are following situations in any, then it alarms:
1) the abnormal aggregation of crowd；
2) the abnormal delay of crowd；
3) more than two people abnormal limbs behavior or contacts；
4) more than two people armed to stand facing each other or have a fist fight；
5) special personnel (thief, illegally attracts customers, wanted criminal at ox) occurs；
6) more people run (in the same direction or incorgruous) simultaneously:
It for entrance area and exit region, is monitored for following situation, any following situations occurs, then carries outAlarm:
1) the abnormal aggregation of crowd；
2) the abnormal delay of crowd；
3) crowded, be lined up out of order；
4) the non-safety check of luggage parcel；
5) more than two people abnormal limbs behavior or contacts；
6) more than two people armed to stand facing each other or have a fist fight；
7) special personnel (thief, illegally attracts customers, wanted criminal at ox) occurs；
8) more people run (in the same direction or incorgruous) simultaneously:
9) incorgruous walking.
For abstinence region, it is monitored for following situation, there are following situations in any, then it alarms:
1) the abnormal aggregation of crowd；
2) the abnormal delay of crowd；
3) special personnel (thief, illegally attracts customers, wanted criminal at ox) occurs；
4) it swarms into without permission.
Example IV
In the present embodiment, long-distance video storage center is made of three parts equipment, normal video memory, suspicious videoMemory and problem video memory.The security level of normal video memory is minimum, and the video content holding time is shorter usuallyIt is -90 days 30 days, the content of preservation includes that analysis module carries out the normal video obtained after primary screening to all videosThe normal video obtained after postsearch screening is carried out to doubtful problem video with analysis module.The safety of suspicious video memoryRank is placed in the middle, and the video content holding time is usually -180 days 90 days, and the content of preservation is analysis module to all videosCarry out the suspicious video obtained after primary screening.The security level highest of problem video memory, video content holding time are logicalIt is often -730 days 180 days, the content of preservation is that analysis module is asked what is obtained after the progress postsearch screening of doubtful problem videoThe problem of topic video and analysis module obtain after the normal video in normal video memory is rechecked and/or inspected by random samplesVideo.
Such as: many people are taken in video and are gathered about the entrance, and analysis module once sieves this section of videoChoosing, it is believed that there may be problems for gathering of people, and just this section of video is stored in suspicious video memory.Analysis module pairThis section of video carries out postsearch screening, and judgement is that personnel are entered the station queuings, and not dangerous presence then stores the video into normallyIn video memory.
Such as: there are many people to get together smoking in video photographs to square, analysis module to this section of video intoRow primary screening, it is believed that there may be problems for gathering of people, and just this section of video is stored in suspicious video memory.Video pointIt analyses module and postsearch screening is carried out to this section of video, it is ox that being identified wherein by face identification method, which has several individuals, is judged asTransaction that ox assembles or ox is attracted customers, there are problems, then store the video in problem video memory, and pass through alarmIt goes to handle before notice station patrol police.
For a better understanding of the present invention, the above combination specific embodiments of the present invention are described in detail, but are notLimitation of the present invention.Any simple modification made to the above embodiment according to the technical essence of the invention, still belongs toIn the range of technical solution of the present invention.In this specification the highlights of each of the examples are it is different from other embodiments itLocate, the same or similar part cross-reference between each embodiment.For system embodiments, due to itself and methodEmbodiment corresponds to substantially, so being described relatively simple, the relevent part can refer to the partial explaination of embodiments of method.