It is a kind of to use public place safety pre-warning system and methodTechnical field
The application belongs to safety-security area, and in particular to a kind of public place safety pre-warning system and method.
Background technique
With the development of science and technology, public place safe early warning technological incorporation machine vision, image procossing and deep learningThe artificial intelligence technology in equal forward positions, the video that many companies are all based on that common camera is shot at present carry out early warning systemSystem exploitation.
But this early warning system may be only available in limited environment, because different scenes exist in actual environmentBackground complexity, the factors such as illumination deficiency, visual angle diversity affect the imaging effect of camera, reduce the spirit of early warning systemQuick property and accuracy.Therefore, the sensitivity and accuracy of above-mentioned security pre-warning system are to be improved.
Summary of the invention
The purpose of the application is to provide a kind of public place safety pre-warning system, can accurately and effectively go out in dangerous situationIt issues warning signal in time now, guarantees the personal safety as well as the property safety of people in public place.
To realize above-mentioned application purpose, the application the technical scheme adopted is that a kind of public place safety pre-warning system,It include: EO-1 hyperion camera apparatus, the first picture pick-up device group, data processor；
The EO-1 hyperion camera apparatus includes EO-1 hyperion camera and image processor；The EO-1 hyperion camera is worked as shootingCrowd in preceding public place obtain include location information and spectral information high spectrum image；Described image processor for pairHigh spectrum image is handled, and obtains comparing most apparent crowd's image as first crowd's image, and extract described the firstThe attitude parameter of each pedestrian, the attitude parameter are used to characterize the posture and profile of pedestrian in group's image；The EO-1 hyperion phaseLocation information is transferred to the first picture pick-up device group by machine equipment；
The first picture pick-up device group includes at least one picture pick-up device, which is used for according to the attitude parameterCrowd in current public place is imaged to obtain the first group's photographed images, is extracted in the first group's photographed imagesCharacterize the location parameter of pedestrian's walking position；
The data processor, for judging out the danger of current public place according to the attitude parameter and location parameterCoefficient issues the first early warning in the case where first danger coefficient is higher than the first safety coefficient as the first danger coefficientSignal.
Optionally, the data processor is also used to according to the first group's photographed images and danger under normal circumstancesThe machine learning and training of the first group's photographed images, obtain the first danger judgement model in dangerous situation；At the dataDevice is managed, specifically for judging out currently according to the attitude parameter and the location parameter and the first danger judgement modelThe danger coefficient of public place is as the first danger coefficient, the case where first danger coefficient is higher than the first safety coefficientUnder, issue the first pre-warning signal.
Optionally, the system also includes the second picture pick-up device groups；The described image of the EO-1 hyperion camera apparatus is handledDevice is also used to handle high spectrum image, obtains second crowd's image, and extract and have danger in second crowd's imageThe coordinate information of the suspicious people of product；The coordinate information is transferred to second picture pick-up device by the EO-1 hyperion camera apparatusGroup；The second picture pick-up device group includes at least one picture pick-up device, and the picture pick-up device is used for according to the coordinate information pairCrowd in current public place is imaged to obtain second crowd's photographed images, extracts table in the second crowd photographed imagesLevy the suspected locations parameter of suspicious figure's walking position；The data processor is also used to be joined according to the suspected locationsThe several and attitude parameter judges out the danger coefficient of the suspicious figure as the second danger coefficient, in the described second dangerous systemIn the case that number is greater than the second safety coefficient, the second pre-warning signal is issued.
Optionally, the data processor is also used to according to the second crowd photographed images and danger under normal circumstancesThe machine learning and training of the second crowd photographed images in dangerous situation, obtain the second danger judgement model；At the dataDevice is managed, specifically for judging out according to the attitude parameter and the suspected locations parameter and the second danger judgement modelThe danger coefficient of current public place is higher than the feelings of the second safety coefficient as the second danger coefficient, in second danger coefficientUnder condition, the second pre-warning signal is issued.
Optionally, the location parameter and the suspected locations parameter include: moving parameter and relative parameter；The movementParameter, for characterizing direction and the speed of pedestrian or suspicious people movement；The relative parameter is changed over time for characterizing, rowPeople or the distance between suspicious people and other pedestrians and direction.
Optionally, the attitude parameter, the parameter including the face orientation for characterizing pedestrian.
On the other hand, present invention also provides a kind of public place safe early warning method, the method is applied to above-mentionedIn public place safety pre-warning system described in one, which comprises
The crowd that EO-1 hyperion camera apparatus shoots in current public place obtains crowd's image of different-waveband, it is described notThe crowd's image for comparing most apparent wave band is shot as first crowd's image with personage is filtered out in crowd's image of wave band, is mentionedThe attitude parameter of each pedestrian in the first crowd image is taken, the attitude parameter is used to characterize the posture and profile of pedestrian,The attitude parameter is transferred to the first picture pick-up device group；
First picture pick-up device group is imaged to obtain first according to the attitude parameter to the crowd in current public placeCrowd's photographed images extract the location parameter that pedestrian's walking position is characterized in the first group's photographed images；
Data processor judges out the danger coefficient conduct of current public place according to the attitude parameter and location parameterFirst danger coefficient issues the first pre-warning signal in the case where first danger coefficient is higher than the first safety coefficient.
Optionally, the method also includes: the data processor is according to the first group shot picture under normal circumstancesThe machine learning and training of the first group's photographed images, obtain the first danger judgement model under image and dangerous situation；NumberThe danger coefficient of current public place is judged out as the first dangerous system according to the attitude parameter and location parameter according to processorNumber, comprising: the data processor is according to the attitude parameter and the location parameter and the first danger judgement modelThe danger coefficient of current public place is judged out as the first danger coefficient.
Optionally, the method also includes: the EO-1 hyperion camera apparatus sieves in crowd's image of the different-wavebandCrowd's image of the most apparent wave band of dangerous goods shooting comparison is selected as second crowd's image, and extracts second crowd's imageIn with dangerous goods suspicious people coordinate information, the coordinate information is transferred to the second picture pick-up device group；It is describedSecond picture pick-up device group images the crowd in current public place according to the coordinate information to obtain the second crowd camera shootingImage extracts the suspected locations parameter that suspicious figure's walking position is characterized in the second crowd photographed images；The numberThe danger coefficient of the suspicious figure is judged out as according to the suspected locations parameter and the attitude parameter according to processorTwo danger coefficients issue the second pre-warning signal in the case where second danger coefficient is greater than the second safety coefficient.
Optionally, the method also includes: the data processors images according to second crowd under normal circumstancesThe machine learning and training of the second crowd photographed images under image and dangerous situation, obtain the second danger judgement model；InstituteData processor is stated to be made according to the danger coefficient that the suspected locations parameter and the attitude parameter judge out the suspicious figureFor the second danger coefficient, comprising: the data processor is according to the attitude parameter and the suspected locations parameter and describedSecond danger judgement model judges out the danger coefficient of current public place as the second danger coefficient, in the described second dangerous systemIn the case that number is higher than the second safety coefficient, the second pre-warning signal is issued.
The application has the advantages that
The public place of the best personage shooting wave band shot by EO-1 hyperion camera is chosen in application scheme firstCrowd's image, such available clearly character contour and posture information, i.e. attitude parameter, picture pick-up device is further according to the postureParameter shoots the location parameter that can further obtain pedestrian to crowd, so can be to avoid the shadow of bad imaging circumstancesIt rings, obtains accurate pedestrian's posture and location information；Final data processor is in conjunction with artificial intelligence or machine learning algorithm pairRelative parameter between attitude parameter, moving parameter and pedestrian carries out danger coefficient judge.In this way, public place can be monitored in real timeSecurity situation, accurately and effectively issued warning signal in time when dangerous situation occurs, guarantee public place in people peopleBody property safety.
Additional aspect and advantage of the invention will be set forth in part in the description, and will partially become from the following descriptionObviously, or practice through the invention is recognized.
Detailed description of the invention
Fig. 1 is a kind of structural schematic diagram of public place safety pre-warning system disclosed in the present application；
Fig. 2 is a kind of shooting schematic diagram of public place safety pre-warning system disclosed in the present application；
Fig. 3 is that the first danger judgement model disclosed in the present application/second danger judgement model establishes schematic diagram；
Fig. 4 is the structural schematic diagram of another public place safety pre-warning system disclosed in the present application；
Fig. 5 is a kind of schematic diagram of public place safe early warning method disclosed in the present application；
Fig. 6 is a kind of schematic diagram of public place safe early warning implementing procedure disclosed in the present application.
Specific embodiment
In order to which the technical problems, technical solutions and beneficial effects solved by the present invention is more clearly understood, below in conjunction withAccompanying drawings and embodiments, the present invention will be described in further detail.It should be appreciated that specific embodiment described herein is only usedTo explain the present invention, it is not intended to limit the present invention.
As shown in FIG. 1, FIG. 1 is a kind of structural schematic diagrams of public place safety pre-warning system disclosed in the present application.The public affairsPlace safety early warning system 100 altogether, comprising: EO-1 hyperion camera apparatus 110, the first picture pick-up device group 120, data processor 103；
EO-1 hyperion camera apparatus 110 includes EO-1 hyperion camera 111 and image processor 112；The EO-1 hyperion camera 111 is usedCrowd's image of different-waveband is obtained in the crowd shot in current public place；Described image processor 112 is used for describedPersonage filtered out in crowd's image of different-waveband shoot compare crowd's image of most apparent wave band as first crowd's image,And the attitude parameter of each pedestrian in the first crowd image is extracted, the attitude parameter is used to characterize the posture and wheel of pedestrianIt is wide；The attitude parameter is transferred to the first picture pick-up device group 120 by the EO-1 hyperion camera apparatus 110；
First picture pick-up device group 120 includes at least one picture pick-up device, these picture pick-up devices are used to be joined according to above-mentioned postureSeveral crowds in current public place are imaged to obtain the first group's photographed images, extract table in the first group's photographed imagesLevy the location parameter of pedestrian's walking position；
Data processor 130, for judging out the danger coefficient of current public place according to attitude parameter and location parameterThe first pre-warning signal is issued in the case where the first danger coefficient is higher than the first safety coefficient as the first danger coefficient.
In the prior art, also have and shot by crowd of the picture pick-up device to public place, and to the picture number of shootingAccording to the scheme issued warning signal after processing, but different scenes there are background complexity, illumination deficiency, visual angle diversity etc. becauseElement, these undesirable environmental factors will affect the imaging of common picture pick-up device.
Single scape is carried out using EO-1 hyperion camera to scan, and can be collected including spatially and spectrally information with single exposureComplete high spectrum image.Certain specific wavelength is more accurate to the imaging of human body, even if still can in complex environment backgroundPedestrian is partitioned into from background environment.In this application, the crowd that EO-1 hyperion camera is shot in current public place obtains differenceCrowd's image of wave band；It is most apparent that image processor filters out personage's shooting comparison in crowd's image of these different-wavebandsCrowd's image of wave band is as first crowd's image, and the attitude parameter for extracting each pedestrian in first crowd's image (is used for tableLevy the posture and profile of pedestrian)；These attitude parameters are transferred to picture pick-up device by EO-1 hyperion camera apparatus, can help to image to setThe standby profile for clearly analyzing pedestrian in image frame, plays help to subsequent acquisition pedestrian position parameter.
In this application, location parameter includes: moving parameter and relative parameter；Moving parameter, for characterizing pedestrian or canDoubt the mobile direction of people and speed；Relative parameter is changed over time for characterizing, between pedestrian or suspicious people and other pedestriansDistance and direction.Under normal circumstances, relative parameter should be more stable, and in the case of fighting, relative parameter willIt is very chaotic.
For example, as shown in Fig. 2, Fig. 2 is that a kind of shooting of public place safety pre-warning system disclosed in the present application is shownIt is intended to.EO-1 hyperion camera 1 shoots the crowd in current public place in figure, obtains crowd's image of different-waveband, filters out peopleThe image that object shooting compares most apparent wave band is split from complex background as first crowd's image, and by pedestrian, is mentionedThe attitude parameter of pedestrian is taken, attitude parameter here can be the parameter of face's direction for characterizing pedestrian.It is multiple in figureGeneral camera is imaged using the multi-angle of view of multiple cameras, according to the scene three-dimensionalreconstruction algorithm of multi-angle of view, available pedestrianMoving parameter and pedestrian between relative parameter.
Data processor can go out current public field according to above-mentioned attitude parameter, moving parameter and relative parameter analysis and judgmentDanger coefficient, danger coefficient be greater than secure threshold in the case where issue warning signal, for example, by prior-warning device noticeNeighbouring Security Personnel rushes for live reinforcement or prior-warning device directly sounds the alarm.Under normal circumstances, the mobile ginseng of pedestrianNumber, relative parameter, attitude parameter (face orientation) etc. are all that comparison is similar, and abnormal behaviour can then show special parameterFeature, such as: pedestrian, which quickly runs, relative parameter is chaotic, pedestrian's moving parameter is slack etc. belongs to abnormal behaviour, needsIt gives warning in advance and pays close attention to.
It is appreciated that crowd's image of the public place shot by EO-1 hyperion camera is chosen in application scheme firstIn, personage shoots the crowd's image for comparing most apparent wave band, such available clearly character contour and posture information, i.e.,Attitude parameter, picture pick-up device shoot the location parameter that can further obtain pedestrian further according to the attitude parameter to crowd,So accurate pedestrian's posture and location information can be obtained to avoid the influence of bad imaging circumstances；Finally again by data processingDevice carries out danger coefficient judge to above-mentioned attitude parameter and location parameter.In this way, passing through the knot of EO-1 hyperion camera and picture pick-up deviceIt closes, accuracy rate more higher than traditional early warning system may be implemented., and the security situation of public place can be monitored in real time, accuratelyIt is effectively issued warning signal in time when dangerous situation occurs, guarantees the personal safety as well as the property safety of people in public place.
As an alternative embodiment, data processor 130, is also used to according to the first group shot under normal circumstancesAs the machine learning and training of group's photographed images the first under image and dangerous situation, the first danger judgement model is obtained；DataProcessor 130, specifically for judging out current public field according to attitude parameter and location parameter and the first danger judgement modelDanger coefficient as the first danger coefficient, in the case where the first danger coefficient is higher than the first safety coefficient, issue firstPre-warning signal.Here safety coefficient can be set as the case may be, be can be and made by machine learning and trainingSetting, this programme is without limitation.
Artificial nerve network model can be selected in the method for machine learning.
As shown in figure 3, Fig. 3 is that a kind of first danger judgement model disclosed in the present application establishes schematic diagram.It is defeated in figureEntering layer, hidden layer and output layer is the common algorithm model of machine learning.Data processor 130 in figure is under normal circumstancesThe moving parameter, relative parameter, attitude parameter of pedestrian carries out machine learning and training in one crowd's photographed images, while at dataIt manages device 130 and machine also is carried out to the moving parameter, relative parameter, attitude parameter of pedestrian in group's photographed images the first under dangerous situationDevice study and training, this large amount of study just will form above-mentioned first danger judgement model with after training.By current pedestrianMoving parameter, relative parameter, attitude parameter inputs this first danger judgement model can obtain current public placeDanger coefficient.
As shown in figure 4, another embodiment of the application also discloses a kind of public place safety pre-warning system 200, this is200 difference compared with above-mentioned public place safety pre-warning system 100 of system is, the image processor of EO-1 hyperion camera apparatus 210212, it is also used to filter out crowd's image that dangerous goods shooting compares most apparent wave band in crowd's image of different-wavebandAs second crowd's image, and extract the coordinate information of the suspicious people in second crowd's image with dangerous goods；EO-1 hyperion phaseThe coordinate information is transferred to the second picture pick-up device equipment group by machine equipment；Second picture pick-up device group 240 includes at least oneA picture pick-up device, these picture pick-up devices to the crowd in current public place according to coordinate information for being imaged to obtain secondCrowd's photographed images extract the suspected locations parameter that suspicious figure's walking position is characterized in second crowd's photographed images；At dataDevice 230 is managed, is also used to judge out the danger coefficient of suspicious figure according to suspected locations parameter and attitude parameter as the second dangerCoefficient issues the second pre-warning signal in the case where the second danger coefficient is greater than the second safety coefficient.
Single scape scanning is carried out using EO-1 hyperion camera 2, can be collected with single exposure including spatially and spectrally informationComplete high spectrum image.A certain specific wavelength to the imagings of a certain dangerous goods (dangerous instrument, liquid fuel etc.) morePrecisely, even if in complex environment background, the dangerous goods can be still partitioned into from background environment.Therefore danger is chosenProduct shooting compares crowd's image of most apparent wave band as second crowd's image, and extracts in second crowd's image with dangerousThe coordinate information of the suspicious people of article；The coordinate information is transferred to picture pick-up device by EO-1 hyperion camera apparatus, that is, passes through dangerProduct carry the suspicious people of the dangerous goods to lock, and coordinate information is transferred to picture pick-up device, it is further to facilitate picture pick-up deviceTrack the moving parameter and relative parameter of the suspicious people.
Data processor can go out work as according to the attitude parameter, moving parameter and relative parameter analysis and judgment of above-mentioned suspicious peopleThe danger coefficient of preceding public place, and be connected with prior-warning device, it is issued in the case where danger coefficient is greater than secure threshold pre-Alert signal, for example live reinforcement or prior-warning device are rushed for by the Security Personnel near prior-warning device notice and directly sounded an alarmSound.
As an alternative embodiment, data processor, is also used to image according to the second crowd under normal circumstancesThe machine learning and training of second crowd's photographed images under image and dangerous situation, obtain the second danger judgement model；At dataDevice is managed, specifically for judging out current public place according to attitude parameter and suspected locations parameter and the second danger judgement modelDanger coefficient as the second danger coefficient, in the case where the second danger coefficient is higher than the second safety coefficient, it is pre- to issue secondAlert signal, safety coefficient here can be set as the case may be, be can be and set by what machine learning training was madeFixed, this programme is without limitation.
Similar with shown in Fig. 3, the data processor 230 in the present embodiment is in second crowd's photographed images under normal circumstancesMoving parameter, relative parameter, the attitude parameter of suspicious people carries out machine learning and training, while data processor 230 is also to dangerThe moving parameter, relative parameter, attitude parameter of suspicious people carry out machine learning and instruction in second crowd's photographed images in dangerous situationPractice, this large amount of study just will form above-mentioned second danger judgement model with after training.By the mobile ginseng of current pedestrianNumber, relative parameter, attitude parameter, which input this second danger judgement model, can obtain the danger coefficient of current public place.
As shown in figure 5, this method is applied to any of the above-described present invention also provides a kind of public place safe early warning methodIn the public place safety pre-warning system of item, this method comprises:
501, the crowd that EO-1 hyperion camera apparatus shoots in current public place obtains crowd's image of different-waveband, upperIt states to filter out personage in crowd's image of different-waveband and shoot and compares crowd's image of most apparent wave band and scheme as the first crowdPicture extracts the attitude parameter of each pedestrian in the first crowd image, which is used to characterize the posture and profile of pedestrian,The attitude parameter is transferred to the first picture pick-up device group；
502, the first picture pick-up device group is imaged to obtain according to above-mentioned attitude parameter to the crowd in current public placeThe first group's photographed images extract the location parameter that pedestrian's walking position is characterized in the first group's photographed images；
503, data processor judges out the danger coefficient of current public place according to above-mentioned attitude parameter and location parameterThe first pre-warning signal is issued in the case where first danger coefficient is higher than the first safety coefficient as the first danger coefficient.
As an alternative embodiment, the above method further include: data processor is according to under normal circumstances firstThe machine learning and training of the first group's photographed images under crowd's photographed images and dangerous situation, obtain the first danger judgement mouldType；Data processor judges out the danger coefficient of current public place as the first dangerous system according to attitude parameter and location parameterNumber, comprising: data processor judges out current public field according to attitude parameter and location parameter and the first danger judgement modelDanger coefficient as the first danger coefficient.
As an alternative embodiment, the above method further include: EO-1 hyperion camera apparatus is in above-mentioned different-wavebandCrowd's image of the most apparent wave band of dangerous goods shooting comparison is filtered out in crowd's image as second crowd's image, and is extractedThe coordinate information of suspicious people in second crowd's image with dangerous goods, is transferred to the second picture pick-up device for the coordinate informationGroup；Second picture pick-up device group is imaged to obtain the second crowd and be taken the photograph according to above-mentioned coordinate information to the crowd in current public placeAs image, the suspected locations parameter that suspicious figure's walking position is characterized in second crowd's photographed images is extracted；Data processor rootJudge out the danger coefficient of above-mentioned suspicious figure as the second danger coefficient according to suspected locations parameter and attitude parameter, this secondIn the case that danger coefficient is greater than the second safety coefficient, the second pre-warning signal is issued.
As an alternative embodiment, the above method further include: data processor is according to under normal circumstances secondThe machine learning and training of second crowd's photographed images under crowd's photographed images and dangerous situation, obtain the second danger judgement mouldType；Data processor judges out the danger coefficient of suspicious figure as the second dangerous system according to suspected locations parameter and attitude parameterNumber, comprising: data processor judges out current public affairs according to attitude parameter and suspected locations parameter and the second danger judgement modelThe danger coefficient in place is issued as the second danger coefficient in the case where the second danger coefficient is higher than the second safety coefficient altogetherSecond pre-warning signal.
A kind of implementing procedure of above-mentioned public place safe early warning method is as shown in fig. 6, EO-1 hyperion camera shoots public fieldThe image of institute crowd, EO-1 hyperion camera can once take multiple band images of public place crowd；Pass through image procossingDevice filters out personage and shoots the crowd's image for comparing most apparent wave band as first crowd's image, filters out dangerous goods shootingCrowd's image of most apparent wave band is compared as second crowd's image；Continue to extract each pedestrian in the first crowd imageAttitude parameter, the attitude parameter are used to characterize the posture and profile of pedestrian, which are transferred to the first picture pick-up device group,The coordinate information for extracting the suspicious people in second crowd's image with dangerous goods, is transferred to the second camera shooting for the coordinate information and setsStandby group；First picture pick-up device group and the second picture pick-up device group are respectively according to above-mentioned attitude parameter and above-mentioned coordinate information to current public affairsThe crowd in place is imaged to obtain the first group's photographed images and second crowd's photographed images altogether；It is the first that this is extracted respectivelyThe location parameter that pedestrian's walking position is characterized in group's photographed images extracts characterization suspicious figure walking in second crowd's photographed imagesThe suspected locations parameter of position；By attitude parameter and location parameter input the first danger judgement model to the first danger coefficient,By suspected locations parameter and attitude parameter input the second danger judgement model to the second danger coefficient；If the first danger coefficient is bigThe first pre-warning signal is then issued in the first safety coefficient, issues the second early warning if the second danger coefficient is greater than the second safety coefficientSignal.
The specific implementation method of above-mentioned public place safe early warning method and corresponding public place safety pre-warning system phaseTogether, which is not described herein again.
Although an embodiment of the present invention has been shown and described, it will be understood by those skilled in the art that: notA variety of change, modification, replacement and modification can be carried out to these embodiments in the case where being detached from the principle of the present invention and objective, thisThe range of invention is defined by the claims and their equivalents.