CN105844659B - The tracking and device of moving component - Google Patents

The tracking and device of moving component Download PDF

Info

Publication number
CN105844659B
CN105844659B CN201510018291.1A CN201510018291A CN105844659B CN 105844659 B CN105844659 B CN 105844659B CN 201510018291 A CN201510018291 A CN 201510018291A CN 105844659 B CN105844659 B CN 105844659B
Authority
CN
China
Prior art keywords
moving component
feature
classification
component
case point
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.)
Active
Application number
CN201510018291.1A
Other languages
Chinese (zh)
Other versions
CN105844659A (en
Inventor
毛文涛
王强
王晓涛
郭萍
王山东
邵光琪
李圭彬
朴根柱
柳賢錫
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Samsung Telecommunications Technology Research Co Ltd
Samsung Electronics Co Ltd
Original Assignee
Beijing Samsung Telecommunications Technology Research Co Ltd
Samsung Electronics Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing Samsung Telecommunications Technology Research Co Ltd, Samsung Electronics Co Ltd filed Critical Beijing Samsung Telecommunications Technology Research Co Ltd
Priority to CN201510018291.1A priority Critical patent/CN105844659B/en
Priority to KR1020150173974A priority patent/KR102595604B1/en
Priority to US14/995,262 priority patent/US10043064B2/en
Publication of CN105844659A publication Critical patent/CN105844659A/en
Application granted granted Critical
Publication of CN105844659B publication Critical patent/CN105844659B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Image Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Human Computer Interaction (AREA)
  • Psychiatry (AREA)
  • Social Psychology (AREA)
  • Databases & Information Systems (AREA)

Abstract

The present invention provides a kind of tracking of moving component and devices, wherein the described method includes: acquiring signal using dynamic visual sensor and exporting the case point of detection;Classification and the position of moving component are identified for the case point that current detection goes out using classifier;Motion profile is determined according to the position of the moving component of the classification successively identified, the tracking result of the moving component as the classification;Wherein, the classifier be according to the dynamic visual sensor acquisition sample signal train in advance come.With the application of the invention, can make tracking process, low energy consumption and can quickly and accurately capture moving target.

Description

The tracking and device of moving component
Technical field
The present invention relates to technical field of intelligent equipment, specifically, the present invention relates to a kind of trackings of moving component And device.
Background technique
Motion target tracking is a research hotspot of computer vision field, is widely used in military surveillance, monitoring system The fields such as system, human-computer interaction.For example, can be captured to the moving target in environment, track, analyze;According to what is analyzed The motion mode of moving target carries out the switching of the operational mode of terminal device, so as to improve man-machine interaction experience.Wherein, terminal Equipment can be mobile terminal, video camera, intelligent glasses, smart television etc..
Currently, existing motion target tracking method, mainly using traditional based on CCD (Charge-coupled Device, charge coupled cell) or CMOS (Complementary Metal Oxide Semiconductor, complementary metal oxygen Compound semiconductor) picture pick-up device, to where moving target scene carry out Image Acquisition;Later, divide from the image of acquisition After cutting out the region and background area moved, moving target is gone out according to the region recognition for the movement being partitioned into, and transported with this The tracking of moving-target.And be partitioned into the moving target in scene consuming time is long, cause existing motion target tracking method without Method is suitable for the occasion with quick track demand.
Moreover, existing motion target tracking method could quickly be caught due to needing picture pick-up device to be constantly in open state Grasp the moving target moved in environment, and the energy consumption of traditional picture pick-up device based on CCD or CMOS is higher, then cause with Track process energy consumption is high, is unfavorable for applying on portable device, such as mobile phone, intelligent glasses equipment.
Therefore, it is necessary to provide one kind, low energy consumption and can quickly capture the motion target tracking method of moving target.
Summary of the invention
The purpose of the present invention aims to solve at least one of above-mentioned technological deficiency, especially tracking process energy consumption height, Yi Jiwu Method quickly and accurately captures the problem of moving target.
The present invention provides a kind of trackings of moving component, comprising:
Signal is acquired using dynamic visual sensor and exports the case point of detection;
Classification and the position of moving component are identified for the case point that current detection goes out using classifier;
Motion profile is determined according to the position of the moving component of the classification successively identified, the fortune as the classification The tracking result of dynamic component;
Wherein, the classifier be according to the dynamic visual sensor acquisition sample signal train in advance come.
The present invention program additionally provides a kind of tracking device of moving component, comprising:
Signal acquisition unit, for acquiring signal using dynamic visual sensor and exporting the case point of detection;
Component recognition unit, for being identified using classifier for the case point that the signal acquisition unit current detection goes out The classification of moving component and position out;Wherein, the classifier is the sample signal acquired according to the dynamic visual sensor It trains in advance;
Motion tracking unit, the moving component of the classification for successively being identified according to the component recognition unit Position determines motion profile, the tracking result of the moving component as the classification.
In the scheme of the present embodiment, for the case point that dynamic visual sensor current detection goes out, preparatory instruction can use Experienced classifier identifies classification and the position of moving component;And it is directed to the other moving component of every type, according to successively identifying The position of moving component of the category determine motion profile, the tracking result of the moving component as the category.
Further, corresponding action command is identified according to tracking result, and respective response is carried out according to action command Operation.
Compared to existing motion target tracking method, on the one hand, the dynamic visual sensor in scheme provided by the invention The moving component that can be fast moved with quick response, and low energy consumption;On the other hand, since dynamic visual sensor is only bright to pixel The degree variation case point above to a certain degree responds, and can directly detect outgoing event in dynamic visual sensor acquisition signal Point, the operation without being partitioned into moving object from scene, can effectively improve the speed and precision of pursuit movement component.
The additional aspect of the present invention and advantage will be set forth in part in the description, these will become from the following description Obviously, or practice through the invention is recognized.
Detailed description of the invention
Fig. 1 is the flow diagram of classifier training method provided in an embodiment of the present invention;
Fig. 2 a is the flow diagram of the tracking of moving component provided in an embodiment of the present invention;
Fig. 2 b is the image schematic diagram of dynamic visual sensor provided in an embodiment of the present invention acquisition;
Fig. 3 is the structural schematic diagram of the tracking device of moving component provided in an embodiment of the present invention;
Fig. 4 is the schematic diagram of internal structure of classifier training unit provided in an embodiment of the present invention;
Fig. 5 is the schematic diagram of internal structure of component recognition unit provided in an embodiment of the present invention;
Fig. 6 is the schematic diagram of internal structure of motor imagination unit provided in an embodiment of the present invention.
Specific embodiment
Clear, complete description is carried out to technical solution of the present invention below with reference to attached drawing, it is clear that described implementation Example is only a part of the embodiments of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, this field is general Logical technical staff obtained all other embodiment without making creative work belongs to the present invention and is protected The range of shield.
The terms such as " module " used in this application, " system " be intended to include with computer-related entity, such as it is but unlimited In hardware, firmware, combination thereof, software or software in execution.For example, module can be, and it is not limited to: processing Process, processor, object, executable program, the thread of execution, program and/or the computer run on device.For example, it counts Calculating the application program run in equipment and this calculating equipment can be module.One or more modules can be located in execution In one process and/or thread, a module can also be located on a computer and/or be distributed in two or more calculating Between machine.
It was found by the inventors of the present invention that existing motion target tracking method can not be suitable for quick track demand The key of occasion is:
In existing motion target tracking method, need to carry out the image of acquisition complicated image procossing, and from shooting Image in by moving target after being partitioned into scene, carry out motion target tracking;And it is partitioned into the movement mesh in scene Consuming time is long for mark.
Therefore, under the occasion with quick track demand, the movement mesh quickly moved can not be captured in time by easily causing Mark.For example, birds quickly fly over during bird appreciation, since above-mentioned traditional picture pick-up device does not often have fast reaction Capturing ability brings many technical requirements to photographer it is thus typically necessary to track Bird Flight track manually.
Further, it was found by the inventors of the present invention that dynamic visual sensor only changes more than to a certain degree pixel intensity Case point respond, and have the characteristics that low energy consumption, illumination condition is wide in range.And low energy consumption that it can be made in mobile device etc. It is in running order when terminal standby, can in time, rapidly, accurately capture moving target;Once user's needs are outer Boundary's condition meets, and the equipment that can automatically control response carries out pattern switching, or when finding unsafe conditions, issues in time pre- It is alert.Illumination condition is wide in range can to make dynamic visual sensor effectively worked in different environmental backgrounds, even at dark The very weak environment of light source, also can capture moving component.
It thereby takes into account, can use dynamic visual sensor acquisition signal, detect that the pixel in the signal of acquisition is bright Degree variation is more than the case point of the range of setting and output.Wherein, pixel intensity variation is more than the case point of the range of setting Position is usually position corresponding to the moving object moved in scene.Therefore, the training based on the case point detected and in advance For determine the affiliated moving component of case point classification classifier, can identify classification and the position of moving component, and Determine the motion profile of moving component.
Compared to existing motion target tracking method, on the one hand, the dynamic visual sensor in scheme provided by the invention The moving object that can be fast moved with quick response, and low energy consumption;On the other hand, dynamic visual sensor only becomes pixel intensity Change the case point above to a certain degree to respond, can directly detect case point in the signal of acquisition, without from It is partitioned into the operation of moving object in scene, can effectively improve the speed and precision of pursuit movement component.
The technical solution that the invention will now be described in detail with reference to the accompanying drawings.
In the embodiment of the present invention, before the tracking for carrying out moving component, case point institute for identification can be trained in advance Belong to the classifier of the classification of moving component, as shown in Figure 1, step can specifically train by the following method:
S101: training sample is generated according to the case point of dynamic visual sensor collecting sample signal output.
It, can be first with the preparatory collecting sample signal of dynamic visual sensor in this step.For example, can be directed to various types of Other moving component is shot the various motion processes of the moving component of the category using dynamic visual sensor, acquisition To sample signal.
Since dynamic visual sensor only responds case point more than pixel intensity variation to a certain degree, transmission is simultaneously The case point of memory response.Therefore, can from the sample signal that dynamic visual sensor acquires direct outgoing event point;And it will The case point of dynamic visual sensor collecting sample signal output is as sample case point.
In practical application, each case point has a corresponding spatial position, position corresponding to a certain case point Neighbouring case point is properly termed as neighbours' point of the case point.In view of the sample signal of dynamic visual sensor acquisition can be one Determine the motion outline for describing object in degree well, and the profile information that these movements generate can also express object itself Shape information, therefore, neighbours' point of case point have good structure description to act on case point, can be in order to judging case point Belong to which part of moving object, i.e., the classification of affiliated moving component.For example, many case points can be generated when human motion, It can judge whether case point belongs to head or hand or body etc. by case point and its a certain range of neighbours point.
Therefore, determine the case point that exports dynamic visual sensor collecting sample signal as sample case point it Afterwards, neighbours' point of the sample case point currently exported can be determined;It is right and according to the position of the sample case point and its neighbours' point The sample case point is classified, that is, judges the classification of moving component belonging to the sample case point.Wherein, moving component Classification specifically can be with head, hand, body etc..
In this way, can be by the sample case point currently exported and neighbours' point of the sample case point, as a training Sample;By the classification of moving component belonging to the sample case point judged to the classification of the moving component of the training sample into Rower is fixed.It specifically, can be to the sampling around the sample case point according to setting for the sample case point currently exported Range is sampled, and neighbours' point of setting quantity is selected;And by the sample case point and the neighbours o'clock selected as one Sample is trained or tests.
S102: using the training sample and its calibration result generated, depth confidence network is trained, is classified Device.
Wherein, the calibration result of training sample refers to the classification for the moving component demarcated for the training sample.
In this step, the multiple training samples composition training sample set that step S101 can be generated utilizes training sample The calibration result for each training sample that collection and the training sample are concentrated is trained depth confidence network, is classified Model.Wherein, on how to be trained to depth confidence network, the common technology hand of those skilled in the art can be used Section.
For example, carrying out successive ignition training to depth confidence network using the training sample and its calibration result that generate.Its In, an iteration training process specifically includes: the training sample set that multiple training samples are formed is as depth confidence network Input;Then, the output of depth confidence network is compared with the calibration result of each training sample;And according to comparison result tune The level parameter of whole depth confidence network continues next iteration, or stops iteration and obtain classifier.
Wherein, the output of depth confidence network is actually the conjecture to the classification of the affiliated moving component of sample case point, In this way, by the way that the classification of the affiliated moving component of conjecture to be compared with calibration result demarcate in advance, accurate, it will The error amount that the two generates is adjusted each level parameter of depth confidence network by the training technique of back-propagating, improves The category division accuracy of finally obtained classifier, with this accurately tracking convenient for moving component.
Based on above-mentioned classifier, the present invention provides a kind of trackings of moving component, as shown in Figure 2 a, process tool Body may include steps of:
S201: acquiring signal using dynamic visual sensor and exports the case point of detection.
Specifically, it can use dynamic visual sensor and acquire signal in real time, and export the case point of detection.Wherein, often The corresponding position of one case point, but the same position may correspond to several case points.Therefore, in the thing of output detection It before part point, needs to get rid of duplicate case point, retain most newly generated case point and exports.
In practical application, there may be noise caused by system, environment etc. in acquisition signal, it therefore, can basis The precedence relationship and space that event occurs close on relationship, remove and make an uproar in the flow of event signal as composed by multiple case points of acquisition Sound.
Further, by accumulation certain time interval (for example, 20ms) in flow of event and binding events point position, The flow of event signal that dynamic visual sensor acquires can be converted into picture signal;As shown in Figure 2 b, the picture signal of conversion In only substantially reflect the profile and texure information of moving target, directly have ignored the object that will not be moved in background.
S202: classification and the position of moving component are identified for the case point that current detection goes out using classifier.
In this step, the classification trained previously according to the sample signal of dynamic visual sensor acquisition can use Device determines the classification of the affiliated moving component of the case point according to the neighbours' point for the case point that current detection goes out.Classifier can be with Neighbours' point of the case point and the case point that are gone out according to the current detection of input, accordingly exports movement portion belonging to the case point The classification of part.
Wherein, neighbours' point of case point can determine in the following way:
Determine dynamic visual sensor acquire the setting before current flow of event signal time interval (for example, All flow of event signals collected in 20ms);And for each flow of event signal for determining, by the flow of event signal Case point in the setting spatial dimension (for example, rectangle of 80 × 80 pixels) on the case point periphery that current detection goes out is determined as this Neighbours' point of case point.
It further, can be with needle after determining the classification of affiliated moving component of all case points detected The category is determined according to the position of each case point for the moving component for belonging to the category to the other moving component of every type The position of moving component.
For example, the center for belonging to the case point of same category of moving component can be calculated;In calculated Position of the heart position as the moving component of the category.In practical application, known in those skilled in the art can be used What common cluster mode obtains the center for belonging to the case point of same category of moving component, the i.e. position of moving component It sets.As an example, K-means clustering method can be used in the embodiment of the present invention, the case point that will test out is given not Same moving component, and the center of moving component is obtained, in order to accurately tracking for subsequent moving component.
S203: determining motion profile according to the position of the moving component successively identified, the tracking knot as moving component Fruit.
It specifically, can be according to successively identifying behind classification and the position for identifying moving component by step S202 The category moving component position, determine the motion profile of the moving component of the category.
In practical application, those skilled in the art can be used, and commonly general track algorithm carries out moving component Motion profile determination, for example, smoothing filter, timing track algorithm etc., this will not be detailed here.
It more preferably,, can also be to identifying after the classification and position for identifying moving component in the embodiment of the present invention The classification of moving component carries out soundness verification, the position of the moving component of false judgment is excluded, so as to improve movement The tracking efficiency of component improves tracking velocity.
Specifically, after the classification of affiliated moving component that all case points detected are determined in step S202, also This can be determined according to the position of each case point for the moving component for belonging to the category for the other moving component of every type The shape and record of the moving component of classification.In turn, it can be determined that whether the shape determined belongs to the moving component of the category Reasonable Shape within the scope of;If so, passing through verifying;Otherwise, verifying does not pass through.Wherein, Reasonable Shape range is according to last time What the shape prior knowledge of the moving component of the shape and category of the moving component of the category of record determined.
More preferably, after step S202 identifies the position of moving component, it can also judge the classification currently identified Moving component position whether in reasonable regional scope;If so, passing through verifying;Otherwise, verifying does not pass through.Wherein, it closes Reason regional scope is the position model according to the moving component of the position and category of the moving component of the category of last registration Enclose what priori knowledge determined.For example, can will work as when moving component is specially the concrete positions such as head or the hand of human body Before the movement portion of the category of the position of the moving component (for example, head or hand) of the category and last registration that identifies The position of part is carried out apart from calculating, if distance meets certain condition, and when meeting the Heuristics of Regular Human's form, is said The position of the moving component of the bright classification currently identified is in reasonable regional scope.
In turn, if the classification of the moving component identified passes through verifying, the classification for the moving component that will identify that and position Set corresponding record.For example, can recorde in the tracking unit list constructed in advance, for movement portion in tracking unit list The position of part is tracked and recorded.In this way, can be according to the moving component of the category successively recorded in tracking unit list Position determines motion profile, and the tracking result as moving component.
In practical application, due to the particularity of dynamic visual sensor imaging, when there is of short duration pause in moving component, The moving component reflected according to the case point that dynamic visual sensor detects is it is possible that the of short duration disappearance of motion profile The case where.Therefore, the continuous tracking to different motion component can be realized by one tracking unit list of maintenance, and Movement position is smoothed.Wherein, common smoothing processing means such as karr can be used in the process of smoothing processing Graceful filtering mode etc..
S204: the identification of action command is carried out according to tracking result.
Specifically, movement segment can be carried out to the obtained tracking result of step S203 to divide, and division is obtained dynamic After making sequence fragment extraction feature, corresponding action command is identified according to the feature of extraction.Wherein, the feature of extraction includes extremely Few following a kind of feature: position feature, route characteristic, moving direction feature, velocity characteristic and acceleration signature etc..
Such as, it can be determined that the moving component that the category whether is directed in instruction database is stored with to match with the feature of extraction Feature;If so, the action command corresponding with this feature that will be recorded in instruction database, is identified as move corresponding with tracking result It instructs.Wherein, instruction database is constructed in advance by technical staff;The other moving component of every type is directed in instruction database in advance, It is stored with the feature of the motion profile to match with the moving component of the category, and corresponding record has preset movement to refer to It enables, such as pattern switching, dangerous prompting etc..
In practical application, some action commands may be needed according to the moving component of multiple classifications all in accordance with certain movement Track carries out athleticism triggering, or needs to be triggered according to the motion profile of entire moving object.
Therefore, more preferably, after different classes of moving component group can also being combined into moving object, according to fortune of all categories The motion profile of dynamic component determines the motion profile of moving object;And the motion profile for the moving object determined is extracted special After sign, corresponding action command is identified according to the feature of extraction.
Wherein, on how to determine the motion profile of moving object according to the motion profile of moving component of all categories, It can specifically be carried out according to the position between component each in tracking unit list apart from calculating, if the distance between two components Meet setting matching condition, then illustrates that the two matches;In this way, the motion profile of each component to match can be intended It closes, obtains the motion profile of moving object.
Wherein, feature is extracted to the motion profile for the moving object determined, specifically includes: the moving object to determining Motion profile carry out movement segment and divide, to the obtained action sequence snippet extraction feature of division.
S205: respective response operation is carried out according to the action command identified.
Specifically, after step S204 identifies action command, corresponding corresponding operating can be carried out according to action command, For example, the recording function of opening terminal apparatus, the camera lens for rotating picture pick-up device etc..
Based on the tracking of above-mentioned moving component, a kind of tracking device of moving component provided in an embodiment of the present invention, As shown in figure 3, can specifically include: signal acquisition unit 301, component recognition unit 302, motion tracking unit 303.
Wherein, signal acquisition unit 301 is used for using dynamic visual sensor acquisition signal and exports the case point of detection.
Component recognition unit 302 is used to know using classifier for the case point that 301 current detection of signal acquisition unit goes out Not Chu moving component classification and position.
Wherein, classifier be according to dynamic visual sensor acquisition sample signal train in advance come.
Motion tracking unit 303 is used to be directed to the other moving component of every type, according to the fortune of the category successively identified The position of dynamic component determines motion profile, the tracking result of the moving component as the category.
More preferably, in the embodiment of the present invention, the tracking device of moving component be can further include: motor imagination unit 304。
Motor imagination unit 304 is used to identify corresponding movement according to the tracking result that motion tracking unit 303 exports Instruction;Respective response operation is carried out according to the action command identified.
In practical application, the sample signal that classifier can be acquired by other devices according to dynamic visual sensor is instructed in advance It practises after coming, is stored in the tracking device of moving component;Alternatively, can also tracked by the tracking device of moving component It trains and in advance before.
Therefore, more preferably, in the embodiment of the present invention, the tracking device of moving component be can further include: classifier Training unit 305.
The case point that classifier training unit 305 is used to be exported according to dynamic visual sensor collecting sample signal generates instruction Practice sample;Using the training sample and its calibration result of generation, depth confidence network is trained, classifier is obtained.Wherein, The calibration result of training sample refers to the classification for the moving component demarcated for the training sample.
Specifically, as shown in figure 4, classifier training unit 305 can specifically include: training sample collection subelement 401, And repetitive exercise subelement 402.
Wherein, training sample collects case point of the subelement 401 for exporting signal acquisition unit 301 as sample thing Part point;Determine neighbours' point of the sample case point currently exported;By the sample case point currently exported and the sample case point Neighbours' point, as a training sample.
Repetitive exercise subelement 402 is used to collect the training sample and its calibration that subelement 401 generates using training sample As a result, carrying out successive ignition training to depth confidence network.
Specifically, repetitive exercise subelement 402 can use training sample collect subelement 401 generate training sample and Its calibration result, when carrying out successive ignition training to depth confidence network, in an iteration training process, by multiple trained samples Input of the training sample set of this composition as depth confidence network;By the output of depth confidence network and the mark of each training sample Determine result to be compared;Continue next iteration according to the level parameter of comparison result percentage regulation confidence network, or stops iteration Obtain classifier.
Preferably, as shown in figure 5, component recognition unit 302 may include: component categories identification subelement 501 and portion Part position identifies subelement 502.
Wherein, component categories identification subelement 501 is used to be gone out using classifier according to 301 current detection of signal acquisition unit Case point neighbours' point, determine the classification of the affiliated moving component of the case point.
Component locations identify that subelement 502 is used to be directed to the other moving component of every type, identify that son is single according to component categories The position of each case point for the moving component for belonging to the category that member 501 is determined, determines the position of the moving component of the category It sets.For example, component locations identification subelement 502 can calculate the centre bit for belonging to the case point of same category of moving component It sets;Using calculated center as the position of the moving component of the category.
More preferably, component recognition unit 302 can also include: soundness verification subelement 503.
The classification for the moving component that soundness verification subelement 503 is used to identify component categories identification subelement 501 Carry out soundness verification;If component categories are identified that subelement 501, component locations identification subelement 502 identify by verifying The classification and position corresponding record of moving component out.
Specifically, soundness verification subelement 503 may determine that the class that component locations identification subelement 502 currently identifies Whether the position of other moving component is in reasonable regional scope;If so, passing through verifying;Otherwise, verifying does not pass through.Wherein, Reasonable regional scope is the position according to the moving component of the position and category of the moving component of the category of last registration What range priori knowledge determined.
Correspondingly, motion tracking unit 303 can be directed to the moving component of each classification, according to soundness verification subelement The position of the moving component of 504 categories successively recorded determines motion profile, and the tracking of the moving component as the category As a result.
Preferably, as shown in fig. 6, motor imagination unit 304 may include: feature extraction subelement 601, action command knowledge Small pin for the case unit 602 and action command respond subelement 603.
Wherein, the tracking result that feature extraction subelement 601 is used to export motion tracking unit 303 carries out movement segment It divides, the action sequence snippet extraction feature that division is obtained.Wherein, the feature of extraction includes a kind of at least following feature: position Set feature, route characteristic, moving direction feature, velocity characteristic and acceleration signature etc..
More preferably, the different classes of movement portion that feature extraction subelement is also used to identify component recognition unit 302 After part group is combined into moving object, the motion profile of the moving component of all categories exported according to motion tracking unit 303 is determined The motion profile of moving object;Feature is extracted to the motion profile of moving object.
Action command identification subelement 602 according to the feature that feature extraction subelement 601 extracts for identifying accordingly Action command.Specifically, action command identifies the moving component other for every type of subelement 602, it can be determined that in instruction database Whether the feature extracted with feature extraction subelement 601 is stored with for the moving component of the category or moving object to match Feature;If so, the action command corresponding with this feature that will be recorded in instruction database, is identified as move corresponding with tracking result It instructs.
Action command response subelement 603 be used to be identified according to action command the action command that identifies of subelement 602 into The operation of row respective response.
In the embodiment of the present invention, the tool of each subelement under each unit and unit in the tracking device of moving component Body function is realized, is referred to the specific steps of the tracking of above-mentioned moving component, this will not be detailed here.
Present inventor has further discovered that intelligent glasses can attract a large amount of attention of user while using, this Sample is easy to cause user to ignore the thing around occurred, includes hazard event, such as the vehicle quickly moved etc..
It is thus preferable to which in practical applications, the tracking and device of above-mentioned moving component can be applied to Brilliant Eyes In mirror.For example, a kind of intelligent glasses for the tracking device for being equipped with above-mentioned moving component, can be used for watching in user virtual Dangerous prompting is carried out when showing picture:
Specifically, can while using intelligent glasses, open the dynamic visual sensor that is arranged on intelligent glasses into Row monitoring.In this way, can be by the tracking device of above-mentioned moving component, when user watches the picture virtually shown, to dynamic The case where periphery of visual sensor real time monitoring, carries out signal acquisition, and according to the tracking device of moving component determine with Track result identifies corresponding action command.Wherein, it is close can be expressed as object for tracking result, and corresponding action command can Think dangerous prompting.Further, when identifying the action command of dangerous prompting, intelligent glasses can promptly exit user Currently watched virtual display picture, and dangerous prompting is issued, inform the arrival of user's danger.
Under normal conditions, the mobile monitoring device in mobile monitoring system is as a mobile device without external power supply, The limitation of the energy and storage space is received, is not suitable for being in open state for a long time.It is contemplated that by above-mentioned moving component Tracking and device are applied in mobile monitoring system.For example, a kind of shifting for the tracking device for being equipped with above-mentioned moving component Dynamic monitoring system, may be implemented to monitor in real time under conditions of low energy consumption:
A dynamic visual sensor can be disposed in mobile monitoring system.Since dynamic visual sensor is with extremely low Energy consumption therefore dynamic visual sensor can be allowed in the open state always, be monitored, acquire signal, simultaneously close off shifting Dynamic monitoring device.In this way, tracked by the moving component that the tracking device of above-mentioned moving component can be moved quickly, And corresponding action command is identified according to the tracking result that the tracking device of moving component is determined.Wherein it is determined that go out with Track result can be expressed as the outsider and enter or leave setting regions;And the corresponding action command identified can for open or Close picture pick-up device.It, can not only be in this way, be equipped with the mobile monitoring system of the tracking device of moving component provided by the invention It realizes under conditions of low energy consumption and monitors always, and can greatly save the storage space of video record, while storage can also be improved Deposit the validity etc. of content.
Further, it is contemplated that photographer such as sport events reporter, Bird Watching fan are equal generally for capture Splendid moment at the time of needing prolonged waiting event to occur, and is captured when this significant instant arrives;And This process is arduous and success rate is lower.
The tracking and device of above-mentioned moving component apply also for quickly moving in shooting system.For example, a kind of peace The quick movement of tracking device equipped with moving component with shooting system, the tracked person quickly moved can be carried out automatically with It claps:
It is set as supervising always with dynamic visual sensor is arranged in shooting system, and by dynamic visual sensor in quick movement Control state is monitored tracked person (such as flying bird, vehicle).When the tracking device of the moving component provided through the invention After identifying tracking result, feature can be extracted from the tracking result identified, and identify accordingly according to the feature of extraction Action command.Wherein, the feature of extraction includes being tracked the moving direction feature and velocity characteristic of person;And corresponding movement refers to Enable to be rotation or dollying equipment.To realize to the automatic with clapping of the tracked person quickly moved, reduce to bat The requirement for the person of taking the photograph.
In technical solution of the present invention, for the case point that dynamic visual sensor current detection goes out, it can use in advance Trained classifier identifies classification and the position of moving component;And it is directed to the other moving component of every type, according to successively identifying The position of the moving component of the category out determines motion profile, the tracking result of the moving component as the category.
Further, after identifying corresponding action command according to tracking result, phase can be carried out according to action command Operation should be responded.
Compared to existing motion target tracking method, on the one hand, the dynamic visual sensor in scheme provided by the invention The moving component that can be fast moved with quick response, and low energy consumption;On the other hand, dynamic visual sensor only becomes pixel intensity Change the case point above to a certain degree to respond, can directly detect case point in dynamic visual sensor acquisition signal, Operation without being partitioned into moving object from scene, can effectively improve the speed and precision of pursuit movement component.
Those skilled in the art of the present technique are appreciated that the present invention includes being related to for executing in operation described herein One or more equipment.These equipment can specially design and manufacture for required purpose, or also may include general Known device in computer.These equipment have the computer program being stored in it, these computer programs are selectively Activation or reconstruct.Such computer program can be stored in equipment (for example, computer) readable medium or be stored in It e-command and is coupled in any kind of medium of bus respectively suitable for storage, the computer-readable medium includes but not Be limited to any kind of disk (including floppy disk, hard disk, CD, CD-ROM and magneto-optic disk), ROM (Read-Only Memory, only Read memory), RAM (Random Access Memory, immediately memory), EPROM (Erasable Programmable Read-Only Memory, Erarable Programmable Read only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory, Electrically Erasable Programmable Read-Only Memory), flash memory, magnetic card or light card Piece.It is, readable medium includes by equipment (for example, computer) with any Jie for the form storage or transmission information that can be read Matter.
Those skilled in the art of the present technique be appreciated that can be realized with computer program instructions these structure charts and/or The combination of each frame and these structure charts and/or the frame in block diagram and/or flow graph in block diagram and/or flow graph.This technology neck Field technique personnel be appreciated that these computer program instructions can be supplied to general purpose computer, special purpose computer or other The processor of programmable data processing method is realized, to pass through the processing of computer or other programmable data processing methods The scheme specified in frame or multiple frames of the device to execute structure chart and/or block diagram and/or flow graph disclosed by the invention.
Those skilled in the art of the present technique have been appreciated that in the present invention the various operations crossed by discussion, method, in process Steps, measures, and schemes can be replaced, changed, combined or be deleted.Further, each with having been crossed by discussion in the present invention Kind of operation, method, other steps, measures, and schemes in process may also be alternated, changed, rearranged, decomposed, combined or deleted. Further, in the prior art to have and the step in various operations, method disclosed in the present invention, process, measure, scheme It may also be alternated, changed, rearranged, decomposed, combined or deleted.
The above is only a preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art For member, without departing from the principle of the present invention, it can also make several improvements and retouch, these improvements and modifications are also answered It is considered as protection scope of the present invention.

Claims (31)

1. a kind of tracking of moving component characterized by comprising
Signal is acquired using dynamic visual sensor and exports the case point of detection;
Classification and the position of moving component are identified for the case point that current detection goes out using classifier;
Motion profile is determined according to the position of the moving component of the classification successively identified, the movement portion as the classification The tracking result of part;
Classification and the position for identifying moving component for the case point that current detection goes out using classifier, comprising:
Using the neighbours' point for the case point that classifier goes out according to current detection, the class of the affiliated moving component of the case point is determined Not;
According to the position of each case point for the moving component for belonging to the category, the position of the moving component of the category is determined.
2. the method as described in claim 1, which is characterized in that the classifier is acquired according to the dynamic visual sensor Sample signal training obtains in advance.
3. method according to claim 2, which is characterized in that the classifier is trained according to the following method to be obtained:
Training sample is generated according to the case point of dynamic visual sensor collecting sample signal output;
Using the training sample and its calibration result of generation, depth confidence network is trained, the classifier is obtained;
Wherein, the calibration result of the training sample refers to the classification for the moving component demarcated for the training sample.
4. method as claimed in claim 3, which is characterized in that described according to the dynamic visual sensor collecting sample signal The case point of output generates training sample, comprising:
Using the case point of dynamic visual sensor collecting sample signal output as sample case point;
Determine neighbours' point of the sample case point currently exported;
By the sample case point currently exported and neighbours' point of the sample case point, as a training sample.
5. method as claimed in claim 4, which is characterized in that it is described using the training sample and its calibration result that generate, it is right Depth confidence network is trained, comprising:
Using the training sample and its calibration result of generation, successive ignition training is carried out to depth confidence network.
6. method as claimed in claim 5, which is characterized in that an iteration training process specifically includes:
The training sample set that multiple training samples are formed is as the input of the depth confidence network;
The output of the depth confidence network is compared with the calibration result of each training sample;
Continue next iteration according to the level parameter that comparison result adjusts the depth confidence network, or stop iteration obtaining it is described Classifier.
7. the method as described in claim 1, which is characterized in that the basis belongs to each case point of the moving component of the category Position, determine the position of the moving component of the category, comprising:
Calculate the center for belonging to the case point of same category of moving component;
Using calculated center as the position of the moving component of the category.
8. the method as described in claim 1-6 is any, which is characterized in that in the classification for identifying moving component and position Afterwards, further includes:
Soundness verification is carried out to the classification of the moving component identified;
If the classification for the moving component that will identify that and position are recorded, and have recorded the class of moving component by verifying Other and position corresponding relationship.
9. method according to claim 8, which is characterized in that the classification of the described pair of moving component identified carries out reasonability Verifying, comprising:
Judge the position of the moving component of the classification currently identified whether in reasonable regional scope;If so, passing through Verifying;Otherwise, verifying does not pass through;
Wherein, the reasonable regional scope is the fortune of the position and the category according to the moving component of the category of last registration What the position range priori knowledge of dynamic component determined.
10. the method as described in claim 1-6 is any, which is characterized in that in the classification that the basis successively identifies The position of moving component determine motion profile, after the tracking result of the moving component as the classification, further includes:
Corresponding action command is identified according to the tracking result;
Corresponding operating is carried out according to the action command.
11. method as claimed in claim 10, which is characterized in that described to identify corresponding movement according to the tracking result Instruction, comprising:
Movement segment is carried out to the tracking result to divide;
The action sequence snippet extraction feature that division is obtained;
Corresponding action command is identified according to the feature of extraction;
Wherein, the feature of the extraction includes at least following a kind of feature: position feature, route characteristic, moving direction feature, speed Spend feature and acceleration signature.
12. method as claimed in claim 10, which is characterized in that described to identify corresponding movement according to the tracking result Instruction, comprising:
After different classes of moving component group is combined into moving object, determined according to the motion profile of moving component of all categories The motion profile of the moving object;
Feature is extracted to the motion profile of the moving object;
Corresponding action command is identified according to the feature of extraction;
Wherein, the feature of the extraction includes at least following a kind of feature: position feature, route characteristic, moving direction feature, speed Spend feature and acceleration signature.
13. the method as described in claim 11 or 12, which is characterized in that the feature according to extraction identifies corresponding dynamic It instructs, comprising:
Whether feature phase with extraction is stored with for the moving component of the classification or the moving object in decision instruction library Matched feature;
If so, the action command corresponding with this feature that will be recorded in described instruction library, is identified as and the tracking result phase The action command answered.
14. method as claimed in claim 10, which is characterized in that
The tracking result indicates that the outsider enters or leave setting regions;And
Corresponding action command is to open or close picture pick-up device;
Alternatively, the tracking result indicates that object is close;And
Corresponding action command is dangerous reminds.
15. the method as described in claim 11 or 12, which is characterized in that
Corresponding action command is rotation or dollying equipment.
16. a kind of tracking device of moving component characterized by comprising
Signal acquisition unit, for acquiring signal using dynamic visual sensor and exporting the case point of detection;
Component recognition unit, the case point for being gone out using classifier for the signal acquisition unit current detection identify fortune The classification of dynamic component and position;
Motion tracking unit, the position of the moving component of the classification for successively being identified according to the component recognition unit Determine motion profile, the tracking result of the moving component as the classification;
The component recognition unit includes:
Component categories identify subelement, for the case point using classifier according to the signal acquisition unit current detection out Neighbours' point, determines the classification of the affiliated moving component of the case point;
Component locations identify subelement, for being directed to the classification of every kind of moving component, identify subelement according to the component categories That determines belongs to the position of each case point of the moving component of the category, determines the position of the moving component of the category.
17. device according to claim 16, which is characterized in that the classifier is according to the dynamic visual sensor The training in advance of the sample signal of acquisition obtains.
18. device as claimed in claim 17, which is characterized in that further include:
Classifier training unit, the case point for being exported according to the dynamic visual sensor collecting sample signal generate training Sample;Using the training sample and its calibration result of generation, depth confidence network is trained, the classifier is obtained;
Wherein, the calibration result of the training sample refers to the classification for the moving component demarcated for the training sample.
19. device as claimed in claim 18, which is characterized in that the classifier training unit includes:
Training sample collects subelement, and the case point for exporting the signal acquisition unit is determined as sample case point Neighbours' point of the sample case point currently exported, by the sample case point currently exported and neighbours' point of the sample case point, As a training sample.
20. device according to claim 19, which is characterized in that the classifier training unit includes:
Repetitive exercise subelement, for collecting the training sample and its calibration result that subelement generates using the training sample, Successive ignition training is carried out to depth confidence network.
21. device according to claim 20, which is characterized in that it is directed to an iteration training process,
The repetitive exercise subelement, the training sample set for forming multiple training samples is as the depth confidence network Input, the output of the depth confidence network is compared with the calibration result of each training sample, according to comparison result tune The level parameter of the whole depth confidence network continues next iteration, or stops iteration and obtain the classifier.
22. device according to claim 16, which is characterized in that component locations identify subelement, belong to for calculating The center of the case point of same category of moving component, using calculated center as the moving component of the category Position.
23. the device as described in claim 16-21 is any, which is characterized in that the component recognition unit further include:
The classification of soundness verification subelement, the moving component for identifying to component categories identification subelement is closed Rationality verifying;And when passing through verifying, the component categories are identified into subelement, component locations identification subelement identification The classification and position corresponding record of moving component out.
24. device according to claim 23, which is characterized in that the soundness verification subelement, it is current for judging Whether the position of the moving component of the classification identified is in reasonable regional scope;And the institute for ought currently identify When stating the position of the moving component of classification in reasonable regional scope, by verifying, otherwise, verifying does not pass through;
Wherein, the reasonable regional scope is the fortune of the position and the category according to the moving component of the category of last registration What the position range priori knowledge of dynamic component determined.
25. the device as described in claim 16-21 is any, which is characterized in that further include:
Motor imagination unit, the tracking result for being exported according to the motion tracking unit identify that corresponding movement refers to It enables, respective response operation is carried out according to the action command.
26. device as claimed in claim 25, which is characterized in that the motor imagination unit includes:
Feature extraction subelement, the tracking result for exporting to the motion tracking unit carry out movement segment and divide, The action sequence snippet extraction feature that division is obtained;Wherein, the feature of the extraction includes a kind of at least following feature: position Feature, route characteristic, moving direction feature, velocity characteristic and acceleration signature;
Action command identifies subelement, and the feature for being extracted according to the feature extraction subelement identifies that corresponding movement refers to It enables;
Action command responds subelement, for identifying that the action command that subelement identifies carries out accordingly according to the action command Response operation.
27. according to the method for claim 25, which is characterized in that
The motor imagination unit, after different classes of moving component group is combined into moving object, according to fortune of all categories The motion profile of dynamic component determines the motion profile of the moving object;And for the motion profile to the moving object Extract feature;And corresponding action command is identified for the feature according to extraction;
Wherein, the feature of the extraction includes at least following a kind of feature: position feature, route characteristic, moving direction feature, speed Spend feature and acceleration signature.
28. device as claimed in claim 26, which is characterized in that
Action command identification subelement in decision instruction library whether for the classification moving component be stored with The feature that the feature that the feature extraction subelement extracts matches;If so, by recorded in described instruction library and this feature Corresponding action command is identified as action command corresponding with the tracking result.
29. device according to claim 25, which is characterized in that
The tracking result indicates that the outsider enters or leave setting regions;And
Corresponding action command is to open or close picture pick-up device;
Alternatively, the tracking result indicates that object is close;And
Corresponding action command is dangerous reminds.
30. the device according to claim 26 or 27, which is characterized in that
Corresponding action command is rotation or dollying equipment.
31. a kind of electronic equipment, which is characterized in that
Processor;
Memory, for storing program;
The processor is used to execute the tracking that described program realizes the described in any item moving components of claim 1 ~ 15.
CN201510018291.1A 2015-01-14 2015-01-14 The tracking and device of moving component Active CN105844659B (en)

Priority Applications (3)

Application Number Priority Date Filing Date Title
CN201510018291.1A CN105844659B (en) 2015-01-14 2015-01-14 The tracking and device of moving component
KR1020150173974A KR102595604B1 (en) 2015-01-14 2015-12-08 Method and apparatus of detecting object using event-based sensor
US14/995,262 US10043064B2 (en) 2015-01-14 2016-01-14 Method and apparatus of detecting object using event-based sensor

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510018291.1A CN105844659B (en) 2015-01-14 2015-01-14 The tracking and device of moving component

Publications (2)

Publication Number Publication Date
CN105844659A CN105844659A (en) 2016-08-10
CN105844659B true CN105844659B (en) 2019-04-26

Family

ID=56579870

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510018291.1A Active CN105844659B (en) 2015-01-14 2015-01-14 The tracking and device of moving component

Country Status (2)

Country Link
KR (1) KR102595604B1 (en)
CN (1) CN105844659B (en)

Families Citing this family (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106384130A (en) * 2016-09-22 2017-02-08 宁波大学 Fault detection method based on data multi-neighbor-local-feature embedding
CN108073929B (en) 2016-11-15 2023-11-24 北京三星通信技术研究有限公司 Object detection method and device based on dynamic vision sensor
CN111149350B (en) * 2017-09-28 2022-02-22 苹果公司 Generating still images using event cameras
EP3543898A1 (en) * 2018-03-21 2019-09-25 Robert Bosch Gmbh Fast detection of secondary objects that may intersect the trajectory of a moving primary object
DE102018211042A1 (en) * 2018-07-04 2020-01-09 Robert Bosch Gmbh Rapid detection of dangerous or endangered objects around a vehicle
WO2020102021A2 (en) 2018-11-13 2020-05-22 Nvidia Corporation Determining associations between objects and persons using machine learning models
CN109544590B (en) * 2018-11-27 2020-05-15 上海芯仑光电科技有限公司 Target tracking method and computing device
CN111988493B (en) * 2019-05-21 2021-11-30 北京小米移动软件有限公司 Interaction processing method, device, equipment and storage medium
US11610330B2 (en) 2019-10-08 2023-03-21 Samsung Electronics Co., Ltd. Method and apparatus with pose tracking
CN110782492B (en) * 2019-10-08 2023-03-28 三星(中国)半导体有限公司 Pose tracking method and device
CN111083354A (en) * 2019-11-27 2020-04-28 维沃移动通信有限公司 Video recording method and electronic equipment
CN112949512B (en) 2021-03-08 2022-07-08 豪威芯仑传感器(上海)有限公司 Dynamic gesture recognition method, gesture interaction method and interaction system

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103533263A (en) * 2012-07-03 2014-01-22 三星电子株式会社 Image sensor chip, operation method, and system having the same
CN103533234A (en) * 2012-07-05 2014-01-22 三星电子株式会社 Image sensor chip, method of operating the same, and system including the image sensor chip
CN103732287A (en) * 2011-05-12 2014-04-16 皮埃尔和玛利居里大学(巴黎第六大学) Method and device for controlling a device for aiding vision
CN103813156A (en) * 2012-11-02 2014-05-21 三星电子株式会社 Motion sensor array device and depth sensing system and methods of using the same
CN104007814A (en) * 2013-02-22 2014-08-27 三星电子株式会社 Apparatus and method for recognizing proximity motion using sensors
CN104272723A (en) * 2011-12-19 2015-01-07 苏黎世大学 Photoarray, particularly for combining sampled brightness sensing with asynchronous detection of time-dependent image data

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP3078166B2 (en) * 1994-02-02 2000-08-21 キヤノン株式会社 Object recognition method
KR101880998B1 (en) * 2011-10-14 2018-07-24 삼성전자주식회사 Apparatus and Method for motion recognition with event base vision sensor
KR102227494B1 (en) * 2013-05-29 2021-03-15 삼성전자주식회사 Apparatus and method for processing an user input using movement of an object
US9696812B2 (en) * 2013-05-29 2017-07-04 Samsung Electronics Co., Ltd. Apparatus and method for processing user input using motion of object

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103732287A (en) * 2011-05-12 2014-04-16 皮埃尔和玛利居里大学(巴黎第六大学) Method and device for controlling a device for aiding vision
CN104272723A (en) * 2011-12-19 2015-01-07 苏黎世大学 Photoarray, particularly for combining sampled brightness sensing with asynchronous detection of time-dependent image data
CN103533263A (en) * 2012-07-03 2014-01-22 三星电子株式会社 Image sensor chip, operation method, and system having the same
CN103533234A (en) * 2012-07-05 2014-01-22 三星电子株式会社 Image sensor chip, method of operating the same, and system including the image sensor chip
CN103813156A (en) * 2012-11-02 2014-05-21 三星电子株式会社 Motion sensor array device and depth sensing system and methods of using the same
CN104007814A (en) * 2013-02-22 2014-08-27 三星电子株式会社 Apparatus and method for recognizing proximity motion using sensors

Also Published As

Publication number Publication date
CN105844659A (en) 2016-08-10
KR102595604B1 (en) 2023-10-30
KR20160087738A (en) 2016-07-22

Similar Documents

Publication Publication Date Title
CN105844659B (en) The tracking and device of moving component
CN109117827B (en) Video-based method for automatically identifying wearing state of work clothes and work cap and alarm system
CN101699862B (en) Acquisition method of high-resolution region-of-interest image of PTZ camera
CN104506819B (en) The mutual feedback tracking system and method for a kind of multi-cam real-time linkage
CN112396658B (en) Indoor personnel positioning method and system based on video
CN108319926A (en) A kind of the safety cap wearing detecting system and detection method of building-site
CN110633612B (en) Monitoring method and system for inspection robot
CN102915638A (en) Surveillance video-based intelligent parking lot management system
WO2021139049A1 (en) Detection method, detection apparatus, monitoring device, and computer readable storage medium
CN102065275B (en) Multi-target tracking method in intelligent video monitoring system
CN111163285A (en) High-altitude falling object monitoring method and system and computer readable storage medium
CN110490043A (en) A kind of forest rocket detection method based on region division and feature extraction
CN108230607B (en) Image fire detection method based on regional characteristic analysis
CN103646250A (en) Pedestrian monitoring method and device based on distance image head and shoulder features
CN104346802A (en) Method and device for monitoring off-job behaviors of personnel
Lengvenis et al. Application of computer vision systems for passenger counting in public transport
US20190096066A1 (en) System and Method for Segmenting Out Multiple Body Parts
US11756303B2 (en) Training of an object recognition neural network
CN110750152A (en) Human-computer interaction method and system based on lip action
CN103514429A (en) Method for detecting specific part of object and image processing equipment
CN109961031A (en) Face fusion identifies identification, target person information display method, early warning supervision method and system
KR101375186B1 (en) Method for detecting disturbance of monitoring camera
CN107547865A (en) Trans-regional human body video frequency object tracking intelligent control method
CN110705453A (en) Real-time fatigue driving detection method
CN109145758A (en) A kind of recognizer of the face based on video monitoring

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant