CN205451095U - A face -identifying device - Google Patents
A face -identifying device Download PDFInfo
- Publication number
- CN205451095U CN205451095U CN201520986440.9U CN201520986440U CN205451095U CN 205451095 U CN205451095 U CN 205451095U CN 201520986440 U CN201520986440 U CN 201520986440U CN 205451095 U CN205451095 U CN 205451095U
- Authority
- CN
- China
- Prior art keywords
- face
- video
- board
- image
- people
- 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
Links
Landscapes
- Collating Specific Patterns (AREA)
Abstract
The utility model relates to a face -identifying device, the device is at least including input module, people's face analysis module and output module. Input module provides desire detection images the sequence for people's face analysis module, and people's face analysis module carries out face identification with the help of degree of depth learning technology, compares assay and merges analysis result based on the parallel retrieval in a plurality of subdata storehouses, and output module closes analysis module 0 and the stack of original video with face detection frame, people's appearance, exports the front end and shows. This disclosed device has automatic, quick, accurate characteristics when discernment people face.
Description
Technical field
It relates to field of video monitoring, particularly to a kind of face identification device.
Background technology
The most a lot of industries, in order to absorb, safeguard more honored guest (VIP) client valuable, potential, are increasingly paid attention to visitant customer providing more high-quality, servicing targetedly.Traditional industries can use the forms such as vip card to distinguish client, but carry card, report mode not humane, the facilitation such as card number.And face recognition technology is based on the characteristic information of face and carries out a kind of technology of identification, has the advantages such as untouchable, concurrency, non-imposed, intuitive.
Utility model content
For above-mentioned subproblem, present disclose provides a kind of face identification device, described device carries out recognition of face by degree of deep learning art, can be bank VIP identification, shops welcome, monitoring scene recognition of face offer service support；If utilizing, device detecting the time of face, place and duration, passenger flow statistics can be realized further, provide help for improving service or inquiry special user.Wherein, modules can use existing video acquisition board, and multimedia analysis board carrys out the module of assembly cost disclosed device, is coupled to relation between modules.
A kind of face identification device, described device at least includes input module, human face analysis module and output module；
Described input module is intended to detect image sequence and sends human face analysis module to；
Described output module includes image output unit and/or message subscribing unit；
The face obtained in human face analysis module is identified by described image output unit, and by information superposition relevant for this face to original video；
Described message subscribing unit sends event message to terminal booking reader；
Wherein, described human face analysis module at least includes following unit:
U100, Face datection tracking cell: to the image received, the face in detecting and tracking image, and carry out Quality estimation, select and meet some frames of requirement as key frame, pass to face alignment unit；
U200, face alignment unit: receive described key frame and extract the face characteristic of each frame, search in User Information Database server and select multiple similar face characteristic to compare；
Wherein: described face characteristic uses multidimensional characteristic vectors to represent；
The first identifier mark that described User Information Database server allows the single M of having to open facial image use identical is stored in.
Accompanying drawing explanation
Apparatus structure pie graph in one embodiment of Fig. 1 disclosure.
Detailed description of the invention
In a basic embodiment, as shown in Figure 1, it is provided that a kind of face identification device, described device at least includes input module, human face analysis module and output module；
Described input module is intended to detect image sequence and sends human face analysis module to；
Described output module includes image output unit and/or message subscribing unit；
The face obtained in human face analysis module is identified by described image output unit, and by information superposition relevant for this face to original video；
Described message subscribing unit sends event message to terminal booking reader；
Wherein, described human face analysis module at least includes following unit:
U100, Face datection tracking cell: to the image received, the face in detecting and tracking image, and carry out Quality estimation, select and meet some frames of requirement as key frame, pass to face alignment unit；
U200, face alignment unit: receive described key frame and extract the face characteristic of each frame, search in User Information Database server and select multiple similar face characteristic to compare；
Wherein: described face characteristic uses multidimensional characteristic vectors to represent；
The first identifier mark that described User Information Database server allows the single M of having to open facial image use identical is stored in.
In this embodiment, wherein, modules can use existing video acquisition board, and multimedia analysis board carrys out the module of assembly cost disclosed device, is coupled to relation between modules.The output of described input module is facial image, and its input can be multimeshed network video source, image sequence, offline video or real-time video, as long as can obtain the image with face after treatment.When image is detected, all images received can be detected, it is also possible to preferably image is detected.In one embodiment, described device carries out a Face datection every 6 frames.When detecting image, extracting the face location in image, face key point information, described face key point information can include the positional informationes such as canthus, the end of eyebrow, the corners of the mouth, nose.When described image sequence is single frames, this image is originally as key frame；When described image sequence is multiframe, from this sequence, select matter measured N frame as key frame.Wherein, after the judgement of quality can be by giving a mark to aftermentioned index, choose the high front N frame of score as key frame.Described index include face picture definition, size, real human face, block, illumination etc..Described face characteristic is represented by multidimensional characteristic vectors, in one embodiment, uses about 180 dimensional feature vectors to represent face characteristic.To having detected that face, it is tracked in subsequent frames.When searching, using N group face characteristic as an entirety, User Information Database server is retrieved similar face, select several faces of highest scoring as returning result.And the most described video of result that comparison is analyzed assembles Dispatching Unit and original image can be sent to image output unit by http agreement or message server.In one embodiment, when at monitoring region recognition face database face, described device makes an announcement immediately or reports to the police, and message includes sending the information such as its special personnel such as recognition result such as VIP, suspicious people, place time, face picture paid close attention to toward subscriber.
In one embodiment, give a kind of method for Quality estimation described in U100, i.e. comprise the steps:
S1010, to each facial image detected, first determine whether that whether two spacing meet and set requirement, requiring if meeting, performing step S1011；Otherwise, this facial image detected is given up；
Whether the face confidence score of the facial image that S1011, calculating detect meets sets requirement, requires if meeting, execution step S1012；Otherwise, this facial image detected is given up；
S1012, calculate positive face score and whether meet and set requirement, then judge that this frame can be used in identifying face as met；Otherwise, this facial image detected is given up.
In one embodiment, it is provided that a kind of implementation specifically selecting key frame.In this embodiment, to the single face followed the tracks of and capture, according to two spacing > 25, face confidence score > 0.95, positive face score, it is judged that whether this frame is used for identifies.Further, additionally providing the method selecting key frame by program realization in this embodiment, i.e. to each image being tracked as same face, one key frame container of internal maintenance, capacity is 10.During beginning, if discontented 10 frames, then every frame is all stored in container；After full 10 frames, be suitable for the frame identified, and and the frame number interval that is finally stored in more than 10, then replace the frame that known quality is worst；Record the frame number that the single image being tracked as same face has been processed, if frame number is more than 20, then terminate to follow the tracks of.
In one embodiment, give one detecting and tracking in described U100 and comprise the steps, the steps include:
S101, carry out a Face datection every some frames, when face being detected, use indicia framing that the part including face is marked the face meeting prescription；
Whether S102, the face area of judge mark overlap with the face area having detected that, when registration meets predetermined threshold value, then it is assumed that be same face with the face having detected that, then enter step S103；Otherwise it is assumed that the face of current markers is new face, tracking terminates；
S103, face to labelling carry out face alignment in indicia framing, detect face key point position, calculate the outer area-encasing rectangle of face key point, and detect before replacing it thinks for the image in the indicia framing of same face.
In this embodiment, using indicia framing to be marked the part including face, the part of institute's labelling can be head, more excellent, it is also possible to include shoulder, in the mark mode including shoulder, can improve discrimination.No matter using which kind of mode, the calculating of registration all can be measured by confidence level, when the confidence level calculated reaches certain scope, it is believed that two objects are same target.And the scope that should reach can determine by the way of test.
Preferably, using many storehouses and parallel search, it may be assumed that the User Information Database server in described U200 includes multiple subdata base, the most described comparison analysis carries out parallel search comparison analysis, and combined analysis result based on multiple subdata bases.This mode is not only supported a large amount of facial images are imported User Information Database server, is not added with again the overall search time simultaneously.Each subdata base imports a certain amount of facial image, and single multiple facial images import identical data base.When retrieval, using multi-threaded parallel to retrieve the mode of each data base in one embodiment, then the result according to comparative analysis merges the result of multiple subdata bases.
In one embodiment, the acquisition methods of face characteristic in the facial image about warehouse-in is given, it may be assumed that in the User Information Database server in described U200, the facial image of storage uses DeepId degree of deep learning algorithm to extract face characteristic.This mode obtains face characteristic and can aid in and identify face accurately.In one embodiment, use this extraction face characteristic extracting mode, extract the characteristic vector of about 180 dimensions.
Use multidimensional characteristic vectors to represent based on face characteristic, in one embodiment, give a kind of when searching similar features vector, reduce number of comparisons speed-up ratio to process approach, it may be assumed that the similar face characteristic in described U200 is obtained by following step:
S2011, set up KD tree: when searching, set up KD tree and search for K neighbour, K >=M；
S2012, traversal KD tree: when traveling through KD tree, every layer choose in face characteristic one-dimensional compare, to determine the branch that next layer is retrieved, finally determine the multiple face characteristics similar to key frame.
This by the way of setting up KD characteristic key tree, when searching for similar features, realized by traversal trie tree, in order to reduce comparison number of times, choose one-dimensional feature at every layer and compare, to determine next layer of branch to be retrieved.
In one embodiment, in described step S102, when the face judging new labelling is same face with the face having detected that, the facial image of this new labelling is used with the face having detected that the second identical identifier mark；And, the comparison analysis in described U200 comprises the steps:
S201, to having the M two field picture of identical second identifier mark, according to whether positive face, one quality score q of sharpness computationi, i ∈ [1, M]；
S202, to the every two field picture in M two field picture, from face database, retrieve comparison respectively find out most like N number of user, corresponding similarity is Si,userj, i ∈ [1, M], j ∈ [1, N]；
S203, to M two field picture retrieval comparison there are K user, calculate the score of the similarity of each user in this K user,
K ∈ [1, K], K=M × N；
S204, basisK user is arranged in descending order, chooses several most like users.
Under this alignments, if User Information Database server includes multiple subdata base, can be in the way of having the final recognition result of multiple acquisition.Such as after multiple subdata bases are carried out parallel search, each subdata base is carried out step S202～S204 then the similarity of all most like users is ranked up after choose return result.For another example, each subdata base is returned score sequence several face characteristics preceding in this subdata base, then the face characteristic returned is re-used Similarity value to be ranked up, select the facial image corresponding to several face characteristics preceding under current sequence as returning result.
Optionally, after comparison is analyzed, described face alignment unit also realizes operations described below:
S2031, use degree of deep learning method carry out face character calculating；
S2032, the face of judgement detection are made whether to be present in User Information Database server；If being present in User Information Database server, then face character result is updated；Otherwise recognition result and face character result of calculation are stored together.
Described face character include user's sex, age, whether wear glasses, the appearance attribute such as medicated cap, mask.The device increasing storage face character can be when externally providing search function, increase retrieval dimension, can temporally, face to be detected with put human face similarity value in storage, appearance attribute, place are filtered, and reduce range of search, accelerate retrieval rate, it is provided that retrieval accuracy.
Optionally, storage face character result of calculation on the basis of can also enclose timing statistics point, place for each result, it may be assumed that described face character result of calculation also include acquisition image time time point and place.The most whether this there is offer data support for positioning certain face in certain region.In one embodiment, described device is that the special personnel such as VIP or suspicious people individually sets up User Information Database server, when user inquires about this kind of personnel, directly face characteristic with the facial image of this library storage can compare, position when whether certain face occurred in certain region easily and fast.
In one embodiment, described device has artificial abortion's statistical function, it may be assumed that described face alignment unit also realizes operating as follows:
S2030, the number of face detected in statistics certain time, time of appearance of each face and duration.In this embodiment, currently detected face and the face detected in certain time can be compared, if being judged as same person by described device, then it is assumed that the number of times that this people occurs is 1 time；So can obtain the different face numbers occurred in a certain time interval, can also obtain in several time intervals simultaneously, the number of times that same face occurs, the time every time occurred and duration.In order to realize such statistical function, described device can be that the face detected sets up temporary User Information database server, and this data base can the most periodically set up.Described device can also set up visitor's feature database specially.When described device is for shops, can be that shops carries out above-mentioned passenger flow statistics.
In one embodiment, during for facial image source for video, when described video can be multimeshed network video source, offline video or real-time video, for obtaining video frame images, described input module also includes that video decoding unit, described output module also include that video assembles Dispatching Unit；
Described video decoding unit is used for reading live video stream or local video file, and parses video frame images；
Described video assembling Dispatching Unit has been used for superposition the former video frame images of Face datection frame and the information relevant to this face and has been re-encoded as video.
In this embodiment, when described input module sends image sequence to be detected to human face analysis module, can realize by the way of queue.
In one embodiment, when video assembles, the information such as Face datection frame that human face analysis module is obtained, the user name identified, it is added in original video frame, then it is re-encoded as video by x264 form, after assembling, it is broadcasted by live555, it is provided that browser or other clients play out.
In one embodiment, described video decoding unit reads rtsp video flowing or local video file, by vlc Video Decoder, explains video frame images, uses buffer queue, image is sent to human face analysis module.
Preferably, described device also includes that photographic head, the most described input module include video configuration unit, and described video configuration unit is for configuring the monitoring parameter of video channel scene.In one embodiment, it is necessary to configuration video channel address.In this manner, can be applied to monitor in real time by the device of realization, and the face in monitoring is carried out Real time identification.Additional alarm function on the basis of the disclosure, can be that the scene having security to need provides the recognition of face monitoring of complete set to service, the artificial abortion gateway such as paid close attention in public security, holdee is carried out automatically face snap, automatically personnel's external appearance characteristic is identified, automatically carrying out fast automatic comparison with warehouse-in a suspect, if finding suspicious people, providing alarm.For another example, community gate inhibition's security it is applied to.
In one embodiment, described device is capable of showing video and recognition result in real time；When user needs to identify its VIP client serviced, system can help client to automatically identify the VIP client entering shops, and provides prompting；Help user to add up every day and enter client's number of shops, add up each client and enter time and the number of times of shops；By the client's face picture entering shop door is preserved, calculate and preserve and include user's sex, age, whether wear glasses, the appearance attribute such as medicated cap, mask, be saved into shop time and duration, facilitate user that client is carried out querying condition filtration；Display is recently entered the customer information of shops, including capturing face picture, access time, place, access times etc..
In one embodiment, described device provides video monitoring service for public security department, at monitoring region, the face that detection enters, and compares with the suspicious people put in storage, when finding the personnel to be monitored, provides prompting.Described prompting can be the combining form of following a kind of or any various ways: static text, pattern or dynamic word, dynamic pattern, sound.
The face recognition technology of the disclosure can quickly recognize visitant customer by recognition of face and associate relevant customer information database.Also can will often patronize, valuable potential client is serviced and recommendation specific aim product as potential visitant customer emphasis by data statistics.Additionally, additional alarm function on the basis of the disclosure, the recognition of face monitoring service of complete set can also be provided for public security, in the artificial abortion gateway that public security is paid close attention to, holdee is carried out automatically face snap, automatically identifies personnel's external appearance characteristic, automatically carry out fast automatic comparison with warehouse-in a suspect, if finding suspicious people, provide alarm.
Being described in detail the disclosure above, specific case principle of this disclosure used herein and embodiment are set forth, and the explanation of above example is only intended to help and understands disclosed method and core concept thereof；Simultaneously for those skilled in the art, according to the thought of the disclosure, the most all will change, in sum, this specification content should not be construed as restriction of this disclosure.
Claims (3)
1. a face identification device, it is characterised in that:
Described device at least includes photographic head, video acquisition board, multimedia analysis board and output module；
Described photographic head connects video acquisition board, in order to the image sequence being intended to detect original video sends video acquisition board to；
Described video acquisition board connects described multimedia analysis board, to the image received, the face in detecting and tracking image, and carry out Quality estimation, selects and meets some frames of requirement as key frame, pass to described multimedia analysis board；
Described multimedia analysis board receives described key frame and extracts the face characteristic of each frame, User Information Database server is searched and selects multiple similar face characteristic to compare, wherein said face characteristic uses multidimensional characteristic vectors to represent, the User Information Database server in described multimedia analysis board is parallel type system；
Described output module connects the terminal of booking reader, and the original video that the result of multimedia analysis board is added to is to export new video.
Device the most according to claim 1, it is characterised in that described photographic head also includes that video decoding unit, described output module also include video encoding unit.
3., according to the arbitrary described device of claim 1-2, it is characterised in that described photographic head also includes video configuration unit, described video configuration unit is for configuring the monitoring parameter of video channel scene.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201520986440.9U CN205451095U (en) | 2015-12-02 | 2015-12-02 | A face -identifying device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201520986440.9U CN205451095U (en) | 2015-12-02 | 2015-12-02 | A face -identifying device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN205451095U true CN205451095U (en) | 2016-08-10 |
Family
ID=57180532
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201520986440.9U Active CN205451095U (en) | 2015-12-02 | 2015-12-02 | A face -identifying device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN205451095U (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107563341A (en) * | 2017-09-15 | 2018-01-09 | 赵立峰 | A kind of face identification device and a kind of face identification system |
| CN107679613A (en) * | 2017-09-30 | 2018-02-09 | 同观科技（深圳）有限公司 | A kind of statistical method of personal information, device, terminal device and storage medium |
| CN108038422A (en) * | 2017-11-21 | 2018-05-15 | 平安科技（深圳）有限公司 | Camera device, the method for recognition of face and computer-readable recording medium |
| CN108241853A (en) * | 2017-12-28 | 2018-07-03 | 深圳英飞拓科技股份有限公司 | A kind of video frequency monitoring method, system and terminal device |
| CN108597065A (en) * | 2018-03-12 | 2018-09-28 | 南京甄视智能科技有限公司 | Passenger flow statistical method based on recognition of face |
| CN109344703A (en) * | 2018-08-24 | 2019-02-15 | 深圳市商汤科技有限公司 | Method for checking object and device, electronic equipment and storage medium |
| CN109711369A (en) * | 2018-12-29 | 2019-05-03 | 深圳英飞拓智能技术有限公司 | Pedestrian count method, apparatus, system, computer equipment and storage medium |
| CN109726680A (en) * | 2018-12-28 | 2019-05-07 | 东方网力科技股份有限公司 | Face recognition method, device, system and electronic equipment |
| CN109978914A (en) * | 2019-03-07 | 2019-07-05 | 北京旷视科技有限公司 | Face tracking methods and device |
| CN110363891A (en) * | 2019-07-04 | 2019-10-22 | 华南理工大学 | A kind of intelligent visitor system suitable for more scenes |
| CN110442606A (en) * | 2019-07-16 | 2019-11-12 | 浙江大华技术股份有限公司 | A kind of processing method of data, equipment and computer storage medium |
| CN111382286A (en) * | 2018-12-27 | 2020-07-07 | 深圳云天励飞技术有限公司 | Data processing method and related product |
-
2015
- 2015-12-02 CN CN201520986440.9U patent/CN205451095U/en active Active
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107563341A (en) * | 2017-09-15 | 2018-01-09 | 赵立峰 | A kind of face identification device and a kind of face identification system |
| CN107679613A (en) * | 2017-09-30 | 2018-02-09 | 同观科技（深圳）有限公司 | A kind of statistical method of personal information, device, terminal device and storage medium |
| CN108038422A (en) * | 2017-11-21 | 2018-05-15 | 平安科技（深圳）有限公司 | Camera device, the method for recognition of face and computer-readable recording medium |
| CN108038422B (en) * | 2017-11-21 | 2021-12-21 | 平安科技（深圳）有限公司 | Camera device, face recognition method and computer-readable storage medium |
| CN108241853A (en) * | 2017-12-28 | 2018-07-03 | 深圳英飞拓科技股份有限公司 | A kind of video frequency monitoring method, system and terminal device |
| CN108597065A (en) * | 2018-03-12 | 2018-09-28 | 南京甄视智能科技有限公司 | Passenger flow statistical method based on recognition of face |
| CN109344703A (en) * | 2018-08-24 | 2019-02-15 | 深圳市商汤科技有限公司 | Method for checking object and device, electronic equipment and storage medium |
| CN109344703B (en) * | 2018-08-24 | 2021-06-25 | 深圳市商汤科技有限公司 | Object detection method and device, electronic equipment and storage medium |
| CN111382286A (en) * | 2018-12-27 | 2020-07-07 | 深圳云天励飞技术有限公司 | Data processing method and related product |
| CN109726680A (en) * | 2018-12-28 | 2019-05-07 | 东方网力科技股份有限公司 | Face recognition method, device, system and electronic equipment |
| CN109711369A (en) * | 2018-12-29 | 2019-05-03 | 深圳英飞拓智能技术有限公司 | Pedestrian count method, apparatus, system, computer equipment and storage medium |
| CN109978914B (en) * | 2019-03-07 | 2021-06-08 | 北京旷视科技有限公司 | Face tracking method and device |
| CN109978914A (en) * | 2019-03-07 | 2019-07-05 | 北京旷视科技有限公司 | Face tracking methods and device |
| CN110363891A (en) * | 2019-07-04 | 2019-10-22 | 华南理工大学 | A kind of intelligent visitor system suitable for more scenes |
| CN110442606A (en) * | 2019-07-16 | 2019-11-12 | 浙江大华技术股份有限公司 | A kind of processing method of data, equipment and computer storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN205451095U (en) | A face -identifying device | |
| CN105488478A (en) | Face recognition system and method | |
| CN110505583B (en) | Trajectory matching method based on bayonet data and signaling data | |
| CN105574506A (en) | Intelligent face tracking system and method based on depth learning and large-scale clustering | |
| US9230547B2 (en) | Metadata extraction of non-transcribed video and audio streams | |
| CN107563343B (en) | FaceID database self-improvement method based on face recognition technology | |
| CN103678417B (en) | Human-machine interaction data treating method and apparatus | |
| CN102342124B (en) | Method and apparatus for providing information related to broadcast programs | |
| WO2017117879A1 (en) | Personal identification processing method, apparatus and system | |
| US20120207356A1 (en) | Targeted content acquisition using image analysis | |
| CN106952112A (en) | A kind of method for delivering advertisement accurately and system | |
| CN105404860A (en) | Method and device for managing information of lost person | |
| US20150019206A1 (en) | Metadata extraction of non-transcribed video and audio streams | |
| JPWO2018180588A1 (en) | Face image matching system and face image search system | |
| CN106355154B (en) | Method for detecting frequent passing of people in surveillance video | |
| CN103606093A (en) | Intelligent chain VIP customer service system based on human characteristics | |
| US20120207349A1 (en) | Targeted content acquisition using image analysis | |
| CN111476183A (en) | Passenger flow information processing method and device | |
| CN108446681B (en) | Pedestrian analysis method, device, terminal and storage medium | |
| CN109377251A (en) | A kind of method and system identifying client properties | |
| CN110717358B (en) | Visitor number counting method and device, electronic equipment and storage medium | |
| CN102486793A (en) | Method and system for searching target user | |
| CN112258254B (en) | Internet advertisement risk monitoring method and system based on big data architecture | |
| CN105589974A (en) | Surveillance video retrieval method and system based on Hadoop platform | |
| CN111914648A (en) | Vehicle detection and identification method and device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |