CN108288280A - Dynamic human face recognition methods based on video flowing and device - Google Patents
Dynamic human face recognition methods based on video flowing and device Download PDFInfo
- Publication number
- CN108288280A CN108288280A CN201711466486.8A CN201711466486A CN108288280A CN 108288280 A CN108288280 A CN 108288280A CN 201711466486 A CN201711466486 A CN 201711466486A CN 108288280 A CN108288280 A CN 108288280A
- Authority
- CN
- China
- Prior art keywords
- human face
- frame
- region
- network
- hierarchical network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 26
- 238000012545 processing Methods 0.000 claims abstract description 10
- 239000013598 vector Substances 0.000 claims description 20
- 238000012216 screening Methods 0.000 claims description 19
- 238000001514 detection method Methods 0.000 abstract description 15
- 238000013528 artificial neural network Methods 0.000 abstract description 7
- 230000005764 inhibitory process Effects 0.000 description 6
- 238000004064 recycling Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/23—Processing of content or additional data; Elementary server operations; Server middleware
- H04N21/234—Processing of video elementary streams, e.g. splicing of video streams or manipulating encoded video stream scene graphs
- H04N21/23418—Processing of video elementary streams, e.g. splicing of video streams or manipulating encoded video stream scene graphs involving operations for analysing video streams, e.g. detecting features or characteristics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
- G06T2207/30201—Face
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Image Analysis (AREA)
Abstract
The dynamic human face recognition methods and device that the invention discloses a kind of based on video flowing, this method uses multitask cascade neural network, entire algorithm network is divided into four levels, first hierarchical network handles input picture, second hierarchical network and third hierarchical network refine candidate region twice, it is returned by frame and obtains human face region, 4th hierarchical network carries out crucial point location, since the network inputs of first three level are sequentially increased with model size, being done for task is also more and more careful, therefore cascade structure processing data are more advantageous to, detection efficiency is maximized with accuracy rate, detection speed is fast, recall rate is high, false drop rate is low.
Description
Technical field
The present invention relates to technical field of face recognition, more particularly to the dynamic human face recognition methods based on video flowing and dress
It sets.
Background technology
Conventional face's detection algorithm and Face datection algorithm based on deep neural network respectively have short length.Conventional face detects
The advantages of algorithm is that speed is fast, the disadvantage is that recall rate is low, false drop rate is high;Face datection algorithm based on deep neural network it is excellent
Point is that recall rate is high, false drop rate is low, the disadvantage is that detection speed is slow.
Invention content
For overcome the deficiencies in the prior art, the purpose of the present invention is to provide the dynamic human face identification sides based on video flowing
Method and device, it is intended to solve the problems, such as that conventional face's detection algorithm recall rate is low, false drop rate is high, and be based on deep neural network
The slow problem of Face datection algorithm speed.
The purpose of the present invention is realized using following technical scheme:
A kind of dynamic human face recognition methods based on video flowing, including:
Image acquisition step obtains a frame of video stream data as input picture;
Preliminary screening step is handled input picture in the first hierarchical network, obtains candidate region and its frame returns
Return Vector Groups, the candidate vector in frame regression vector group is assessed and calibrated, then inhibits removal to repeat by non-maximization
Candidate region;
The candidate region that preliminary screening step obtains is sent into the second hierarchical network and is once refined by refinement step
Processing, excludes undesirable candidate region, executes calibration by frame recurrence, non-maximization is recycled to inhibit to carry out candidate regions
The merging in domain;
Secondary refinement step, the candidate region that a refinement step is obtained are sent into third hierarchical network and carry out secondary refinement
Processing, excludes undesirable candidate region, executes calibration by frame recurrence, non-maximization is recycled to inhibit to carry out candidate regions
The merging in domain obtains human face region;
Key point positioning step, the human face region that secondary refinement step is obtained are sent into the 4th hierarchical network, carry out crucial
Point location.
On the basis of the above embodiments, it is preferred that after the secondary refinement step, further include:
The human face region of present frame is amplified by face tracking step, and amplified human face region is sent into third layer
Grade network, the face tracking for carrying out next frame is returned by frame.
On the basis of the above embodiments, it is preferred that before the face tracking step, further include:
Markers step marks the ID of each human face region.
On the basis of above-mentioned any embodiment, it is preferred that in the preliminary screening step, make in the first hierarchical network
Input picture is handled with full volume machine network.
A kind of dynamic human face identification device based on video flowing, including:
Image collection module, for obtaining a frame of video stream data as input picture;
Preliminary screening module, for being handled input picture in the first hierarchical network, obtain candidate region and its
Frame regression vector group is assessed and is calibrated to the candidate vector in frame regression vector group, then inhibits to remove by non-maximization
The candidate region repeated;
Refinement module, the candidate region for obtaining preliminary screening module are sent into the second hierarchical network and are carried out once
Micronization processes exclude undesirable candidate region, and calibration is executed by frame recurrence, and non-maximization inhibition is recycled to be waited
The merging of favored area;
Secondary refinement module, it is secondary that the progress of third hierarchical network is sent into the candidate region for obtaining a refinement module
Micronization processes exclude undesirable candidate region, and calibration is executed by frame recurrence, and non-maximization inhibition is recycled to be waited
The merging of favored area obtains human face region;
Key point locating module, the human face region for obtaining secondary refinement module are sent into the 4th hierarchical network, are carried out
Crucial point location.
On the basis of the above embodiments, it is preferred that further include:
Amplified human face region is sent into the by face tracking module for being amplified the human face region of present frame
Three hierarchical networks return the face tracking for carrying out next frame by frame.
On the basis of the above embodiments, it is preferred that further include:
Mark module, the ID for marking each human face region.
On the basis of above-mentioned any embodiment, it is preferred that the preliminary screening module is used in the first hierarchical network
Input picture is handled using full volume machine network.
Compared with prior art, the beneficial effects of the present invention are:
The dynamic human face recognition methods and device that the invention discloses a kind of based on video flowing, this method use multitask grade
Join neural network, entire algorithm network is divided into four levels, and the first hierarchical network handles input picture, the second level net
Network and third hierarchical network refine candidate region twice, by frame return obtain human face region, the 4th hierarchical network into
Row key point location, since the network inputs of first three level are sequentially increased with model size, being done for task is also to get over
It is more careful to come, therefore is more advantageous to cascade structure processing data, maximizes detection efficiency with accuracy rate, detection speed is fast, detection
Rate is high, false drop rate is low.
Description of the drawings
Present invention will be further explained below with reference to the attached drawings and examples.
Fig. 1 shows a kind of flow signal of dynamic human face recognition methods based on video flowing provided in an embodiment of the present invention
Figure;
Fig. 2 shows a kind of structural representations of the dynamic human face identification device based on video flowing provided in an embodiment of the present invention
Figure.
Specific implementation mode
In the following, in conjunction with attached drawing and specific implementation mode, the present invention is described further, it should be noted that not
Under the premise of conflicting, new implementation can be formed between various embodiments described below or between each technical characteristic in any combination
Example.
Specific embodiment one
As shown in Figure 1, an embodiment of the present invention provides a kind of dynamic human face recognition methods based on video flowing, including it is following
Step.
Image acquisition step S101 obtains a frame of video stream data as input picture.
Preliminary screening step S102, is handled input picture in the first hierarchical network, obtain candidate region and its
Frame regression vector group is assessed and is calibrated to the candidate vector in frame regression vector group, then inhibits to remove by non-maximization
The candidate region repeated.In the step, the first hierarchical network quickly handles entire input picture, obtains candidate frame and this
The frame regression vector group of a little candidate frames;Then these candidate vectors are assessed, and are calibrated;Non- maximization inhibition is reused to make a return journey
Except the candidate region largely repeated.
Refinement step S103, the candidate region that preliminary screening step S102 is obtained are sent into the second hierarchical network and are carried out
Micronization processes, exclude undesirable candidate region, and calibration is executed by frame recurrence, recycling is non-maximize inhibit into
The merging of row candidate region.The processing that the second hierarchical network refines is sent into candidate region by the step, excludes largely not
Satisfactory candidate region executes calibration by frame recurrence, and non-maximization inhibition is recycled to merge.
Secondary refinement step S104, the candidate region that a refinement step S103 is obtained are sent into third hierarchical network and are carried out
Secondary micronization processes exclude undesirable candidate region, and calibration is executed by frame recurrence, recycling is non-maximize inhibit into
The merging of row candidate region obtains human face region.Third level is sent into the remaining candidate region of second hierarchical network by the step
The processing that network is more refined excludes last remaining undesirable candidate region, and calibration is executed by frame recurrence,
It recycles non-maximization inhibition to merge, is finally left human face region.Since the network of first three level is sequentially increased,
Being done for task is also more and more careful, is more advantageous to cascade structure in this way, maximizes detection efficiency with accuracy rate.
Key point positioning step S106, the human face region that secondary refinement step S104 is obtained are sent into the 4th hierarchical network,
Carry out crucial point location.The human face region that the step exports third hierarchical network inputs the 4th hierarchical network, carries out key point
Location tasks.
The embodiment of the present invention uses multitask cascade neural network, entire algorithm network to be divided into four levels, the first level
Network handles input picture, and the second hierarchical network and third hierarchical network refine candidate region twice, pass through
Frame, which returns, obtains human face region, and the 4th hierarchical network carries out crucial point location, since the network inputs of first three level are with model
Size is sequentially increased, and being done for task is also more and more careful, therefore is more advantageous to cascade structure processing data, is maximized
For detection efficiency with accuracy rate, detection speed is fast, and recall rate is high, false drop rate is low.
Requirement based on the human face detection tech of video flowing to timeliness in practical application scene is relatively high, therefore detects
Speed becomes the important indicator for weighing an algorithm model quality.Preferably, the embodiment of the present invention is in the secondary refinement step
After S104, can also include:Face tracking step S105, the human face region of present frame is amplified, by amplified face
Third hierarchical network is sent into region, and the face tracking for carrying out next frame is returned by frame.The advantage of doing so is that in recognition of face
Face tracking algorithm, very big optimizing detection speed are added later.Using Face datection model third hierarchical network to needing to track
Target frame make frame and return to achieve the effect that robust tracking, advantage is that tracking velocity is very fast, very accurate.By inventor's
Many experiments show that this method is highly effective.Specific method can be:Directly the human face region of previous frame is amplified, then
Secondary input third hierarchical network returns task to carry out the tracking of face, to return the position of frame by the frame of third hierarchical network
Set the tracking result as the human face region.
Preferably, the embodiment of the present invention can also include before the face tracking step S105:Markers step, label
The ID of each human face region.The advantage of doing so is that the ID of each target is marked, using Face datection model third level net
The target frame corresponding with each ID that network tracks needs does frame recurrence.
The embodiment of the present invention does not limit the mode for handling input picture, it is preferred that the preliminary screening step S102
In, input picture is handled using full volume machine network in the first hierarchical network.The embodiment of the present invention can use one
Full volume machine network quickly handles entire input picture, obtains the frame regression vector group of candidate frame and these candidate frames, the
One hierarchical network is very small, and preliminary screening operation can be quickly carried out very much to candidate frame by rolling up machine operation entirely.
Preferably, the embodiment of the present invention can be applied to the occasions such as access control and attendance, meeting signature and member management.
In above-mentioned specific embodiment one, the dynamic human face recognition methods based on video flowing is provided, is corresponded
, the application also provides the dynamic human face identification device based on video flowing.Implement since device embodiment is substantially similar to method
Example, so describing fairly simple, the relevent part can refer to the partial explaination of embodiments of method.Device described below is implemented
Example is only schematical.
Specific embodiment two
As shown in Fig. 2, an embodiment of the present invention provides a kind of dynamic human face identification device based on video flowing, including:
Image collection module 201, for obtaining a frame of video stream data as input picture;
Preliminary screening module 202, for being handled input picture in the first hierarchical network, obtain candidate region and
Its frame regression vector group, is assessed and is calibrated to the candidate vector in frame regression vector group, then is gone by non-maximization inhibition
Except the candidate region repeated;
Refinement module 203, candidate region for obtaining preliminary screening module 202 be sent into the second hierarchical network into
Micronization processes of row, exclude undesirable candidate region, execute calibration by frame recurrence, non-maximization is recycled to inhibit
Carry out the merging of candidate region;
Secondary refinement module 204, candidate region for obtaining a refinement module 203 be sent into third hierarchical network into
The secondary micronization processes of row, exclude undesirable candidate region, execute calibration by frame recurrence, non-maximization is recycled to inhibit
The merging of candidate region is carried out, human face region is obtained;
Key point locating module 206, the human face region for obtaining secondary refinement module 204 are sent into the 4th level net
Network carries out crucial point location.
The embodiment of the present invention uses multitask cascade neural network, entire algorithm network to be divided into four levels, the first level
Network handles input picture, and the second hierarchical network and third hierarchical network refine candidate region twice, pass through
Frame, which returns, obtains human face region, and the 4th hierarchical network carries out crucial point location, since the network inputs of first three level are with model
Size is sequentially increased, and being done for task is also more and more careful, therefore is more advantageous to cascade structure processing data, is maximized
For detection efficiency with accuracy rate, detection speed is fast, and recall rate is high, false drop rate is low.
Preferably, the embodiment of the present invention can also include:Face tracking module 205 is used for the human face region of present frame
It is amplified, amplified human face region is sent into third hierarchical network, the face tracking for carrying out next frame is returned by frame.
Preferably, the embodiment of the present invention can also include:Mark module, the ID for marking each human face region.
Preferably, the preliminary screening module 202 is used in the first hierarchical network scheme input using full volume machine network
As being handled.
It will be apparent to those skilled in the art that technical solution that can be as described above and design, make various other
Corresponding change and deformation, and all these changes and deformation should all belong to the protection domain of the claims in the present invention
Within.
Claims (8)
1. a kind of dynamic human face recognition methods based on video flowing, which is characterized in that including:
Image acquisition step obtains a frame of video stream data as input picture;
Preliminary screening step is handled input picture in the first hierarchical network, obtain candidate region and its frame return to
Amount group is assessed and is calibrated to the candidate vector in frame regression vector group, then inhibits the time that removal repeats by non-maximization
Favored area;
Refinement step, the candidate region that preliminary screening step is obtained are sent into the second hierarchical network and are carried out at primary refinement
Reason, excludes undesirable candidate region, executes calibration by frame recurrence, non-maximization is recycled to inhibit to carry out candidate region
Merging;
Secondary refinement step, the candidate region that a refinement step is obtained are sent into third hierarchical network and are carried out at secondary refinement
Reason, excludes undesirable candidate region, executes calibration by frame recurrence, non-maximization is recycled to inhibit to carry out candidate region
Merging, obtain human face region;
Key point positioning step, the human face region that secondary refinement step is obtained are sent into the 4th hierarchical network, and it is fixed to carry out key point
Position.
2. the dynamic human face recognition methods according to claim 1 based on video flowing, which is characterized in that the secondary refinement
After step, further include:
The human face region of present frame is amplified by face tracking step, and amplified human face region is sent into third level net
Network returns the face tracking for carrying out next frame by frame.
3. the dynamic human face recognition methods according to claim 2 based on video flowing, which is characterized in that the face tracking
Before step, further include:
Markers step marks the ID of each human face region.
4. the dynamic human face recognition methods according to claim 1 or 2 based on video flowing, which is characterized in that described preliminary
In screening step, input picture is handled using full volume machine network in the first hierarchical network.
5. a kind of dynamic human face identification device based on video flowing, which is characterized in that including:
Image collection module, for obtaining a frame of video stream data as input picture;
Preliminary screening module obtains candidate region and its frame returns for being handled input picture in the first hierarchical network
Return Vector Groups, the candidate vector in frame regression vector group is assessed and calibrated, then inhibits removal to repeat by non-maximization
Candidate region;
Refinement module, the candidate region for obtaining preliminary screening module are sent into the second hierarchical network and are once refined
Processing, excludes undesirable candidate region, executes calibration by frame recurrence, non-maximization is recycled to inhibit to carry out candidate regions
The merging in domain;
Secondary refinement module, the candidate region for obtaining a refinement module are sent into third hierarchical network and carry out secondary refinement
Processing, excludes undesirable candidate region, executes calibration by frame recurrence, non-maximization is recycled to inhibit to carry out candidate regions
The merging in domain obtains human face region;
Key point locating module, the human face region for obtaining secondary refinement module are sent into the 4th hierarchical network, carry out crucial
Point location.
6. the dynamic human face identification device according to claim 5 based on video flowing, which is characterized in that further include:
Amplified human face region is sent into third layer by face tracking module for being amplified the human face region of present frame
Grade network, the face tracking for carrying out next frame is returned by frame.
7. the dynamic human face identification device according to claim 6 based on video flowing, which is characterized in that further include:
Mark module, the ID for marking each human face region.
8. the dynamic human face identification device according to claim 5 or 6 based on video flowing, which is characterized in that described preliminary
Screening module is used in the first hierarchical network handle input picture using full volume machine network.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711466486.8A CN108288280A (en) | 2017-12-28 | 2017-12-28 | Dynamic human face recognition methods based on video flowing and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711466486.8A CN108288280A (en) | 2017-12-28 | 2017-12-28 | Dynamic human face recognition methods based on video flowing and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108288280A true CN108288280A (en) | 2018-07-17 |
Family
ID=62832438
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711466486.8A Pending CN108288280A (en) | 2017-12-28 | 2017-12-28 | Dynamic human face recognition methods based on video flowing and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108288280A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108960206A (en) * | 2018-08-07 | 2018-12-07 | 北京字节跳动网络技术有限公司 | Video frame treating method and apparatus |
CN109214303A (en) * | 2018-08-14 | 2019-01-15 | 北京工商大学 | A kind of multithreading dynamic human face based on cloud API is registered method |
CN109446892A (en) * | 2018-09-14 | 2019-03-08 | 杭州宇泛智能科技有限公司 | Human eye notice positioning method and system based on deep neural network |
CN109948450A (en) * | 2019-02-22 | 2019-06-28 | 深兰科技(上海)有限公司 | A kind of user behavior detection method, device and storage medium based on image |
CN110059645A (en) * | 2019-04-23 | 2019-07-26 | 杭州智趣智能信息技术有限公司 | A kind of face identification method, system and electronic equipment and storage medium |
CN110969189A (en) * | 2019-11-06 | 2020-04-07 | 杭州宇泛智能科技有限公司 | Face detection method and device and electronic equipment |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107358223A (en) * | 2017-08-16 | 2017-11-17 | 上海荷福人工智能科技(集团)有限公司 | A kind of Face datection and face alignment method based on yolo |
CN107506707A (en) * | 2016-11-30 | 2017-12-22 | 奥瞳系统科技有限公司 | Using the Face datection of the small-scale convolutional neural networks module in embedded system |
-
2017
- 2017-12-28 CN CN201711466486.8A patent/CN108288280A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107506707A (en) * | 2016-11-30 | 2017-12-22 | 奥瞳系统科技有限公司 | Using the Face datection of the small-scale convolutional neural networks module in embedded system |
CN107358223A (en) * | 2017-08-16 | 2017-11-17 | 上海荷福人工智能科技(集团)有限公司 | A kind of Face datection and face alignment method based on yolo |
Non-Patent Citations (4)
Title |
---|
CHRISTOPH FEICHTENHOFER ET AL.: "Detect to Track and Track to Detect", 《2017 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV)》 * |
KAIPENG ZHANG ET AL.: "Joint Face Detection and Alignment Using Multitask Cascaded Convolutional Networks", 《IEEE SIGNAL PROCESSING LETTERS》 * |
NIAN CAI ET AL.: "Detecting facial landmarks in the video based on a hybrid framework", 《HTTPS://ARXIV.ORG/ABS/1609.06441》 * |
徐光柱 等: "《实用性目标检测与跟踪算法原理与应用》", 30 June 2015, 国防工业出版社 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108960206A (en) * | 2018-08-07 | 2018-12-07 | 北京字节跳动网络技术有限公司 | Video frame treating method and apparatus |
CN109214303A (en) * | 2018-08-14 | 2019-01-15 | 北京工商大学 | A kind of multithreading dynamic human face based on cloud API is registered method |
CN109214303B (en) * | 2018-08-14 | 2021-10-01 | 北京工商大学 | Multithreading dynamic face sign-in method based on cloud API |
CN109446892A (en) * | 2018-09-14 | 2019-03-08 | 杭州宇泛智能科技有限公司 | Human eye notice positioning method and system based on deep neural network |
CN109446892B (en) * | 2018-09-14 | 2023-03-24 | 杭州宇泛智能科技有限公司 | Human eye attention positioning method and system based on deep neural network |
CN109948450A (en) * | 2019-02-22 | 2019-06-28 | 深兰科技(上海)有限公司 | A kind of user behavior detection method, device and storage medium based on image |
CN110059645A (en) * | 2019-04-23 | 2019-07-26 | 杭州智趣智能信息技术有限公司 | A kind of face identification method, system and electronic equipment and storage medium |
CN110969189A (en) * | 2019-11-06 | 2020-04-07 | 杭州宇泛智能科技有限公司 | Face detection method and device and electronic equipment |
CN110969189B (en) * | 2019-11-06 | 2023-07-25 | 杭州宇泛智能科技有限公司 | Face detection method and device and electronic equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108288280A (en) | Dynamic human face recognition methods based on video flowing and device | |
WO2022252398A1 (en) | Ship trajectory feature point extraction-based spatio-temporal dp method | |
CN105405154B (en) | Target object tracking based on color-structure feature | |
Gao et al. | Trustful internet of surveillance things based on deeply represented visual co-saliency detection | |
CN106125087A (en) | Dancing Robot indoor based on laser radar pedestrian tracting method | |
CN109886245A (en) | A kind of pedestrian detection recognition methods based on deep learning cascade neural network | |
CN101567046B (en) | Target recognition method of unmanned aerial vehicle based on minimum circle-cover matching | |
JP6453488B2 (en) | Statistical method and apparatus for passersby based on identification of human head top | |
CN105894542A (en) | Online target tracking method and apparatus | |
CN105931269A (en) | Tracking method for target in video and tracking device thereof | |
CN104680187B (en) | Naval vessel track Hierarchical Cluster Analysis based on buffering area similarity measurement | |
CN107993250A (en) | A kind of fast multi-target pedestrian tracking and analysis method and its intelligent apparatus | |
CN106156750A (en) | A kind of based on convolutional neural networks to scheme to search car method | |
CN104835175A (en) | Visual attention mechanism-based method for detecting target in nuclear environment | |
CN103020614B (en) | Based on the human motion identification method that space-time interest points detects | |
CN103440667A (en) | Automatic device for stably tracing moving targets under shielding states | |
Liu et al. | Extended faster R-CNN for long distance human detection: Finding pedestrians in UAV images | |
CN108805902A (en) | A kind of space-time contextual target tracking of adaptive scale | |
CN109857097A (en) | A kind of monitoring method and system of road surface roller compaction process | |
WO2023201938A1 (en) | Missing trajectory filling method and system | |
CN110531618A (en) | Closed loop based on effective key frame detects robot self-localization error cancelling method | |
CN103945227B (en) | Video semantic block partition method based on light stream clustering | |
Wu | Monitoring system of key technical features of male tennis players based on internet of things security technology | |
Yang et al. | An improved helmet detection algorithm based on yolo v4 | |
CN108882152A (en) | A kind of privacy of user guard method reported based on Path selection |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180717 |