CN106845432A - The method and apparatus that a kind of face is detected jointly with human body - Google Patents
The method and apparatus that a kind of face is detected jointly with human body Download PDFInfo
- Publication number
- CN106845432A CN106845432A CN201710067572.5A CN201710067572A CN106845432A CN 106845432 A CN106845432 A CN 106845432A CN 201710067572 A CN201710067572 A CN 201710067572A CN 106845432 A CN106845432 A CN 106845432A
- Authority
- CN
- China
- Prior art keywords
- face
- frame
- information
- loss function
- pedestrian
- 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.)
- Granted
Links
Classifications
-
- 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
- G06V40/161—Detection; Localisation; Normalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/94—Hardware or software architectures specially adapted for image or video understanding
-
- 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
- G06V40/103—Static body considered as a whole, e.g. static pedestrian or occupant recognition
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Human Computer Interaction (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Software Systems (AREA)
- Traffic Control Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses the method and apparatus that a kind of face and human body are detected jointly, wherein, the method includes:Obtain normal data;Wherein, the position frame information of the position with human body of the face for marking each pedestrian is included in normal data;Common identification model is modified by the position frame information in the normal data;Wherein, the common identification model is generated based on Faster RCNN;The common detection of face and human body is carried out to image in video to be identified based on revised common identification model, to export real-time structured message;Include the face information and human body information of each pedestrian in wherein described structured message simultaneously.Recognized while realizing efficient with this, saved resource, and improve practicality and ensure that the corresponding relation of face and human body.
Description
Technical field
The present invention relates to recognize field, the method and apparatus that more particularly to a kind of face is detected jointly with human body.
Background technology
In the prior art, it is individually to carry out to carry out recognition of face and human bioequivalence, and simple target type can only be entered
Row detection, if carrying out the independent multiple models of detection needs to the target of polymorphic type, for example, work as to need to detect face and detection
Pedestrian, that is accomplished by a human-face detector and a human body detector, more moneys can be taken using two sets of independent detectors
Source.
In addition, the detection of target and the correction of testing result are separately carried out, it is necessary to what is repeated carries out category filter candidate
Frame and correction candidate frame position and size, exist it is substantial amounts of compute repeatedly, making the practicality of system reduces;
It is follow-up in real time information structurizing process, independent detection to face and human body lack corresponding relation, but
, it is necessary to face and human body are all corresponded to some specific individual goal in actual demand, based on existing Face datection
Done with human detection result, it is necessary to expend extra calculating to be matched.
The content of the invention
For defect of the prior art, the present invention proposes the method and apparatus that a kind of face is detected jointly with human body,
It is used to overcome defect of the prior art.
The present invention proposes embodiment in detail below:
The embodiment of the present invention proposes a kind of method that face is detected jointly with human body, including:
Obtain normal data;Wherein, the position of the position with human body of the face for marking each pedestrian is included in normal data
Put frame information;
Common identification model is modified by the position frame information in the normal data;Wherein, the common knowledge
Other model is generated based on Faster RCNN;
The common detection of face and human body is carried out to image in video to be identified based on revised common identification model, with
Export real-time structured message;Include the face information and human body information of each pedestrian in wherein described structured message simultaneously.
In a specific embodiment, the acquisition normal data, including:
Obtain the monitor video under different scenes;
Each pedestrian in the image in monitor video is identified;
For identify each pedestrian carry out face where position and human body position be labeled as different position frames,
To generate position frame information;
Integrate recognized image and frame information in position therein generation normal data.
In a specific embodiment, the position frame information by the normal data is to common identification model
It is modified, including:
The parameter of the common identification model of step A, random initializtion;Wherein, the parameter includes:The ginseng of correspondence loss function
Number;The loss function is used to be modified common identification model;
Step B, the position frame information in the normal data is passed sequentially through the common identification mould after being initialized
Type is calculated, to obtain the loss of each loss function of correspondence;
Step C, gradient is obtained by carrying out derivation to each loss function, and reversely passed by chain rule
Broadcast, each parameter after being updated;
Step D, repeat step B and step C, until loss no longer declines, to obtain final parameter;
Step E, the amendment by final parameter completion to common identification model.
In a specific embodiment, the loss function includes:Classification Loss function, position return loss function,
Relative position constraint loss function;
Wherein, the Classification Loss function, the classification for carrying out face/pedestrian/background to the candidate frame in image, with
Face frame and pedestrian's frame are obtained, and gets rid of the candidate frame in background;
The position returns loss function, and position and size for correcting face frame and pedestrian's frame improve the standard of positioning
Exactness;
The relative position constraint loss function, the relative position relation for ensureing face frame and pedestrian's frame is normal.
It is described image in video to be identified is entered based on revised common identification model in a specific embodiment
The common detection of pedestrian's face and human body, to export real-time structured message, including:
Video to be identified is decoded, to obtain the every two field picture in video to be identified;
If the disposal ability of the common identification module exceedes preset value, face is carried out with human body for every two field picture
Common detection, the structured message of the face information, human body information and satellite information of each pedestrian is included with generation in real time simultaneously;
Wherein, the satellite information includes:The point position information of the corresponding camera of video to be identified, the temporal information of video to be identified.
The embodiment of the present invention also proposed the equipment that a kind of face is detected jointly with human body, including:
Acquisition module, for obtaining normal data;Wherein, the face and people for marking each pedestrian is included in normal data
The position frame information of the position of body;
Correcting module, for being modified to common identification model by the position frame information in the normal data;Its
In, the common identification model is generated based on Faster RCNN;
Detection module, for carrying out face and human body to image in video to be identified based on revised common identification model
Common detection, to export real-time structured message;Include that the face of each pedestrian is believed in wherein described structured message simultaneously
Breath and human body information.
In a specific embodiment, the acquisition module is used for:
Obtain the monitor video under different scenes;
Each pedestrian in the image in monitor video is identified;
For identify each pedestrian carry out face where position and human body position be labeled as different position frames,
To generate position frame information;
Integrate recognized image and frame information in position therein generation normal data.
In a specific embodiment, the correcting module, for performing operations described below:
The parameter of the common identification model of step A, random initializtion;Wherein, the parameter includes:The ginseng of correspondence loss function
Number;The loss function is used to be modified common identification model;
Step B, the position frame information in the normal data is passed sequentially through the common identification mould after being initialized
Type is calculated, to obtain the loss of each loss function of correspondence;
Step C, gradient is obtained by carrying out derivation to each loss function, and reversely passed by chain rule
Broadcast, each parameter after being updated;
Step D, repeat step B and step C, until loss no longer declines, to obtain final parameter;
Step E, the amendment by final parameter completion to common identification model.
In a specific embodiment, the loss function includes:Classification Loss function, position return loss function,
Relative position constraint loss function;
Wherein, the Classification Loss function, the classification for carrying out face/pedestrian/background to the candidate frame in image, with
Face frame and pedestrian's frame are obtained, and gets rid of the candidate frame in background;
The position returns loss function, and position and size for correcting face frame and pedestrian's frame improve the standard of positioning
Exactness;
The relative position constraint loss function, the relative position relation for ensureing face frame and pedestrian's frame is normal.
In a specific embodiment, the detection module is used for:
Video to be identified is decoded, to obtain the every two field picture in video to be identified;
If the disposal ability of the common identification module exceedes preset value, face is carried out with human body for every two field picture
Common detection, the structured message of the face information, human body information and satellite information of each pedestrian is included with generation in real time simultaneously;
Wherein, the satellite information includes:The point position information of the corresponding camera of video to be identified, the temporal information of video to be identified.
With this, the invention discloses the method and apparatus that a kind of face and human body are detected jointly, wherein, the method includes:
Obtain normal data;Wherein, the position frame information of the position with human body of the face for marking each pedestrian is included in normal data;
Common identification model is modified by the position frame information in the normal data;Wherein, the common identification model is
Based on Faster RCNN generations;Face and people are carried out to image in video to be identified based on revised common identification model
The common detection of body, to export real-time structured message;Include the face of each pedestrian in wherein described structured message simultaneously
Information and human body information.Recognized while realizing efficient with this, saved resource, and improve practicality and ensure that people
The corresponding relation of face and human body.
Brief description of the drawings
Technical scheme in order to illustrate more clearly the embodiments of the present invention, below will be attached to what is used needed for embodiment
Figure is briefly described, it will be appreciated that the following drawings illustrate only certain embodiments of the present invention, thus be not construed as it is right
The restriction of scope, for those of ordinary skill in the art, on the premise of not paying creative work, can also be according to this
A little accompanying drawings obtain other related accompanying drawings.
Fig. 1 is the schematic flow sheet of the method that a kind of face that the embodiment of the present invention is proposed is detected jointly with human body;
Fig. 2 is the signal of calibration position frame in the method that a kind of face that the embodiment of the present invention is proposed is detected jointly with human body
Figure;
Fig. 3 is showing for the structured message in the method that a kind of face that the embodiment of the present invention is proposed is detected jointly with human body
It is intended to;
Fig. 4 is the structural representation of the equipment that a kind of face that the embodiment of the present invention is proposed is detected jointly with human body.
Specific embodiment
Hereinafter, the various embodiments of the disclosure will be described more fully.The disclosure can have various embodiments, and
Can wherein adjust and change.It should be understood, however, that:It is limited to spy disclosed herein in the absence of by the various embodiments of the disclosure
Determine the intention of embodiment, but in the spirit and scope that should be interpreted as cover the various embodiments for falling into the disclosure by the disclosure
All adjustment, equivalent and/or alternative.
Hereinafter, can be used in the various embodiments of the disclosure term " including " or " may include " indicate it is disclosed
Function, operation or the presence of element, and do not limit the increase of one or more functions, operation or element.Additionally, such as existing
Used in the various embodiments of the disclosure, term " including ", " having " and its cognate be meant only to represent special characteristic, number
The combination of word, step, operation, element, component or foregoing item, and be understood not to exclude first it is one or more other
The presence of the combination of feature, numeral, step, operation, element, component or foregoing item or increase one or more features, numeral,
The possibility of the combination of step, operation, element, component or foregoing item.
In the various embodiments of the disclosure, statement "or" or " at least one of A or/and B " include what is listed file names with
Any combinations of word or all combinations.For example, statement " A or B " or " at least one of A or/and B " may include A, may include
B may include A and B both.
The statement (" first ", " second " etc.) used in the various embodiments of the disclosure can be modified in various implementations
Various element in example, but corresponding element can not be limited.For example, presented above be not intended to limit the suitable of the element
Sequence and/or importance.The purpose for being only used for differentiating an element and other elements presented above.For example, first user is filled
Put and indicate different user device with second user device, although the two is all user's set.For example, not departing from each of the disclosure
In the case of planting the scope of embodiment, the first element is referred to alternatively as the second element, and similarly, the second element is also referred to as first
Element.
It should be noted that:If an element ' attach ' to another element by description, can be by the first composition unit
Part is directly connected to the second element, and " connection " the 3rd can be constituted between the first element and the second element
Element.On the contrary, when an element " being directly connected to " is arrived into another element, it will be appreciated that be in the first element
And second do not exist the 3rd element between element.
The term " user " used in the various embodiments of the disclosure may indicate that the people that uses electronic installation or use electricity
The device (for example, artificial intelligence electronic installation) of sub-device.
The term used in the various embodiments of the disclosure is only used for describing the purpose of specific embodiment and not anticipating
In the various embodiments of the limitation disclosure.As used herein, singulative is intended to also including plural form, unless context is clear
Chu ground is indicated otherwise.Unless otherwise defined, all terms (including the technical term and scientific terminology) tool being otherwise used herein
There is the implication identical implication being generally understood that with the various embodiment one skilled in the art of the disclosure.The term
(term limited such as in the dictionary for generally using) is to be interpreted as to be had and the situational meaning in correlative technology field
Identical implication and will be not construed as with Utopian implication or excessively formal implication, unless in the various of the disclosure
It is clearly defined in embodiment.
Embodiment 1
The embodiment of the invention discloses a kind of method that face and human body are detected jointly, as shown in figure 1, including:
Step 101, acquisition normal data;Wherein, include the face that marks each pedestrian in normal data with human body
The position frame information of position;
Step 102, common identification model is modified by the position frame information in the normal data;Wherein, institute
Common identification model is stated to be generated based on Faster RCNN;
Step 103, face to image in video to be identified is carried out based on revised common identification model and human body is total to
With detection, to export real-time structured message;In wherein described structured message simultaneously include each pedestrian face information and
Human body information.
In a specific embodiment, normal data is obtained described in step 101, including:
Obtain the monitor video under different scenes;
Each pedestrian in the image in monitor video is identified;
For identify each pedestrian carry out face where position and human body position be labeled as different position frames,
To generate position frame information;
Integrate recognized image and frame information in position therein generation normal data.
Specifically, collecting data, the monitor video comprising pedestrian is (such as such as Fig. 2 examples under can collecting different scenes
Scape), in order to lift the stability across scene performance of detection algorithm, the data under more scenes are collected as far as possible;And one
(specifically, can individually be marked using mode of the prior art) is marked in frame picture and goes out human body and correspondence face
Position frame (as shown in frame different in Fig. 2), but in order to avoid repeat demarcate, reply every section of video demarcate (example every frame
Such as took a frame every 3 seconds to demarcate).
In a specific embodiment, the position frame information pair by the normal data in step 102
Common identification model is modified, including:
The parameter of the common identification model of step A, random initializtion;Wherein, the parameter includes:The ginseng of correspondence loss function
Number;The loss function is used to be modified common identification model;
Step B, the position frame information in the normal data is passed sequentially through the common identification mould after being initialized
Type is calculated, to obtain the loss of each loss function of correspondence;
Step C, gradient is obtained by carrying out derivation to each loss function, and reversely passed by chain rule
Broadcast, each parameter after being updated;
Step D, repeat step B and step C, until loss no longer declines, to obtain final parameter;
Step E, the amendment by final parameter completion to common identification model.
In the particular embodiment, the loss function includes:Classification Loss function, position return loss function, relative
Position constraint loss function;
Wherein, the Classification Loss function, the classification for carrying out face/pedestrian/background to the candidate frame in image, with
Face frame and pedestrian's frame are obtained, and gets rid of the candidate frame in background;
The position returns loss function, and position and size for correcting face frame and pedestrian's frame improve the standard of positioning
Exactness;
The relative position constraint loss function, the relative position relation for ensureing face frame and pedestrian's frame is normal.
Classification Loss function, mainly carries out the classification of face/pedestrian/background to candidate frame, obtains face frame and pedestrian's frame,
Get rid of the candidate frame in background;And position returns loss function, position and size for correcting face frame and pedestrian's frame make
It is more accurate to position;Relative position constraint loss function, statistics mark picture in face relative to human body position, in human body
Inframe determines an anchor point for face location, and the European side-play amount of face prediction block and human body anchor point position is calculated during training, with
This is used as constraint face frame and the loss function of the relative position relation of pedestrian's frame so that face frame and pedestrian's frame are not in not
Normal relative position relation.
In a specific embodiment, described in step 103 is based on revised common identification model to be identified
Image carries out the common detection of face and human body in video, to export real-time structured message, including:
Video to be identified is decoded, to obtain the every two field picture in video to be identified;
If the disposal ability of the common identification module exceedes preset value, face is carried out with human body for every two field picture
Common detection, the structured message of the face information, human body information and satellite information of each pedestrian is included with generation in real time simultaneously;
Wherein, the satellite information includes:The point position information of the corresponding camera of video to be identified, the temporal information of video to be identified.
Specifically, when common identification model is trained, sample is sent into mould by the parameter of random initializtion model first one by one
Type is calculated, and obtains each loss of each loss function, and the target of training is to minimize every loss, therefore, it is asked
Lead and obtain gradient, backpropagation is carried out using chain rule, update model parameter;This process is repeated to each sample, until testing
The loss for demonstrate,proving collection no longer declines, namely reaches minimum value.
Follow-up to obtain image by decoding when live video stream is accessed, the algorithm performance according to common identification model is determined
When detect, if algorithm speed meet, can detect frame by frame, by technical scheme to a frame treatment be once
The detection to face and human body can be simultaneously completed, specific testing result is as shown in Figure 3.
With this, this programme realizes following technique effect:
1st, while completing the detection and the correction of testing result to target, it is to avoid largely compute repeatedly, raising efficiency;
2nd, polymorphic type target is integrated, with a unified model, (framework i.e. using Faster RCNN is simultaneously complete
Into the detection and correction of target frame) target detection of multiple types is completed, reduce the occupancy of computing redundancy and system resource;
3rd, the association of structured message (face and human body) is automatically performed when real-time detection, complete knot is directly obtained
Structure information.
Embodiment 2
The embodiment of the invention also discloses the equipment that a kind of face and human body are detected jointly, as illustrated, including:
Acquisition module 201, for obtaining normal data;Wherein, the face that marks each pedestrian is included in normal data
With the position frame information of the position of human body;
Correcting module 202, for being modified to common identification model by the position frame information in the normal data;
Wherein, the common identification model is generated based on Faster RCNN;
Detection module 203, for based on revised common identification model image in video to be identified is carried out face with
The common detection of human body, to export real-time structured message;Include the people of each pedestrian in wherein described structured message simultaneously
Face information and human body information.
In a specific embodiment, the acquisition module 201 is used for:
Obtain the monitor video under different scenes;
Each pedestrian in the image in monitor video is identified;
For identify each pedestrian carry out face where position and human body position be labeled as different position frames,
To generate position frame information;
Integrate recognized image and frame information in position therein generation normal data.
In a specific embodiment, the correcting module 202, for performing operations described below:
The parameter of the common identification model of step A, random initializtion;Wherein, the parameter includes:The ginseng of correspondence loss function
Number;The loss function is used to be modified common identification model;
Step B, the position frame information in the normal data is passed sequentially through the common identification mould after being initialized
Type is calculated, to obtain the loss of each loss function of correspondence;
Step C, gradient is obtained by carrying out derivation to each loss function, and reversely passed by chain rule
Broadcast, each parameter after being updated;
Step D, repeat step B and step C, until loss no longer declines, to obtain final parameter;
Step E, the amendment by final parameter completion to common identification model.
In a specific embodiment, the loss function includes:Classification Loss function, position return loss function,
Relative position constraint loss function;
Wherein, the Classification Loss function, the classification for carrying out face/pedestrian/background to the candidate frame in image, with
Face frame and pedestrian's frame are obtained, and gets rid of the candidate frame in background;
The position returns loss function, and position and size for correcting face frame and pedestrian's frame improve the standard of positioning
Exactness;
The relative position constraint loss function, the relative position relation for ensureing face frame and pedestrian's frame is normal.
In a specific embodiment, the detection module 203 is used for:
Video to be identified is decoded, to obtain the every two field picture in video to be identified;
If the disposal ability of the common identification module exceedes preset value, face is carried out with human body for every two field picture
Common detection, the structured message of the face information, human body information and satellite information of each pedestrian is included with generation in real time simultaneously;
Wherein, the satellite information includes:The point position information of the corresponding camera of video to be identified, the temporal information of video to be identified.
With this, the invention discloses the method and apparatus that a kind of face and human body are detected jointly, wherein, the method includes:
Obtain normal data;Wherein, the position frame information of the position with human body of the face for marking each pedestrian is included in normal data;
Common identification model is modified by the position frame information in the normal data;Wherein, the common identification model is
Based on Faster RCNN generations;Face and people are carried out to image in video to be identified based on revised common identification model
The common detection of body, to export real-time structured message;Include the face of each pedestrian in wherein described structured message simultaneously
Information and human body information.Recognized while realizing efficient with this, saved resource, and improve practicality and ensure that people
The corresponding relation of face and human body.
It will be appreciated by those skilled in the art that accompanying drawing is a schematic diagram for being preferable to carry out scene, module in accompanying drawing or
Flow is not necessarily implemented necessary to the present invention.
It will be appreciated by those skilled in the art that module in device in implement scene can according to implement scene describe into
Row is distributed in the device of implement scene, it is also possible to carry out one or more dresses that respective change is disposed other than this implement scene
In putting.The module of above-mentioned implement scene can merge into a module, it is also possible to be further split into multiple submodule.
The invention described above sequence number is for illustration only, and the quality of implement scene is not represented.
Disclosed above is only several specific implementation scenes of the invention, but, the present invention is not limited to this, Ren Heben
What the technical staff in field can think change should all fall into protection scope of the present invention.
Claims (10)
1. a kind of method that face is detected jointly with human body, it is characterised in that including:
Obtain normal data;Wherein, the position frame of the position with human body of the face for marking each pedestrian is included in normal data
Information;
Common identification model is modified by the position frame information in the normal data;Wherein, the common identification mould
Type is generated based on Faster RCNN;
The common detection of face and human body is carried out to image in video to be identified based on revised common identification model, to export
Real-time structured message;Include the face information and human body information of each pedestrian in wherein described structured message simultaneously.
2. the method for claim 1, it is characterised in that the acquisition normal data, including:
Obtain the monitor video under different scenes;
Each pedestrian in the image in monitor video is identified;
For identify each pedestrian carry out face where position and human body position be labeled as different position frames, with life
Into position frame information;
Integrate recognized image and frame information in position therein generation normal data.
3. the method for claim 1, it is characterised in that the position frame information by the normal data is to altogether
It is modified with identification model, including:
The parameter of the common identification model of step A, random initializtion;Wherein, the parameter includes:The parameter of correspondence loss function;
The loss function is used to be modified common identification model;
Step B, the position frame information in the normal data is passed sequentially through into the common identification model after being initialized enter
Row is calculated, to obtain the loss of each loss function of correspondence;
Step C, gradient is obtained by carrying out derivation to each loss function, and backpropagation is carried out by chain rule, obtained
Each parameter after to renewal;
Step D, repeat step B and step C, until loss no longer declines, to obtain final parameter;
Step E, the amendment by final parameter completion to common identification model.
4. method as claimed in claim 3, it is characterised in that the loss function includes:Classification Loss function, position return
Loss function, relative position constraint loss function;
Wherein, the Classification Loss function, the classification for carrying out face/pedestrian/background to the candidate frame in image, to obtain
Face frame and pedestrian's frame, and get rid of the candidate frame in background;
The position returns loss function, and position and size for correcting face frame and pedestrian's frame improve the degree of accuracy of positioning;
The relative position constraint loss function, the relative position relation for ensureing face frame and pedestrian's frame is normal.
5. the method for claim 1, it is characterised in that described to be regarded to be identified based on revised common identification model
Image carries out the common detection of face and human body in frequency, to export real-time structured message, including:
Video to be identified is decoded, to obtain the every two field picture in video to be identified;
If the disposal ability of the common identification module exceedes preset value, face is carried out for every two field picture common with human body
Detection, the structured message of the face information, human body information and satellite information of each pedestrian is included with generation in real time simultaneously;Its
In, the satellite information includes:The point position information of the corresponding camera of video to be identified, the temporal information of video to be identified.
6. the equipment that a kind of face is detected jointly with human body, it is characterised in that including:
Acquisition module, for obtaining normal data;Wherein, include the face that marks each pedestrian in normal data with human body
The position frame information of position;
Correcting module, for being modified to common identification model by the position frame information in the normal data;Wherein, institute
Common identification model is stated to be generated based on Faster RCNN;
Detection module, is total to for carrying out face to image in video to be identified based on revised common identification model with human body
With detection, to export real-time structured message;In wherein described structured message simultaneously include each pedestrian face information and
Human body information.
7. equipment as claimed in claim 6, it is characterised in that the acquisition module, is used for:
Obtain the monitor video under different scenes;
Each pedestrian in the image in monitor video is identified;
For identify each pedestrian carry out face where position and human body position be labeled as different position frames, with life
Into position frame information;
Integrate recognized image and frame information in position therein generation normal data.
8. equipment as claimed in claim 6, it is characterised in that the correcting module, for performing operations described below:
The parameter of the common identification model of step A, random initializtion;Wherein, the parameter includes:The parameter of correspondence loss function;
The loss function is used to be modified common identification model;
Step B, the position frame information in the normal data is passed sequentially through into the common identification model after being initialized enter
Row is calculated, to obtain the loss of each loss function of correspondence;
Step C, gradient is obtained by carrying out derivation to each loss function, and backpropagation is carried out by chain rule, obtained
Each parameter after to renewal;
Step D, repeat step B and step C, until loss no longer declines, to obtain final parameter;
Step E, the amendment by final parameter completion to common identification model.
9. equipment as claimed in claim 8, it is characterised in that the loss function includes:Classification Loss function, position return
Loss function, relative position constraint loss function;
Wherein, the Classification Loss function, the classification for carrying out face/pedestrian/background to the candidate frame in image, to obtain
Face frame and pedestrian's frame, and get rid of the candidate frame in background;
The position returns loss function, and position and size for correcting face frame and pedestrian's frame improve the degree of accuracy of positioning;
The relative position constraint loss function, the relative position relation for ensureing face frame and pedestrian's frame is normal.
10. equipment as claimed in claim 6, it is characterised in that the detection module, is used for:
Video to be identified is decoded, to obtain the every two field picture in video to be identified;
If the disposal ability of the common identification module exceedes preset value, face is carried out for every two field picture common with human body
Detection, the structured message of the face information, human body information and satellite information of each pedestrian is included with generation in real time simultaneously;Its
In, the satellite information includes:The point position information of the corresponding camera of video to be identified, the temporal information of video to be identified.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710067572.5A CN106845432B (en) | 2017-02-07 | 2017-02-07 | A kind of method and apparatus that face detects jointly with human body |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710067572.5A CN106845432B (en) | 2017-02-07 | 2017-02-07 | A kind of method and apparatus that face detects jointly with human body |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106845432A true CN106845432A (en) | 2017-06-13 |
CN106845432B CN106845432B (en) | 2019-09-17 |
Family
ID=59121520
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710067572.5A Expired - Fee Related CN106845432B (en) | 2017-02-07 | 2017-02-07 | A kind of method and apparatus that face detects jointly with human body |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106845432B (en) |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108121951A (en) * | 2017-12-11 | 2018-06-05 | 北京小米移动软件有限公司 | Characteristic point positioning method and device |
CN108197605A (en) * | 2018-01-31 | 2018-06-22 | 电子科技大学 | Yak personal identification method based on deep learning |
CN108537165A (en) * | 2018-04-08 | 2018-09-14 | 百度在线网络技术(北京)有限公司 | Method and apparatus for determining information |
CN108921008A (en) * | 2018-05-14 | 2018-11-30 | 深圳市商汤科技有限公司 | Portrait identification method, device and electronic equipment |
CN109684956A (en) * | 2018-12-14 | 2019-04-26 | 深源恒际科技有限公司 | A kind of vehicle damage detection method and system based on deep neural network |
CN109740516A (en) * | 2018-12-29 | 2019-05-10 | 深圳市商汤科技有限公司 | A kind of user identification method, device, electronic equipment and storage medium |
CN109948494A (en) * | 2019-03-11 | 2019-06-28 | 深圳市商汤科技有限公司 | Image processing method and device, electronic equipment and storage medium |
CN110197113A (en) * | 2019-03-28 | 2019-09-03 | 杰创智能科技股份有限公司 | A kind of method for detecting human face of high-precision anchor point matching strategy |
CN110705469A (en) * | 2019-09-30 | 2020-01-17 | 重庆紫光华山智安科技有限公司 | Face matching method and device and server |
CN110889314A (en) * | 2018-09-10 | 2020-03-17 | 北京市商汤科技开发有限公司 | Image processing method, device, electronic equipment, server and system |
CN111027474A (en) * | 2019-12-09 | 2020-04-17 | Oppo广东移动通信有限公司 | Face area acquisition method and device, terminal equipment and storage medium |
CN111063083A (en) * | 2019-12-16 | 2020-04-24 | 腾讯科技(深圳)有限公司 | Access control method and device, computer readable storage medium and computer equipment |
CN112488057A (en) * | 2020-12-17 | 2021-03-12 | 北京航空航天大学 | Single-camera multi-target tracking method utilizing human head point positioning and joint point information |
CN113326773A (en) * | 2021-05-28 | 2021-08-31 | 北京百度网讯科技有限公司 | Recognition model training method, recognition method, device, equipment and storage medium |
CN114783043A (en) * | 2022-06-24 | 2022-07-22 | 杭州安果儿智能科技有限公司 | Child behavior track positioning method and system |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102609684A (en) * | 2012-01-16 | 2012-07-25 | 宁波江丰生物信息技术有限公司 | Human body posture detection method and device |
CN104899575A (en) * | 2015-06-19 | 2015-09-09 | 南京大学 | Human body assembly dividing method based on face detection and key point positioning |
-
2017
- 2017-02-07 CN CN201710067572.5A patent/CN106845432B/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102609684A (en) * | 2012-01-16 | 2012-07-25 | 宁波江丰生物信息技术有限公司 | Human body posture detection method and device |
CN102609684B (en) * | 2012-01-16 | 2013-12-18 | 宁波江丰生物信息技术有限公司 | Human body posture detection method and device |
CN104899575A (en) * | 2015-06-19 | 2015-09-09 | 南京大学 | Human body assembly dividing method based on face detection and key point positioning |
Non-Patent Citations (2)
Title |
---|
SHAOQING REN: "Faster R-CNN: Towards real-time object detection with region proposal networks", 《IEEE TRANS PATTERN ANAL MACH INTELL》 * |
XIAOFEI LI: "A Unified Framework for Concurrent Pedestrian and Cyclist Detection", 《IEEE INTELLIGENT TRANSPORTATION SYSTEMS SOCIETY》 * |
Cited By (25)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108121951A (en) * | 2017-12-11 | 2018-06-05 | 北京小米移动软件有限公司 | Characteristic point positioning method and device |
CN108197605A (en) * | 2018-01-31 | 2018-06-22 | 电子科技大学 | Yak personal identification method based on deep learning |
CN108537165A (en) * | 2018-04-08 | 2018-09-14 | 百度在线网络技术(北京)有限公司 | Method and apparatus for determining information |
CN108921008A (en) * | 2018-05-14 | 2018-11-30 | 深圳市商汤科技有限公司 | Portrait identification method, device and electronic equipment |
CN108921008B (en) * | 2018-05-14 | 2024-06-11 | 深圳市商汤科技有限公司 | Portrait identification method and device and electronic equipment |
CN110889314A (en) * | 2018-09-10 | 2020-03-17 | 北京市商汤科技开发有限公司 | Image processing method, device, electronic equipment, server and system |
CN109684956A (en) * | 2018-12-14 | 2019-04-26 | 深源恒际科技有限公司 | A kind of vehicle damage detection method and system based on deep neural network |
CN109740516A (en) * | 2018-12-29 | 2019-05-10 | 深圳市商汤科技有限公司 | A kind of user identification method, device, electronic equipment and storage medium |
CN109740516B (en) * | 2018-12-29 | 2021-05-14 | 深圳市商汤科技有限公司 | User identification method and device, electronic equipment and storage medium |
TWI702544B (en) * | 2019-03-11 | 2020-08-21 | 大陸商深圳市商湯科技有限公司 | Method, electronic device for image processing and computer readable storage medium thereof |
KR102446687B1 (en) | 2019-03-11 | 2022-09-23 | 선전 센스타임 테크놀로지 컴퍼니 리미티드 | Image processing method and apparatus, electronic device and storage medium |
WO2020181728A1 (en) * | 2019-03-11 | 2020-09-17 | 深圳市商汤科技有限公司 | Image processing method and apparatus, electronic device, and storage medium |
KR20200110642A (en) * | 2019-03-11 | 2020-09-24 | 선전 센스타임 테크놀로지 컴퍼니 리미티드 | Image processing method and device, electronic device and storage medium |
CN109948494B (en) * | 2019-03-11 | 2020-12-29 | 深圳市商汤科技有限公司 | Image processing method and device, electronic equipment and storage medium |
CN109948494A (en) * | 2019-03-11 | 2019-06-28 | 深圳市商汤科技有限公司 | Image processing method and device, electronic equipment and storage medium |
US11288531B2 (en) | 2019-03-11 | 2022-03-29 | Shenzhen Sensetime Technology Co., Ltd. | Image processing method and apparatus, electronic device, and storage medium |
CN110197113A (en) * | 2019-03-28 | 2019-09-03 | 杰创智能科技股份有限公司 | A kind of method for detecting human face of high-precision anchor point matching strategy |
CN110197113B (en) * | 2019-03-28 | 2021-06-04 | 杰创智能科技股份有限公司 | Face detection method of high-precision anchor point matching strategy |
CN110705469A (en) * | 2019-09-30 | 2020-01-17 | 重庆紫光华山智安科技有限公司 | Face matching method and device and server |
CN111027474A (en) * | 2019-12-09 | 2020-04-17 | Oppo广东移动通信有限公司 | Face area acquisition method and device, terminal equipment and storage medium |
CN111027474B (en) * | 2019-12-09 | 2024-03-15 | Oppo广东移动通信有限公司 | Face region acquisition method and device, terminal equipment and storage medium |
CN111063083A (en) * | 2019-12-16 | 2020-04-24 | 腾讯科技(深圳)有限公司 | Access control method and device, computer readable storage medium and computer equipment |
CN112488057A (en) * | 2020-12-17 | 2021-03-12 | 北京航空航天大学 | Single-camera multi-target tracking method utilizing human head point positioning and joint point information |
CN113326773A (en) * | 2021-05-28 | 2021-08-31 | 北京百度网讯科技有限公司 | Recognition model training method, recognition method, device, equipment and storage medium |
CN114783043A (en) * | 2022-06-24 | 2022-07-22 | 杭州安果儿智能科技有限公司 | Child behavior track positioning method and system |
Also Published As
Publication number | Publication date |
---|---|
CN106845432B (en) | 2019-09-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106845432A (en) | The method and apparatus that a kind of face is detected jointly with human body | |
CN111414887A (en) | Secondary detection mask face recognition method based on YO L OV3 algorithm | |
CN102982336B (en) | Model of cognition generates method and system | |
CN107194396A (en) | Method for early warning is recognized based on the specific architecture against regulations in land resources video monitoring system | |
CN110633610B (en) | Student state detection method based on YOLO | |
CN111582234B (en) | Large-scale oil tea tree forest fruit intelligent detection and counting method based on UAV and deep learning | |
CN106097346A (en) | A kind of video fire hazard detection method of self study | |
CN111126325A (en) | Intelligent personnel security identification statistical method based on video | |
CN110084165A (en) | The intelligent recognition and method for early warning of anomalous event under the open scene of power domain based on edge calculations | |
CN108154110A (en) | A kind of intensive people flow amount statistical method based on the detection of the deep learning number of people | |
CN110569772A (en) | Method for detecting state of personnel in swimming pool | |
CN105208325A (en) | Territorial resource monitoring and early warning method based on image fixed-point snapshot and comparative analysis | |
CN111914636A (en) | Method and device for detecting whether pedestrian wears safety helmet | |
CN114049325A (en) | Construction method and application of lightweight face mask wearing detection model | |
CN106612457B (en) | Video sequence alignment schemes and system | |
CN115951014A (en) | CNN-LSTM-BP multi-mode air pollutant prediction method combining meteorological features | |
CN111091110A (en) | Wearing identification method of reflective vest based on artificial intelligence | |
CN109389016A (en) | A kind of method and system that the number of people counts | |
IL257092A (en) | A method and system for tracking objects between cameras | |
CN110543828A (en) | Student attention analysis system based on wearable device and multi-mode intelligent analysis | |
CN117726991B (en) | High-altitude hanging basket safety belt detection method and terminal | |
CN117808374B (en) | Intelligent acceptance management method and system for building engineering quality | |
CN112686152A (en) | Crop pest and disease identification method with multi-size input and multi-size targets | |
CN102867214B (en) | Counting management method for people within area range | |
CN112417974A (en) | Public health monitoring method |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190917 Termination date: 20220207 |