CN109522790A - Human body attribute recognition approach, device, storage medium and electronic equipment - Google Patents
Human body attribute recognition approach, device, storage medium and electronic equipment Download PDFInfo
- Publication number
- CN109522790A CN109522790A CN201811168498.7A CN201811168498A CN109522790A CN 109522790 A CN109522790 A CN 109522790A CN 201811168498 A CN201811168498 A CN 201811168498A CN 109522790 A CN109522790 A CN 109522790A
- Authority
- CN
- China
- Prior art keywords
- human body
- image
- monitored
- layer
- identified
- 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
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of human body attribute recognition approach, device, storage medium and electronic equipment.Human body attribute recognition approach provided by the invention, it include: first to obtain image to be monitored, and the outline data information of human body to be identified is obtained from image to be monitored by preset human testing model, human testing frame is determined further according to outline data information, and the corresponding images to be recognized of human body to be identified is intercepted from image to be monitored using human testing frame, then feature extraction is carried out to images to be recognized by preset human body attribute Recognition Model, to obtain the attributive character information of the human body to be identified in images to be recognized.Human body attribute recognition approach provided by the invention may be implemented the function of carrying out identification screening to target person in security protection application scenarios, greatly accelerate recognition speed and accuracy, and reduce the investment of human cost.
Description
Technical field
The present invention relates to technical field of computer vision more particularly to a kind of human body attribute recognition approach, device, storage Jie
Matter and electronic equipment.
Background technique
With the continuous development of internet and artificial intelligence technology, more and more fields start to be related to computer vision certainly
Dynamicization calculates and analysis, wherein monitoring safety-security area is then wherein mostly important one of application scenarios.
In order to be identified using the monitoring video in security protection scene to the attribute of suspect, to realize to suspicion
The identification and tracking for doubting people, for example, public security organ often needs attributive character (such as gender, age bracket, the clothes according to suspect
Deng) reduce search range.In addition, can also be the scenes such as shop, museum, many businessmans in other application scenarios
It needs to count major customer group according to customer people's feature is carried out.And in the prior art, it is all to be known using eye-observation mostly
Others finds out target person at the mode of body attributive character from monitoring video.
But human body Attribute Recognition is carried out using traditional human eye observation, error is larger, and needs to expend a large amount of people
Power and material resources, it is difficult to adapt to characteristics of human body's identification in practical application scene.
Summary of the invention
The present invention provides a kind of human body attribute recognition approach, device, storage medium and electronic equipment, to be directed to figure to be monitored
As exporting the human testing frame of human body to be identified and the corresponding attributive character information of the human body to be identified, to realize in reality
The function of identification screening is carried out in border security protection application scenarios to target person.
In a first aspect, the present invention provides a kind of human body attribute recognition approach, comprising:
Image to be monitored is obtained, and people to be identified is obtained from the image to be monitored by preset human testing model
The outline data information of body；
Human testing frame is determined according to the outline data information, and utilizes the human testing frame from the figure to be monitored
The corresponding images to be recognized of the human body to be identified is intercepted as in；
Feature extraction is carried out to the images to be recognized by preset human body attribute Recognition Model, it is described wait know to obtain
The attributive character information of the human body to be identified in other image.
In a kind of possible design, it is described by preset human body attribute Recognition Model to the images to be recognized into
Row feature extraction, after the attributive character information to obtain the human body to be identified in the images to be recognized, further includes:
Human testing frame described in Overlapping display and the attributive character information on the image to be monitored, wherein institute
Human testing frame is stated for accommodating the human body to be identified in the image to be monitored.
In a kind of possible design, it is described obtained from the image to be monitored by preset human testing model to
Identify the outline data information of human body, comprising:
The processing of multiple convolution layer and the processing of pond layer are carried out by the image to be monitored, to form full articulamentum；
Classification and Identification is carried out to the full articulamentum using target detection network, to obtain the profile of the human body to be identified
Data information.
It is described that human testing frame is determined according to the outline data information in a kind of possible design, comprising:
Determine the top left corner apex of the human testing frame first in preset coordinate system according to the outline data information
Coordinate value, the first frame long value and the first frame height value, wherein the abscissa value in first coordinate value is less than or equal to described
Minimum abscissa value in outline data information, the ordinate value in first coordinate value be greater than or equal to the number of contours it is believed that
Maximum ordinate value in breath, the first frame long value be greater than or equal in the outline data information maximum abscissa value with it is described
The difference of minimum abscissa value, the first frame height value be greater than or equal to the outline data information in maximum ordinate value with it is described
The difference of minimum ordinate value.
In a kind of possible design, the preset human body attribute Recognition Model is tensor recurrent neural networks model.
In a kind of possible design, the tensor recurrent neural networks model includes input layer, the first convolutional layer, first
Correcting layer, the first pond layer, the second convolutional layer, the second correcting layer, the second pond layer, third convolutional layer, tensor recurrence layer and defeated
Layer out；
Wherein, the input layer, first convolutional layer, first correcting layer, first pond layer, described second
Convolutional layer, second correcting layer, second pond layer and the third convolutional layer are sequentially connected, the tensor recurrence layer
It is connected to the third convolutional layer entirely, the output layer is connected to the tensor recurrence layer entirely.
In a kind of possible design, obtained from the image to be monitored described by preset human testing model
Before the outline data information of human body to be identified, further includes:
First pretreatment is carried out to the image to be monitored, so that the image to be monitored meets preset dimension requirement.
In a kind of possible design, after carrying out the first pretreatment to the image to be monitored, further includes:
Second pretreatment is carried out to the image to be monitored, so that the image to be monitored meets preset RGB color value
It is required that.
In a kind of possible design, the attributive character information includes at least at least one in following attributive character:
Physiological characteristic information, dress up characteristic information and behavior characteristic information.
In a kind of possible design, the physiological characteristic information includes at least at least one in following attributive character:
Sex character information, age characteristics information and send out characteristic information long；
The characteristic information of dressing up is including at least at least one in following attributive character:
Garment ornament information and accessories characteristic information；
The behavior characteristic information includes at least at least one in following attributive character:
Motion characteristic information, belongings characteristic information and vehicles characteristic information.
Second aspect, the present invention also provides a kind of human body property recognition means, comprising:
Module is obtained, for obtaining image to be monitored, and by preset human testing model from the image to be monitored
The middle outline data information for obtaining human body to be identified；
Determining module for determining human testing frame according to the outline data information, and utilizes the human testing frame
The corresponding images to be recognized of the human body to be identified is intercepted from the image to be monitored；
Extraction module, for carrying out feature extraction to the images to be recognized by preset human body attribute Recognition Model,
To obtain the attributive character information of the human body to be identified in the images to be recognized.
In a kind of possible design, the human body property recognition means, further includes:
Display module, for human testing frame and the attributive character described in the Overlapping display on the image to be monitored
Information, wherein the human testing frame is used to accommodate the human body to be identified in the image to be monitored.
In a kind of possible design, the acquisition module is specifically used for:
The processing of multiple convolution layer and the processing of pond layer are carried out by the image to be monitored, to form full articulamentum；
Classification and Identification is carried out to the full articulamentum using target detection network, to obtain the profile of the human body to be identified
Data information.
It is described that human testing frame is determined according to the outline data information in a kind of possible design, comprising:
Determine the top left corner apex of the human testing frame first in preset coordinate system according to the outline data information
Coordinate value, the first frame long value and the first frame height value, wherein the abscissa value in first coordinate value is less than or equal to described
Minimum abscissa value in outline data information, the ordinate value in first coordinate value be greater than or equal to the number of contours it is believed that
Maximum ordinate value in breath, the first frame long value be greater than or equal in the outline data information maximum abscissa value with it is described
The difference of minimum abscissa value, the first frame height value be greater than or equal to the outline data information in maximum ordinate value with it is described
The difference of minimum ordinate value.
In a kind of possible design, the preset human body attribute Recognition Model is tensor recurrent neural networks model.
In a kind of possible design, the tensor recurrent neural networks model includes input layer, the first convolutional layer, first
Correcting layer, the first pond layer, the second convolutional layer, the second correcting layer, the second pond layer, third convolutional layer, tensor recurrence layer and defeated
Layer out；
Wherein, the input layer, first convolutional layer, first correcting layer, first pond layer, described second
Convolutional layer, second correcting layer, second pond layer and the third convolutional layer are sequentially connected, the tensor recurrence layer
It is connected to the third convolutional layer entirely, the output layer is connected to the tensor recurrence layer entirely.
In a kind of possible design, the human body property recognition means, further includes:
Preprocessing module, for carrying out the first pretreatment to the image to be monitored, so that the image to be monitored meets
Preset dimension requirement.
In a kind of possible design, the preprocessing module is also used to:
Second pretreatment is carried out to the image to be monitored, so that the image to be monitored meets preset RGB color value
It is required that.
In a kind of possible design, the attributive character information includes at least at least one in following attributive character:
Physiological characteristic information, dress up characteristic information and behavior characteristic information.
In a kind of possible design, the physiological characteristic information includes at least at least one in following attributive character:
Sex character information, age characteristics information and send out characteristic information long；
The characteristic information of dressing up is including at least at least one in following attributive character:
Garment ornament information and accessories characteristic information；
The behavior characteristic information includes at least at least one in following attributive character:
Motion characteristic information, belongings characteristic information and vehicles characteristic information.
The third aspect, the present invention also provides a kind of computer readable storage mediums, are stored thereon with computer program, the journey
Any one possible human body attribute recognition approach in first aspect is realized when sequence is executed by processor.
Fourth aspect, the present invention also provides a kind of electronic equipment, comprising:
Processor；And
Memory, for storing the executable instruction of the processor；
Wherein, be configured to execute any one in first aspect via the executable instruction is executed can for the processor
The human body attribute recognition approach of energy.
A kind of human body attribute recognition approach, device, storage medium and electronic equipment provided by the invention are first passed through using pre-
If human testing model the outline data information of human body to be identified is obtained from image to be monitored, further according to outline data information
It determines human testing frame, and intercepts the corresponding images to be recognized of human body to be identified from image to be monitored using human testing frame,
Then by preset human body attribute Recognition Model to images to be recognized carry out feature extraction, with obtain in images to be recognized to
Identify the attributive character information of human body, thus realize the function of carrying out identification screening to target person in security protection application scenarios,
Recognition speed and accuracy are greatly accelerated, and reduces the investment of human cost.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this hair
Bright some embodiments for those of ordinary skill in the art without any creative labor, can be with
It obtains other drawings based on these drawings.
Fig. 1 is the application scenario diagram of human body attribute recognition approach shown according to an exemplary embodiment；
Fig. 2 is the flow chart of human body attribute recognition approach shown according to an exemplary embodiment；
Fig. 3 is a kind of possible application scene figure of embodiment illustrated in fig. 2；
Fig. 4 is the flow chart of the human body attribute recognition approach shown according to another exemplary embodiment；
Fig. 5 is the structural schematic diagram of human body property recognition means shown according to an exemplary embodiment；
Fig. 6 is the structural schematic diagram of the human body property recognition means shown according to another exemplary embodiment；
Fig. 7 is the structural schematic diagram of present invention electronic equipment shown according to an exemplary embodiment.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without creative efforts, shall fall within the protection scope of the present invention.
Fig. 1 is the application scenario diagram of human body attribute recognition approach shown according to an exemplary embodiment.As shown in Figure 1,
Human body attribute recognition approach provided in this embodiment is in use, can be by input image data either video data
Single-frame images carries out Attribute Recognition, wherein the single-frame images in input image data either video data, which can be, passes through peace
The terminal devices such as anti-camera, smart phone, tablet computer, personal computer obtain.And when being inputted is video data,
Frame decoding processing first can be carried out to video data, first every frame image is individually identified, is finally realized to all picture frames
Identification.
Wherein, for applying human body attribute recognition approach provided by the present embodiment in police's suspect's investigation field
It is illustrated.If the police know the artificial male of suspicion, youth, long hair from the information that reporter provides at this time and wear glasses
Attributive character information at this point, can may be haunted the security protection video in section by obtaining the suspect, and is known it
Not, meet the target person of clue to help the police that can quickly filter out in a large amount of data, to improve target person
Recognition efficiency is conducive to practical popularization and application.
Fig. 2 is the flow chart of human body attribute recognition approach shown according to an exemplary embodiment.As shown in Fig. 2, this reality
The human body attribute recognition approach of example offer is provided, comprising:
Step 101 obtains image to be monitored, and is obtained from image to be monitored wait know by preset human testing model
The outline data information of others' body.
Specifically, image to be detected can pass through the single-frame images in input image data either video data, wherein
Single-frame images of the input image data either in video data can be by security protection camera, smart phone, tablet computer,
The terminal devices such as personal computer obtain.And when being inputted is video data, first video data can be carried out at frame decoding
Reason, is first individually identified every frame image, finally realizes the identification to all picture frames, and real after every frame completes identification
When by human testing frame and human body attributive character information superposition in this frame image.
After getting image to be monitored, can be obtained from image to be monitored by preset human testing model to
The outline data information for identifying human body, and for target detection therein, can be based on SSD target detection frame or
RefineDet target detection frame, in the present embodiment, not to the specific calculation of the outline data acquisition of information of human body to be identified
Method is defined, and need to only guarantee the outline data information that human body to be identified is obtained from image to be monitored.
In a kind of possible design, the processing of multiple convolution layer can be carried out by image to be monitored and pond layer is handled,
To form full articulamentum, then target detection network is recycled to carry out Classification and Identification to full articulamentum, to obtain human body to be identified
Outline data information, wherein the target detection network can be based on SSD target detection frame or RefineDet target inspection
Survey frame.
Step 102 determines human testing frame according to outline data information, and using human testing frame from image to be monitored
Intercept the corresponding images to be recognized of human body to be identified.
Specifically, it is based on SSD target detection frame or RefineDet target detection frame, can be exported as several human bodies
Detection block, wherein the position of each frame and the first coordinate value is used with size, wherein the first coordinate value includes 4 coordinates
(xmin, ymin, w, h), xmin indicate that x coordinate of the human testing frame upper left corner in preset coordinate system, ymin indicate human body
Y-coordinate of the detection block upper left corner in preset coordinate system, w indicate that the first frame long value of human testing frame, h indicate human testing
First frame height value of frame, wherein above-mentioned value respectively corresponds a human region as unit of pixel.
In order to distinguish each car separately through human testing frame, people can be determined according to outline data information
The top left corner apex of body detection block the first coordinate value, the first frame long value and first frame height value in preset coordinate system, wherein the
Abscissa value in one coordinate value is less than or equal to minimum abscissa value in outline data information, the ordinate in the first coordinate value
Value is greater than or equal to maximum ordinate value in outline data information, and the first frame long value is greater than or equal to maximum in outline data information
The difference of abscissa value and minimum abscissa value, the first frame height value is more than or equal to maximum ordinate value in outline data information and most
The difference of small ordinate value.
After determining human testing frame, it can use human testing frame and intercept human body pair to be identified from image to be monitored
The images to be recognized answered.In addition, corresponding to be identified intercepting human body to be identified from image to be monitored using human testing frame
After image, unification can also be carried out to its size again, such as can scale it again to fixed dimension, such as 224*224
Pixel, the images to be recognized input human body attribute Recognition Model after it will scale carries out feature extraction later.
Step 103 carries out feature extraction to images to be recognized by preset human body attribute Recognition Model, to obtain wait know
The attributive character information of human body to be identified in other image.
It, can be by being based on moving-vision network (MobileNet), to images to be recognized after inputting images to be recognized
Feature extraction is carried out, to obtain the attributive character information of the human body to be identified in images to be recognized.And for getting wait know
The attributive character information of others' body, can be directly displayed on image to be monitored, be also possible to input corresponding statistical forecast mould
Type is screened, not special to the attribute of accessed human body to be identified to determine suspicion human body, and in the present embodiment
The concrete application scene of reference breath is specifically limited.
In a kind of possible design, above-mentioned human body attribute Recognition Model can also be tensor recurrent neural network mould
Type.Specifically, tensor recurrent neural networks model includes input layer, the first convolutional layer, the first correcting layer, the first pond layer,
Two convolutional layers, the second correcting layer, the second pond layer, third convolutional layer, tensor recurrence layer and output layer；Wherein, input layer, first
Convolutional layer, the first correcting layer, the first pond layer, the second convolutional layer, the second correcting layer, the second pond layer and third convolutional layer according to
Secondary connection, tensor recurrence layer are connected to third convolutional layer entirely, and output layer is connected to tensor recurrence layer entirely.
And at least one in following attributive character is included at least for attributive character information:
Physiological characteristic information, dress up characteristic information and behavior characteristic information.
In a kind of possible design, above-mentioned physiological characteristic information includes at least at least one in following attributive character
:
Sex character information, age characteristics information and send out characteristic information long；
Above-mentioned characteristic information of dressing up is including at least at least one in following attributive character:
Garment ornament information and accessories characteristic information；
Above-mentioned behavior characteristic information includes at least at least one in following attributive character:
Motion characteristic information, belongings characteristic information and vehicles characteristic information.
In one specifically application scenarios, when carrying out feature identification using MobileNet, it can be believed with defined attribute feature
Breath, specifically as described in table one:
In the present embodiment, it first passes through and obtains human body to be identified from image to be monitored using preset human testing model
Outline data information, determine human testing frame further according to outline data information, and using human testing frame from image to be monitored
The corresponding images to be recognized of middle interception human body to be identified, then by preset human body attribute Recognition Model to images to be recognized into
Row feature extraction, to obtain the attributive character information of the human body to be identified in images to be recognized, to realize in security protection applied field
The function of carrying out identification screening in scape to target person, greatly accelerates recognition speed and accuracy, and reduce human cost
Investment.
On the basis of embodiment shown in Fig. 2, images to be recognized is being carried out by preset human body attribute Recognition Model
Feature extraction can also be in image to be monitored after the attributive character information to obtain the human body to be identified in images to be recognized
Upper Overlapping display human testing frame and attributive character information.
Fig. 3 is a kind of possible application scene figure of embodiment illustrated in fig. 2.As shown in figure 3, get human testing frame with
And after attributive character information, can on image to be monitored Overlapping display human testing frame and attributive character information, wherein
Human testing frame is used to accommodate the human body to be identified in image to be monitored.
In the present embodiment, it first passes through and obtains human body to be identified from image to be monitored using preset human testing model
Outline data information, determine human testing frame further according to outline data information, and using human testing frame from image to be monitored
The corresponding images to be recognized of middle interception human body to be identified, then by preset human body attribute Recognition Model to images to be recognized into
Row feature extraction is finally folded on image to be monitored with obtaining the attributive character information of the human body to be identified in images to be recognized
Add display human testing frame and attributive character information, identification sieve is carried out to suspicion human body in security protection application scenarios to realize
The function of choosing greatly accelerates recognition speed and accuracy, and reduces the investment of human cost.
Fig. 4 is the flow chart of the human body attribute recognition approach shown according to another exemplary embodiment.As shown in figure 4, this
The human body attribute recognition approach that embodiment provides, comprising:
Step 201 treats the first pretreatment of monitoring image progress, so that image to be monitored meets preset dimension requirement.
In order to enable preset human testing model to get various sizes of figure to be monitored to different modes
As being identified, monitoring image can also be treated and carry out the first pretreatment, so that image to be monitored meets preset dimension requirement.Example
It such as, can be by the image scaling to be monitored of input at fixed dimension, such as 512*512 pixel.
Step 202 treats the second pretreatment of monitoring image progress, so that image to be monitored meets preset RGB color value
It is required that.
In order to enable preset human testing model to get different color, brightness and right to different modes
Image to be monitored than degree is identified, can also be treated monitoring image and be carried out the second pretreatment, so that image to be monitored meets
Preset RGB color value requirement.For example, image to be monitored can be subtracted to unified RGB mean value, such as [104,117,123].
Step 203 obtains image to be monitored, and is obtained from image to be monitored wait know by preset human testing model
The outline data information of others' body.
Step 204 determines human testing frame according to outline data information, and using human testing frame from image to be monitored
Intercept the corresponding images to be recognized of human body to be identified.
Step 205 carries out feature extraction to images to be recognized by preset human body attribute Recognition Model, to obtain wait know
The attributive character information of human body to be identified in other image.
It is worth explanatorily, the specific implementation of step 203-205 is referring to step 101-103 in embodiment illustrated in fig. 2
Description, which is not described herein again.
Step 206, Overlapping display human testing frame and attributive character information on image to be monitored.
After getting human testing frame and attributive character information, can on image to be monitored Overlapping display human body
Detection block and attributive character information, wherein human testing frame is used to accommodate the human body to be identified in image to be monitored.
Fig. 5 is the structural schematic diagram of human body property recognition means shown according to an exemplary embodiment.As shown in figure 5,
Human body property recognition means provided in this embodiment, comprising:
Module 301 is obtained, for obtaining image to be monitored, and by preset human testing model from the figure to be monitored
The outline data information of human body to be identified is obtained as in；
Determining module 302 for determining human testing frame according to the outline data information, and utilizes the human testing
Frame intercepts the corresponding images to be recognized of the human body to be identified from the image to be monitored；
Extraction module 303 is mentioned for carrying out feature to the images to be recognized by preset human body attribute Recognition Model
It takes, to obtain the attributive character information of the human body to be identified in the images to be recognized.
On the basis of embodiment shown in Fig. 5, Fig. 6 is the human body Attribute Recognition dress shown according to another exemplary embodiment
The structural schematic diagram set.As shown in fig. 6, human body property recognition means provided by the embodiment, further includes:
Display module 304, for human testing frame and the attribute described in the Overlapping display on the image to be monitored
Characteristic information, wherein the human testing frame is used to accommodate the human body to be identified in the image to be monitored.
In a kind of possible design, the acquisition module 301 is specifically used for:
The processing of multiple convolution layer and the processing of pond layer are carried out by the image to be monitored, to form full articulamentum；
Classification and Identification is carried out to the full articulamentum using target detection network, to obtain the profile of the human body to be identified
Data information.
It is described that human testing frame is determined according to the outline data information in a kind of possible design, comprising:
Determine the top left corner apex of the human testing frame first in preset coordinate system according to the outline data information
Coordinate value, the first frame long value and the first frame height value, wherein the abscissa value in first coordinate value is less than or equal to described
Minimum abscissa value in outline data information, the ordinate value in first coordinate value be greater than or equal to the number of contours it is believed that
Maximum ordinate value in breath, the first frame long value be greater than or equal in the outline data information maximum abscissa value with it is described
The difference of minimum abscissa value, the first frame height value be greater than or equal to the outline data information in maximum ordinate value with it is described
The difference of minimum ordinate value.
In a kind of possible design, the preset human body attribute Recognition Model is tensor recurrent neural networks model.
In a kind of possible design, the tensor recurrent neural networks model includes input layer, the first convolutional layer, first
Correcting layer, the first pond layer, the second convolutional layer, the second correcting layer, the second pond layer, third convolutional layer, tensor recurrence layer and defeated
Layer out；
Wherein, the input layer, first convolutional layer, first correcting layer, first pond layer, described second
Convolutional layer, second correcting layer, second pond layer and the third convolutional layer are sequentially connected, the tensor recurrence layer
It is connected to the third convolutional layer entirely, the output layer is connected to the tensor recurrence layer entirely.
Please continue to refer to Fig. 6, human body property recognition means provided by the embodiment, further includes:
Preprocessing module 305, for carrying out the first pretreatment to the image to be monitored, so that the image to be monitored is full
Sufficient preset dimension requirement.
In a kind of possible design, the preprocessing module 305 is also used to:
Second pretreatment is carried out to the image to be monitored, so that the image to be monitored meets preset RGB color value
It is required that.
In a kind of possible design,
The attributive character information includes at least at least one in following attributive character:
Physiological characteristic information, dress up characteristic information and behavior characteristic information.
In a kind of possible design, the physiological characteristic information includes at least at least one in following attributive character:
Sex character information, age characteristics information and send out characteristic information long；
The characteristic information of dressing up is including at least at least one in following attributive character:
Garment ornament information and accessories characteristic information；
The behavior characteristic information includes at least at least one in following attributive character:
Motion characteristic information, belongings characteristic information and vehicles characteristic information.
It is worth explanatorily, the terminal device in Fig. 5 and embodiment illustrated in fig. 6 can be used for executing shown in above-mentioned Fig. 2 and Fig. 4
Method in embodiment, specific implementation is similar with technical effect, and which is not described herein again.
The present invention also provides a kind of computer readable storage mediums, are stored thereon with computer program, and the program is processed
Device realizes the technical solution of any one of aforementioned embodiment of the method when executing, it is similar that the realization principle and technical effect are similar, herein no longer
It repeats.
Fig. 7 is the structural schematic diagram of present invention electronic equipment shown according to an exemplary embodiment.As shown in fig. 7, this
The electronic equipment that embodiment provides, comprising:
Processor 401；And
Memory 402, for storing the executable instruction of the processor；
Wherein, the processor is configured to execute any one of aforementioned embodiment of the method via the executable instruction is executed
The technical solution, it is similar that the realization principle and technical effect are similar, and details are not described herein again.
Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention., rather than its limitations；To the greatest extent
Pipe present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: its according to
So be possible to modify the technical solutions described in the foregoing embodiments, or part of or all technical features are carried out etc.
With replacement；And these modifications or substitutions, technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution
Range.
Claims (13)
1. a kind of human body attribute recognition approach characterized by comprising
Image to be monitored is obtained, and human body to be identified is obtained from the image to be monitored by preset human testing model
Outline data information；
Human testing frame is determined according to the outline data information, and using the human testing frame from the image to be monitored
Intercept the corresponding images to be recognized of the human body to be identified；
Feature extraction is carried out to the images to be recognized by preset human body attribute Recognition Model, to obtain the figure to be identified
The attributive character information of the human body to be identified as in.
2. human body attribute recognition approach according to claim 1, which is characterized in that pass through preset human body attribute described
Identification model carries out feature extraction to the images to be recognized, to obtain the human body to be identified in the images to be recognized
After attributive character information, further includes:
Human testing frame described in Overlapping display and the attributive character information on the image to be monitored, wherein the people
Body detection block is used to accommodate the human body to be identified in the image to be monitored.
3. human body attribute recognition approach according to claim 1, which is characterized in that described to pass through preset human testing mould
Type obtains the outline data information of human body to be identified from the image to be monitored, comprising:
The processing of multiple convolution layer and the processing of pond layer are carried out by the image to be monitored, to form full articulamentum；
Classification and Identification is carried out to the full articulamentum using target detection network, to obtain the outline data of the human body to be identified
Information.
4. human body attribute recognition approach according to claim 3, which is characterized in that described according to the outline data information
Determine human testing frame, comprising:
The top left corner apex of the human testing frame the first coordinate in preset coordinate system is determined according to the outline data information
Value, the first frame long value and the first frame height value, wherein the abscissa value in first coordinate value is less than or equal to the profile
Minimum abscissa value in data information, the ordinate value in first coordinate value are greater than or equal in the outline data information
Maximum ordinate value, the first frame long value are greater than or equal to maximum abscissa value and the minimum in the outline data information
The difference of abscissa value, the first frame height value are greater than or equal to maximum ordinate value and the minimum in the outline data information
The difference of ordinate value.
5. human body attribute recognition approach according to claim 4, which is characterized in that the preset human body Attribute Recognition mould
Type is tensor recurrent neural networks model.
6. human body attribute recognition approach according to claim 5, which is characterized in that the tensor recurrent neural networks model
Including input layer, the first convolutional layer, the first correcting layer, the first pond layer, the second convolutional layer, the second correcting layer, the second pond layer,
Third convolutional layer, tensor recurrence layer and output layer；
Wherein, the input layer, first convolutional layer, first correcting layer, first pond layer, second convolution
Layer, second correcting layer, second pond layer and the third convolutional layer are sequentially connected, and the tensor recurrence layer connects entirely
It is connected to the third convolutional layer, the output layer is connected to the tensor recurrence layer entirely.
7. human body attribute recognition approach according to claim 1, which is characterized in that pass through preset human testing described
Model is before the outline data information for obtaining human body to be identified in the image to be monitored, further includes:
First pretreatment is carried out to the image to be monitored, so that the image to be monitored meets preset dimension requirement.
8. human body attribute recognition approach according to claim 7, which is characterized in that carrying out the to the image to be monitored
After one pretreatment, further includes:
Second pretreatment is carried out to the image to be monitored, so that the image to be monitored meets preset RGB color value requirement.
9. human body attribute recognition approach described in any one of -8 according to claim 1, which is characterized in that the attributive character
Information includes at least at least one in following attributive character:
Physiological characteristic information, dress up characteristic information and behavior characteristic information.
10. human body attribute recognition approach according to claim 9, which is characterized in that the physiological characteristic information is at least wrapped
Include at least one in following attributive character:
Sex character information, age characteristics information and send out characteristic information long；
The characteristic information of dressing up is including at least at least one in following attributive character:
Garment ornament information and accessories characteristic information；
The behavior characteristic information includes at least at least one in following attributive character:
Motion characteristic information, belongings characteristic information and vehicles characteristic information.
11. a kind of human body property recognition means characterized by comprising
Module is obtained, is obtained from the image to be monitored for obtaining image to be monitored, and by preset human testing model
Take the outline data information of human body to be identified；
Determining module for determining human testing frame according to the outline data information, and utilizes the human testing frame from institute
It states and intercepts the corresponding images to be recognized of the human body to be identified in image to be monitored；
Extraction module, for carrying out feature extraction to the images to be recognized by preset human body attribute Recognition Model, to obtain
Obtain the attributive character information of the human body to be identified in the images to be recognized.
12. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
Claim 1-9 described in any item human body attribute recognition approaches are realized when execution.
13. a kind of electronic equipment characterized by comprising
Processor；And
Memory, for storing the executable instruction of the processor；
Wherein, the processor is configured to require 1-9 described in any item via executing the executable instruction and carry out perform claim
Human body attribute recognition approach.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811168498.7A CN109522790A (en) | 2018-10-08 | 2018-10-08 | Human body attribute recognition approach, device, storage medium and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811168498.7A CN109522790A (en) | 2018-10-08 | 2018-10-08 | Human body attribute recognition approach, device, storage medium and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109522790A true CN109522790A (en) | 2019-03-26 |
Family
ID=65769733
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811168498.7A Pending CN109522790A (en) | 2018-10-08 | 2018-10-08 | Human body attribute recognition approach, device, storage medium and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109522790A (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110110693A (en) * | 2019-05-17 | 2019-08-09 | 北京字节跳动网络技术有限公司 | The method and apparatus of face character for identification |
| CN110287856A (en) * | 2019-06-21 | 2019-09-27 | 上海闪马智能科技有限公司 | A kind of security personnel's behavior analysis system, method and device |
| CN110674756A (en) * | 2019-09-25 | 2020-01-10 | 普联技术有限公司 | Human body attribute recognition model training method, human body attribute recognition method and device |
| CN110716634A (en) * | 2019-08-28 | 2020-01-21 | 北京市商汤科技开发有限公司 | Interaction method, device, equipment and display equipment |
| CN111145215A (en) * | 2019-12-25 | 2020-05-12 | 北京迈格威科技有限公司 | Target tracking method and device |
| CN111274945A (en) * | 2020-01-19 | 2020-06-12 | 北京百度网讯科技有限公司 | Pedestrian attribute identification method and device, electronic equipment and storage medium |
| CN111414812A (en) * | 2020-03-03 | 2020-07-14 | 平安科技（深圳）有限公司 | Human body attribute identification method, system, computer device and storage medium |
| CN111598164A (en) * | 2020-05-15 | 2020-08-28 | 北京百度网讯科技有限公司 | Method and device for identifying attribute of target object, electronic equipment and storage medium |
| CN111783619A (en) * | 2020-06-29 | 2020-10-16 | 北京百度网讯科技有限公司 | Human body attribute identification method, device, equipment and storage medium |
| CN112417205A (en) * | 2019-08-20 | 2021-02-26 | 富士通株式会社 | Target retrieval device and method and electronic equipment |
| CN112836549A (en) * | 2019-11-22 | 2021-05-25 | 虹软科技股份有限公司 | User information detection method and system and electronic equipment |
| CN113537126A (en) * | 2021-07-28 | 2021-10-22 | 浙江大华技术股份有限公司 | Method and device for determining human body wearing violation, storage medium and electronic device |
| CN114332915A (en) * | 2021-12-02 | 2022-04-12 | 深圳云天励飞技术股份有限公司 | Human body attribute detection method and device, computer equipment and storage medium |
| CN114424258A (en) * | 2019-12-23 | 2022-04-29 | 深圳市欢太科技有限公司 | Attribute identification method and device, storage medium and electronic equipment |
| CN115131825A (en) * | 2022-07-14 | 2022-09-30 | 北京百度网讯科技有限公司 | Human body attribute identification method and device, electronic equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103324907A (en) * | 2012-03-22 | 2013-09-25 | 中国科学院计算技术研究所 | Human body appearance model learning method and system for human body reappearance detection |
| CN106127173A (en) * | 2016-06-30 | 2016-11-16 | 北京小白世纪网络科技有限公司 | A kind of human body attribute recognition approach based on degree of depth study |
| CN106203458A (en) * | 2015-04-29 | 2016-12-07 | 杭州海康威视数字技术股份有限公司 | Crowd's video analysis method and system |
| CN107153812A (en) * | 2017-03-31 | 2017-09-12 | 深圳先进技术研究院 | A kind of exercising support method and system based on machine vision |
| CN108256567A (en) * | 2018-01-12 | 2018-07-06 | 环球大数据科技有限公司 | A kind of target identification method and system based on deep learning |
-
2018
- 2018-10-08 CN CN201811168498.7A patent/CN109522790A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103324907A (en) * | 2012-03-22 | 2013-09-25 | 中国科学院计算技术研究所 | Human body appearance model learning method and system for human body reappearance detection |
| CN106203458A (en) * | 2015-04-29 | 2016-12-07 | 杭州海康威视数字技术股份有限公司 | Crowd's video analysis method and system |
| CN106127173A (en) * | 2016-06-30 | 2016-11-16 | 北京小白世纪网络科技有限公司 | A kind of human body attribute recognition approach based on degree of depth study |
| CN107153812A (en) * | 2017-03-31 | 2017-09-12 | 深圳先进技术研究院 | A kind of exercising support method and system based on machine vision |
| CN108256567A (en) * | 2018-01-12 | 2018-07-06 | 环球大数据科技有限公司 | A kind of target identification method and system based on deep learning |
Cited By (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110110693A (en) * | 2019-05-17 | 2019-08-09 | 北京字节跳动网络技术有限公司 | The method and apparatus of face character for identification |
| CN110287856A (en) * | 2019-06-21 | 2019-09-27 | 上海闪马智能科技有限公司 | A kind of security personnel's behavior analysis system, method and device |
| CN112417205A (en) * | 2019-08-20 | 2021-02-26 | 富士通株式会社 | Target retrieval device and method and electronic equipment |
| CN110716634A (en) * | 2019-08-28 | 2020-01-21 | 北京市商汤科技开发有限公司 | Interaction method, device, equipment and display equipment |
| CN110674756A (en) * | 2019-09-25 | 2020-01-10 | 普联技术有限公司 | Human body attribute recognition model training method, human body attribute recognition method and device |
| CN110674756B (en) * | 2019-09-25 | 2022-07-05 | 普联技术有限公司 | Human body attribute recognition model training method, human body attribute recognition method and device |
| WO2021098855A1 (en) * | 2019-11-22 | 2021-05-27 | 虹软科技股份有限公司 | User information detection method and system, and electronic device |
| CN112836549A (en) * | 2019-11-22 | 2021-05-25 | 虹软科技股份有限公司 | User information detection method and system and electronic equipment |
| CN114424258A (en) * | 2019-12-23 | 2022-04-29 | 深圳市欢太科技有限公司 | Attribute identification method and device, storage medium and electronic equipment |
| CN114424258B (en) * | 2019-12-23 | 2024-10-01 | 深圳市欢太科技有限公司 | Attribute identification method, attribute identification device, storage medium and electronic equipment |
| CN111145215A (en) * | 2019-12-25 | 2020-05-12 | 北京迈格威科技有限公司 | Target tracking method and device |
| CN111145215B (en) * | 2019-12-25 | 2023-09-05 | 北京迈格威科技有限公司 | Target tracking method and device |
| CN111274945B (en) * | 2020-01-19 | 2023-08-08 | 北京百度网讯科技有限公司 | Pedestrian attribute identification method and device, electronic equipment and storage medium |
| CN111274945A (en) * | 2020-01-19 | 2020-06-12 | 北京百度网讯科技有限公司 | Pedestrian attribute identification method and device, electronic equipment and storage medium |
| WO2021174941A1 (en) * | 2020-03-03 | 2021-09-10 | 平安科技（深圳）有限公司 | Physical attribute recognition method, system, computer device, and storage medium |
| CN111414812A (en) * | 2020-03-03 | 2020-07-14 | 平安科技（深圳）有限公司 | Human body attribute identification method, system, computer device and storage medium |
| CN111598164A (en) * | 2020-05-15 | 2020-08-28 | 北京百度网讯科技有限公司 | Method and device for identifying attribute of target object, electronic equipment and storage medium |
| CN111783619A (en) * | 2020-06-29 | 2020-10-16 | 北京百度网讯科技有限公司 | Human body attribute identification method, device, equipment and storage medium |
| CN111783619B (en) * | 2020-06-29 | 2023-08-11 | 北京百度网讯科技有限公司 | Human body attribute identification method, device, equipment and storage medium |
| CN113537126A (en) * | 2021-07-28 | 2021-10-22 | 浙江大华技术股份有限公司 | Method and device for determining human body wearing violation, storage medium and electronic device |
| CN114332915A (en) * | 2021-12-02 | 2022-04-12 | 深圳云天励飞技术股份有限公司 | Human body attribute detection method and device, computer equipment and storage medium |
| CN115131825A (en) * | 2022-07-14 | 2022-09-30 | 北京百度网讯科技有限公司 | Human body attribute identification method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109522790A (en) | Human body attribute recognition approach, device, storage medium and electronic equipment | |
| US10943145B2 (en) | Image processing methods and apparatus, and electronic devices | |
| US11170210B2 (en) | Gesture identification, control, and neural network training methods and apparatuses, and electronic devices | |
| CN111178183B (en) | Face detection method and related device | |
| KR102629380B1 (en) | Method for Distinguishing a Real Three-Dimensional Object from a Two-Dimensional Spoof of the Real Object | |
| CN105518709B (en) | The method, system and computer program product of face for identification | |
| CN109508636A (en) | Vehicle attribute recognition methods, device, storage medium and electronic equipment | |
| CN108427927B (en) | Object re-recognition method and apparatus, electronic device, program, and storage medium | |
| WO2020078119A1 (en) | Method, device and system for simulating user wearing clothing and accessories | |
| CN110427905A (en) | Pedestrian tracting method, device and terminal | |
| CN110991261A (en) | Interactive behavior recognition method and device, computer equipment and storage medium | |
| CN111753782B (en) | False face detection method and device based on double-current network and electronic equipment | |
| CN109815843A (en) | Object detection method and Related product | |
| CN111160202A (en) | AR equipment-based identity verification method, AR equipment-based identity verification device, AR equipment-based identity verification equipment and storage medium | |
| CN111353385B (en) | Pedestrian re-identification method and device based on mask alignment and attention mechanism | |
| CN105303163B (en) | A kind of method and detection device of target detection | |
| CN114758362A (en) | Clothing changing pedestrian re-identification method based on semantic perception attention and visual masking | |
| CN109670517A (en) | Object detection method, device, electronic equipment and target detection model | |
| CN109583364A (en) | Image-recognizing method and equipment | |
| CN112633222B (en) | Gait recognition method, device, equipment and medium based on countermeasure network | |
| CN112464690A (en) | Living body identification method, living body identification device, electronic equipment and readable storage medium | |
| CN109784230A (en) | A kind of facial video image quality optimization method, system and equipment | |
| CN112487844A (en) | Gesture recognition method, electronic device, computer-readable storage medium, and chip | |
| CN113763440A (en) | Image processing method, device, equipment and storage medium | |
| CN109977875A (en) | Gesture identification method and equipment based on deep learning |
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 |
Application publication date: 20190326 |
|
| RJ01 | Rejection of invention patent application after publication |