CN109784140A - Driver attributes' recognition methods and Related product - Google Patents
Driver attributes' recognition methods and Related product Download PDFInfo
- Publication number
- CN109784140A CN109784140A CN201811375755.4A CN201811375755A CN109784140A CN 109784140 A CN109784140 A CN 109784140A CN 201811375755 A CN201811375755 A CN 201811375755A CN 109784140 A CN109784140 A CN 109784140A
- Authority
- CN
- China
- Prior art keywords
- attribute
- driver
- image
- input picture
- input
- 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
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
This application provides a kind of driver attributes' recognition methods and Related product, method includes: acquisition input picture, includes driver's area image in the input picture;The input picture is input to default convolutional neural networks model to handle, obtains the first attribute-bit collection, the default convolutional neural networks model includes characteristics of image network and embedded network;The input picture is input to pre-set zoom network to handle, obtains target area image, the target area image is the behavior region of the driver;The target area image is input to the default convolutional neural networks model to handle, obtains the second attribute-bit collection;The corresponding objective attribute target attribute mark of the driver is determined according to the first attribute-bit collection and the second attribute-bit collection.Using the application can the attribute to driver precisely identified.
Description
Technical field
This application involves technical field of image processing, and in particular to a kind of driver attributes' recognition methods and Related product.
Background technique
With the high speed development of road traffic, quick, comfortable, convenient is the generally impression of people, but thus bring
Traffic accident also becomes serious problem of society today concerned by people.Many traffic accidents are due to the bad violation driving behavior of driver
Caused by.Some research achievements are had been achieved for the identification bad unlawful practice of driver both at home and abroad.Therefore, how to driver
Attribute the problem of being identified it is urgently to be resolved.
Summary of the invention
The embodiment of the present application provides a kind of driver attributes' recognition methods and Related product, can precisely identify driver
Attribute.
The embodiment of the present application first aspect provides a kind of driver attributes' recognition methods, comprising:
Input picture is obtained, includes driver's area image in the input picture;
The input picture is input to default convolutional neural networks model to handle, obtains the first attribute-bit collection,
The default convolutional neural networks model includes characteristics of image network and embedded network;
The input picture is input to pre-set zoom network to handle, obtains target area image, the target area
Area image is the behavior region of the driver;
The target area image is input to the default convolutional neural networks model to handle, obtains the second attribute
Identification sets;
The corresponding target category of the driver is determined according to the first attribute-bit collection and the second attribute-bit collection
Property mark.
The embodiment of the present application second aspect provides a kind of driver attributes' identification device, comprising:
Acquiring unit includes driver's area image in the input picture for obtaining input picture;
First processing units are handled for the input picture to be input to default convolutional neural networks model, are obtained
To the first attribute-bit collection, the default convolutional neural networks model includes characteristics of image network and embedded network;
The second processing unit handles for the input picture to be input to pre-set zoom network, obtains target area
Area image, the target area image are the behavior region of the driver;
The first processing units are input to the default convolutional Neural net also particularly useful for by the target area image
Network model is handled, and the second attribute-bit collection is obtained;
Determination unit, for determining the driver according to the first attribute-bit collection and the second attribute-bit collection
Corresponding objective attribute target attribute mark.
The third aspect, the embodiment of the present application provide a kind of electronic equipment, comprising: processor and memory;And one
Or multiple programs, one or more of programs are stored in the memory, and are configured to be held by the processor
Row, described program includes the instruction for the step some or all of as described in first aspect.
Fourth aspect, the embodiment of the present application provide a kind of computer readable storage medium, wherein described computer-readable
Storage medium is for storing computer program, wherein the computer program executes computer such as the embodiment of the present application the
The instruction of step some or all of described in one side.
5th aspect, the embodiment of the present application provide a kind of computer program product, wherein the computer program product
Non-transient computer readable storage medium including storing computer program, the computer program are operable to make to calculate
Machine executes the step some or all of as described in the embodiment of the present application first aspect.The computer program product can be one
A software installation packet.
Implement the embodiment of the present application, has the following beneficial effects:
Using above-mentioned driver attributes' recognition methods provided by the embodiments of the present application and Related product, input picture is obtained,
Include driver's area image in input picture, input picture is input to default convolutional neural networks model and is handled, is obtained
To the first attribute-bit collection, default convolutional neural networks model includes characteristics of image network and embedded network, and input picture is defeated
Enter to pre-set zoom network and handled, obtains target area image, target area image is the behavior region of driver, by mesh
Mark area image is input to default convolutional neural networks model and is handled, and the second attribute-bit collection is obtained, according to the first attribute
Identification sets and the second attribute-bit collection determine the corresponding objective attribute target attribute mark of driver, can carry out to the attribute of driver accurate
Identification.
Detailed description of the invention
In order to more clearly explain the technical solutions in the embodiments of the present application, make required in being described below to embodiment
Attached drawing is briefly described, it should be apparent that, the accompanying drawings in the following description is some embodiments of the present application, for ability
For the those of ordinary skill of domain, without creative efforts, it can also be obtained according to these attached drawings other attached
Figure.
Figure 1A is a kind of embodiment flow diagram of driver attributes' recognition methods provided by the embodiments of the present application;
Figure 1B is a kind of demonstration schematic diagram of driver attributes' recognition methods provided by the embodiments of the present application;
Fig. 2 is a kind of example structure schematic diagram of driver attributes' identification device provided by the embodiments of the present application;
Fig. 3 is the example structure schematic diagram of a kind of electronic equipment provided by the embodiments of the present application.
Specific embodiment
Below in conjunction with the attached drawing in the embodiment of the present application, technical solutions in the embodiments of the present application carries out clear, complete
Site preparation description, it is clear that described embodiment is some embodiments of the present application, instead of all the embodiments.Based on this Shen
Please in embodiment, every other implementation obtained by those of ordinary skill in the art without making creative efforts
Example, shall fall in the protection scope of this application.
The description and claims of this application and term " first ", " second ", " third " and " in the attached drawing
Four " etc. are not use to describe a particular order for distinguishing different objects.In addition, term " includes " and " having " and it
Any deformation, it is intended that cover and non-exclusive include.Such as it contains the process, method of a series of steps or units, be
System, product or equipment are not limited to listed step or unit, but optionally further comprising the step of not listing or list
Member, or optionally further comprising other step or units intrinsic for these process, methods, product or equipment.
Referenced herein " embodiment " is it is meant that a particular feature, structure, or characteristic described can wrap in conjunction with the embodiments
It is contained at least one embodiment of the application.It is identical that each position in the description shows that the phrase might not be each meant
Embodiment, nor the independent or alternative embodiment with other embodiments mutual exclusion.Those skilled in the art explicitly and
Implicitly understand, embodiment described herein can be combined with other embodiments.
Row electronic equipment described by the embodiment of the present application may include smart phone (such as Android phone, iOS mobile phone,
Windows Phone mobile phone etc.), tablet computer, video matrix, monitor supervision platform, mobile unit, satellite, palm PC, notebook
Computer, mobile internet device (MID, Mobile Internet Devices) or wearable device etc., above-mentioned is only citing,
And it is non exhaustive, including but not limited to above-mentioned apparatus, certainly, above-mentioned electronic equipment can also be server.
Target identification method in the related technology is that image tag is assigned to trained concentration training, to predict object instance
Presence, such method by the image for largely having certain feature, exist for small sample even no specimen it is obvious not
Foot.Different with correlation technique, the target of joint various definitions attribute forecast is the new category having never seen before identification.Phase
The learning method of pass, which primarily focuses on, introduces non-linear or linear modeling method, uses specific different with design of various targets
Regularization term learns vision-Semantic mapping.The chief motivation of mapping matrix study is between visual space and semantic space
It is aligned the minimum of loss, still, final goal is the unseen classification of classification.Therefore, visual signature and characterizing semantics should
It can be distinguished to identify different targets, and in existing field, this problem is all ignored always.Correlation technique is universal
There are some defects.It is exactly the CNN model from pre-training firstly, characteristics of image is not engineer, in this way to having no class
May not have enough characterization abilities for identification mission;Again, user-defined attribute is semantic description type, but not
Detailed, which limit act in classificatory identification;Finally, it is special to merge low layer face in existing research seldom in Unified frame
Sign is extracted and embedded space building.
Figure 1A is please referred to, for a kind of embodiment process signal of driver attributes' recognition methods provided by the embodiments of the present application
Figure.Driver attributes' recognition methods as described in this embodiment, comprising the following steps:
101, input picture is obtained, includes driver's area image in the input picture.
Wherein, the available input picture of electronic equipment, input picture can be shone by the whole body for the pedestrian that camera is shot,
Input picture can shoot gained by the camera of public place.It include driver's area image in input picture, i.e., by driving
The person's of sailing area image can identify the behavior property of driver.
Optionally, can also include the following steps: between above-mentioned steps 101- step 102
The input picture is pre-processed, target image is obtained;
Then, step 102, the input picture default convolutional neural networks model is input to handle, then it can be by
Implement according to such as under type:
The target image is input to default convolutional neural networks model to handle.
Wherein, electronic equipment can pre-process input picture, for example, size adjusting, background removal, etc. behaviour
Make, obtains target image.
Optionally, above-mentioned steps pre-process the input picture, obtain target image, it may include following steps:
A1, processing is zoomed in and out to the input picture, so that in the input picture after scaling processing and image library
Image size it is the same;
A2, image segmentation is carried out to the input picture after scaling processing, obtains the target image.
Wherein, electronic equipment can zoom in and out processing to input picture, so that input picture and figure after scaling processing
As the size of the image in library, then, processing is zoomed in and out to the input picture after the scaling processing, obtains target
Image, so, it is possible to reduce invalid data promote subsequent accuracy of identification.
Optionally, above-mentioned steps pre-process the input picture, obtain target image, it may include following steps:
B1, FIG pull handle is carried out to the input picture, obtains pedestrian area image;
B2, processing is zoomed in and out to the pedestrian area image, obtains the target image, the target figure after scaling processing
As the size of the image in image library.
Wherein it is possible to first carry out FIG pull handle to input picture, pedestrian area image, the specific algorithm of FIG pull handle are obtained
It can be image segmentation algorithm, for example, the image segmentation algorithm based on comentropy, cutting based on GraphCuts figure the image of algorithm
Partitioning algorithm, image segmentation algorithm based on watershed algorithm etc., are not limited thereto, secondly, to the pedestrian area image
Processing is zoomed in and out, so that it is consistent with the size of the image in image library, obtains target image.
102, the input picture is input to default convolutional neural networks model to handle, obtains the first attribute-bit
Collection, the default convolutional neural networks model includes characteristics of image network and embedded network.
Wherein, the default convolutional neural networks model can be by user's self-setting or system default.First attribute
Identification sets include at least an attribute-bit, and attribute-bit can be following at least one: hand, body, face, mobile phone, safety
Band cigarette, mask, mask, child, makes a phone call, smokes, having fastened the safety belt, wearing masks, embracing child etc..Above-mentioned default convolution
Neural network model may include characteristics of image network and embedded network, and characteristics of image network is for realizing feature extraction, insertion
For network for realizing Attribute Recognition, characteristics of image network and embedded network can be two-stage cascade network.
Optionally, in the embodiment of the present application, study characteristics of image can be responsible for from one by presetting convolutional neural networks model
Convolutional network starts, the selection for this network be it is flexible, can choose conv1 to the fc7 of VGG19 network, can also be with
Select conv1 to the pool5 of GoogLeNet.φ (x) can be expressed as by providing an image or its zoom area x, characteristics of image
=WIF*x.Wherein WIFThe entire parameter of character network is represented, * represents the sequence of operations of character network.In the embodiment of the present application
Model in, character network is the joint training in entire frame, and feature obtained can be very good to adjust with embedded part
Section.
Optionally, above-mentioned embedded network is based primarily upon the embedded space of one connection vision and semantic information of study.If language
Justice is characterized as ψ (y), i.e. user's Custom Attributes is Α, then mapping function may be defined as:
F(x,y;W)=φ (x)TWay
Attribute space and the compatibility score of embedded space are defined by inner product:
S=< WTφ(x),ay>
Similitude between compatibility score test image and the attribute annotations of class.Therefore, for learning matrix, can make
With the softmax loss of standard:
In this way, this basic incorporation model is suitable for most of existing learning methods, and achieve good performance.
However, it is to be limited size, usually not distinctive based on user-defined attribute.In order to solve this problem, it introduces
The attribute space of one enhancing, wherein image is mapped to user's defined attribute (UA) and implicit attribute (LA).
Optionally, the input picture is input to default convolutional neural networks model and handled by above-mentioned steps 102,
Obtain the first attribute-bit collection, it may include following steps:
21, feature extraction is carried out to the input picture by described image character network, obtains the first set of image characteristics;
22, Attribute Recognition is carried out to the first image feature set by the embedded network, obtains at least one attribute
Mark, at least one described attribute-bit are user's Custom Attributes mark;
23, attribute forecast is carried out by least one described attribute-bit, obtains target implicit attribute mark;
24, at least one described attribute-bit and implicit attribute mark are regard as the first attribute-bit collection.
Wherein, electronic equipment can carry out feature extraction to the input picture by above-mentioned characteristics of image network, obtain
First set of image characteristics, the first characteristics of image concentrate include multiple characteristic points, by embedded network to the first set of image characteristics into
Row Attribute Recognition obtains at least one attribute-bit, at least one attribute-bit is user's Custom Attributes mark, by least
One attribute-bit carries out attribute forecast, obtains target implicit attribute mark, for example, attribute-bit includes hand, mobile phone, then implies
Attribute can be to make a phone call, and regard at least one attribute-bit and implicit attribute mark as the first attribute-bit collection.
It is alternatively possible to enhance embedded images feature
The purpose of this way is to make embedded images feature φe(x) associated with user's defined attribute and implicit attribute.This
Apply for that embodiment can be by φe(x) two parts k-dim are divided into:
φe(x)=[φatt(x);φlat(x)],
Then, first k-dim is embedded in feature φatt(x) association and user's defined attribute, second k-dim component
φlat(x) association and implicit attribute.Based on this it is assumed that for φ similar with basic modelatt(x), softmax loss is utilized
To train this model.It is usual:
For second insertion feature φlat(x), it is therefore an objective to make the feature of study that there is distinctive for target identification.This
Application embodiment learns implicit attribute using triplet loss, while adjusting between the class of implicit attribute feature/inter- object distance:
Llat=max (0, m+d (φlat(xi),φlat(xk))-d(φlat(xi),φlat(xj)))
Wherein, xi,xkIt is similar image, xjIt is inhomogeneous image.D (x, y) is the Euclidean distance between x and y
Square.M is the edge of triplet loss.
User's defined attribute and implicit attribute are characterized in mapping from the expression of identical image, but use two differences
Matrix:
To each size, when network training includes softmax loss and triplet loss.For a double ruler
Network is spent, entirely implicit distinctive characteristic model is trained by following loss function:
The final goal function of multiple dimensioned network can be by polymerizeing all loss functions of all scales similarly come structure
It makes.
Optionally, above-mentioned steps 23 carry out attribute forecast by least one described attribute-bit, it may include following step
It is rapid:
According to the mapping relations between preset user's Custom Attributes and implicit attribute mark, determine it is described at least one
The corresponding target implicit attribute mark of attribute-bit.
Wherein, the mapping relations between Custom Attributes and implicit attribute mark can be stored in advance in electronic equipment, into
And the corresponding target implicit attribute mark of at least one attribute-bit can be determined according to the mapping relations.
103, the input picture is input to pre-set zoom network to handle, obtains target area image, the mesh
Mark the behavior region that area image is the driver.
Wherein, above-mentioned pre-set zoom network can be able to be a depth by user's self-setting or system default
Learn convolutional neural networks, can using the last convolutional layer of characteristics of image network output as scale network input, certainly,
It can also be using input picture as the input of pre-set zoom network.In order to effectively calculate, candidate region is assumed to be square, it
Position can be indicated with three parameters:
[zx,zy,zs]=WZ*φ(x)conv
Wherein, zx,zyIt represents and searches for square x-axis and y-axis coordinate, zsRepresent square length.φ(x)convRepresentative image
The output of the last convolutional layer of character network.Scaling network is the full convolution for following two layers of stacking of sigmoid activation primitive
Layer, WZRepresent the parameter of scaling network.
After obtaining the coordinate of square, region of search can directly cut from original image (i.e. input picture) and obtain.
Optionally, the pre-set zoom network includes sigmoid function;Above-mentioned steps 103 input the input picture
It is handled to pre-set zoom network, obtains target area image, it may include following steps:
31, the corresponding coordinate position in behavior region of the driver is determined;
32, a continuous mask of two dimension is generated by the sigmoid function;
33, it is based on the coordinate position, operation is carried out by the mask and the input picture, obtains the target area
Area image.
Wherein, the corresponding coordinate position in behavior region of driver can be realized by target identification, for example, identification is specified
Object, specified object can be following at least one: mobile phone, child, safety belt, mask, mask (facial mask) etc. are not made herein
It limits.In turn, a part of region comprising specified object can determine the corresponding coordinate position in behavior region of driver, electricity
Sub- equipment can generate a continuous mask of two dimension by sigmoid function, be based on coordinate position, schemed by mask and input
As carrying out operation, target area image is obtained.In the specific implementation, optimization of the discontinuous trimming operation in backpropagation is not
Easily.
In the specific implementation, generating the continuous mask M (x, y) of a bidimensional first with sigmoid function:
Mx=f (x-zx+0.5zs)-f(x-zx-0.5zs)
My=f (y-zy+0.5zs)-f(y-zy-0.5zs)
It is then possible to obtain clipping region by executing between original image x and continuous mask M by element multiplication ⊙:
xcrop=x ⊙ M
Finally, can also further use bilinear interpolation to obtain the more preferable expression of finer local clipping region
Adaptive clipping region is zoomed into size identical with original image, then, zoom area can be sent to next scale
Characteristics of image network in extract have more distinctive feature.
104, the target area image is input to the default convolutional neural networks model to handle, obtains second
Attribute-bit collection.
Wherein, the specific descriptions of above-mentioned steps 104 are referred to above-mentioned steps 102, and details are not described herein.
105, the corresponding mesh of the driver is determined according to the first attribute-bit collection and the second attribute-bit collection
Mark attribute-bit.
In the specific implementation, the second attribute-bit collection also corresponds to one since the first attribute-bit collection corresponds to some attribute-bits
A little attribute-bits can then be identified intersection between the two as objective attribute target attribute.
Optionally, in the implicit distinctive characteristic model proposed, input picture can be mapped to user and define category
In property and implicit attribute.Therefore, it can be predicted with federated user defined attribute and implicit attribute space.
User's defined attribute is predicted.A test image x is given, it can be mapped to the expression of user's defined attribute
For φatt(x), in order to predict its class label, compatibility score can be used to select most matched unseen classification:
Implicit attribute prediction, test image x, which also may map to implicit attribute, indicates φlat(x)。
Firstly, calculating the implicit attribute prototype for having no class.Specifically, all sample x of Cong Yijian class siMap theirs
Implicit attribute feature, and use implicit attribute prototype of the characteristic mean as class s, that is,
For having no class u, the embodiment of the present application calculates class u and the relationship between class s has been seen in user's defined attribute space.This
Kind relationship can be obtained by solving following ridge regression problem:
By to the identical relationship of implicit attribute space application, the available prototype for having no class u:
φ is indicated with implicit attributelat(x) classification results of test image x may be implemented as follows:
Secondly, merging multiple spaces.User's defined attribute and implicit attribute space are connected to carry out various definitions attribute connection
Close prediction:
Finally, fusion is multiple dimensioned.For double scales imply distinctive model, obtained on each scale user property and
The Analysis On Multi-scale Features of acquisition are combined to obtain further improvement, drive in this way, can precisely identify by implicit attribute feature
The person's of sailing Attribute Recognition.
Under illustration, as shown in Figure 1B, first order scale space can be input to default volume based on entire input picture
Product neural network model, i.e., successively handle input picture by characteristics of image network and embedded network, obtain the first attribute
Identification sets, second level dimensional space can be input to default convolutional neural networks model with Behavior-based control area image, i.e., by image
Character network and embedded network are successively handled behavior area image, obtain the second attribute-bit collection, are finally belonged to by first
Property identification sets, the second attribute-bit collection merge multiple dimensioned realizations driver attributes and identify.
Using above-mentioned driver attributes' recognition methods provided by the embodiments of the present application, input picture is obtained, in input picture
Comprising driver's area image, input picture is input to default convolutional neural networks model and is handled, the first attribute is obtained
Identification sets, presetting convolutional neural networks model includes characteristics of image network and embedded network, and input picture is input to default contracting
It puts network to be handled, obtains target area image, target area image is the behavior region of driver, by target area image
It is input to default convolutional neural networks model to be handled, obtains the second attribute-bit collection, according to the first attribute-bit collection and
Two attribute-bit collection determine driver's corresponding objective attribute target attribute mark, can the attribute to driver precisely identified.
Consistent with the abovely, specific as follows the following are the device for implementing above-mentioned driver attributes' recognition methods:
Referring to Fig. 2, for a kind of example structure signal of driver attributes' identification device provided by the embodiments of the present application
Figure.Driver attributes' identification device as described in this embodiment, comprising: acquiring unit 201, first processing units 202, second
Processing unit 203 and determination unit 204, specific as follows:
Acquiring unit 201 includes driver's area image in the input picture for obtaining input picture;
First processing units 202 are handled for the input picture to be input to default convolutional neural networks model,
The first attribute-bit collection is obtained, the default convolutional neural networks model includes characteristics of image network and embedded network;
The second processing unit 203 handles for the input picture to be input to pre-set zoom network, obtains target
Area image, the target area image are the behavior region of the driver;
The first processing units 201 are input to the default convolution mind also particularly useful for by the target area image
It is handled through network model, obtains the second attribute-bit collection;
Determination unit 204, for being driven according to the first attribute-bit collection and the second attribute-bit collection determination
The corresponding objective attribute target attribute mark of the person of sailing.
Optionally, the pre-set zoom network includes sigmoid function;
It the input picture is input to pre-set zoom network handles described, in terms of obtaining target area image,
Described the second processing unit 202 is specifically used for:
Determine the corresponding coordinate position in behavior region of the driver;
A continuous mask of two dimension is generated by the sigmoid function;
Based on the coordinate position, operation is carried out by the mask and the input picture, obtains the target area
Image.
Optionally, in terms of the corresponding coordinate position in behavior region of the determination driver, the second processing
Unit 202 is specifically used for:
Target detection is carried out to the target image, obtains goal-selling;
Determine the rectangular area of the pre-set dimension size comprising the goal-selling;
Obtain behavior region corresponding coordinate position of the apex coordinate as the driver of the rectangular area.
Optionally, it the input picture is input to default convolutional neural networks model handles described, obtain
In terms of one attribute-bit collection, the first processing units 201 are specifically used for:
Feature extraction is carried out to the input picture by described image character network, obtains the first set of image characteristics;
Attribute Recognition is carried out to the first image feature set by the embedded network, obtains at least one attribute mark
Know, at least one described attribute-bit is user's Custom Attributes mark;
Attribute forecast is carried out by least one described attribute-bit, obtains target implicit attribute mark;
It regard at least one described attribute-bit and implicit attribute mark as the first attribute-bit collection.
Optional, in terms of the progress attribute forecast by least one described attribute-bit, first processing is single
Member 201 is specifically used for:
According to the mapping relations between preset user's Custom Attributes and implicit attribute mark, determine it is described at least one
The corresponding target implicit attribute mark of attribute-bit.
Using above-mentioned driver attributes' identification device provided by the embodiments of the present application, input picture is obtained, in input picture
Comprising driver's area image, input picture is input to default convolutional neural networks model and is handled, the first attribute is obtained
Identification sets, presetting convolutional neural networks model includes characteristics of image network and embedded network, and input picture is input to default contracting
It puts network to be handled, obtains target area image, target area image is the behavior region of driver, by target area image
It is input to default convolutional neural networks model to be handled, obtains the second attribute-bit collection, according to the first attribute-bit collection and
Two attribute-bit collection determine driver's corresponding objective attribute target attribute mark, can the attribute to driver precisely identified.
Consistent with the abovely, referring to Fig. 3, the example structure for a kind of electronic equipment provided by the embodiments of the present application is shown
It is intended to.Electronic equipment as described in this embodiment, comprising: at least one input equipment 1000;At least one output equipment
2000;At least one processor 3000, such as CPU;With memory 4000, above-mentioned input equipment 1000, output equipment 2000, place
Reason device 3000 and memory 4000 are connected by bus 5000.
Wherein, above-mentioned input equipment 1000 concretely touch panel, physical button or mouse.
Above-mentioned output equipment 2000 concretely display screen.
Above-mentioned memory 4000 can be high speed RAM memory, can also be nonvolatile storage (non-volatile
), such as magnetic disk storage memory.Above-mentioned memory 4000 is used to store a set of program code, above-mentioned input equipment 1000, defeated
Equipment 2000 and processor 3000 are used to call the program code stored in memory 4000 out, perform the following operations:
Above-mentioned processor 3000, is used for:
Input picture is obtained, includes driver's area image in the input picture;
The input picture is input to default convolutional neural networks model to handle, obtains the first attribute-bit collection,
The default convolutional neural networks model includes characteristics of image network and embedded network;
The input picture is input to pre-set zoom network to handle, obtains target area image, the target area
Area image is the behavior region of the driver;
The target area image is input to the default convolutional neural networks model to handle, obtains the second attribute
Identification sets;
The corresponding target category of the driver is determined according to the first attribute-bit collection and the second attribute-bit collection
Property mark.
Using above-mentioned electronic equipment provided by the embodiments of the present application, input picture is obtained, includes driver in input picture
Input picture is input to default convolutional neural networks model and handled, obtains the first attribute-bit collection by area image, is preset
Convolutional neural networks model includes characteristics of image network and embedded network, and input picture is input at pre-set zoom network
Reason obtains target area image, and target area image is the behavior region of driver, and target area image is input to default volume
Product neural network model is handled, and the second attribute-bit collection is obtained, according to the first attribute-bit collection and the second attribute-bit collection
Determine driver's corresponding objective attribute target attribute mark, can the attribute to driver precisely identified.
Optionally, the pre-set zoom network includes sigmoid function;
It the input picture is input to pre-set zoom network handles described, in terms of obtaining target area image,
Above-mentioned processor 3000 is specifically used for:
Determine the corresponding coordinate position in behavior region of the driver;
A continuous mask of two dimension is generated by the sigmoid function;
Based on the coordinate position, operation is carried out by the mask and the input picture, obtains the target area
Image.
Optionally, in terms of the corresponding coordinate position in behavior region of the determination driver, above-mentioned processor
3000 are specifically used for:
Target detection is carried out to the target image, obtains goal-selling;
Determine the rectangular area of the pre-set dimension size comprising the goal-selling;
Obtain behavior region corresponding coordinate position of the apex coordinate as the driver of the rectangular area.
It is optional, it the input picture is input to default convolutional neural networks model handles described, obtain the
In terms of one attribute-bit collection, above-mentioned processor 3000 is specifically used for:
Feature extraction is carried out to the input picture by described image character network, obtains the first set of image characteristics;
Attribute Recognition is carried out to the first image feature set by the embedded network, obtains at least one attribute mark
Know, at least one described attribute-bit is user's Custom Attributes mark;
Attribute forecast is carried out by least one described attribute-bit, obtains target implicit attribute mark;
It regard at least one described attribute-bit and implicit attribute mark as the first attribute-bit collection.
Optionally, in terms of the progress attribute forecast by least one described attribute-bit, above-mentioned processor 3000
It is specifically used for:
According to the mapping relations between preset user's Custom Attributes and implicit attribute mark, determine it is described at least one
The corresponding target implicit attribute mark of attribute-bit.
The embodiment of the present application also provides a kind of computer storage medium, wherein the computer storage medium can be stored with journey
Sequence, the program include the part or complete for any driver attributes' recognition methods recorded in above method embodiment when executing
Portion's step.
The embodiment of the present application also provides a kind of computer program product, and the computer program product includes storing calculating
The non-transient computer readable storage medium of machine program, the computer program are operable to that computer is made to execute such as above-mentioned side
Some or all of any driver attributes' recognition methods recorded in method embodiment step.
Although the application is described in conjunction with each embodiment herein, however, implementing the application claimed
In the process, those skilled in the art are by checking the attached drawing, disclosure and the appended claims, it will be appreciated that and it is real
Other variations of the existing open embodiment.In the claims, " comprising " (comprising) word is not excluded for other compositions
Part or step, "a" or "an" are not excluded for multiple situations.Claim may be implemented in single processor or other units
In several functions enumerating.Mutually different has been recited in mutually different dependent certain measures, it is not intended that these are arranged
It applies to combine and generates good effect.
It will be understood by those skilled in the art that embodiments herein can provide as method, apparatus (equipment) or computer journey
Sequence product.Therefore, complete hardware embodiment, complete software embodiment or combining software and hardware aspects can be used in the application
The form of embodiment.Moreover, it wherein includes the calculating of computer usable program code that the application, which can be used in one or more,
The computer program implemented in machine usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.)
The form of product.Computer program is stored/distributed in suitable medium, is provided together with other hardware or as the one of hardware
Part can also use other distribution forms, such as pass through the wired or wireless telecommunication system of Internet or other.
The application be referring to the embodiment of the present application method, apparatus (equipment) and computer program product flow chart with/
Or block diagram describes.It should be understood that each process that can be realized by computer program instructions in flowchart and/or the block diagram and/
Or the combination of the process and/or box in box and flowchart and/or the block diagram.It can provide these computer program instructions
To general purpose computer, special purpose computer, Embedded Processor or other programmable License Plate equipment processor to generate one
A machine so that by instructions that computer or processors of other programmable License Plate equipment execute generate for realizing
The device for the function of being specified in one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable License Plate equipment with spy
Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,
Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or
The function of being specified in multiple boxes.
These computer program instructions can also be loaded into computer or other programmable License Plate equipment, so that counting
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or
The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one
The step of function of being specified in a box or multiple boxes.
Although the application is described in conjunction with specific features and embodiment, it is clear that, do not departing from this Shen
In the case where spirit and scope please, it can be carry out various modifications and is combined.Correspondingly, the specification and drawings are only institute
The exemplary illustration for the application that attached claim is defined, and be considered as covered within the scope of the application any and all and repair
Change, change, combining or equivalent.Obviously, those skilled in the art the application can be carried out various modification and variations without
It is detached from spirit and scope.If in this way, these modifications and variations of the application belong to the claim of this application and its
Within the scope of equivalent technologies, then the application is also intended to include these modifications and variations.
Claims (10)
1. a kind of driver attributes' recognition methods characterized by comprising
Input picture is obtained, includes driver's area image in the input picture;
The input picture is input to default convolutional neural networks model to handle, obtains the first attribute-bit collection, it is described
Default convolutional neural networks model includes characteristics of image network and embedded network;
The input picture is input to pre-set zoom network to handle, obtains target area image, the target area figure
Behavior region as being the driver;
The target area image is input to the default convolutional neural networks model to handle, obtains the second attribute-bit
Collection;
The corresponding objective attribute target attribute mark of the driver is determined according to the first attribute-bit collection and the second attribute-bit collection
Know.
2. the method according to claim 1, wherein the pre-set zoom network includes sigmoid function;
It is described the input picture is input to pre-set zoom network to handle, obtain target area image, comprising:
Determine the corresponding coordinate position in behavior region of the driver;
A continuous mask of two dimension is generated by the sigmoid function;
Based on the coordinate position, operation is carried out by the mask and the input picture, obtains the target area image.
3. according to the method described in claim 2, it is characterized in that, the corresponding seat in behavior region of the determination driver
Cursor position, comprising:
Target detection is carried out to the target image, obtains goal-selling;
Determine the rectangular area of the pre-set dimension size comprising the goal-selling;
Obtain behavior region corresponding coordinate position of the apex coordinate as the driver of the rectangular area.
4. method according to any one of claims 1 to 3, which is characterized in that it is described the input picture is input to it is pre-
If convolutional neural networks model is handled, the first attribute-bit collection is obtained, comprising:
Feature extraction is carried out to the input picture by described image character network, obtains the first set of image characteristics;
Attribute Recognition is carried out to the first image feature set by the embedded network, obtains at least one attribute-bit, institute
At least one attribute-bit is stated as user's Custom Attributes mark;
Attribute forecast is carried out by least one described attribute-bit, obtains target implicit attribute mark;
It regard at least one described attribute-bit and implicit attribute mark as the first attribute-bit collection.
5. according to the method described in claim 4, it is characterized in that, described carry out attribute by least one described attribute-bit
Prediction, comprising:
According to the mapping relations between preset user's Custom Attributes and implicit attribute mark, at least one described attribute is determined
Identify corresponding target implicit attribute mark.
6. a kind of driver attributes' identification device characterized by comprising
Acquiring unit includes driver's area image in the input picture for obtaining input picture;
First processing units are handled for the input picture to be input to default convolutional neural networks model, obtain
One attribute-bit collection, the default convolutional neural networks model includes characteristics of image network and embedded network;
The second processing unit handles for the input picture to be input to pre-set zoom network, obtains target area figure
Picture, the target area image are the behavior region of the driver;
The first processing units are input to the default convolutional neural networks mould also particularly useful for by the target area image
Type is handled, and the second attribute-bit collection is obtained;
Determination unit, for determining that the driver is corresponding according to the first attribute-bit collection and the second attribute-bit collection
Objective attribute target attribute mark.
7. device according to claim 6, which is characterized in that the pre-set zoom network includes sigmoid function;
It the input picture is input to pre-set zoom network handles described, it is described in terms of obtaining target area image
The second processing unit is specifically used for:
Determine the corresponding coordinate position in behavior region of the driver;
A continuous mask of two dimension is generated by the sigmoid function;
Based on the coordinate position, operation is carried out by the mask and the input picture, obtains the target area image.
8. device according to claim 7, which is characterized in that corresponding in the behavior region of the determination driver
In terms of coordinate position, described the second processing unit is specifically used for:
Target detection is carried out to the target image, obtains goal-selling;
Determine the rectangular area of the pre-set dimension size comprising the goal-selling;
Obtain behavior region corresponding coordinate position of the apex coordinate as the driver of the rectangular area.
9. according to the described in any item devices of claim 6 to 8, which is characterized in that be input to the input picture described
Default convolutional neural networks model is handled, and in terms of obtaining the first attribute-bit collection, the first processing units are specifically used for:
Feature extraction is carried out to the input picture by described image character network, obtains the first set of image characteristics;
Attribute Recognition is carried out to the first image feature set by the embedded network, obtains at least one attribute-bit, institute
At least one attribute-bit is stated as user's Custom Attributes mark;
Attribute forecast is carried out by least one described attribute-bit, obtains target implicit attribute mark;
It regard at least one described attribute-bit and implicit attribute mark as the first attribute-bit collection.
10. a kind of computer readable storage medium, which is characterized in that storage is used for the computer program of electronic data interchange,
In, the computer program makes computer execute the method according to claim 1 to 5.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811375755.4A CN109784140A (en) | 2018-11-19 | 2018-11-19 | Driver attributes' recognition methods and Related product |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811375755.4A CN109784140A (en) | 2018-11-19 | 2018-11-19 | Driver attributes' recognition methods and Related product |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109784140A true CN109784140A (en) | 2019-05-21 |
Family
ID=66496052
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811375755.4A Pending CN109784140A (en) | 2018-11-19 | 2018-11-19 | Driver attributes' recognition methods and Related product |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109784140A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111222566A (en) * | 2020-01-02 | 2020-06-02 | 平安科技(深圳)有限公司 | User attribute identification method, device 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 |
CN111832526A (en) * | 2020-07-23 | 2020-10-27 | 浙江蓝卓工业互联网信息技术有限公司 | Behavior detection method and device |
CN112926427A (en) * | 2021-02-18 | 2021-06-08 | 浙江智慧视频安防创新中心有限公司 | Target user dressing attribute identification method and device |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106682602A (en) * | 2016-12-16 | 2017-05-17 | 深圳市华尊科技股份有限公司 | Driver behavior identification method and terminal |
JP2017102599A (en) * | 2015-11-30 | 2017-06-08 | 日本電信電話株式会社 | Estimation device, parameter learning device, method, and program |
CN106815574A (en) * | 2017-01-20 | 2017-06-09 | 博康智能信息技术有限公司北京海淀分公司 | Set up detection model, detect the method and apparatus for taking mobile phone behavior |
CN107273870A (en) * | 2017-07-07 | 2017-10-20 | 郑州航空工业管理学院 | The pedestrian position detection method of integrating context information under a kind of monitoring scene |
CN108388880A (en) * | 2018-03-15 | 2018-08-10 | 广东工业大学 | A kind of method and device that monitoring driver drives using mobile phone |
CN108764306A (en) * | 2018-05-15 | 2018-11-06 | 深圳大学 | Image classification method, device, computer equipment and storage medium |
-
2018
- 2018-11-19 CN CN201811375755.4A patent/CN109784140A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2017102599A (en) * | 2015-11-30 | 2017-06-08 | 日本電信電話株式会社 | Estimation device, parameter learning device, method, and program |
CN106682602A (en) * | 2016-12-16 | 2017-05-17 | 深圳市华尊科技股份有限公司 | Driver behavior identification method and terminal |
CN106815574A (en) * | 2017-01-20 | 2017-06-09 | 博康智能信息技术有限公司北京海淀分公司 | Set up detection model, detect the method and apparatus for taking mobile phone behavior |
CN107273870A (en) * | 2017-07-07 | 2017-10-20 | 郑州航空工业管理学院 | The pedestrian position detection method of integrating context information under a kind of monitoring scene |
CN108388880A (en) * | 2018-03-15 | 2018-08-10 | 广东工业大学 | A kind of method and device that monitoring driver drives using mobile phone |
CN108764306A (en) * | 2018-05-15 | 2018-11-06 | 深圳大学 | Image classification method, device, computer equipment and storage medium |
Non-Patent Citations (1)
Title |
---|
YAN LI 等: "Discriminative Learning of Latent Features for Zero-Shot Recognition", 《2018 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111222566A (en) * | 2020-01-02 | 2020-06-02 | 平安科技(深圳)有限公司 | User attribute identification method, device 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 |
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 |
CN111832526A (en) * | 2020-07-23 | 2020-10-27 | 浙江蓝卓工业互联网信息技术有限公司 | Behavior detection method and device |
CN111832526B (en) * | 2020-07-23 | 2024-06-11 | 浙江蓝卓工业互联网信息技术有限公司 | Behavior detection method and device |
CN112926427A (en) * | 2021-02-18 | 2021-06-08 | 浙江智慧视频安防创新中心有限公司 | Target user dressing attribute identification method and device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109657533B (en) | Pedestrian re-identification method and related product | |
CN109784140A (en) | Driver attributes' recognition methods and Related product | |
CN111832468B (en) | Gesture recognition method and device based on biological recognition, computer equipment and medium | |
CN110033018B (en) | Graph similarity judging method and device and computer readable storage medium | |
KR20180104609A (en) | Method, system, apparatus and readable storage medium for realizing insurance claims fraud prevention based on a plurality of image correspondence | |
CN110555481A (en) | Portrait style identification method and device and computer readable storage medium | |
CN106030610B (en) | The real-time 3D gesture recognition and tracking system of mobile device | |
CN110197146A (en) | Facial image analysis method, electronic device and storage medium based on deep learning | |
Bertoni et al. | Perceiving humans: from monocular 3d localization to social distancing | |
CN109800682A (en) | Driver attributes' recognition methods and Related product | |
CN111242030A (en) | Video data processing method, device, equipment and computer readable storage medium | |
CN110851641A (en) | Cross-modal retrieval method and device and readable storage medium | |
CN110738650B (en) | Infectious disease infection identification method, terminal device and storage medium | |
US20230334893A1 (en) | Method for optimizing human body posture recognition model, device and computer-readable storage medium | |
CN109670517A (en) | Object detection method, device, electronic equipment and target detection model | |
CN112132032A (en) | Traffic sign detection method and device, electronic equipment and storage medium | |
CN110147731A (en) | Vehicle type recognition method and Related product | |
CN111507285A (en) | Face attribute recognition method and device, computer equipment and storage medium | |
CN114241524A (en) | Human body posture estimation method and device, electronic equipment and readable storage medium | |
CN114168768A (en) | Image retrieval method and related equipment | |
CN114332484A (en) | Key point detection method and device, computer equipment and storage medium | |
CN112819510A (en) | Fashion trend prediction method, system and equipment based on clothing multi-attribute recognition | |
CN110502961A (en) | A kind of facial image detection method and device | |
CN114792401A (en) | Training method, device and equipment of behavior recognition model and storage medium | |
CN115131826B (en) | Article detection and identification method, and network model training method and device |
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: 20190521 |
|
RJ01 | Rejection of invention patent application after publication |