Summary of the invention
The embodiment of the present application proposes image processing method and device.
In a first aspect, the embodiment of the present application provides a kind of image processing method, comprising: the image comprising target is obtained,
Change of scale is carried out to image, the image at least one scale that obtains that treated;By acquired image and treated figure
As input convolutional neural networks, the candidate frame of the position of characteristic pattern and multiple instruction targets is obtained, wherein each target is corresponding extremely
Few two candidate frames;In the candidate frame in each image, candidate frame of the size in preset size range is determined, wherein
The size range of candidate frame corresponding to the image of different scale is different;Determine at least one in the candidate frame in size range
A candidate frame region corresponding in characteristic pattern, obtains feature corresponding to region, inputs the full connection of convolutional neural networks
Layer.
In some embodiments, determining at least one candidate frame in the candidate frame in size range in characteristic pattern
Before corresponding region, method further include: non-maxima suppression is carried out to the candidate frame in preset size range, with
To at least one candidate frame.
In some embodiments, change of scale is carried out to image, comprising: up-sampling and/or down-sampling are carried out to image,
In, the size range of candidate frame corresponding to the image that down-sampling obtains is to up-sample more than or equal to the first preset threshold
The size range of candidate frame corresponding to the image arrived is less than or equal to the second preset threshold, and the first preset threshold is greater than second
Preset threshold.
In some embodiments, the size range of candidate frame corresponding to acquired image is in third predetermined threshold value and
Between four preset thresholds, wherein third predetermined threshold value is greater than the 4th preset threshold, and it is pre- that third predetermined threshold value is greater than or equal to first
If threshold value, the 4th preset threshold is less than or equal to the second preset threshold.
In some embodiments, in response to there are at least two graphical rules to be greater than acquired figure in treated image
The scale of picture, at least two images, the size range of candidate frame corresponding to the lesser image of scale is specified less than first
The size range of threshold value, candidate frame corresponding to the biggish image of scale is less than the second specified threshold, and the first specified threshold is big
In the second specified threshold.
In some embodiments, in response in treated image there are more than two graphical rules be less than it is acquired
The scale of image, in more than two images, the size range of candidate frame corresponding to the lesser image of scale is greater than third
The size range of specified threshold, candidate frame corresponding to the biggish image of scale is greater than the 4th specified threshold, and third specifies threshold
Value is greater than the 4th specified threshold.
Second aspect, the embodiment of the present application provide a kind of image processing apparatus, comprising: acquiring unit is configured to obtain
The image comprising target is taken, change of scale is carried out to image, the image at least one scale that obtains that treated;Input unit,
It is configured to acquired image and treated image input convolutional neural networks obtaining characteristic pattern and multiple instruction targets
Position candidate frame, wherein corresponding at least two candidate frames of each target;Determination unit is configured in each image
Candidate frame in, determine candidate frame of the size in preset size range, wherein candidate corresponding to the image of different scale
The size range of frame is different;Area determination unit is configured to determine at least one of candidate frame in size range time
The region for selecting frame corresponding in characteristic pattern obtains feature corresponding to region, inputs the full articulamentum of convolutional neural networks.
In some embodiments, device further include: selection unit is configured to the time in preset size range
Frame is selected to carry out non-maxima suppression, to obtain at least one candidate frame.
In some embodiments, the acquiring unit, is further configured to: up-sampling and/or down-sampling are carried out to image,
Wherein, the size range of candidate frame corresponding to the image that down-sampling obtains is more than or equal to the first preset threshold, up-sampling
The size range of candidate frame corresponding to obtained image is less than or equal to the second preset threshold, and the first preset threshold is greater than the
Two preset thresholds.
In some embodiments, the size range of candidate frame corresponding to acquired image is in third predetermined threshold value and
Between four preset thresholds, wherein third predetermined threshold value is greater than the 4th preset threshold, and it is pre- that third predetermined threshold value is greater than or equal to first
If threshold value, the 4th preset threshold is less than or equal to the second preset threshold.
In some embodiments, in response to there are at least two graphical rules to be greater than acquired figure in treated image
The scale of picture, at least two images, the size range of candidate frame corresponding to the lesser image of scale is specified less than first
The size range of threshold value, candidate frame corresponding to the biggish image of scale is less than the second specified threshold, and the first specified threshold is big
In the second specified threshold.
In some embodiments, in response in treated image there are more than two graphical rules be less than it is acquired
The scale of image, in more than two images, the size range of candidate frame corresponding to the lesser image of scale is greater than third
The size range of specified threshold, candidate frame corresponding to the biggish image of scale is greater than the 4th specified threshold, and third specifies threshold
Value is greater than the 4th specified threshold.
The third aspect, the embodiment of the present application provide a kind of electronic equipment, comprising: one or more processors;Storage dress
It sets, for storing one or more programs, when one or more programs are executed by one or more processors, so that one or more
A processor realizes the method such as any embodiment in image processing method.
Fourth aspect, the embodiment of the present application provide a kind of computer readable storage medium, are stored thereon with computer journey
Sequence realizes the method such as any embodiment in image processing method when the program is executed by processor.
Image procossing scheme provided by the embodiments of the present application carries out scale to image firstly, obtaining the image comprising target
Transformation, the image at least one scale that obtains that treated.Later, by acquired image, image inputs convolution with treated
Neural network obtains the candidate frame of the position of characteristic pattern and multiple instruction targets, wherein each target corresponding at least two is candidate
Frame.Then, in the candidate frame in each image, candidate frame of the size in preset size range is determined, wherein different rulers
The size range of candidate frame corresponding to the image of degree is different.Finally, determining at least one in the candidate frame in size range
A candidate frame region corresponding in characteristic pattern, obtains feature corresponding to region, inputs the full connection of convolutional neural networks
Layer.Method provided by the embodiments of the present application can by determining the candidate frames of different size ranges from the image of different scale, with
To different size of Target Acquisition feature more abundant.
Specific embodiment
The application is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining related invention, rather than the restriction to the invention.It also should be noted that in order to
Convenient for description, part relevant to related invention is illustrated only in attached drawing.
It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present application can phase
Mutually combination.The application is described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.
Fig. 1 is shown can be using the exemplary system of the embodiment of the image processing method or image processing apparatus of the application
System framework 100.
As shown in Figure 1, system architecture 100 may include terminal device 101,102,103, network 104 and server 105.
Network 104 between terminal device 101,102,103 and server 105 to provide the medium of communication link.Network 104 can be with
Including various connection types, such as wired, wireless communication link or fiber optic cables etc..
User can be used terminal device 101,102,103 and be interacted by network 104 with server 105, to receive or send out
Send message etc..Various telecommunication customer end applications can be installed on terminal device 101,102,103, such as image processing application,
Video class application, live streaming application, instant messaging tools, mailbox client, social platform software etc..
Here terminal device 101,102,103 can be hardware, be also possible to software.When terminal device 101,102,
103 be hardware when, can be the various electronic equipments with display screen, including but not limited to smart phone, tablet computer, electronics
Book reader, pocket computer on knee and desktop computer etc..It, can be with when terminal device 101,102,103 is software
It is mounted in above-mentioned cited electronic equipment.Multiple softwares or software module may be implemented into (such as providing distribution in it
The multiple softwares or software module of formula service), single software or software module also may be implemented into.It is not specifically limited herein.
Server 105 can be to provide the server of various services, such as provide support to terminal device 101,102,103
Background server.Background server can carry out analyzing etc. to data such as the images received processing, and by processing result (example
Such as feature) feed back to terminal device.
It should be noted that image processing method provided by the embodiment of the present application can be by server 105 or terminal
Equipment 101,102,103 executes, correspondingly, image processing apparatus can be set in server 105 or terminal device 101,
102, in 103.
It should be understood that the number of terminal device, network and server in Fig. 1 is only schematical.According to realization need
It wants, can have any number of terminal device, network and server.
With continued reference to Fig. 2, the process 200 of one embodiment of the image processing method according to the application is shown.The figure
As processing method, comprising the following steps:
Step 201, obtain include target image, change of scale is carried out to image, at least one ruler that obtains that treated
The image of degree.
In the present embodiment, the executing subject (such as server shown in FIG. 1 or terminal device) of image processing method can
To obtain the image for including target, and change of scale is carried out to acquired image, at least one ruler that obtains that treated
The image of degree.The object for having certain meaning that target is presented by image, such as trees, house.It may include phase in the picture
Same target or various targets not of uniform size, that pattern is different.
Herein, scale refers to the pixel number of image.For example, the scale of acquired image is 224 × 224, scale becomes
The graphical rule obtained after changing is 256 × 256.Specifically, change of scale can be using at least one in up-sampling and down-sampling
Kind.
Step 202, by acquired image and treated that image inputs convolutional neural networks, characteristic pattern and multiple is obtained
Indicate the candidate frame of the position of target, wherein corresponding at least two candidate frames of each target.
In the present embodiment, acquired image can be inputted convolutional neural networks by above-mentioned executing subject, also, can also
Convolutional neural networks are inputted with the image for obtaining change of scale, obtain the candidate frame of the position of multiple instruction targets
(proposals) and characteristic pattern (feature map).Specifically, above-mentioned executing subject can be determined using various ways and be waited
Select frame.Such as in the case where convolutional neural networks include region candidate network (Region Proposal Network RPN),
Candidate frame can be determined using region candidate network.Further, it is also possible to true using selective search (Selective Search)
Determine candidate frame.Characteristic pattern can be obtained by the convolutional layer of convolutional neural networks, and the characteristic pattern obtained by different images convolution is not
Together.Here candidate frame can be expressed as positions and dimensions.Position can indicate by the coordinate of some point of candidate frame, than
Such as midpoint or top left corner apex.Size can pass through area, perimeter or width, high expression.
Step 203, in the candidate frame in each image, candidate frame of the size in preset size range is determined,
In, the size range of candidate frame corresponding to the image of different scale is different.
In the present embodiment, above-mentioned executing subject can determine size in preset ruler in the candidate frame of each image
Candidate frame in very little range.Because the size range of candidate frame corresponding to the image of different scale is different, determining ruler
When candidate frame in very little range, the size of candidate frame determined by the image to different scale is not quite similar.Corresponding to image
Candidate frame, which refers to, inputs the obtained candidate frame of convolutional neural networks for image.
For example, the available scale of above-mentioned executing subject is 224 × 224 original image, and carries out down-sampling, obtains one
The small image that a scale is 112 × 112.The corresponding candidate frame of original image and the corresponding candidate frame of small image can be set respectively in advance
Determine size range: less than 8 × 8 and be greater than 8 × 8, or less than 9 × 9 and greater than 8 × 8 etc..
In some optional implementations of the present embodiment, in response to there are at least two images in treated image
Scale is greater than the scale of acquired image, at least two images, the size of candidate frame corresponding to the lesser image of scale
Range is less than the first specified threshold, and the size range of candidate frame corresponding to the biggish image of scale is less than the second specified threshold
Value, the first specified threshold are greater than the second specified threshold.
In response in treated image there are the scale that more than two graphical rules are less than acquired image, two
In above image, the size range of candidate frame corresponding to the lesser image of scale be greater than third specified threshold, scale compared with
The size range of candidate frame corresponding to big image is greater than the 4th specified threshold, and third specified threshold is greater than the 4th specified threshold
Value.
Numerical value in these optional implementations, in the size range of candidate frame corresponding to the biggish image of scale
Smaller, and the numerical value in the size range of candidate frame corresponding to the lesser image of scale is larger, the two size ranges can be with
It partially overlaps.
For example, original image scale is 128 × 128, and after up-sampling, obtained image is the A figure that scale is 224 × 224
The B image that picture and scale are 256 × 256.The size range of candidate frame corresponding to A image can be less than 6 × 6 (here two
A 6 be respectively wide and high pixel numbers), the size range of candidate frame corresponding to B image can be less than 5 × 5.
Clarification of objective in the biggish image of the scale of these implementations is easier to be acquired, and can embody target
More details.And the target in the lesser image of scale is more able to reflect the global feature of target.Therefore, it is possible to from scale compared with
Emphasis determines lesser target in big image, and emphasis determines biggish target in the lesser image of scale, with more acurrate
Ground obtains different size of clarification of objective.
Corresponding to step 204, determining at least one candidate frame in the candidate frame in size range in characteristic pattern
Region obtains the feature in region, inputs the full articulamentum of convolutional neural networks.
In the present embodiment, above-mentioned executing subject can determine at least one of the candidate frame in size range candidate
Frame region corresponding in characteristic pattern.Later, the feature in region is obtained, and acquired feature is inputted into convolutional neural networks
Full articulamentum (Connected Layer), (such as can be to full articulamentum to carry out the subsequent processing of convolutional neural networks
As a result classified and returned), obtain the final output of convolutional neural networks.Above-mentioned executing subject is in the feature for obtaining region
When, the eigenmatrix of part corresponding to above-mentioned zone can be determined from the eigenmatrix corresponding to characteristic pattern, and extract
Come.
Characteristic pattern corresponding to different images is different.Candidate frame in the size range corresponding to each image is multiple
In the case where, it can determine each candidate frame corresponding different zones of institute in characteristic pattern.
Above-mentioned steps 204 can be real by the specific pond layer (ROI Pooling Layer) in convolutional neural networks
It is existing.
In some optional implementations of the present embodiment, before above-mentioned steps 204, this method can also include:
Non-maxima suppression is carried out to the candidate frame in preset size range, to obtain at least one above-mentioned candidate
Frame.
In these optional implementations, above-mentioned executing subject can to the candidate frame in preset size range into
Row non-maxima suppression (Non-Maximum Suppression, NMS), to be generated by above-mentioned non-maxima suppression process
At least one candidate frame stated.Then, above-mentioned executing subject can determine at least one candidate frame generated in characteristic pattern
Corresponding region.Non-maxima suppression can screen candidate frame, obtain and the callout box place for label target
The candidate frame being positioned relatively close to.
These implementations can remove the poor candidate frame of accuracy by non-maxima suppression, and increase obtains target
The accuracy of the feature taken.
With continued reference to the schematic diagram that Fig. 3, Fig. 3 are according to the application scenarios of the image processing method of the present embodiment.?
In the application scenarios of Fig. 3, the available image 302 comprising target of executing subject 301 carries out change of scale to image 302, obtains
The image 303 of at least one scale to treated;By acquired image, image inputs convolutional neural networks with treated,
Obtain the candidate frame 305 of the position of characteristic pattern 304 and multiple instruction targets, wherein corresponding at least two candidate frames of each target;
In the candidate frame in each image, candidate frame 306 of the size in preset size range is determined, wherein different scale
The size range of candidate frame corresponding to image is different;Determine that at least one candidate frame in the candidate frame in size range exists
Corresponding region 307 in characteristic pattern, obtains feature 308 corresponding to region, inputs the full articulamentum of convolutional neural networks.
The method provided by the above embodiment of the application can be by determining different size ranges from the image of different scale
Candidate frame, more abundant and accurate feature can be obtained with the target to all size.
With further reference to Fig. 4, it illustrates the processes 400 of another embodiment of image processing method.The image procossing
The process 400 of method, comprising the following steps:
Step 401, obtain include target image, up-sampling and/or down-sampling are carried out to image, obtain that treated extremely
The image of few a kind of scale, wherein the size range of candidate frame corresponding to the image that down-sampling obtains is more than or equal to the
One preset threshold, the size range of candidate frame corresponding to the image up-sampled be less than or equal to the second preset threshold,
First preset threshold is greater than the second preset threshold.
In the present embodiment, executing subject (such as server shown in FIG. 1 or the end of image processing method operation thereon
End equipment) the available image comprising target, up-sampling and down-sampling are carried out to image, the image that obtains that treated.Processing
Image afterwards includes at least two scales.Specifically, the size model of candidate frame corresponding to the large scale image up-sampled
Numerical value in enclosing is smaller, and the numerical value in the size range of candidate frame corresponding to the small scale image that down-sampling obtains is larger.
In some optional implementations of the present embodiment, the size range of candidate frame corresponding to acquired image
Between third predetermined threshold value and the 4th preset threshold, wherein third predetermined threshold value is greater than the 4th preset threshold, and third presets threshold
Value is greater than or equal to the first preset threshold, and the 4th preset threshold is less than or equal to the second preset threshold.
In these implementations, the numerical value of the size range of candidate frame corresponding to acquired original image is placed in the middle.In this way,
The target of some moderate dimensions can be determined from original image, these targets are obtained from original image with the size for these targets
Feature, thus the relatively accurately moderate target of detecting size.
Step 402, by acquired image and treated that image inputs convolutional neural networks, characteristic pattern and multiple is obtained
Indicate the candidate frame of the position of target, wherein corresponding at least two candidate frames of each target.
In the present embodiment, acquired image can be inputted convolutional neural networks by above-mentioned executing subject, also, can also
Convolutional neural networks are inputted with the image for obtaining change of scale, to obtain the candidate frame and feature of the position of multiple instruction targets
Figure.Specifically, above-mentioned executing subject can determine candidate frame using various ways.
Step 403, in the candidate frame in each image, candidate frame of the size in preset size range is determined,
In, the size range of candidate frame corresponding to the image of different scale is different.
In the present embodiment, above-mentioned executing subject can determine size in preset ruler in the candidate frame of each image
Candidate frame in very little range.Because the size range of candidate frame corresponding to the image of different scale is different, determining ruler
When candidate frame in very little range, the size of candidate frame determined by the image to different scale is not quite similar.Corresponding to image
Candidate frame, which refers to, inputs the obtained candidate frame of convolutional neural networks for image.
Corresponding to step 404, determining at least one candidate frame in the candidate frame in size range in characteristic pattern
Region obtains the feature in region, inputs the full articulamentum of convolutional neural networks.
In the present embodiment, above-mentioned executing subject can determine at least one of the candidate frame in size range candidate
Frame, the corresponding region in characteristic pattern.Later, the feature in region is obtained, and acquired feature is inputted into convolutional Neural net
The full articulamentum of network obtains the final output of convolutional neural networks to carry out the subsequent processing of convolutional neural networks.Above-mentioned execution
Main body can determine spy corresponding to target area from the eigenmatrix corresponding to characteristic pattern when obtaining the feature in region
The part of matrix is levied, and is extracted.
The present embodiment can obtain the image of different scale by up-sampling, down-sampling, get to different size of mesh
Mark obtains feature abundant.Further, the present embodiment can be by the candidate frame of at least three kinds size ranges, more accurately
Get different size clarification of objective in image.
With further reference to Fig. 5, as the realization to method shown in above-mentioned each figure, this application provides a kind of image procossing dresses
The one embodiment set, the Installation practice is corresponding with embodiment of the method shown in Fig. 2, which specifically can be applied to respectively
In kind electronic equipment.
As shown in figure 5, the image processing apparatus 500 of the present embodiment includes: acquiring unit 501, input unit 502, determines
Unit 503 and area determination unit 504.Wherein, acquiring unit 501, be configured to obtain include target image, to image into
Row change of scale, the image at least one scale that obtains that treated;Input unit 502 is configured to acquired image
Image inputs convolutional neural networks with treated, obtains the candidate frame of the position of characteristic pattern and multiple instruction targets, wherein every
Corresponding at least two candidate frames of a target;Determination unit 503 is configured in the candidate frame in each image, determines size
Candidate frame in preset size range, wherein the size range of candidate frame corresponding to the image of different scale is different;Area
Domain determination unit 504 is configured to determine the institute in characteristic pattern of at least one candidate frame in the candidate frame in size range
Corresponding region obtains feature corresponding to region, inputs the full articulamentum of convolutional neural networks.
In some embodiments, the available image comprising target of acquiring unit 501, and to acquired image into
Row change of scale, with the image at least one scale that obtains that treated.Pair for having certain meaning that target is presented by image
As, such as trees, house.
In some embodiments, acquired image can be inputted convolutional neural networks by input unit 502, also,
The image that change of scale obtains convolutional neural networks be can be inputted into, candidate frame and the spy of the position of multiple instruction targets obtained
Sign figure.Specifically, above-mentioned executing subject can determine candidate frame using various ways.
In some embodiments, determination unit 503 can determine size in preset ruler in the candidate frame of each image
Candidate frame in very little range.Because the size range of candidate frame corresponding to the image of different scale is different, determining ruler
When candidate frame in very little range, the size of candidate frame determined by the image to different scale is not quite similar.Corresponding to image
Candidate frame, which refers to, inputs the obtained candidate frame of convolutional neural networks for image.
In some embodiments, area determination unit 504 can determine at least one in the candidate frame in size range
A candidate frame region corresponding in characteristic pattern.Later, the feature in region is obtained, and acquired feature input convolution is refreshing
Full articulamentum through network obtains the final output of convolutional neural networks to carry out the subsequent processing of convolutional neural networks.
In some optional implementations of the present embodiment, the device further include: selection unit is configured to pre-
If size range in candidate frame carry out non-maxima suppression, to obtain at least one candidate frame.
In some optional implementations of the present embodiment, which is further configured to: being carried out to image
Up-sampling and/or down-sampling, wherein the size range of candidate frame corresponding to the image that down-sampling obtains is more than or equal to the
One preset threshold, the size range of candidate frame corresponding to the image up-sampled be less than or equal to the second preset threshold,
First preset threshold is greater than the second preset threshold.
In some optional implementations of the present embodiment, the size range of candidate frame corresponding to acquired image
Between third predetermined threshold value and the 4th preset threshold, wherein third predetermined threshold value is greater than the 4th preset threshold, and third presets threshold
Value is greater than or equal to the first preset threshold, and the 4th preset threshold is less than or equal to the second preset threshold.
In some optional implementations of the present embodiment, in response to there are at least two images in treated image
Scale is greater than the scale of acquired image, at least two images, the size of candidate frame corresponding to the lesser image of scale
Range is less than the first specified threshold, and the size range of candidate frame corresponding to the biggish image of scale is less than the second specified threshold
Value, the first specified threshold are greater than the second specified threshold.
In some optional implementations of the present embodiment, in response to there are more than two figures in treated image
As scale is less than the scale of acquired image, in more than two images, candidate frame corresponding to the lesser image of scale
Size range is greater than third specified threshold, and the size range of candidate frame corresponding to the biggish image of scale is to refer to greater than the 4th
Determine threshold value, third specified threshold is greater than the 4th specified threshold.
Below with reference to Fig. 6, it illustrates the computer systems 600 for the electronic equipment for being suitable for being used to realize the embodiment of the present application
Structural schematic diagram.Electronic equipment shown in Fig. 6 is only an example, function to the embodiment of the present application and should not use model
Shroud carrys out any restrictions.
As shown in fig. 6, computer system 600 includes central processing unit (CPU and/or GPU) 601, it can be according to depositing
Storage is loaded into random access storage device (RAM) 603 in the program in read-only memory (ROM) 602 or from storage section 608
Program and execute various movements appropriate and processing.In RAM 603, also it is stored with system 600 and operates required various journeys
Sequence and data.Central processing unit 601, ROM 602 and RAM 603 are connected with each other by bus 604.Input/output (I/O)
Interface 605 is also connected to bus 604.
I/O interface 605 is connected to lower component: the importation 606 including keyboard, mouse etc.;It is penetrated including such as cathode
The output par, c 607 of spool (CRT), liquid crystal display (LCD) etc. and loudspeaker etc.;Storage section 608 including hard disk etc.;
And the communications portion 609 of the network interface card including LAN card, modem etc..Communications portion 609 via such as because
The network of spy's net executes communication process.Driver 610 is also connected to I/O interface 605 as needed.Detachable media 611, such as
Disk, CD, magneto-optic disk, semiconductor memory etc. are mounted on as needed on driver 610, in order to read from thereon
Computer program be mounted into storage section 608 as needed.
Particularly, in accordance with an embodiment of the present disclosure, it may be implemented as computer above with reference to the process of flow chart description
Software program.For example, embodiment of the disclosure includes a kind of computer program product comprising be carried on computer-readable medium
On computer program, which includes the program code for method shown in execution flow chart.In such reality
It applies in example, which can be downloaded and installed from network by communications portion 609, and/or from detachable media
611 are mounted.When the computer program is executed by central processing unit 601, limited in execution the present processes above-mentioned
Function.It should be noted that the computer-readable medium of the application can be computer-readable signal media or computer can
Read storage medium either the two any combination.Computer readable storage medium for example can be --- but it is unlimited
In system, device or the device of --- electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor, or any above combination.It calculates
The more specific example of machine readable storage medium storing program for executing can include but is not limited to: have the electrical connection, portable of one or more conducting wires
Formula computer disk, hard disk, random access storage device (RAM), read-only memory (ROM), erasable programmable read only memory
(EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), light storage device, magnetic memory device or
The above-mentioned any appropriate combination of person.In this application, computer readable storage medium can be it is any include or storage program
Tangible medium, which can be commanded execution system, device or device use or in connection.And in this Shen
Please in, computer-readable signal media may include in a base band or as carrier wave a part propagate data-signal,
In carry computer-readable program code.The data-signal of this propagation can take various forms, including but not limited to
Electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be computer-readable
Any computer-readable medium other than storage medium, the computer-readable medium can send, propagate or transmit for by
Instruction execution system, device or device use or program in connection.The journey for including on computer-readable medium
Sequence code can transmit with any suitable medium, including but not limited to: wireless, electric wire, optical cable, RF etc. are above-mentioned
Any appropriate combination.
Flow chart and block diagram in attached drawing are illustrated according to the system of the various embodiments of the application, method and computer journey
The architecture, function and operation in the cards of sequence product.In this regard, each box in flowchart or block diagram can generation
A part of one module, program segment or code of table, a part of the module, program segment or code include one or more use
The executable instruction of the logic function as defined in realizing.It should also be noted that in some implementations as replacements, being marked in box
The function of note can also occur in a different order than that indicated in the drawings.For example, two boxes succeedingly indicated are actually
It can be basically executed in parallel, they can also be executed in the opposite order sometimes, and this depends on the function involved.Also it to infuse
Meaning, the combination of each box in block diagram and or flow chart and the box in block diagram and or flow chart can be with holding
The dedicated hardware based system of functions or operations as defined in row is realized, or can use specialized hardware and computer instruction
Combination realize.
Being described in unit involved in the embodiment of the present application can be realized by way of software, can also be by hard
The mode of part is realized.Described unit also can be set in the processor, for example, can be described as: a kind of processor packet
Include acquiring unit, input unit, determination unit and area determination unit.Wherein, the title of these units is under certain conditions simultaneously
The restriction to the unit itself is not constituted, for example, acquiring unit is also described as " image comprising target being obtained, to figure
As carrying out change of scale, the unit of the image at least one scale that obtains that treated ".
As on the other hand, present invention also provides a kind of computer-readable medium, which be can be
Included in device described in above-described embodiment;It is also possible to individualism, and without in the supplying device.Above-mentioned calculating
Machine readable medium carries one or more program, when said one or multiple programs are executed by the device, so that should
Device: obtain include target image, change of scale is carried out to image, the image at least one scale that obtains that treated;It will
Image inputs convolutional neural networks to acquired image with treated, obtains the time of the position of characteristic pattern and multiple instruction targets
Select frame, wherein corresponding at least two candidate frames of each target;In the candidate frame in each image, determine size preset
Candidate frame in size range, wherein the size range of candidate frame corresponding to the image of different scale is different;It determines in size
At least one candidate frame in candidate frame in range region corresponding in characteristic pattern, obtains feature corresponding to region,
Input the full articulamentum of convolutional neural networks.
Above description is only the preferred embodiment of the application and the explanation to institute's application technology principle.Those skilled in the art
Member is it should be appreciated that invention scope involved in the application, however it is not limited to technology made of the specific combination of above-mentioned technical characteristic
Scheme, while should also cover in the case where not departing from foregoing invention design, it is carried out by above-mentioned technical characteristic or its equivalent feature
Any combination and the other technical solutions formed.Such as features described above has similar function with (but being not limited to) disclosed herein
Can technical characteristic replaced mutually and the technical solution that is formed.