CN108537224A - Image detecting method and device - Google Patents
Image detecting method and device Download PDFInfo
- Publication number
- CN108537224A CN108537224A CN201810367525.7A CN201810367525A CN108537224A CN 108537224 A CN108537224 A CN 108537224A CN 201810367525 A CN201810367525 A CN 201810367525A CN 108537224 A CN108537224 A CN 108537224A
- Authority
- CN
- China
- Prior art keywords
- region
- candidate region
- group
- candidate
- target object
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The disclosure is directed to image detecting method and devices.This method includes：Determine the candidate region in target image；Candidate region is converted, obtains candidate region group, candidate region group includes the candidate region after candidate region and at least one transformation；Obtain the destination probability that each region in the group of candidate region includes target object；Determine that the region that destination probability meets preset condition in the group of candidate region is target object region.The technical solution can reduce the difficulty of determining objects in images position, so as to improve user experience.
Description
Technical field
This disclosure relates to image processing field more particularly to image detecting method and device.
Background technology
Image detection refers to being handled image, analyzed and being identified, to determine the classification and the object of objects in images
The position of body in the picture, image detection can be applied in the scenes such as vehicle assistant drive, video frequency searching.In the related technology,
When carrying out image detection, generally first by objects in images may extracted region out be used as candidate region, then to extraction
All candidate regions out carry out object identification respectively, to determine the region where object.
Invention content
To overcome the problems in correlation technique, a kind of image detecting method of embodiment of the disclosure offer and device.
Technical solution is as follows：
It is according to an embodiment of the present disclosure in a first aspect, provide a kind of image detecting method, including：
Determine the candidate region in target image；
Candidate region is converted, candidate region group is obtained, candidate region group includes candidate region and at least one
Candidate region after transformation；
Obtain the destination probability that each region in the group of candidate region includes target object；
Determine that the region that destination probability meets preset condition in the group of candidate region is target object region.
In the technical scheme provided by this disclosed embodiment, by obtaining the original candidates region in target image, to original
Beginning candidate region is converted, and obtains candidate region group, candidate region group includes original candidates region and at least one transformation
Original candidates region afterwards, when original candidates region may include multiple objects mutually blocked, candidate regions after the conversion
Object mutually blocks in domain part may be less, thus carried out in candidate region after the conversion the difficulty of object identification compared with
Low, object identification is more accurate, includes the destination probability of target object by obtaining each region in the group of candidate region, and determine
The region that destination probability meets preset condition in the group of candidate region is target object region, so that it is determined that target object
Position, therefore said program reduces the difficulty of determining objects in images position.
In one embodiment, candidate region is converted, including：
Candidate region is zoomed in or out centered on the center of candidate region.
In one embodiment, candidate region is converted, including：
One side of candidate region is moved in parallel to the another side of candidate region in parallel.
In one embodiment, the destination probability that each region in the group of candidate region includes target object is obtained, including：
Obtain the feature vector in each region in the group of candidate region；
The feature vector in each region in the group of candidate region is inputted into multilayer perceptron MLP neural networks, obtains candidate regions
The destination probability in each region in the group of domain.
In one embodiment, method further includes：
Obtain the position adjustment amount in each region in the group of candidate region corresponding with target object；
According to the position adjustment amount in each region in the group of candidate region, the position in each region in the group of candidate region is carried out
Adjustment；
Determine that the region that destination probability meets preset condition in the group of candidate region is target object region, including：
Determine that the region that destination probability meets preset condition in the candidate region group after the adjustment of position is target object place
Region.
Second aspect according to an embodiment of the present disclosure provides a kind of image detection device, including：
Candidate region determining module, for determining the candidate region in target image；
Candidate region group acquisition module obtains candidate region group, candidate region group packet for being converted to candidate region
Include the candidate region after candidate region and at least one transformation；
Destination probability acquisition module, for obtaining the destination probability that each region in the group of candidate region includes target object；
Target area determining module, for determining that the region that destination probability meets preset condition in the group of candidate region is target
Object region.
In one embodiment, candidate region group acquisition module, including：
Submodule is scaled, for zooming in or out candidate region centered on the center of candidate region.
In one embodiment, candidate region group acquisition module, including：
Submodule is moved in parallel, is used for the parallel shifting in the another side of one side of candidate region to candidate region in parallel
It is dynamic.
In one embodiment, destination probability acquisition module, including：
Feature vector acquisition module, the feature vector for obtaining each region in the group of candidate region；
Destination probability acquisition module, by the feature vector input multilayer perceptron MLP god in each region in the group of candidate region
Through network, the destination probability in each region in the group of candidate region is obtained.
In one embodiment, device further includes：
Position adjustment amount acquisition submodule, the position for obtaining each region in the group of candidate region corresponding with target object
Set adjustment amount；
Position adjustment submodule, for the position adjustment amount according to each region in the group of candidate region, to candidate region group
In the position in each region be adjusted；
Target area determining module, including：
Target area determination sub-module, for determine position adjustment after candidate region group in destination probability meet preset item
The region of part is target object region.
The third aspect according to an embodiment of the present disclosure provides a kind of image detection device, including：
Processor；
Memory for storing processor-executable instruction；
Wherein, processor is configured as：
Determine the candidate region in target image；
Candidate region is converted, candidate region group is obtained, candidate region group includes candidate region and at least one
Candidate region after transformation；
Obtain the destination probability that each region in the group of candidate region includes target object；
Determine that the region that destination probability meets preset condition in the group of candidate region is target object region.
Fourth aspect according to an embodiment of the present disclosure provides a kind of computer readable storage medium, is stored thereon with meter
Calculation machine instructs, when which is executed by processor the step of any one of first aspect of realization embodiment of the disclosure method.
It should be understood that above general description and following detailed description is only exemplary and explanatory, not
The disclosure can be limited.
Description of the drawings
The drawings herein are incorporated into the specification and forms part of this specification, and shows the implementation for meeting the disclosure
Example, and together with specification for explaining the principles of this disclosure.
Fig. 1 a are the flow diagrams 1 according to the image detecting method shown in an exemplary embodiment；
Fig. 1 b are the flow diagrams 2 according to the image detecting method shown in an exemplary embodiment；
Fig. 2 a are structural representation Fig. 1 according to the image detection device shown in an exemplary embodiment；
Fig. 2 b are structural representation Fig. 2 according to the image detection device shown in an exemplary embodiment；
Fig. 2 c are structural representation Fig. 3 according to the image detection device shown in an exemplary embodiment；
Fig. 2 d are structural representation Fig. 4 according to the image detection device shown in an exemplary embodiment；
Fig. 2 e are structural representation Fig. 5 according to the image detection device shown in an exemplary embodiment；
Fig. 3 is a kind of block diagram of device shown according to an exemplary embodiment；
Fig. 4 is a kind of block diagram of device shown according to an exemplary embodiment；
Fig. 5 is a kind of block diagram of device shown according to an exemplary embodiment.
Specific implementation mode
Example embodiments are described in detail here, and the example is illustrated in the accompanying drawings.Following description is related to
When attached drawing, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements.Following exemplary embodiment
Described in embodiment do not represent all implementations consistent with this disclosure.On the contrary, they be only with it is such as appended
The example of the consistent device and method of some aspects be described in detail in claims, the disclosure.
With the high speed development and people's living standards continue to improve of science and technology, in recent years, image detecting technique is wide
It is general to be applied in the scenes such as vehicle assistant drive, video frequency searching.When carrying out image detection, sliding window or aobvious is generally first passed through
The modes such as work property detection by objects in images may extracted region be out used as candidate region, then the office based on candidate region
Portion's feature carries out object identification respectively to all candidate regions extracted, to determine the position where respective objects.
Although said program can determine that the position where respective objects, said program are the parts based on candidate region
Feature is identified, but when factor included in candidate region is excessive, in fact it could happen that multiple objects were mutually blocked
Situation, the difficulty that object identification is carried out to the candidate region is higher, improves the difficulty of determining object position, for example work as time
When favored area includes the automobile blocked by trees, when carrying out automobile identification to the candidate region, trees may be to identification
Automobile impacts, and improves the difficulty for determining the automobile position.
To solve the above-mentioned problems, in the technical scheme provided by this disclosed embodiment, by obtaining in target image
Original candidates region converts original candidates region, obtains candidate region group, candidate region group includes original candidates region
And the original candidates region after at least one transformation, when original candidates region may include multiple objects mutually blocked,
The part that object mutually blocks in candidate region after the conversion may be less, therefore object is carried out in candidate region after the conversion
The difficulty of body identification is relatively low, and object identification is more accurate, includes target object by obtaining each region in the group of candidate region
Destination probability, and determine that the region that destination probability meets preset condition in the group of candidate region is target object region, to
The position of target object is determined, therefore said program reduces the difficulty of determining objects in images position.
Embodiment of the disclosure provides a kind of image detecting method, as shown in Figure 1a, includes the following steps 101 to step
104：
In a step 101, the original candidates region in target image is determined.
It is exemplary, it determines the original candidates region in target image, can be to use selective search (selective
Search) it is notable to suggest that network (Region Proposal Network, RPN) algorithm carries out target image for algorithm or region
Property detection, and find out original candidates region R={ r according to testing result1,r2,...,rn, wherein r=(x, y, w, h) is exemplary
, the quantity n in the original candidates region in a target image is hundreds of to thousands of.
In a step 102, original candidates region is converted, obtains candidate region group.
Wherein, candidate region group includes the original candidates region after original candidates region and at least one transformation.
Illustratively, original candidates region is converted, can is to be amplified, reduce to original candidates region, is left
Right random shearing, upper and lower random shearing etc..Can be in being with the center of candidate region for example, being converted to candidate region
The heart zooms in or out candidate region；Candidate region is converted, or by one side of candidate region in parallel
The another side of candidate region moves in parallel.
For example, can be to each candidate region riDoing a series of transformation, (amplification is reduced, left and right random shearing, up and down at random
Shearing), it is one group of candidate region { r by its augmentationi 1,ri 2,...,ri k}。
In step 103, the destination probability that each region in the group of candidate region includes target object is obtained.
Illustratively, the destination probability that each region in the group of candidate region includes target object is obtained, can be to obtain to wait
The feature vector in each region in favored area group, and the feature vector in each region in the group of candidate region is inputted into multilayer perceptron
MLP neural networks obtain the destination probability in each region in the group of candidate region.
For example, the CNN convolution algorithms of multilayer can be carried out to target image, the characteristic pattern F of target image is obtainedc, and will
Each candidate region r after augmentationiIt is mapped to characteristic pattern FcIn, obtain each candidate region riIn FcIn corresponding region rc
=(xc,yc,wc,hc)=(sc*x,sc*y,sc*w,sc* h), wherein scFor input image size to the scaling of its characteristic pattern size
Coefficient, to each candidate region riIn FcIn corresponding region rcThe operation of maximum value pondization is carried out, designated length is mapped as
Feature vector fc, by the feature vector f corresponding to candidate regioncIt inputs in MLP neural networks, calculating the candidate region is
The probability of target object.Wherein, target object can be the object of a classification, or the object of multiple classifications.
It should be noted that the number of plies of CNN networks, each layer of convolution kernel size and number can be according to actual here
It needs, to adjust, tradeoff to be made on algorithm speed and precision.For example, if thinking the speed of raising algorithm, it is possible to reduce convolution
The quantity of the number of plies and each layer of convolution kernel, correspondingly, algorithm can lose certain precision.Under normal circumstances, CNN networks are being carried out
Calculating process in, the operation of down-sampling is had, so the characteristic pattern finally obtained can reduce many times than input picture.
At step 104, determine that the region that destination probability meets preset condition in the group of candidate region is target object place
Region.
Illustratively, determine that the region that destination probability meets preset condition in the group of candidate region is target object location
Domain can be to carry out threshold filtering to the destination probability of each candidate region first, destination probability is less than to the candidate regions of threshold value
Domain is filtered, and then non-maximum value that remaining candidate region is carried out to IOU=0.5 inhibits, and removes the candidate region of repetition,
Final remaining candidate region is the testing result of the object category, as target object region.
In the technical scheme provided by this disclosed embodiment, by obtaining the original candidates region in target image, to original
Beginning candidate region is converted, and obtains candidate region group, candidate region group includes original candidates region and at least one transformation
Original candidates region afterwards, when original candidates region may include multiple objects mutually blocked, candidate regions after the conversion
Object mutually blocks in domain part may be less, thus carried out in candidate region after the conversion the difficulty of object identification compared with
Low, object identification is more accurate, includes the destination probability of target object by obtaining each region in the group of candidate region, and determine
The region that destination probability meets preset condition in the group of candidate region is target object region, so that it is determined that target object
Position, therefore said program reduces the difficulty of determining objects in images position.
In one embodiment, as shown in Figure 1 b, image detecting method further includes step 105 to step 106：
In step 105, the position adjustment amount in each region in the group of candidate region corresponding with target object is obtained.
In step 106, according to the position adjustment amount in each region in the group of candidate region, to each area in the group of candidate region
The position in domain is adjusted.
At step 104, determine that the region that destination probability meets preset condition in the group of candidate region is target object place
Region can be realized by step 1041：
In step 1041, determine that the region that destination probability meets preset condition in the candidate region group after the adjustment of position is
Target object region.
It illustratively, can be by the feature vector f corresponding to candidate regioncInput position adjusts in MLP network, calculates
Coordinate position adjustment amount Δ r=(Δ x, Δ y, the Δ w, Δ h), and according to the coordinate bit of obtained candidate region of the candidate region
Adjustment amount is set to be adjusted the coordinate of candidate region：
rnew=r+ Δs r=(x+ Δs x, y+ Δ y, w+ Δ w, h+ Δ h).
It should be noted that can also in obtaining candidate region group each region include target object destination probability
When, by by the feature vector f corresponding to candidate regioncIt is input in a MLP network, while calculating the candidate region and being
The probability of target object and coordinate position adjustment amount Δ r=(the Δ x, Δ y, Δ w, Δ h) of the candidate region.
By obtaining the position adjustment amount in each region in the group of candidate region corresponding with target object, and according to candidate regions
The position adjustment amount in each region in the group of domain is adjusted the position in each region in the group of candidate region, determines that position adjusts
The region that destination probability meets preset condition in candidate region group afterwards is target object region, can improve candidate region
Include the probability of target object, reduces the difficulty for determining objects in images position.
Following is embodiment of the present disclosure, can be used for executing embodiments of the present disclosure.
Fig. 2 a are a kind of block diagram of the image detection device 20 shown according to an exemplary embodiment, image detection device
20 can be a part for server or server, or a part for terminal or terminal, image detection device 20 can be with
Pass through being implemented in combination with as some or all of of electronic equipment for software, hardware or both.As shown in Figure 2 a, which examines
Surveying device 200 includes：
Candidate region determining module 201, for determining the candidate region in target image.
Candidate region group acquisition module 202 obtains candidate region group, candidate region for being converted to candidate region
Group includes the candidate region after candidate region and at least one transformation.
Destination probability acquisition module 203 includes that the target of target object is general for obtaining each region in the group of candidate region
Rate.
Target area determining module 204, for determining that the region that destination probability meets preset condition in the group of candidate region is
Target object region.
In one embodiment, as shown in Figure 2 b, candidate region group acquisition module 202, including：
Submodule 2021 is scaled, for zooming in or out candidate region centered on the center of candidate region.
In one embodiment, as shown in Figure 2 c, candidate region group acquisition module 202, including：
Submodule 2022 is moved in parallel, for putting down one side of candidate region to the another side of candidate region in parallel
Row movement.
In one embodiment, as shown in Figure 2 d, destination probability acquisition module 203, including：
Feature vector acquisition submodule 2031, the feature vector for obtaining each region in the group of candidate region.
The feature vector in each region in the group of candidate region is inputted multilayer perceptron by destination probability acquisition submodule 2032
MLP neural networks obtain the destination probability in each region in the group of candidate region.
In one embodiment, as shown in Figure 2 e, image detection device 200 further includes：
Position adjustment amount acquisition module 205, for obtaining each region in the group of candidate region corresponding with target object
Position adjustment amount.
Position adjusting type modules 206, for the position adjustment amount according to each region in the group of candidate region, to candidate region group
In the position in each region be adjusted.
Target area determining module 204, including：
Target area determination sub-module 2041, for determining, destination probability meets pre- in the candidate region group after the adjustment of position
If the region of condition is target object region.
Embodiment of the disclosure provides a kind of image detection device, which can be by obtaining target image
In original candidates region, original candidates region is converted, obtain candidate region group, candidate region group includes original candidates
Original candidates region behind region and at least one transformation, when original candidates region may include multiple objects mutually blocked
When, the part that object mutually blocks in candidate region after the conversion may be less, therefore in candidate region after the conversion into
The difficulty of row object identification is relatively low, and object identification is more accurate, includes object by obtaining each region in the group of candidate region
The destination probability of body, and determine that the region that destination probability meets preset condition in the group of candidate region is target object region,
So that it is determined that the position of target object, therefore said program reduces the difficulty of determining objects in images position.
Fig. 3 is a kind of block diagram of image detection device 30 shown according to an exemplary embodiment, the image detection device
30 can be a part for server or server, or a part for terminal or terminal, image detection device 30 include：
Processor 301；
Memory 302 for storing 301 executable instruction of processor；
Wherein, processor 301 is configured as：
Determine the candidate region in target image；
Candidate region is converted, candidate region group is obtained, candidate region group includes candidate region and at least one
Candidate region after transformation；
Obtain the destination probability that each region in the group of candidate region includes target object；
Determine that the region that destination probability meets preset condition in the group of candidate region is target object region.
In one embodiment, above-mentioned processor 301 can be additionally configured to：
Candidate region is converted, including：
Candidate region is zoomed in or out centered on the center of candidate region.
In one embodiment, above-mentioned processor 301 can be additionally configured to：
Candidate region is converted, including：
One side of candidate region is moved in parallel to the another side of candidate region in parallel.
In one embodiment, above-mentioned processor 301 can be additionally configured to：
The destination probability that each region in the group of candidate region includes target object is obtained, including：
Obtain the feature vector in each region in the group of candidate region；
The feature vector in each region in the group of candidate region is inputted into multilayer perceptron MLP neural networks, obtains candidate regions
The destination probability in each region in the group of domain.
In one embodiment, above-mentioned processor 301 can be additionally configured to：
Obtain the position adjustment amount in each region in the group of candidate region corresponding with target object；
According to the position adjustment amount in each region in the group of candidate region, the position in each region in the group of candidate region is carried out
Adjustment；
Determine that the region that destination probability meets preset condition in the group of candidate region is target object region, including：
Determine that the region that destination probability meets preset condition in the candidate region group after the adjustment of position is target object place
Region.
Embodiment of the disclosure provides a kind of image detection device, which can be by obtaining target image
In original candidates region, original candidates region is converted, obtain candidate region group, candidate region group includes original candidates
Original candidates region behind region and at least one transformation, when original candidates region may include multiple objects mutually blocked
When, the part that object mutually blocks in candidate region after the conversion may be less, therefore in candidate region after the conversion into
The difficulty of row object identification is relatively low, and object identification is more accurate, includes object by obtaining each region in the group of candidate region
The destination probability of body, and determine that the region that destination probability meets preset condition in the group of candidate region is target object region,
So that it is determined that the position of target object, therefore said program reduces the difficulty of determining objects in images position.
Fig. 4 is a kind of block diagram of device 400 for detection image shown according to an exemplary embodiment, for example, dress
It can be mobile phone, computer, digital broadcast terminal, messaging devices, game console, tablet device, medical treatment to set 400
Equipment, body-building equipment, personal digital assistant etc..
Device 400 may include following one or more components：Processing component 402, memory 404, power supply module 406,
Multimedia component 408, audio component 410, the interface 412 of input/output (I/O), sensor module 414 and communication component
416。
The integrated operation of 402 usual control device 400 of processing component, such as with display, call, data communication, phase
Machine operates and record operates associated operation.Processing element 402 may include that one or more processors 420 refer to execute
It enables, to perform all or part of the steps of the methods described above.In addition, processing component 402 may include one or more modules, just
Interaction between processing component 402 and other assemblies.For example, processing component 402 may include multi-media module, it is more to facilitate
Interaction between media component 408 and processing component 402.
Memory 404 is configured not store various types of data to support the operation in device 400.These data are shown
Example includes instruction for any application program or method that are operated on device 400, contact data, and telephone book data disappears
Breath, image, video etc..Memory 404 can be by any kind of volatibility or non-volatile memory device or their group
It closes and realizes, such as static RAM (SRAM), electrically erasable programmable read-only memory (EEPROM) is erasable to compile
Journey read-only memory (EPROM), programmable read only memory (PROM), read-only memory (ROM), magnetic memory, flash
Device, disk or CD.
Power supply module 406 provides electric power for the various assemblies of device 400.Power supply module 406 may include power management system
System, one or more power supplys and other generated with for device 400, management and the associated component of distribution electric power.
Multimedia component 408 is included in the screen of one output interface of offer between device 400 and user.In some realities
It applies in example, screen may include liquid crystal display (LCD) and touch panel (TP).If screen includes touch panel, screen can
To be implemented as touch screen, to receive input signal from the user.Touch panel include one or more touch sensors with
Sense the gesture on touch, slide, and touch panel.The touch sensor can not only sense the side of touch or sliding action
Boundary, but also detect duration and pressure associated with the touch or slide operation.In some embodiments, multimedia group
Part 408 includes a front camera and/or rear camera.When device 400 is in operation mode, such as screening-mode or video
When pattern, front camera and/or rear camera can receive external multi-medium data.Each front camera and postposition
Camera can be a fixed optical lens system or have focusing and optical zoom capabilities.
Audio component 410 is configured as output and/or input audio signal.For example, audio component 410 includes a Mike
Wind (MIC), when device 400 is in operation mode, when such as call model, logging mode and speech recognition mode, microphone by with
It is set to reception external audio signal.The received audio signal can be further stored in memory 404 or via communication set
Part 416 is sent.In some embodiments, audio component 410 further includes a loud speaker, is used for exports audio signal.
I/O interfaces 412 provide interface between processing component 402 and peripheral interface module, and above-mentioned peripheral interface module can
To be keyboard, click wheel, button etc..These buttons may include but be not limited to：Home button, volume button, start button and lock
Determine button.
Sensor module 414 includes one or more sensors, and the state for providing various aspects for device 400 is commented
Estimate.For example, sensor module 414 can detect the state that opens/closes of device 400, and the relative positioning of component, for example, it is described
Component is the display and keypad of device 400, and sensor module 414 can be with 400 1 components of detection device 400 or device
Position change, the existence or non-existence that user contacts with device 400,400 orientation of device or acceleration/deceleration and device 400
Temperature change.Sensor module 414 may include proximity sensor, be configured to detect without any physical contact
Presence of nearby objects.Sensor module 414 can also include optical sensor, such as CMOS or ccd image sensor, at
As being used in application.In some embodiments, which can also include acceleration transducer, gyro sensors
Device, Magnetic Sensor, pressure sensor or temperature sensor.
Communication component 416 is configured to facilitate the communication of wired or wireless way between device 400 and other equipment.Device
400 can access the wireless network based on communication standard, such as WiFi, 2G or 3G or combination thereof.In an exemplary implementation
In example, communication component 416 receives broadcast singal or broadcast related information from external broadcasting management system via broadcast channel.
In one exemplary embodiment, the communication component 416 further includes near-field communication (NFC) module, to promote short range communication.Example
Such as, NFC module can be based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra wide band (UWB) technology,
Bluetooth (BT) technology and other technologies are realized.
In the exemplary embodiment, device 400 can be believed by one or more application application-specific integrated circuit (ASIC), number
Number processor (DSP), digital signal processing appts (DSPD), programmable logic device (PLD), field programmable gate array
(FPGA), controller, microcontroller, microprocessor or other electronic components are realized, for executing the above method.
In the exemplary embodiment, it includes the non-transitorycomputer readable storage medium instructed, example to additionally provide a kind of
Such as include the memory 404 of instruction, above-metioned instruction can be executed by the processor 420 of device 400 to complete the above method.For example,
The non-transitorycomputer readable storage medium can be ROM, random access memory (RAM), CD-ROM, tape, floppy disk
With optical data storage devices etc..
A kind of non-transitorycomputer readable storage medium, when the instruction in the storage medium is by the processing of device 400
When device executes so that device 400 is able to carry out above-mentioned image detecting method, the method includes：
Determine the candidate region in target image；
Candidate region is converted, candidate region group is obtained, candidate region group includes candidate region and at least one
Candidate region after transformation；
Obtain the destination probability that each region in the group of candidate region includes target object；
Determine that the region that destination probability meets preset condition in the group of candidate region is target object region.
In one embodiment, candidate region is converted, including：
Candidate region is zoomed in or out centered on the center of candidate region.
In one embodiment, candidate region is converted, including：
One side of candidate region is moved in parallel to the another side of candidate region in parallel.
In one embodiment, the destination probability that each region in the group of candidate region includes target object is obtained, including：
Obtain the feature vector in each region in the group of candidate region；
The feature vector in each region in the group of candidate region is inputted into multilayer perceptron MLP neural networks, obtains candidate regions
The destination probability in each region in the group of domain.
In one embodiment, the method further includes：
Obtain the position adjustment amount in each region in the group of candidate region corresponding with target object；
According to the position adjustment amount in each region in the group of candidate region, the position in each region in the group of candidate region is carried out
Adjustment；
Determine that the region that destination probability meets preset condition in the group of candidate region is target object region, including：
Determine that the region that destination probability meets preset condition in the candidate region group after the adjustment of position is target object place
Region.
Fig. 5 is a kind of block diagram of device 500 being used for shown according to an exemplary embodiment.For example, device 500 can be with
It is provided as a server.Device 500 includes processing component 522, further comprises one or more processors, and by depositing
Memory resource representated by reservoir 532, can be by the instruction of the execution of processing component 522, such as application program for storing.It deposits
The application program stored in reservoir 532 may include it is one or more each correspond to one group of instruction module.This
Outside, processing component 522 is configured as executing instruction, to execute the above method.
Device 500 can also include the power management that a power supply module 526 is configured as executive device 500, and one has
Line or radio network interface 550 are configured as device 500 being connected to network and input and output (I/O) interface 558.Dress
Setting 500 can operate based on the operating system for being stored in memory 532, such as Windows ServerTM, Mac OS XTM,
UnixTM, LinuxTM,
FreeBSDTM or similar.
A kind of non-transitorycomputer readable storage medium, when the instruction in the storage medium is by the processing of device 500
When device executes so that device 500 is able to carry out image detecting method, the method includes：
Determine the candidate region in target image；
Candidate region is converted, candidate region group is obtained, candidate region group includes candidate region and at least one
Candidate region after transformation；
Obtain the destination probability that each region in the group of candidate region includes target object；
Determine that the region that destination probability meets preset condition in the group of candidate region is target object region.
In one embodiment, candidate region is converted, including：
Candidate region is zoomed in or out centered on the center of candidate region.
In one embodiment, candidate region is converted, including：
One side of candidate region is moved in parallel to the another side of candidate region in parallel.
In one embodiment, the destination probability that each region in the group of candidate region includes target object is obtained, including：
Obtain the feature vector in each region in the group of candidate region；
The feature vector in each region in the group of candidate region is inputted into multilayer perceptron MLP neural networks, obtains candidate regions
The destination probability in each region in the group of domain.
In one embodiment, the method further includes：
Obtain the position adjustment amount in each region in the group of candidate region corresponding with target object；
According to the position adjustment amount in each region in the group of candidate region, the position in each region in the group of candidate region is carried out
Adjustment；
Determine that the region that destination probability meets preset condition in the group of candidate region is target object region, including：
Determine that the region that destination probability meets preset condition in the candidate region group after the adjustment of position is target object place
Region.
Those skilled in the art will readily occur to its of the disclosure after considering specification and putting into practice disclosure disclosed herein
Its embodiment.This application is intended to cover any variations, uses, or adaptations of the disclosure, these modifications, purposes or
Person's adaptive change follows the general principles of this disclosure and includes the undocumented common knowledge in the art of the disclosure
Or conventional techniques.The description and examples are only to be considered as illustrative, and the true scope and spirit of the disclosure are by following
Claim is pointed out.
It should be understood that the present disclosure is not limited to the precise structures that have been described above and shown in the drawings, and
And various modifications and changes may be made without departing from the scope thereof.The scope of the present disclosure is only limited by the accompanying claims.
Claims (12)
1. a kind of image detecting method, which is characterized in that including：
Determine the candidate region in target image；
The candidate region is converted, obtain candidate region group, the candidate region group include the candidate region and
Candidate region after at least one transformation；
Obtain the destination probability that each region in the candidate region group includes target object；
Determine that the region that destination probability meets preset condition in the candidate region group is the target object region.
2. image processing method according to claim 1, which is characterized in that it is described that the candidate region is converted,
Including：
The candidate region is zoomed in or out centered on the center of the candidate region.
3. image processing method according to claim 1, which is characterized in that it is described that the candidate region is converted,
Including：
One side of the candidate region is moved in parallel to the another side of the candidate region in parallel.
4. image detecting method according to claim 1, which is characterized in that described to obtain in the candidate region group each
Region includes the destination probability of target object, including：
Obtain the feature vector in each region in the candidate region group；
The feature vector in each region in the candidate region group is inputted into multilayer perceptron MLP neural networks, obtains the time
The destination probability in each region in favored area group.
5. image detecting method according to claim 1, which is characterized in that the method further includes：
Obtain the position adjustment amount in each region in the candidate region group corresponding with the target object；
According to the position adjustment amount in each region in the candidate region group, to the position in each region in the candidate region group
It is adjusted；
The region that destination probability meets preset condition in the determination candidate region group is the target object region,
Including：
Determine that the region that destination probability meets preset condition in the candidate region group after the adjustment of position is target object place
Region.
6. a kind of image detection device, which is characterized in that including：
Candidate region determining module, for determining the candidate region in target image；
Candidate region group acquisition module obtains candidate region group, the candidate region for being converted to the candidate region
Group includes the candidate region after the candidate region and at least one transformation；
Destination probability acquisition module, for obtaining the destination probability that each region in the candidate region group includes target object；
Target area determining module, for determining that the region that destination probability meets preset condition in the candidate region group is described
Target object region.
7. image processing apparatus according to claim 6, which is characterized in that the candidate region group acquisition module, including：
Submodule is scaled, for zooming in or out the candidate region centered on the center of the candidate region.
8. image processing apparatus according to claim 6, which is characterized in that the candidate region group acquisition module, including：
Submodule is moved in parallel, for putting down one side of the candidate region to the another side of the candidate region in parallel
Row movement.
9. image detection device according to claim 6, which is characterized in that the destination probability acquisition module, including：
Feature vector acquisition submodule, the feature vector for obtaining each region in the candidate region group；
Destination probability acquisition submodule, by the feature vector input multilayer perceptron MLP in each region in the candidate region group
Neural network obtains the destination probability in each region in the candidate region group.
10. image detection device according to claim 6, which is characterized in that described device further includes：
Position adjustment amount acquisition module, for obtaining each region in the candidate region group corresponding with the target object
Position adjustment amount；
Position adjusting type modules, for the position adjustment amount according to each region in the candidate region group, to the candidate region
The position in each region is adjusted in group；
The target area determining module, including：
Target area determination sub-module, for determine position adjustment after candidate region group in destination probability meet preset condition
Region is the target object region.
11. a kind of image detection device, which is characterized in that including：
Processor；
Memory for storing processor-executable instruction；
Wherein, the processor is configured as：
Determine the candidate region in target image；
The candidate region is converted, obtain candidate region group, the candidate region group include the candidate region and
Candidate region after at least one transformation；
Obtain the destination probability that each region in the candidate region group includes target object；
Determine that the region that destination probability meets preset condition in the candidate region group is the target object region.
12. a kind of computer readable storage medium, is stored thereon with computer instruction, which is characterized in that the instruction is by processor
The step of any one of claim 1-5 the methods are realized when execution.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810367525.7A CN108537224A (en) | 2018-04-23 | 2018-04-23 | Image detecting method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810367525.7A CN108537224A (en) | 2018-04-23 | 2018-04-23 | Image detecting method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108537224A true CN108537224A (en) | 2018-09-14 |
Family
ID=63479390
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810367525.7A Pending CN108537224A (en) | 2018-04-23 | 2018-04-23 | Image detecting method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108537224A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106326817A (en) * | 2015-07-03 | 2017-01-11 | 佳能株式会社 | Method and device for detecting object from image |
| CN106682586A (en) * | 2016-12-03 | 2017-05-17 | 北京联合大学 | Method for real-time lane line detection based on vision under complex lighting conditions |
| CN107292306A (en) * | 2017-07-07 | 2017-10-24 | 北京小米移动软件有限公司 | Object detection method and device |
| CN107527029A (en) * | 2017-08-18 | 2017-12-29 | 卫晨 | A kind of improved Faster R CNN method for detecting human face |
| CN107729880A (en) * | 2017-11-15 | 2018-02-23 | 北京小米移动软件有限公司 | Method for detecting human face and device |
-
2018
- 2018-04-23 CN CN201810367525.7A patent/CN108537224A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106326817A (en) * | 2015-07-03 | 2017-01-11 | 佳能株式会社 | Method and device for detecting object from image |
| CN106682586A (en) * | 2016-12-03 | 2017-05-17 | 北京联合大学 | Method for real-time lane line detection based on vision under complex lighting conditions |
| CN107292306A (en) * | 2017-07-07 | 2017-10-24 | 北京小米移动软件有限公司 | Object detection method and device |
| CN107527029A (en) * | 2017-08-18 | 2017-12-29 | 卫晨 | A kind of improved Faster R CNN method for detecting human face |
| CN107729880A (en) * | 2017-11-15 | 2018-02-23 | 北京小米移动软件有限公司 | Method for detecting human face and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107492115B (en) | Target object detection method and device | |
| CN108256555B (en) | Image content identification method and device and terminal | |
| CN111310616A (en) | Image processing method and device, electronic equipment and storage medium | |
| CN106205628B (en) | Voice signal optimization method and device | |
| CN108510987A (en) | Method of speech processing and device | |
| CN108121952A (en) | Face key independent positioning method, device, equipment and storage medium | |
| CN105184313B (en) | Disaggregated model construction method and device | |
| CN105631803B (en) | The method and apparatus of filter processing | |
| CN108010060A (en) | Object detection method and device | |
| CN106845377A (en) | Face key independent positioning method and device | |
| CN105335684B (en) | Face detection method and device | |
| CN106980840A (en) | Shape of face matching process, device and storage medium | |
| CN108062547A (en) | Character detecting method and device | |
| CN107784279A (en) | Method for tracking target and device | |
| CN106778531A (en) | Face detection method and device | |
| CN107992841A (en) | The method and device of identification objects in images, electronic equipment, readable storage medium storing program for executing | |
| CN108846795A (en) | Image processing method and device | |
| CN107729880A (en) | Method for detecting human face and device | |
| EP4287068A1 (en) | Model training method, scene recognition method, and related device | |
| CN108717542A (en) | Identify the method, apparatus and computer readable storage medium of character area | |
| CN108154466A (en) | Image processing method and device | |
| CN109034150A (en) | Image processing method and device | |
| CN107748867A (en) | The detection method and device of destination object | |
| CN108154093B (en) | Face information identification method and device, electronic equipment and machine-readable storage medium | |
| CN108540653A (en) | Terminal device and interaction control 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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180914 |