CN110059594A - A kind of environment sensing adapting to image recognition methods and device - Google Patents
A kind of environment sensing adapting to image recognition methods and device Download PDFInfo
- Publication number
- CN110059594A CN110059594A CN201910261185.4A CN201910261185A CN110059594A CN 110059594 A CN110059594 A CN 110059594A CN 201910261185 A CN201910261185 A CN 201910261185A CN 110059594 A CN110059594 A CN 110059594A
- Authority
- CN
- China
- Prior art keywords
- model
- scene
- image
- library
- feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/35—Categorising the entire scene, e.g. birthday party or wedding scene
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
This disclosure relates to a kind of environment sensing adapting to image recognition methods and device, wherein method includes: acquisition model step, obtains multiple models, and multiple models adapt to different scenes respectively;Each image in the library of bottom is passed through the progress feature extraction of each model respectively and obtains planting modes on sink characteristic on earth, bottom planting modes on sink characteristic is respectively stored into each model bottom library corresponding with each model by bottom library initialization step;Information acquiring step obtains testing image;Scene judgment step judges the scene type of testing image, chooses one or more models as judgment models according to scene type;Characteristic extraction step extracts the feature to be measured of testing image by one or more judgment models;Information compares step, and feature to be measured and the bottom planting modes on sink characteristic in the library of model bottom are compared, judge whether to match.The disclosure improves the accuracy of image recognition result by a kind of environment sensing adapting to image recognition methods.
Description
Technical field
This disclosure relates to field of image recognition, it is specifically related to a kind of environment sensing adapting to image recognition methods and dress
It sets.
Background technique
As computer science is in the fast development of field of human-computer interaction, the application of image recognition technology, which receives, generally to be weighed
Depending on.Image recognition technology includes two stages of feature extraction and Characteristic Contrast, is traditionally extracted not frequently with unified model
With the characteristics of image in situation, then extracted characteristics of image is compared with feature to be compared, to realize to image
Identification.
Summary of the invention
In order to overcome problems of the prior art, the disclosure provides a kind of environment sensing adapting to image recognition methods
And device.
In a first aspect, the embodiment of the present disclosure provides a kind of environment sensing adapting to image recognition methods comprising obtain mould
Type step, obtains multiple models, and multiple models adapt to different scenes respectively;Bottom library initialization step, by each figure in the library of bottom
Planting modes on sink characteristic on earth is obtained as carrying out feature extraction by each model respectively, bottom planting modes on sink characteristic is respectively stored into and each model
Each corresponding model bottom library;Information acquiring step obtains testing image;Scene judgment step judges the field of testing image
Scape type chooses one or more models as judgment models according to scene type;Characteristic extraction step passes through one or more
Judgment models extract the feature to be measured of testing image;Information compares step, by the bottom planting modes on sink characteristic in feature to be measured and model bottom library
It compares, judges whether to match.
In one example, model includes scene single dimension adaptive model or scene various dimensions adaptive model, and scene single dimension is suitable
Model is answered to adapt to distinguish the scene type that the factor divides by a scene;Scene various dimensions adaptive model adapts to pass through multiple fields
The scene type that scenic spot molecular group divides.
In one example, it includes: time, light, density of stream of people, weather, region that scene, which distinguishes the factor,.
In one example, scene judgment step distinguishes the factor according to scene, judges scene type.
In one example, in information comparison step: model bottom library is model corresponding with judgment models bottom library.
In one example, in information comparison step: model bottom library is whole model bottom library.
In one example, a kind of environment sensing adapting to image recognition methods further include: alarm step, according to matching result,
When there are when at least one bottom planting modes on sink characteristic and characteristic matching to be measured, alarm step issues alarm signal.
In one example, scene judgment step further include: scene type described in real-time judge;Information acquiring step further include:
By acquiring in real time, testing image is obtained;And acquisition parameter is adjusted to adapt to corresponding scene type according to scene type.
In one example, a kind of environment sensing adapting to image recognition methods further include: storing step, by characteristic storage to be measured
To the aspect of model corresponding with judgment models library.
In one example, storing step further include: store the corresponding testing image of feature to be measured.
Second aspect, the embodiment of the present disclosure provide a kind of environment sensing adapting to image identification device, and the environment sensing is certainly
Adapting to pattern recognition device has the function of realizing the environment sensing adapting to image recognition methods that above-mentioned first aspect is related to.Institute
Corresponding software realization can also be executed by hardware realization by hardware by stating function.The hardware or software include one
A or multiple modules corresponding with above-mentioned function.
In one example, environment sensing adapting to image identification device includes: acquisition model module, for obtaining multiple moulds
Type, multiple models adapt to different scenes respectively;Bottom library preset module, it is multiple for passing through each image in the library of bottom respectively
Model carries out feature extraction and obtains planting modes on sink characteristic on earth, and bottom planting modes on sink characteristic is respectively stored into multiple model bottoms corresponding with multiple models
Library;Data obtaining module, for obtaining testing image;Scene judgment module, for judging the scene type of testing image, according to
Scene type chooses one or more models as judgment models;Characteristic extracting module, for judging mould by one or more
Type extracts the feature to be measured of testing image;Information contrast module, for making the bottom planting modes on sink characteristic in feature to be measured and model bottom library
It compares, judges whether to match.
The third aspect, the embodiment of the present disclosure provide a kind of electronic equipment, wherein electronic equipment includes: memory, for depositing
Storage instruction;And processor, the instruction execution environment sensing adapting to image recognition methods for calling memory to store.
Fourth aspect, the embodiment of the present disclosure provide a kind of computer readable storage medium, wherein computer-readable storage medium
Matter is stored with computer executable instructions, computer executable instructions when executed by the processor, performing environment perception self-adaption
Image-recognizing method.
A kind of environment sensing adapting to image recognition methods and device that the disclosure provides, on the one hand, environment sensing is adaptive
It answers image-recognizing method to carry out feature extraction to each image in the library of bottom by multiple models respectively and obtains planting modes on sink characteristic on earth, i.e.,
Obtain the feature of each image in the library of different scenes type bottom;On the other hand, environment sensing adapting to image identifies
For method also directed to the Variation Features of scene type, the model of the feature to be measured of testing image is extracted in adaptive adjustment, and then improves
The accuracy of feature extraction to be measured under corresponding scene type, by the feature to be measured extracted according to corresponding scene type characteristic with
Bottom planting modes on sink characteristic of the bottom library image in corresponding scene type compares, and the final accuracy for improving image recognition result is effectively kept away
Exempt from false recognition rate and leaks the generation of discrimination.
Detailed description of the invention
The following detailed description is read with reference to the accompanying drawings, above-mentioned and other purposes, the feature of disclosure embodiment
It will become prone to understand with advantage.In the accompanying drawings, embodiment of the present disclosure is shown by way of example rather than limitation,
Wherein:
Fig. 1 shows a kind of environment sensing adapting to image recognition methods schematic diagram of embodiment of the present disclosure offer;
Fig. 2 shows another environment sensing adapting to image recognition methods schematic diagrames that the embodiment of the present disclosure provides;
Fig. 3 shows a kind of environment sensing adapting to image identification device schematic diagram of embodiment of the present disclosure offer;
Fig. 4 shows another environment sensing adapting to image identification device schematic diagram of embodiment of the present disclosure offer;
Fig. 5 shows a kind of electronic equipment schematic diagram of embodiment of the present disclosure offer.
Specific embodiment
The principle and spirit of the disclosure are described below with reference to several illustrative embodiments.It should be appreciated that providing this
A little embodiments are used for the purpose of making those skilled in the art can better understand that realizing the disclosure in turn, and be not with any
Mode limits the scope of the present disclosure.
Although being noted that the statements such as " first " used herein, " second " to describe implementation of the disclosure mode not
Same module, step and data etc., still the statement such as " first ", " second " is merely in different modules, step and data etc.
Between distinguish, and be not offered as specific sequence or significance level.In fact, the statements such as " first ", " second " are complete
It may be used interchangeably.
Fig. 1 is a kind of schematic diagram for environment sensing adapting to image recognition methods 10 that the embodiment of the present disclosure provides.Such as Fig. 1
Shown, environment sensing adapting to image recognition methods 10 includes obtaining model step 110, bottom library initialization step 120, acquisition of information
Step 130, scene judgment step 140, characteristic extraction step 150, information compare step 160.Below to each step in Fig. 1
It is described in detail.
Model step 110 is obtained, multiple models can be obtained in advance, multiple models adapt to different scenes respectively.In reality
In the application of border, with the increase of scene type, corresponding increase therewith is also needed with the model that scene is adapted.For example, scene can be with
For scene on daytime, night scenes, high density stream of people scene and low-density stream of people's scene, therefore the corresponding model obtained is
Daytime model, night model, high density stream of people model and low-density stream of people's model;Model can also be set accordingly are as follows: daytime
High density stream of people model, low-density stream of people model on daytime, night high density stream of people model, night low-density stream of people model.It needs
Illustrate, different models is obtained by different training sample training.For example, high density model height on daytime on daytime
The image of density is trained, and in the training process, calculates loss letter with the difference between calculated classification and actual classification
Number, and calculated classification results are corrected according to loss function, and then continue to optimize training pattern.
In addition, model used in each scene is also different, difference is mainly reflected in the size of model.Large-sized model
Using the Deep model without beta pruning, for example, we can be improved in the large-sized model of night scenes application resnet101
Component superposition also can be deeper;And it is directed to the identification of scene on daytime, we can be using the transformation based on resnet50
Model can additionally distill the model for different subdivision scenes by large-sized model.
Each image in the library of bottom is passed through each model respectively and carries out feature extraction by bottom library initialization step 120, with
Planting modes on sink characteristic on earth, and bottom planting modes on sink characteristic is respectively stored into each model bottom library corresponding with each model.If
Obtain whole models that 110 stage of model step obtains be respectively daytime model, night model, high density stream of people model and low close
Stream of people's model is spent, then, in bottom library initialization step 120, environment sensing adapting to image recognition methods 10 can will be in the library of bottom
Each image extracts corresponding figure with model on daytime, night model, high density stream of people model and low-density stream of people model respectively
The feature of picture, with the bottom planting modes on sink characteristic of each image formed under corresponding scene, and by each image daytime scene bottom
Planting modes on sink characteristic stores into the model bottom library of scene on daytime, stores the bottom planting modes on sink characteristic of night scenes to the model bottom library of night scenes
In, the bottom planting modes on sink characteristic of high density stream of people's scene stored into the model bottom library of high density stream of people's scene, by the low-density stream of people
The bottom planting modes on sink characteristic of scape is stored into the model bottom library of low-density stream of people's scene.
Information acquiring step 130 obtains testing image.Information acquiring step 130 can be obtained by image storage equipment
The testing image being stored in image storage apparatus can also carry out acquisition in real time by monitoring device and obtain.
Scene judgment step 140 judges the scene type of testing image, chooses one or more models according to scene type
As judgment models.For example, judging that the scene type of testing image is scene on daytime, therefore by scene judgment step 140
Model on daytime is chosen as judgment models according to scene on daytime;If judging the field of testing image by scene judgment step 140
When scape type is high density on daytime stream of people's scene, therefore model on daytime and height can be chosen respectively according to high density stream of people scene on daytime
Density stream of people's model is as judgment models;Or high density on daytime stream of people model conduct can be chosen according to the setting of realistic model
Judgment models.
Characteristic extraction step 150 extracts the feature to be measured of testing image by one or more judgment models.For example,
In an example, according to above-mentioned scene judgment step 140, model on daytime and the high density stream of people are had chosen as judgment models, in feature
150 stage of extraction step, testing image pass through the spy to be measured of model on daytime and high density stream of people's model extraction testing image respectively
Sign.
Information compares step 160, and feature to be measured and the bottom planting modes on sink characteristic in the library of model bottom are compared, judge whether to match.
In information comparison step 160, the bottom planting modes on sink characteristic by feature to be measured respectively with the image in the library of model bottom is compared, and then judges
Whether feature to be measured matches with the bottom planting modes on sink characteristic of the image in the library of model bottom.
A kind of environment sensing adapting to image recognition methods 10 that the disclosure provides, on the one hand, environment sensing is adaptively schemed
Planting modes on sink characteristic on earth is obtained as recognition methods 10 carries out feature extraction to each image in the library of bottom by each model respectively, i.e.,
Obtain the feature of each image in the library of different scenes type bottom;On the other hand, environment sensing adapting to image identifies
Variation Features of the method 10 also directed to scene type, the model of the feature to be measured of adaptive adjustment extraction testing image, Jin Erti
The accuracy of to be measured feature extraction of the height under corresponding scene type, the feature to be measured that will be extracted according to corresponding scene type characteristic
It compares to bottom planting modes on sink characteristic of the bottom library image in corresponding scene type, the final accuracy for improving image recognition result, effectively
It avoids false recognition rate and leaks the generation of discrimination.
In one example, model includes scene single dimension adaptive model or scene various dimensions adaptive model, wherein scene one-dimensional
Adaptive model is spent to adapt to distinguish the scene type that the factor divides by a scene;Scene various dimensions adaptive model is adapted to by more
A scene distinguishes the scene type that the factor divides.
Model is divided into scene single dimension adaptive model and scene various dimensions adaptive model, is obtaining model stage 110, ring
Border perception self-adaption image-recognizing method 10 can be fitted according to model selection scene single dimension is confirmed the characteristics of the scene for implementing ground
Answer model or scene various dimensions adaptive model.
When the scene for implementing ground includes that a small amount of scene is distinguished because of the period of the day from 11 p.m. to 1 a.m, in the acquisition model stage 110, the model of acquisition can be set
It is set to scene single dimension adaptive model, by the model of setting single dimension, model on daytime and night model are such as divided into according to light,
It is divided into high density stream of people model and low-density stream of people's model according to flow of the people, environment sensing adapting to image recognition methods can be enabled
10 characteristic extraction step 150 and information comparison step 160 execution pattern it is simpler, efficient;When the scene packet for implementing ground
It distinguishes containing more scene because of the period of the day from 11 p.m. to 1 a.m, is obtaining the model stage 110, the model of acquisition may be configured as scene various dimensions adaptive model,
By the way that the model of various dimensions is arranged, several scene single dimension adaptive models can be consolidated into a scene various dimensions and adapt to mould
Type e.g. according to light and flow of the people, is divided into high density stream of people model on daytime, low-density stream of people model on daytime, night high density people
Flow model, night low-density stream of people model.Environment sensing adapting to image recognition methods 10 is adapted to by scene set various dimensions
Model, on the one hand, memory space occupied by model can be effectively reduced, on the other hand, when the scene for implementing ground includes several fields
When the molecular group of scenic spot, spy to be measured that scene various dimensions adaptive model can will need scene single dimension adaptive model to extract in batches
Sign, disposable extract are completed, and then the extraction rate of feature to be measured and the speed of subsequent information comparison can be improved.
It may include time, light, density of stream of people, weather and region etc. that scene, which distinguishes the factor,.With Novel Temporal Scenario distinguish because
For son, scene single dimension adaptive model can be divided into model on daytime and night model, wherein divide daytime and night when
Between point can according to implement ground concrete condition be manually set.
If, can be by scene various dimensions adaptive model by taking Novel Temporal Scenario distinguishes the factor and density of stream of people scene distinguishes the factor as an example
Be divided into high density stream of people model on daytime, daytime low-density stream of people model, night high density stream of people model and night low-density people
Flow model, wherein the cut-point for dividing the flow of the people of the high density stream of people and the low-density stream of people can be according to the concrete condition for implementing ground
It is manually set.
In one example, scene judgment step 140 can judge scene type according to the scene differentiation factor.By scene judge because
Son judges scene type, can be by complicated scene partitioning at different clearly scene types.
In one example, in information comparison step 160: model bottom library is model corresponding with judgment models bottom library.Also
It is to say, in information comparison step 160, according to the corresponding judgment models of feature to be measured, to transfer mould corresponding with judgment models
Bottom planting modes on sink characteristic in the library of type bottom, for being compared with feature to be measured.For example, being white with judgment models corresponding to feature to be measured
For its model, in information comparison step 160, environment sensing adapting to image recognition methods 10 is by transferring the figure in the library of bottom
As the bottom planting modes on sink characteristic according to model extraction on daytime, that is, the bottom planting modes on sink characteristic in the library of model corresponding with model on daytime bottom is transferred, come
It is compared with feature to be measured, judges whether feature to be measured and the bottom planting modes on sink characteristic of a certain image in the library of model bottom match.
In information comparison step 160, model bottom library is model corresponding with judgment models bottom library, that is, according to be measured
The corresponding judgment models of feature, to transfer the bottom planting modes on sink characteristic in the library of model corresponding with judgment models bottom, then with feature to be measured
It compares, the operand of environment sensing adapting to image recognition methods 10 can be reduced, improve information versus speed.
In one example, in information comparison step 160, model bottom library is whole model bottom library.That is, in information
It compares in step 160, feature to be measured compares parallel with the bottom planting modes on sink characteristic in the library of whole model bottoms.It is same by feature parallel to be measured
When in whole model bottoms library carry out retrieval analysis, can more comprehensively comparative analysis feature to be measured, it is adaptive to reduce environment sensing
Leakage knowledge rate of the image-recognizing method 10 in information comparison process.For example, being respectively daytime with model corresponding to the planting modes on sink characteristic of bottom
For model, night model, high density stream of people model and low-density stream of people's model, then it will form in bottom library initialization step 120
The bottom planting modes on sink characteristic of corresponding model, i.e. model on daytime bottom planting modes on sink characteristic, night model bottom planting modes on sink characteristic, high density stream of people bottom planting modes on sink characteristic and low
Density stream of people bottom planting modes on sink characteristic.In information comparison step 160, environment sensing adapting to image recognition methods 10 is by by spy to be measured
Sign respectively with model on the daytime bottom planting modes on sink characteristic of each image, night model bottom planting modes on sink characteristic, high density stream of people bottom Al Kut in the library of bottom
Low-density of seeking peace stream of people bottom planting modes on sink characteristic compares, and judges the bottom planting modes on sink characteristic of a certain image under feature to be measured is corresponding with a model
Whether match.
In information comparison step 160, model bottom library is whole model bottom library, that is, by feature to be measured and whole moulds
Bottom planting modes on sink characteristic in the library of type bottom compares, and realizes the complete contrast of feature to be measured Yu bottom planting modes on sink characteristic, that is, realizes feature to be measured
Comparison of each image under each scene with the library of bottom, thereby reduces environment sensing adapting to image recognition methods
The 10 leakage knowledge rate in information comparison process.
In one example, environment sensing adapting to image recognition methods 10 further includes alarm step 170, is compared according in information
The matching result judged in step 160, when there are at least one bottom planting modes on sink characteristic with characteristic matching to be measured, alarm step 170
Issue alarm signal.The stage 160 is compared in information, when at least finding a bottom planting modes on sink characteristic, in other words, when there is some image
When one bottom planting modes on sink characteristic and characteristic matching to be measured, alarm signal occurs immediately for alarm step 170, on the one hand, can effectively reduce ring
The leakage knowledge rate of border perception self-adaption image-recognizing method 10, on the other hand, environment sensing adapting to image recognition methods 10 passes through
Alarm step 170 issues alarm signal in time, to cause the attention of user or other people.
In one example, scene judgment step 140 further includes real-time judge scene type, and scene judgment step 140 can also be with
By period setting steps (being not shown) come real-time judge scene type, i.e., by period setting steps, it will implement ground
Different scene types is divided into according to the time, when implementing being in the scene type of corresponding time, scene judgment step
140 can directly judge, and determine scene type, and then choose one or more models according to scene type and be used as and judge mould
Type.Wherein, the specific setting of period setting steps, can be artificially depending on the case where implementing ground.For example, can be 6 points by morning
Model on daytime is set as at 7 points to afternoon, sets night mould at 7 points in afternoon (being free of at 7 points) to 6 points of morning (being free of at 6 points)
Type.During running, period judgment step automatically switches model according to current system time clock, and current fortune is arranged
Row model is placed in etcd to system global variables and carries out distributed storage, consistent convenient for each Server assistance, and etcd is certainly
Dynamic dissemination system configuration change is into each cluster server.
Information acquiring step 130 further includes, by acquiring in real time, obtaining testing image, and adopt according to scene type adjustment
Collect parameter, to fit corresponding scene type.Information acquiring step 130 can acquire equipment by front-end image and be acquired, preceding
End equipment can be photographing device and be also possible to video capture device.Acquisition parameter may include the clarity of image, image
Exposure, visual angle of image etc..
In information gathering process, by adjusting acquisition parameter according to scene type, to adapt to corresponding scene type,
When can make when implementing in any scene type, the image of acquisition can be enabled more preferable to meet environment sensing adaptive
Answer the requirement of the image of 10 pairs of image-recognizing method acquisitions.For example, when the scene type for implementing ground is night scenes, in information
In collection process, environment sensing adapting to image recognition methods 10 can pass through by adjusting the exposure of collected picture
Increase the exposure of collected image be more clear the image of acquisition, to meet the requirement for closing Image Acquisition.
Fig. 2 is the schematic diagram of another environment sensing adapting to image recognition methods 10 of the embodiment of the present disclosure.Such as Fig. 2 institute
Show, in one example, environment sensing adapting to image recognition methods 10 further include: storing step 180, by characteristic storage to be measured to
In the corresponding aspect of model library of judgment models.When the image in the library of bottom increases, correspondingly, being needed in bottom library initialization step 120
Feature extraction is carried out to the image newly increased, to obtain newly increasing the bottom planting modes on sink characteristic of image, by characteristic storage to be measured to phase
In the aspect of model library answered, feature to be measured can be extracted again and compared with the bottom planting modes on sink characteristic for newly increasing image, come judge to
Survey whether feature matches with the bottom planting modes on sink characteristic for newly increasing image.
In another example, storing step 180 can also store the corresponding image of feature to be measured, when feature to be measured and bottom Al Kut
When sign matching, the corresponding image of feature to be measured stored by storing step 180, can find out and match with bottom planting modes on sink characteristic immediately
Image.
A kind of environment sensing adapting to image recognition methods 10 of the disclosure can be applied to recognition of face, article identification etc.
Aspect.
Based on identical inventive concept, the embodiment of the present disclosure also provides a kind of environment sensing adapting to image identification device
20。
Fig. 3 shows a kind of schematic diagram of environment sensing adapting to image identification device 20 of embodiment of the present disclosure offer.
As shown in figure 3, a kind of environment sensing adapting to image identification device 20 includes obtaining model module 210, bottom library preset module
220, data obtaining module 230, scene judgment module 240, characteristic extracting module 250, information contrast module 260.
Wherein, model module 210 is obtained, for obtaining multiple models, multiple models adapt to different scenes respectively;Bottom library
Preset module 220 obtains planting modes on sink characteristic on earth for each image in the library of bottom to be carried out feature extraction by multiple models respectively, will
Bottom planting modes on sink characteristic is respectively stored into multiple model bottoms library corresponding with multiple models;Data obtaining module 230, for obtaining to mapping
Picture, data obtaining module 230, which can be, obtains the testing image being stored in image storage apparatus by image storage equipment,
It can be video flowing camera, or other front-end information acquisition devices, directly acquisition testing image.As a kind of deformation,
Scene judgment module 240 can also be directly installed on data obtaining module 230, be obtained in data obtaining module 230 to be measured
During image, scene type locating for testing image is judged;Scene judgment module 240, for judging the field of testing image
Scape type chooses one or more models as judgment models according to scene type.Scene judgment module can be a sensing
Device judges the type of scene by sensor, is also possible to a period setting device, device is arranged by the period, will be real
Different scene types is divided into according to the time with applying, when implementing being in the scene type of corresponding time, scene judgement
Module 240 can be judged directly, determine scene type, and then choose one or more models according to scene type and be used as and sentence
Disconnected model;Characteristic extracting module 250, for extracting the feature to be measured of testing image by one or more judgment models;Information
Contrast module 260 judges whether to match for feature to be measured and the bottom planting modes on sink characteristic in the library of model bottom to compare.
In one example, information comparison module 260 is also used to: according to the corresponding judgment models of feature to be measured, being transferred and is judged
Bottom planting modes on sink characteristic in the library of model corresponding model bottom, for being compared with feature to be measured.
In another example, information comparison module 260 is also used to: by the bottom planting modes on sink characteristic in feature to be measured and whole model bottoms library
It compares.
In one example, environment sensing adapting to image identification device 20 further includes alarm module 270, according to matching result,
When there are when at least one bottom planting modes on sink characteristic and characteristic matching to be measured, alarm module 270 issues alarm signal.
Fig. 4 shows another 20 schematic diagram of environment sensing adapting to image identification device of embodiment of the present disclosure offer.
As shown in figure 4, a kind of environment sensing adapting to image identification device 20 further includes memory module 280, for depositing feature to be measured
It stores up in the aspect of model corresponding with judgment models library.
In another example, memory module 280 can also store the corresponding image of feature to be measured, when feature to be measured and bottom Al Kut
When sign matching, the corresponding image of feature to be measured is stored by memory module 280, can extract and match with bottom planting modes on sink characteristic immediately
Image.
Fig. 5 shows a kind of electronic equipment 30 that an embodiment of the disclosure provides.As shown in figure 5, the disclosure
The a kind of electronic equipment 30 that one embodiment provides, wherein the electronic equipment 30 includes memory 310, processor 320, defeated
Enter/export (Input/Output, I/O) interface 330.Wherein, memory 310, for storing instruction.Processor 320, for adjusting
The instruction execution embodiment of the present disclosure stored with memory 310 is used for environment sensing adapting to image recognition methods 10.Wherein,
Processor 320 is connect with memory 310, I/O interface 330 respectively, such as can pass through the connection of bus system and/or other forms
Mechanism (not shown) is attached.Memory 310 can be used for storing program and data, including use involved in the embodiment of the present disclosure
In the program of environment sensing adapting to image identification, processor 320 by operation be stored in the program of memory 310 thereby executing
The various function application and data processing of electronic equipment 30.
Processor 320 can use digital signal processor (Digital Signal in the embodiment of the present disclosure
Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), programmable patrol
At least one of volume array (Programmable Logic Array, PLA) example, in hardware realizes, the processor 320
It can be central processing unit (Central Processing Unit, CPU) or there is data-handling capacity and/or instruction
The combination of one or more of the processing unit of other forms of executive capability.
Memory 310 in the embodiment of the present disclosure may include one or more computer program products, the computer
Program product may include various forms of computer readable storage mediums, such as volatile memory and/or non-volatile deposit
Reservoir.The volatile memory for example may include random access memory (Random Access Memory, RAM) and/
Or cache memory (cache) etc..The nonvolatile memory for example may include read-only memory (Read-Only
Memory, ROM), flash memory (Flash Memory), hard disk (Hard Disk Drive, HDD) or solid state hard disk
(Solid-State Drive, SSD) etc..
In the embodiment of the present disclosure, I/O interface 330 can be used for receiving input instruction (such as number or character information, and
Generate key signals input related with the user setting of electronic equipment 30 and function control etc.), it can also be output to the outside various
Information (for example, image or sound etc.).In the embodiment of the present disclosure I/O interface 330 may include physical keyboard, function button (such as
Volume control button, switch key etc.), mouse, operating stick, trace ball, microphone, one in loudspeaker and touch panel etc.
It is a or multiple.
In some embodiments, present disclose provides a kind of computer readable storage medium, the computer-readable storages
Media storage has computer executable instructions, and computer executable instructions when executed by the processor, execute described above appoint
Where method.
Although description operation in a particular order in the accompanying drawings should not be construed as requiring specific shown in
Sequence or serial order operate to execute these operations, or shown in requirement execution whole to obtain desired result.?
In specific environment, multitask and parallel processing be may be advantageous.
Disclosed method and device can be completed using standard programming technology, using rule-based logic or its
His logic realizes various method and steps.It should also be noted that herein and the terms used in the claims " device "
" module " is intended to include using the realization of a line or multirow software code and/or hardware realization and/or for receiving input
Equipment.
One or more combined individually or with other equipment can be used in any step, operation or program described herein
A hardware or software module are executed or are realized.In one embodiment, software module use includes comprising computer program
The computer program product of the computer-readable medium of code is realized, can be executed by computer processor any for executing
Or whole described step, operation or programs.
For the purpose of example and description, the preceding description of disclosure implementation is had been presented for.Preceding description is not poor
The disclosure is restricted to exact form disclosed by also not the really wanting of act property, according to the above instruction there is likely to be various modifications and
Modification, or various changes and modifications may be obtained from the practice of the disclosure.Select and describe these embodiments and be in order to
Illustrate the principle and its practical application of the disclosure, so that those skilled in the art can be to be suitable for the special-purpose conceived
Come in a variety of embodiments with various modifications and using the disclosure.
Claims (13)
1. a kind of environment sensing adapting to image recognition methods, wherein the described method includes:
Model step is obtained, obtains multiple models, multiple models adapt to different scenes respectively;
Each image in the library of bottom is passed through each described model progress feature extraction respectively and obtains library on earth by bottom library initialization step
The bottom planting modes on sink characteristic is respectively stored into each model bottom library corresponding with model described in each by feature;
Information acquiring step obtains testing image;
Scene judgment step judges the scene type of the testing image, chooses one or more institutes according to the scene type
Model is stated as judgment models;
Characteristic extraction step extracts the feature to be measured of the testing image by one or more judgment models;
Information compares step, and the feature to be measured and the bottom planting modes on sink characteristic in the library of the model bottom are compared, judged whether
Matching.
2. according to the method described in claim 1, wherein,
The model includes scene single dimension adaptive model or scene various dimensions adaptive model,
The scene single dimension adaptive model adapts to distinguish the scene type that the factor divides by a scene;
The scene various dimensions adaptive model adapts to distinguish the scene type that the factor divides by multiple scenes.
3. according to the method described in claim 2, wherein,
It includes: time, light, density of stream of people, weather, region that the scene, which distinguishes the factor,.
4. according to the method in claim 2 or 3, wherein
The scene judgment step distinguishes the factor according to the scene, judges the scene type.
5. according to the method described in claim 1, wherein,
In information comparison step: model bottom library is model bottom library corresponding with the judgment models.
6. according to the method described in claim 1, wherein,
In information comparison step: model bottom library is whole model bottom library.
7. method according to claim 5 or 6, wherein
The method also includes alarm step, according to matching result, when there are bottom planting modes on sink characteristic described at least one with it is described to be measured
When characteristic matching, the alarm step issues alarm signal.
8. according to the method described in claim 1, wherein,
The scene judgment step further include: scene type described in real-time judge;
The information acquiring step further include: by acquiring in real time, obtain the testing image;And according to the scene type tune
Whole acquisition parameter is to adapt to the corresponding scene type.
9. according to the method described in claim 1, wherein,
The method also includes: storing step, by the characteristic storage to be measured to the aspect of model corresponding with the judgment models
Library.
10. according to the method described in claim 9, wherein,
The storing step further include: store the feature to be measured testing image accordingly.
11. a kind of environment sensing adapting to image identification device, wherein the environment sensing adapting to image identification device packet
It includes:
Model module is obtained, for obtaining multiple models, multiple models adapt to different scenes respectively;
Bottom library preset module obtains on earth for each image in the library of bottom to be carried out feature extraction by multiple models respectively
The bottom planting modes on sink characteristic is respectively stored into multiple model bottoms library corresponding with multiple models by planting modes on sink characteristic;
Data obtaining module, for obtaining testing image;
Scene judgment module chooses one or more according to the scene type for judging the scene type of the testing image
A model is as judgment models;
Characteristic extracting module, for extracting the feature to be measured of the testing image by one or more judgment models;
Information contrast module judges for the feature to be measured and the bottom planting modes on sink characteristic in the library of the model bottom to compare
Whether match.
12. a kind of electronic equipment, wherein the electronic equipment includes:
Memory, for storing instruction;And
Processor, environment sensing described in any one of instruction execution claim 1-10 for calling memory storage
Adapting to image recognition methods.
13. a kind of computer readable storage medium, wherein
The computer-readable recording medium storage has computer executable instructions, and the computer executable instructions are by handling
When device executes, perform claim requires environment sensing adapting to image recognition methods described in any one of 1-10.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910261185.4A CN110059594B (en) | 2019-04-02 | 2019-04-02 | Environment perception self-adaptive image recognition method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910261185.4A CN110059594B (en) | 2019-04-02 | 2019-04-02 | Environment perception self-adaptive image recognition method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110059594A true CN110059594A (en) | 2019-07-26 |
CN110059594B CN110059594B (en) | 2021-10-22 |
Family
ID=67318219
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910261185.4A Active CN110059594B (en) | 2019-04-02 | 2019-04-02 | Environment perception self-adaptive image recognition method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110059594B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112016465A (en) * | 2020-08-28 | 2020-12-01 | 北京百度网讯科技有限公司 | Scene recognition method, device and system |
EP4012606A4 (en) * | 2019-08-07 | 2023-08-02 | ZTE Corporation | Method for estimating and presenting passenger flow, system, and computer readable storage medium |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101916374A (en) * | 2010-08-20 | 2010-12-15 | 浙江大学 | Characteristic selection method based on tracking time prediction |
CN102043953A (en) * | 2011-01-27 | 2011-05-04 | 北京邮电大学 | Real-time-robust pedestrian detection method aiming at specific scene |
US20120314031A1 (en) * | 2011-06-07 | 2012-12-13 | Microsoft Corporation | Invariant features for computer vision |
CN103617432A (en) * | 2013-11-12 | 2014-03-05 | 华为技术有限公司 | Method and device for recognizing scenes |
CN106886789A (en) * | 2015-12-16 | 2017-06-23 | 芋头科技(杭州)有限公司 | A kind of image recognition sorter and method |
CN107169450A (en) * | 2017-05-15 | 2017-09-15 | 中国科学院遥感与数字地球研究所 | The scene classification method and system of a kind of high-resolution remote sensing image |
WO2018106524A1 (en) * | 2016-12-07 | 2018-06-14 | Alcatel-Lucent Usa Inc. | Feature detection in compressive imaging |
CN108229498A (en) * | 2017-08-30 | 2018-06-29 | 黄建龙 | A kind of slider of slide fastener recognition methods, device and equipment |
CN108764022A (en) * | 2018-04-04 | 2018-11-06 | 链家网(北京)科技有限公司 | A kind of image-recognizing method and system |
CN109543628A (en) * | 2018-11-27 | 2019-03-29 | 北京旷视科技有限公司 | A kind of face unlock, bottom library input method, device and electronic equipment |
-
2019
- 2019-04-02 CN CN201910261185.4A patent/CN110059594B/en active Active
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101916374A (en) * | 2010-08-20 | 2010-12-15 | 浙江大学 | Characteristic selection method based on tracking time prediction |
CN102043953A (en) * | 2011-01-27 | 2011-05-04 | 北京邮电大学 | Real-time-robust pedestrian detection method aiming at specific scene |
US20120314031A1 (en) * | 2011-06-07 | 2012-12-13 | Microsoft Corporation | Invariant features for computer vision |
CN103617432A (en) * | 2013-11-12 | 2014-03-05 | 华为技术有限公司 | Method and device for recognizing scenes |
CN106886789A (en) * | 2015-12-16 | 2017-06-23 | 芋头科技(杭州)有限公司 | A kind of image recognition sorter and method |
WO2018106524A1 (en) * | 2016-12-07 | 2018-06-14 | Alcatel-Lucent Usa Inc. | Feature detection in compressive imaging |
CN107169450A (en) * | 2017-05-15 | 2017-09-15 | 中国科学院遥感与数字地球研究所 | The scene classification method and system of a kind of high-resolution remote sensing image |
CN108229498A (en) * | 2017-08-30 | 2018-06-29 | 黄建龙 | A kind of slider of slide fastener recognition methods, device and equipment |
CN108764022A (en) * | 2018-04-04 | 2018-11-06 | 链家网(北京)科技有限公司 | A kind of image-recognizing method and system |
CN109543628A (en) * | 2018-11-27 | 2019-03-29 | 北京旷视科技有限公司 | A kind of face unlock, bottom library input method, device and electronic equipment |
Non-Patent Citations (5)
Title |
---|
AGOURIS P ET AL: "《An environment for Content-based image retrieval from large spatial databases》", 《SCIENCEDIRECT》 * |
刘鹏程: "《基于特征调制的SAR欺骗干扰方法研究》", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
吕凝: "《基于内容的视频数据库多模式检索方法研究》", 《中国博士学位论文全文数据库 信息科技辑》 * |
许真等: "《基于纹理特征库的微光图像色彩纹理传递》", 《红外技术》 * |
马媛媛: "《基于纹理分类的图像检索技术研究》", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP4012606A4 (en) * | 2019-08-07 | 2023-08-02 | ZTE Corporation | Method for estimating and presenting passenger flow, system, and computer readable storage medium |
US11816875B2 (en) | 2019-08-07 | 2023-11-14 | Xi'an Zhongxing New Software Co., Ltd. | Method for estimating and presenting passenger flow, system, and computer readable storage medium |
CN112016465A (en) * | 2020-08-28 | 2020-12-01 | 北京百度网讯科技有限公司 | Scene recognition method, device and system |
CN112016465B (en) * | 2020-08-28 | 2024-06-28 | 北京至为恒通企业管理有限公司 | Scene recognition method, device and system |
Also Published As
Publication number | Publication date |
---|---|
CN110059594B (en) | 2021-10-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110428388B (en) | Image data generation method and device | |
Yi et al. | ASSD: Attentive single shot multibox detector | |
US20200004777A1 (en) | Image Retrieval with Deep Local Feature Descriptors and Attention-Based Keypoint Descriptors | |
US11537891B2 (en) | Intelligent recognition and alert methods and systems | |
US10565699B2 (en) | Apparatus and method for detecting anomaly in plant pipe using multiple meta-learning | |
CN110268423A (en) | The system and method for distribution training for deep learning model | |
Obrist et al. | BatScope manages acoustic recordings, analyses calls, and classifies bat species automatically | |
CN108681746A (en) | A kind of image-recognizing method, device, electronic equipment and computer-readable medium | |
CN110288015A (en) | A kind for the treatment of method and apparatus of portrait retrieval | |
Mann et al. | Automatic flower detection and phenology monitoring using time‐lapse cameras and deep learning | |
CN110059594A (en) | A kind of environment sensing adapting to image recognition methods and device | |
CN114324336B (en) | Nondestructive measurement method for biomass of soybean in whole growth period | |
CN111783812B (en) | Forbidden image recognition method, forbidden image recognition device and computer readable storage medium | |
JP7204786B2 (en) | Visual search method, device, computer equipment and storage medium | |
CN110334716A (en) | Characteristic pattern processing method, image processing method and device | |
CN113297486B (en) | Click rate prediction method and related device | |
CN113361194A (en) | Sensor drift calibration method based on deep learning, electronic equipment and storage medium | |
CN110442749A (en) | Video frame processing method and processing device | |
CN116958267B (en) | Pose processing method and device, electronic equipment and storage medium | |
CN115018081A (en) | Feature selection method, application program prediction method and device | |
CN111652168A (en) | Group detection method, device and equipment based on artificial intelligence and storage medium | |
CN115578796A (en) | Training method, device, equipment and medium for living body detection model | |
TWI814623B (en) | Method for identifying images, computer device and storage medium | |
CN113223017A (en) | Training method of target segmentation model, target segmentation method and device | |
US20240144633A1 (en) | Image recognition method, electronic device and storage medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |