CN107153844A - The accessory system being improved to flowers identifying system and the method being improved - Google Patents
The accessory system being improved to flowers identifying system and the method being improved Download PDFInfo
- Publication number
- CN107153844A CN107153844A CN201710338697.7A CN201710338697A CN107153844A CN 107153844 A CN107153844 A CN 107153844A CN 201710338697 A CN201710338697 A CN 201710338697A CN 107153844 A CN107153844 A CN 107153844A
- Authority
- CN
- China
- Prior art keywords
- flowers
- classification
- picture
- identifying system
- identification
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
- G06F18/2178—Validation; Performance evaluation; Active pattern learning techniques based on feedback of a supervisor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/98—Detection or correction of errors, e.g. by rescanning the pattern or by human intervention; Evaluation of the quality of the acquired patterns
- G06V10/987—Detection or correction of errors, e.g. by rescanning the pattern or by human intervention; Evaluation of the quality of the acquired patterns with the intervention of an operator
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention is a kind of accessory system being improved to flowers identifying system, the accessory system includes flowers identifying system, is not identified result output module, correction module, flower classification identifier, flowers automatic classification collection device, the flowers identifying system, is identified for plant photo or picture to acquisition;It is described not to be identified result output module, the plant photo or picture for exporting the flowers identifying system nonrecognition;The correction module, flowers name is given for the plant photo or picture to output;The flower classification identifier, for the photo or picture submitted according to the correction module and its described flowers name, carries out Classification and Identification at least one times;The flowers automatic classification collection device, for classifying, collecting, storing the plant photo or picture verified by the flower classification identifier, and feeds back in the flowers identifying system.
Description
Technical field
Present patent application is related to network technique field and picture contrast identification technology field, is specifically related to auxiliary improved
System tool.
Background technology
With the development of computer technology, deep learning algorithm can be based on, the image in sample image database is entered
Row training, which obtains image recognition model, (can such as be based on convolutional neural networks algorithm, the image in sample image database is entered
Row training obtains CNN (Convolutional Neural Network, convolutional neural networks) training pattern)), image recognition mould
Type can recognize the classification of picture, and add corresponding class label for picture.
The recognition accuracy of general pattern identification model is not up to absolutely, if what image recognition model to be recognized
The picture that error category label is added in substantial amounts of picture, corresponding substantial amounts of picture also compares many, and technical staff needs
The picture that with the addition of error category label is found out from substantial amounts of picture, so as to cause the effect for obtaining the picture of class label mistake
Rate is than relatively low.
With the progress of artificial intelligence technology, occur in that some apply the knowledge of artificial intelligence technology flower SEV/APP.Should
Kind knowing flower SEV/APP mainly applies the depth learning technology of artificial intelligence branch, with the plant image storehouse collected in advance
Flower chart picture is trained as material to machine recognition model.Finally flowers identification is carried out using the identification model trained.
Because when carrying out model training using material, simple machine learning is can not to know flowers title and species point
Category, it is therefore desirable to reference to human knowledge, be labeled in advance to picture in picture library, determine its corresponding colored name and kind.
Chinese Patent Application No. is CN201310461551.3, a kind of deep learning real-time performance major class of the disclosure of the invention
In the method for other image recognition, including training and identification process, training process, the Gabor characteristic of samples pictures is extracted simultaneously first
Maximum selection is carried out, hereafter linear local code is carried out using the feature code book Jing Guo clustering processing, finally using the golden word in space
Tower method carries out the export of characteristic vector, and is trained using support vector machine classifier;In identification process, by test pictures
Characteristic vector be identified using the support vector machine classifier trained.The present invention overcomes traditional method for extracting local feature
When semantic information shortage, the discrimination of multi-class image recognition can be obviously improved.
Chinese Patent Application No. is CN201510457979.X, and this application proposes kind of image-recognizing method and device, wherein,
The image-recognizing method comprises the following steps:The various dimensions local feature of image is extracted, and extracts the deep learning feature of image;
Various dimensions local feature and deep learning feature are spliced, and by metric learning spliced feature is learnt with
Obtain metric learning feature;Image is identified according to metric learning feature.The image-recognizing method of the embodiment of the present application, energy
Enough greatly improve the precision of image recognition.
Spend SEV/APP accuracys rate relatively high although knowing now, do not reach also completely correctly, can also often go out
Mistake, but the situation without a kind of mechanism for error is improved;
It is trained additionally, due to identification model according to material in picture library, is limited to the quantity and kind of picture library
Class, train come identification model can not cover the species in all real worlds, while in picture library picture number compared with
Few species, it recognizes that accuracy can be very low.
The content of the invention
In order to solve the technical problem that above-mentioned prior art is present, the present invention provides a kind of mechanism, in user's identification flowers
When, when known identification mistake occur or the species that can not recognized occur, new information can be collected, new information pair is reused
Model is trained improvement;By constantly collecting simultaneously sustained improvement model, higher and higher recognition accuracy is reached.The present invention is
It is achieved through the following technical solutions:
A kind of method being improved to flowers identifying system, methods described step includes:
S100:Obtain plant photo or picture;
S200:The flowers identifying system recognizes the plant photo or picture and classified;
S300:The flowers identifying system shows recognition result;
S400:Accessory system is confirmed to not identified result or feeds back to the flowers again after being modified identification
Identifying system.
Further, the method being improved to flowers identifying system, the S300 steps also comprise the following steps:
S310:The flowers identifying system stores the sorted identified plant photo or picture to knowledge
Other Flower Database;
S320:The flowers identifying system is exported to the accessory system to not being identified result.
Further, the method being improved to flowers identifying system, the S300 steps also comprise the following steps:
S500:The flowers identifying system exports the plant photo or figure of the flowers identifying system nonrecognition
Piece;
S600:Flowers name is given to the plant photo or picture of output, is then submitted;
S700:To submitting information (including plant photo or picture and its described flowers name) to carry out at least one times
Classification and Identification, then by the submission information (including plant photo or picture and its described flowers are named) by checking
Feed back in the flowers identifying system.
Further, the method being improved to flowers identifying system, the S700 steps also comprise the following steps:
S710:The Classification and Identification at least includes Inception-v3, ResNet-152, ResNet-101 and VGG-19 tetra-
Cover Classification and Identification model;
S720:When in multiple Classification and Identification models at least one described Classification and Identification model inspection result with it is described
Submit information consistent, it is assumed that submitting information by checking, corresponding photo or picture and classification information are collected auxiliary to first
Help identification Flower Database;
S730:By the submission information by checking, and feed back in step S200.
Further, the method being improved to flowers identifying system, the S700 steps also comprise the following steps:
S750:The plant photo or picture that can not assert for the Classification and Identification and its described flowers life
Name, carries out supplement identification and certification;
S760:To having classified and the plant photo or picture of certification and its described flowers name are collected to second
Assist in identifying Flower Database;
S770:By the submission information by checking, and feed back in step S200.
The present invention also provides a kind of intelligent flower identification assistant learning system:
A kind of accessory system being improved to flowers identifying system, the accessory system includes flowers identifying system, no
Identified result output module, correction module, flower classification identifier, flowers automatic classification collection device,
The flowers identifying system, is identified for plant photo or picture to acquisition;
It is described not to be identified result output module, the plant photo for exporting the flowers identifying system nonrecognition
Or picture;
The correction module, flowers name is given for the plant photo or picture to output;
The flower classification identifier, for the photo or picture submitted according to the correction module and its institute
Flowers name is stated, Classification and Identification at least one times is carried out;
The flowers automatic classification collection device, for classifying, collecting, storing what is verified by the flower classification identifier
The plant photo or picture, and feed back in the flowers identifying system.
Further, the accessory system being improved to flowers identifying system, it is described not to be identified result output mould
Block exports the plant photo or picture, and gives possible former plant ranking.
Further, the accessory system being improved to flowers identifying system, what the correction module was given
The flowers name includes flowers item name, flowers scientific name and/or flowers common first names.
Further, the accessory system being improved to flowers identifying system, the flower classification identifier is at least
Including tetra- sets of Classification and Identification preliminary examination modules of Inception-v3, ResNet-152, ResNet-101 and VGG-19, when (multiple institutes
State in Classification and Identification preliminary examination module at least) wherein there are a Classification and Identification preliminary examination module testing result and the error correcting
Module submits information consistent, it is assumed that submitting information by checking, just collects corresponding photo or picture and classification information
The flowers automatic classification collection device.
Further, the accessory system being improved to flowers identifying system, the accessory system also includes flowers
Classification annotation module, flower classification collection module,
The flower classification labeling module, for the plant photo that can not be assert for the flower classification identifier
Or picture and its described flowers name, carry out supplement identification and certification;
The flower classification collection module, for classified and certification the plant photo or picture and its institute
State flowers name be collected, store, and feed back in the flowers identifying system.
The present invention one of at least has the advantages that:
1. it will be rejected instant invention overcomes the picture or photo not confirmed by learning system originally, it is impossible to effectively learn again
The technical problem of habit.
There can be two kinds of modes to feed back help the accessory system that flowers identifying system is improved 2. the present invention is assigned
Flowers identifying system is learnt again.
3. present invention greatly enhances the learning ability of flowers identifying system and recognition effect, improve system user
Consumer's Experience.
4. the present invention can accomplish to make up the learning ability of original flowers identifying system and the deficiency of resolution capability.
5th, the accessory system being improved to flowers identifying system that the present invention is provided, reliability is high, while execution efficiency
It is high, have a wide range of application.
Brief description of the drawings
The present invention is described in further detail with reference to the accompanying drawings and detailed description:
Fig. 1 is first embodiment of the invention schematic flow sheet;
Fig. 2 is first embodiment of the invention system structure diagram;
Fig. 3 is first embodiment of the invention accessory system module diagram;
Fig. 4 is second embodiment of the invention accessory system module diagram;
Fig. 5 is third embodiment of the invention data flow diagram.
Description of reference numerals
Flowers identifying system -1000, (flowers identification) accessory system -2000, be not identified result output module -
2100th, correction module -2200, flower classification identifier -2300, flowers automatic classification collection device -2400, flowers point
Class labeling module -2500, flower classification collection module -2600.
Embodiment
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
There is the accompanying drawing used required in technology description to be briefly described, it should be apparent that, explanation and accompanying drawing are for the present invention below
It is exemplary, and is understood not to the limitation present invention.Numerous details are following description described to facilitate to this hair
Sensible solution.However, in some instances, well known or conventional details is not specified simultaneously, to meet the requirement that specification is succinct.
In one typical computing hardware configuration of the application, client/terminal, the network equipment and trusted party include one
Individual or multiple processors (CPU), input/output interface, network interface and internal memory.
Client, mobile terminal or the network equipment in the present invention include processor, are handled containing single core processor or multinuclear
Device.Processor is alternatively referred to as one or more microprocessors, CPU (CPU) etc..More specifically, processor can be
Complicated instruction set calculates (CISC) microprocessor, Jing Ke Cao Neng (RISC) microprocessor, very long instruction word (VLIW)
Microprocessor, the processor for realizing other instruction set, or realize the processor of instruction set combination.Processor can also be one or many
Individual application specific processor, such as application specific integrated circuit (ASIC), field programmable gate array (FPGA), digital signal processor
(DSP), network processing unit, graphics processor, network processing unit, communication processor, cipher processor, coprocessor, embedded
Processor or be capable of process instruction any other type logical block.Processor is used to perform the behaviour that the present invention is discussed
Make the instruction with step.
Client, mobile terminal or the network equipment in the present invention include memory, for storing big data, it may include one
It is individual or multiple volatile storage devices, such as random access memory (RAM), dynamic ram (DRAM), synchronous dram (SDRAM), quiet
State RAM (SRAM) or other kinds of storage device.Memory can store including by processor or any other equipment execution
The information of command sequence.For example, several operation systems, device driver, firmware (for example, input and output fundamental system or
) and/or the executable code and/or data of application program can be loaded in memory and by computing device BIOS.
The operating system of client, mobile terminal or the network equipment in the present invention can be any kind of operating system,
Windows, Windows Phone of such as Microsoft, Apple Inc. IOS, the Android of Google, and Linux,
Unix operating systems or other real-time or embedded OS VxWorks etc..
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
There is the accompanying drawing used required in technology description to be briefly described, it should be apparent that, explanation and accompanying drawing are for the present invention below
It is exemplary, and is understood not to the limitation present invention.Numerous details are following description described to facilitate to this hair
Sensible solution.However, in some instances, well known or conventional details is not specified simultaneously, to meet the requirement that specification is succinct.This
The equipment and control method of invention are referring to following embodiments:
First embodiment
If Fig. 1 is shown in first embodiment of the invention schematic flow sheet:
A kind of method being improved to flowers identifying system, methods described step includes:
S100:Obtain plant photo or picture;
S200:The flowers identifying system recognizes the plant photo or picture and classified;
S300:The flowers identifying system shows recognition result;
S400:Accessory system is confirmed to not identified result or feeds back to the flowers again after being modified identification
Identifying system.
Preferably, the method being improved to flowers identifying system, the S300 steps also comprise the following steps:
S310:The flowers identifying system stores the sorted identified plant photo or picture to knowledge
Other Flower Database;
S320:The flowers identifying system is exported to the accessory system to not being identified result.
Preferably, the method being improved to flowers identifying system, the S300 steps also comprise the following steps:
S500:The flowers identifying system exports the plant photo or figure of the flowers identifying system nonrecognition
Piece;
S600:Flowers name is given to the plant photo or picture of output, is then submitted;
S700:To submitting information (including plant photo or picture and its described flowers name) to carry out at least one times
Classification and Identification, then by the submission information (including plant photo or picture and its described flowers are named) by checking
Feed back in the flowers identifying system.
Preferably, the method being improved to flowers identifying system, the S700 steps also comprise the following steps:
S710:The Classification and Identification at least includes Inception-v3, ResNet-152, ResNet-101 and VGG-19 tetra-
Cover Classification and Identification model;I.e. flower classification identifier 2300 at least includes as above four sets of (flowers) Classification and Identification models.
S720:When in multiple Classification and Identification models at least one described Classification and Identification model inspection result with it is described
Submit information consistent, it is assumed that submitting information by checking, corresponding photo or picture and classification information are collected auxiliary to first
Help identification Flower Database;
S730:By the submission information by checking, and feed back in step S200.
The present embodiment also provides a kind of intelligent flower identification assistant learning system, and its relation with flowers identifying system is as schemed
2 be shown in first embodiment of the invention system structure diagram:
A kind of accessory system being improved to flowers identifying system, such as accessory system 2000, Fig. 3 are the present invention the
Shown in one embodiment accessory system module diagram, including be not identified result output module 2100, correction module 2200,
Flower classification identifier 2300, flowers automatic classification collection device 2400,
The flowers identifying system 1000, is identified for plant photo or picture to acquisition;
It is described not to be identified result output module 2100, the plant for exporting the flowers identifying system nonrecognition
Photo or picture;
The correction module 2200, flowers name is given for the plant photo or picture to output;
The flower classification identifier 2300, for the photo or picture submitted according to the correction module and
Its described flowers name, carries out Classification and Identification at least one times;
The flowers automatic classification collection device 2400, is tested for classifying, collecting, store by the flower classification identifier
The plant photo or picture of card, and feed back in the flowers identifying system.
Preferably, the accessory system being improved to flowers identifying system, it is described not to be identified result output mould
Block 2100 exports the plant photo or picture, and gives possible former plant ranking.
Preferably, the accessory system being improved to flowers identifying system, the correction module 2200 is given
The flowers name given includes flowers item name, flowers scientific name and/or flowers common first names.
Preferably, the accessory system being improved to flowers identifying system, the flower classification identifier 2300
At least include tetra- sets of Classification and Identification preliminary examination modules of Inception-v3, ResNet-152, ResNet-101 and VGG-19, when (many
In the individual Classification and Identification preliminary examination module at least) wherein there are a Classification and Identification preliminary examination module testing result and the mistake
Correcting module submits information consistent, it is assumed that submitting information by checking, just returns corresponding photo or picture and classification information
Collect the flowers automatic classification collection device 2400.
Second embodiment
On the basis of embodiment one, it is the present invention second that further preferred flowers identification, which improves accessory system such as Fig. 4,
Shown in embodiment accessory system module diagram:
It is preferred that the accessory system being improved to flowers identifying system, the accessory system also include flower classification mark
Module 2500, flower classification collection module 2600,
The flower classification labeling module 2500, for the plant that can not be assert for the flower classification identifier
Photo or picture and its described flowers name, carry out supplement identification and certification;
The flower classification collection module 2600, for classified and certification the plant photo or picture and
Its described flowers name is collected, stored, and feeds back in the flowers identifying system 1000.
The further preferred method being improved to flowers identifying system, the S700 steps also comprise the following steps:
S750:The plant photo or picture that can not assert for the Classification and Identification and its described flowers life
Name, carries out supplement identification and certification;
S760:To having classified and the plant photo or picture of certification and its described flowers name are collected to second
Assist in identifying Flower Database;
S770:By the submission information by checking, and feed back in step S200.
3rd embodiment
Convolutional neural networks are to be used for the important algorithm of image recognition in deep learning.Present convolutional neural networks depth
Learning method can make full use of mass data to learn the characteristic information to great amount of images automatically, it is not necessary to prior hand-designed tool
Body sample characteristics, the parameter of model from data learning by obtaining, wherein the parameter of thousands of can be included, and can band
Carry out higher identification accuracy.Training convolutional neural networks need the data set of mass data, and data set needs correspondence classification
The label of title.
First, the identification model based on convolutional neural networks is carried out using the existing data set comprising a large amount of flower classifications
Training.Then classification is identified in the flowers picture taken pictures or chosen to user using the model trained.Whole process is such as
Under:
1) the convolutional neural networks model for recognizing flowers is set up, model can be used what existing google increased income
Inception-v4 models
2) prepare the data set of flower chart picture, the plant classification picture library of Largest In China can be used --- Chinese Plants figure
The training of model is carried out as the flowers photo of storehouse (PlantPhoto Bank ofChina, PPBC).These images have been labelled with
Item name.Data set can be divided into training dataset and test validation data set, and the model obtained after the completion of training is also needed to
Using test validation data set checking accuracy, only reach that just calculation obtains last identification model after certain accuracy.
3) identification model trained is put into server end (SEV)
4) when user is using flower SEV/APP is known, flowers are taken pictures or selected with flower chart picture, SEV/APP is pre- by picture compression
Handle and end of uploading onto the server.
5) server end is calculated the picture received automatically using identification model, and finally draws classification index value
And probable value, and identification is obtained into the corresponding item name of index value and Credibility probability value return SEV/APP display results.
User is optional when mistake is recognized as known to occurring or can not recognized when recognizing flowers using SEV/APP
Submission recognition failures feedback is selected, to selecting correct classification in the classification that can be provided from system of correct classification known to user, to being
The classification not having in system can directly fill in correct classification;Correct classification is not known to user and system identification does not go out or recognition accuracy
Low-down situation, directly submits feedback.
Server end corresponding with service module is received after request, will above upload image corresponding with the feedback request, with
And the information submitted is given to preliminary examination module together.Multiple different Classification and Identification models are had in preliminary examination module, these classification are known
Other model has used the algorithms different from system identification model, such as Inception-v3, ResNet-152, ResNet-101,
VGG-19 etc..Preliminary examination module can give the feedback information including category result many Classification and Identification model treatment checkings, when multiple
At least one Classification and Identification model inspection result is consistent with submitting information in Classification and Identification model, it is assumed that submitting information to pass through
Checking, automatic classification based training collection is collected by respective image and classification information.
Only need to set a threshold value (such as 50%) when being detected using multiple Classification and Identification models, as long as a certain point
Class testing result Credibility probability reaches threshold value and submits classification information consistent with its, so that it may pass through checking at last.Use many points
The purpose that class identification model is detected, one is the accuracy that information is submitted in checking, and two be to reduce artificial participation, is realized automatically whole
Individual process.
For without the information and the information not over many Classification and Identification model preliminary examinations for submitting corrigendum category result, entering
Row collects, and by manually participating in taxonomic revision, addition or confirmation mark, forms manual sort's training set.
The identification model of system is instructed again using the automatic classification based training collection and manual sort's training set after arrangement
Practice and improve, after checking accuracy is improved after the completion of new model training, then more new system uses the identification model after new training.
The continuous persistent loop of whole flow process is carried out, and is constantly carried out retrofit identification model, is improved its recognition accuracy.
The present embodiment is further illustrated, Fig. 5 is referred to for third embodiment of the invention data flow diagram:
200 identification models represented are initially the flower for being trained according to existing plant picture classification storehouse and completing and obtaining
Grass identification model;
100-200-300 process is user using the main process for knowing flower SEV/APP:User is adjusted using SEV/APP
With taking pictures or selecting photo, SEV/APP pre-processes compression of images and end of uploading onto the server, and server end uses identification model
The picture received is calculated by model, and draws flowers item name and Credibility probability value, and result is returned into SEV/APP
It is shown to user;
400-500 represents that user is seen after the knowledge flower result that SEV/APP is shown, if it is confirmed that the recognition result that system is given is wrong
By mistake, identification error feedback can be submitted in SEV/APP, if while user knows correct flowers item name, also user people
System is together submitted to for correct item name;
600 expression server-side systems receive the error feedback of user's submission, using preliminary examination module to submitting result to carry out
Automatic checking;Preliminary examination module uses multiple different Classification and Identification models, and these Classification and Identification models have been used and system identification
The different algorithm of model, when at least one Classification and Identification model inspection result in multiple Classification and Identification models and submission information one
Cause, it is assumed that submitting information by checking, respective image and classification information are collected into automatic classification based training collection.
700 automatic classification based training collection store the image and correspondence contingency table signature that system is verified.
700-200 processes are that system is automatic by training set according to certain strategy (as regular or training set reaches certain amount)
Data are as new material, and the identification model in system 200 is trained again, by the continuous training of this new material come
Continue to optimize identification model.Its continuous training process is still using the algorithm of former deep learning, and training process can be automatically constantly excellent
Change its inner parameter.
800-900 processes are a supplements to the automatic verification process of system, and when system, checking user submits feedback information
When inconsistent, this kind of user can be listed in background management system and submits image and information, by system manager or specified knowledge
Expert reaffirms and done classification annotation.System can also serve as the image reaffirmed and marked addition training set new
Train material.
900-200 process is similar with 700-200 process.
It is obvious to a person skilled in the art that the invention is not restricted to the details of above-mentioned one exemplary embodiment, Er Qie
In the case of without departing substantially from spirit or essential attributes of the invention, the present invention can be realized in other specific forms.Therefore, no matter
From the point of view of which point, embodiment all should be regarded as exemplary, and be nonrestrictive, the scope of the present invention is by appended power
Profit is required rather than described above is limited, it is intended that all in the implication and scope of the equivalency of claim by falling
Change is included in the present invention.Any reference in claim should not be considered as to the claim involved by limitation.This
Outside, it is clear that the word of " comprising " one is not excluded for other units or step, and odd number is not excluded for plural number.That is stated in device claim is multiple
Unit or device can also be realized by a unit or device by software or hardware.The first, the second grade word is used for table
Show title, and be not offered as any specific order.
Claims (10)
1. a kind of method being improved to flowers identifying system, it is characterised in that methods described step includes:
S100:Obtain plant photo or picture;
S200:The flowers identifying system recognizes the plant photo or picture and classified;
S300:The flowers identifying system shows recognition result;
S400:Accessory system is confirmed or is modified after identification to feed back to the flowers identification again to not being identified result
System.
2. the method according to claim 1 being improved to flowers identifying system, it is characterised in that the S300 steps
Also comprise the following steps:
S310:The sorted identified plant photo or picture are stored to identification and spent by the flowers identifying system
Grass database;
S320:The flowers identifying system is exported to the accessory system to not being identified result.
3. the method according to claim 1 being improved to flowers identifying system, it is characterised in that the S300 steps
Also comprise the following steps:
S500:The flowers identifying system exports the plant photo or picture of the flowers identifying system nonrecognition;
S600:Flowers name is given to the plant photo or picture of output, is then submitted;
S700:To submitting the Classification and Identification of information progress at least one times, then it will be fed back to by the submission information of checking
In the flowers identifying system.
4. the method according to claim 3 being improved to flowers identifying system, it is characterised in that the S700 steps
Also comprise the following steps:
S710:The Classification and Identification at least includes tetra- sets points of Inception-v3, ResNet-152, ResNet-101 and VGG-19
Class identification model;
S720:When at least one described Classification and Identification model inspection result is submitted with described in multiple Classification and Identification models
Information is consistent, and corresponding photo or picture and classification information are collected to first and assist in identifying Flower Database;
S730:By the submission information by checking, and feed back in step S200.
5. the method according to claim 3 being improved to flowers identifying system, it is characterised in that the S700 steps
Also comprise the following steps:
S750:The plant photo or picture that can not assert for the Classification and Identification and its described flowers name, enter
The identification of row supplement and certification;
S760:To having classified and the plant photo or picture of certification and its described flowers name are collected to the second auxiliary
Recognize Flower Database;
S770:By the submission information by checking, and feed back in step S200.
6. a kind of accessory system being improved to flowers identifying system, the accessory system include flowers identifying system, not by
Recognition result output module, correction module, flower classification identifier, flowers automatic classification collection device, it is characterised in that
The flowers identifying system, is identified for plant photo or picture to acquisition;
It is described not to be identified result output module, for export the flowers identifying system nonrecognition the plant photo or
Picture;
The correction module, flowers name is given for the plant photo or picture to output;
The flower classification identifier, for the photo or picture and its described flower submitted according to the correction module
Grass is named, and carries out Classification and Identification at least one times;
The flowers automatic classification collection device, for classifying, collecting, store by described in flower classification identifier checking
Plant photo or picture, and feed back in the flowers identifying system.
7. the accessory system according to claim 6 being improved to flowers identifying system, it is characterised in that it is described not by
Recognition result output module exports the plant photo or picture, and gives possible former plant ranking.
8. the accessory system according to claim 6 being improved to flowers identifying system, it is characterised in that the mistake
The flowers name that correction module is given includes flowers item name, flowers scientific name and/or flowers common first names.
9. the accessory system according to claim 6 being improved to flowers identifying system, it is characterised in that the flowers
Classification and Identification device at least includes tetra- sets of Classification and Identification preliminary examinations of Inception-v3, ResNet-152, ResNet-101 and VGG-19
Module, when wherein thering is a Classification and Identification preliminary examination module testing result and the correction module to submit information consistent,
Corresponding photo or picture and classification information are just collected into the flowers automatic classification collection device.
10. the accessory system according to claim 6 being improved to flowers identifying system, it is characterised in that described auxiliary
Auxiliary system also includes flower classification labeling module, flower classification collection module,
The flower classification labeling module, for the plant photo that can not assert for the flower classification identifier or
Picture and its described flowers name, carry out supplement identification and certification;
The flower classification collection module, for classified and certification the plant photo or picture and its described flower
Grass name is collected, stored, and feeds back in the flowers identifying system.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710338697.7A CN107153844A (en) | 2017-05-12 | 2017-05-12 | The accessory system being improved to flowers identifying system and the method being improved |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710338697.7A CN107153844A (en) | 2017-05-12 | 2017-05-12 | The accessory system being improved to flowers identifying system and the method being improved |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107153844A true CN107153844A (en) | 2017-09-12 |
Family
ID=59793677
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710338697.7A Pending CN107153844A (en) | 2017-05-12 | 2017-05-12 | The accessory system being improved to flowers identifying system and the method being improved |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107153844A (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108171275A (en) * | 2018-01-17 | 2018-06-15 | 百度在线网络技术(北京)有限公司 | For identifying the method and apparatus of flowers |
CN108229979A (en) * | 2018-01-24 | 2018-06-29 | 亳州职业技术学院 | Based on wechat small routine and the American Ginseng of depth learning technology tasting assistant |
CN109101547A (en) * | 2018-07-05 | 2018-12-28 | 北京泛化智能科技有限公司 | Management method and device for wild animal |
CN109147093A (en) * | 2018-09-04 | 2019-01-04 | 惠州市德赛西威智能交通技术研究院有限公司 | A kind of picture sample acquisition device and method |
CN109190712A (en) * | 2018-09-21 | 2019-01-11 | 福州大学 | A kind of line walking image automatic classification system of taking photo by plane based on deep learning |
CN109389161A (en) * | 2018-09-28 | 2019-02-26 | 广州大学 | Rubbish identification evolutionary learning method, apparatus, system and medium based on deep learning |
CN109409690A (en) * | 2018-09-29 | 2019-03-01 | 广州绿桦环保科技有限公司 | A kind of plant conservation system |
CN109492684A (en) * | 2018-10-31 | 2019-03-19 | 西安同瑞恒达电子科技有限公司 | Data processing method and device |
CN109886324A (en) * | 2019-02-01 | 2019-06-14 | 广州云测信息技术有限公司 | Icon-based programming method and apparatus |
CN110287882A (en) * | 2019-06-26 | 2019-09-27 | 北京林业大学 | A kind of big chrysanthemum kind image-recognizing method based on deep learning |
CN113420196A (en) * | 2021-06-07 | 2021-09-21 | 青岛海信智慧生活科技股份有限公司 | Commodity category determination method, device, equipment and medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1806501A (en) * | 2005-01-17 | 2006-07-26 | 厦门市汇阳科技有限公司 | Marine phytoplankton automatic distinguishing method and apparatus |
US20080075361A1 (en) * | 2006-09-21 | 2008-03-27 | Microsoft Corporation | Object Recognition Using Textons and Shape Filters |
CN102354366A (en) * | 2011-09-23 | 2012-02-15 | 上海合合信息科技发展有限公司 | Network based image recognition method and system |
CN103336974A (en) * | 2013-06-21 | 2013-10-02 | 华南理工大学 | Flower and plant category recognition method based on local constraint sparse characterization |
CN105844286A (en) * | 2016-03-11 | 2016-08-10 | 博康智能信息技术有限公司 | Newly added vehicle logo identification method and apparatus |
-
2017
- 2017-05-12 CN CN201710338697.7A patent/CN107153844A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1806501A (en) * | 2005-01-17 | 2006-07-26 | 厦门市汇阳科技有限公司 | Marine phytoplankton automatic distinguishing method and apparatus |
US20080075361A1 (en) * | 2006-09-21 | 2008-03-27 | Microsoft Corporation | Object Recognition Using Textons and Shape Filters |
CN102354366A (en) * | 2011-09-23 | 2012-02-15 | 上海合合信息科技发展有限公司 | Network based image recognition method and system |
CN103336974A (en) * | 2013-06-21 | 2013-10-02 | 华南理工大学 | Flower and plant category recognition method based on local constraint sparse characterization |
CN105844286A (en) * | 2016-03-11 | 2016-08-10 | 博康智能信息技术有限公司 | Newly added vehicle logo identification method and apparatus |
Non-Patent Citations (1)
Title |
---|
微软亚洲研究院 等: "微软识花", 《HTTPS://WWW.IFANR.COM/APP/730436》 * |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108171275A (en) * | 2018-01-17 | 2018-06-15 | 百度在线网络技术(北京)有限公司 | For identifying the method and apparatus of flowers |
CN108229979A (en) * | 2018-01-24 | 2018-06-29 | 亳州职业技术学院 | Based on wechat small routine and the American Ginseng of depth learning technology tasting assistant |
CN109101547A (en) * | 2018-07-05 | 2018-12-28 | 北京泛化智能科技有限公司 | Management method and device for wild animal |
CN109101547B (en) * | 2018-07-05 | 2021-11-12 | 北京泛化智能科技有限公司 | Management method and device for wild animals |
CN109147093B (en) * | 2018-09-04 | 2021-01-12 | 惠州市德赛西威智能交通技术研究院有限公司 | Picture sample acquisition device and method |
CN109147093A (en) * | 2018-09-04 | 2019-01-04 | 惠州市德赛西威智能交通技术研究院有限公司 | A kind of picture sample acquisition device and method |
CN109190712A (en) * | 2018-09-21 | 2019-01-11 | 福州大学 | A kind of line walking image automatic classification system of taking photo by plane based on deep learning |
CN109389161A (en) * | 2018-09-28 | 2019-02-26 | 广州大学 | Rubbish identification evolutionary learning method, apparatus, system and medium based on deep learning |
CN109409690A (en) * | 2018-09-29 | 2019-03-01 | 广州绿桦环保科技有限公司 | A kind of plant conservation system |
CN109492684A (en) * | 2018-10-31 | 2019-03-19 | 西安同瑞恒达电子科技有限公司 | Data processing method and device |
CN109886324A (en) * | 2019-02-01 | 2019-06-14 | 广州云测信息技术有限公司 | Icon-based programming method and apparatus |
CN110287882A (en) * | 2019-06-26 | 2019-09-27 | 北京林业大学 | A kind of big chrysanthemum kind image-recognizing method based on deep learning |
CN113420196A (en) * | 2021-06-07 | 2021-09-21 | 青岛海信智慧生活科技股份有限公司 | Commodity category determination method, device, equipment and medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107153844A (en) | The accessory system being improved to flowers identifying system and the method being improved | |
CN111444236A (en) | Mobile terminal user portrait construction method and system based on big data | |
CN104899579A (en) | Face recognition method and face recognition device | |
CN109284675A (en) | A kind of recognition methods of user, device and equipment | |
CN113707323B (en) | Disease prediction method, device, equipment and medium based on machine learning | |
CN111062260A (en) | Automatic generation method of facial cosmetic recommendation scheme | |
US11527102B2 (en) | Systems and methods of automated biometric identification reporting | |
CN110414298A (en) | A kind of more attribute joint recognition methods of monkey face | |
CN114372532B (en) | Method, device, equipment, medium and product for determining label labeling quality | |
CN113707304A (en) | Triage data processing method, device, equipment and storage medium | |
CN113779295A (en) | Retrieval method, device, equipment and medium for abnormal cell image features | |
CN117130938A (en) | Method and device for generating test cases based on knowledge graph | |
CN116313021B (en) | Medical treatment management system and method for hospital | |
CN116469505A (en) | Data processing method, device, computer equipment and readable storage medium | |
CN116416522A (en) | Plant species determination method, plant species determination device and computer readable storage medium | |
JP2021015549A (en) | Information processing method and information processing device | |
CN111866376B (en) | Feature recognition method and device, pan-tilt camera and storage medium | |
CN105989094A (en) | Image retrieval method based on middle layer expression of hidden layer semantics | |
CN112016515A (en) | File cabinet vacancy detection method and device | |
CN116364248B (en) | Automatic pathological image uploading and matching method based on automatic slide label identification technology | |
CN118135333B (en) | Medical picture intelligent sorting method and device, electronic equipment and readable storage medium | |
TWI852643B (en) | Artificial intelligence facial emotion recognition system | |
TW202022771A (en) | Species data analysis method, system and computer program product | |
KR102614062B1 (en) | Method and apparatus for quarantine of imported ornamental fish through data preprocessing and deep neural network-based image detection and classification | |
CN117496323B (en) | Multi-scale second-order pathological image classification method and system based on transducer |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20170912 |
|
WD01 | Invention patent application deemed withdrawn after publication |