CN110472609A - A kind of birds image-recognizing method, device, equipment and storage medium - Google Patents
A kind of birds image-recognizing method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN110472609A CN110472609A CN201910775414.4A CN201910775414A CN110472609A CN 110472609 A CN110472609 A CN 110472609A CN 201910775414 A CN201910775414 A CN 201910775414A CN 110472609 A CN110472609 A CN 110472609A
- Authority
- CN
- China
- Prior art keywords
- birds
- images
- recognized
- target
- genius loci
- 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
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
- G06V10/464—Salient features, e.g. scale invariant feature transforms [SIFT] using a plurality of salient features, e.g. bag-of-words [BoW] representations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
This application discloses a kind of birds image-recognizing method, device, equipment and storage mediums, and wherein method includes: the images to be recognized that acquisition includes birds target；The region where birds target is obtained after carrying out local area localization to images to be recognized based on preset location algorithm；According to multiple location Feature Selection Model, feature extraction is carried out to the region where birds target, obtains multiple genius locis of birds target；Identify that obtain the corresponding similarity score of each genius loci, verifying has one-to-one relationship between genius loci and genius loci to each genius loci using classifier and verifying genius loci；The recognition result that birds target in images to be recognized is calculated according to all similarity scores, solves the low technical problem of existing image-recognizing method recognition efficiency.
Description
Technical field
This application involves technical field of image processing more particularly to a kind of birds image-recognizing method, device, equipment and deposit
Storage media.
Background technique
With the rapid development of aircraft industry, bird knocks machine event away and gradually increases, and not only brings huge economy to aircraft industry
Loss, while the life security of pilot and passenger have been injured, therefore effectively the collision of aircraft and flying bird is avoided to cause
The extensive concern of people.
Statistics shows that most of plane collision incident is taken off and landing period, existing main to utilize image procossing skill
Art identifies the birds of low-latitude flying above airport and its near zone, and wherein image recognition is the important ring in image procossing,
Its correctness directly affects recognition result.But existing image-recognizing method recognition efficiency is low.
Summary of the invention
In view of this, being solved this application provides a kind of birds image-recognizing method, device, equipment and storage medium
The low technical problem of existing image-recognizing method recognition efficiency.
The application first aspect provides a kind of birds image-recognizing method, comprising:
Acquisition includes the images to be recognized of birds target；
The birds target is obtained after carrying out local area localization to the images to be recognized based on preset location algorithm
The region at place；
According to multiple location Feature Selection Model, feature extraction is carried out to the region where the birds target, is obtained described
Multiple genius locis of birds target；
Each genius loci is identified using classifier and verifying genius loci, obtains each genius loci
Corresponding similarity score has one-to-one relationship between the verifying genius loci and the genius loci；
The recognition result of birds target described in images to be recognized is calculated according to all similarity scores.
Optionally, described to be based on preset location algorithm, after carrying out local area localization to the images to be recognized, obtain institute
Region where stating birds target specifically includes:
Classification activation mapping is carried out to images to be recognized, obtains the corresponding thermodynamic chart of the images to be recognized；
The pixel average of the thermodynamic chart is calculated according to all pixels value in the thermodynamic chart；
The all pixels point in the thermodynamic chart is subjected to binary conversion treatment according to the pixel average, and by pixel value
For 1 the region that surrounds of pixel as the region where the birds target.
Optionally, the multiple location Feature Selection Model includes convolutional neural networks model,
It is then described according to multiple location Feature Selection Model, feature extraction is carried out to the region where the birds target, is obtained
Multiple genius locis to the birds target specifically include:
Convolution operation is carried out to the region where the birds target according to the preset convolutional neural networks model, and will
Positioned at different convolutional layers convolution results successively up-sampled, down-sampling, weighting be multiplied after, obtain corresponding sampled result,
Using a sampled result as a genius loci.
Optionally, the identification knot that birds target described in images to be recognized is calculated according to all similarity scores
Fruit specifically includes:
To all similarity score averageds, similarity score average value is obtained, judges the identification score
Whether average value is greater than preset threshold, if the verifying genius loci is then made the bird kind of corresponding verifying birds as institute
The recognition result of birds target is stated, if it is not, then to executing after re-starting feature extraction to the region where the birds target
Subsequent step.
Optionally, described that each genius loci is identified using classifier and verifying genius loci, it obtains each
The genius loci corresponding similarity score has to correspond between the verifying genius loci and the genius loci and close
System specifically includes:
It obtains for verifying the verifying genius loci for taking each genius loci；
According to the classifier, the corresponding genius loci is identified using the verifying genius loci, is obtained
Corresponding similarity score.
Optionally, described to be based on preset location algorithm, after carrying out local area localization to the images to be recognized, obtain institute
Before stating the region where birds target further include:
The images to be recognized is normalized.
Optionally, the normalized specifically includes:
According to prediction picture size, the images to be recognized is cut, so that the images to be recognized and described preset
Image etc. is big；
Scaled down is carried out to the corresponding pixel value of each pixel in the images to be recognized after cutting.
The application second aspect provides a kind of birds pattern recognition device, comprising:
Acquiring unit, for obtain include birds target images to be recognized；
Positioning unit after carrying out local area localization to the images to be recognized, is obtained for being based on preset location algorithm
Region where the birds target；
Feature extraction unit, for being carried out to the region where the birds target according to multiple location Feature Selection Model
Feature extraction obtains multiple genius locis of the birds target；
First recognition unit, for being identified using classifier and verifying genius loci to each genius loci,
The corresponding similarity score of each genius loci is obtained, is had one by one between the verifying genius loci and the genius loci
Corresponding relationship；
Second recognition unit, for calculating birds target described in images to be recognized according to all similarity scores
Recognition result.
The application third aspect provides a kind of birds image recognition apparatus, comprising: processor and memory；
Said program code is transferred to the processor for storing program code by the memory；
The processor is used for the birds image recognition side according to the instruction execution first aspect of said program code
Method.
The application fourth aspect provides a kind of storage medium, and the storage medium is for storing program code, the journey
Sequence code is for executing birds image-recognizing method described in first aspect.
As can be seen from the above technical solutions, the application has the following advantages:
This application provides a kind of birds image-recognizing methods, comprising: acquisition includes the images to be recognized of birds target；
The region where birds target is obtained after carrying out local area localization to images to be recognized based on preset location algorithm；According to more
Genius loci extracts model, carries out feature extraction to the region where birds target, obtains multiple genius locis of birds target；
Each genius loci is identified using classifier and verifying genius loci, the corresponding similarity of each genius loci is obtained and comments
Point, verifying has one-to-one relationship between genius loci and genius loci；Figure to be identified is calculated according to all similarity scores
The recognition result of birds target as in.
In the application, the overall space in the region where entire birds target is directly positioned in images to be recognized, then
Feature extraction is carried out simultaneously using multiple positions of the multiple location Feature Selection Model to birds target, what final basis extracted
After feature is identified, recognition result is obtained, the implementation method settled at one go is used in positioning and feature extraction phases, solves
Existing image-recognizing method recognition efficiency low technical problem.
Detailed description of the invention
Fig. 1 is a kind of flow diagram of the embodiment one of birds image-recognizing method in the embodiment of the present application；
Fig. 2 is a kind of flow diagram of the embodiment two of birds image-recognizing method in the embodiment of the present application；
Fig. 3 is a kind of structural schematic diagram of the embodiment of birds pattern recognition device of the embodiment of the present application；
Fig. 4 is CAM schematic diagram in the embodiment of the present application；
Fig. 5 is the calculating schematic diagram that thermodynamic chart weighting is multiplied in the embodiment of the present application；
Fig. 6 is the structural schematic diagram of preset convolutional neural networks model in the embodiment of the present application.
Specific embodiment
The embodiment of the present application provides a kind of birds image-recognizing method, device, equipment and storage medium, solves existing
The low technical problem of image-recognizing method recognition efficiency.
In order to make those skilled in the art more fully understand application scheme, below in conjunction in the embodiment of the present application
Attached drawing, the technical scheme in the embodiment of the application is clearly and completely described, it is clear that described embodiment is only this
Apply for a part of the embodiment, instead of all the embodiments.Based on the embodiment in the application, those of ordinary skill in the art exist
Every other embodiment obtained under the premise of creative work is not made, shall fall in the protection scope of this application.
Referring to Fig. 1, a kind of flow diagram of the embodiment one of birds image-recognizing method in the embodiment of the present application, packet
It includes:
Step 101, acquisition include the images to be recognized of birds target.
It should be noted that carry out Object identifying to the birds target in images to be recognized, obtaining first includes bird
Classification target images to be recognized.
Step 102 is based on preset location algorithm, after carrying out local area localization to images to be recognized, obtains birds target
The region at place.
It should be noted that after obtaining images to be recognized, to the birds target in images to be recognized according to preset positioning
Algorithm directly extracts the region where entire birds target, is positioned step by step to different parts compared to existing
Method, calculating speed are more quick.
Step 103 obtains the region progress feature extraction where birds target according to multiple location Feature Selection Model
Multiple genius locis of birds target.
It should be noted that after obtaining region shared by the entire birds target in images to be recognized, directly according to more
Genius loci extracts model, while extracting multiple genius locis of entire birds target, is no longer according to previous one one
A mode extracted in turn, extraction rate are relatively rapid.
Step 104 identifies each genius loci using classifier and verifying genius loci, obtains each genius loci
Corresponding similarity score, verifying has one-to-one relationship between genius loci and genius loci.
It should be noted that obtain in entire birds target after multiple genius locis, it is special to each position using classifier
Corresponding verifying genius loci of seeking peace is identified, the corresponding similarity score of the genius loci is obtained.
Step 105, the recognition result that birds target in images to be recognized is calculated according to all similarity scores.
It should be noted that after obtaining all similarity scores, according to all similarity scores can calculate to
Identify the recognition result of birds target in image.
In the present embodiment, the overall space in the region where entire birds target is directly positioned in images to be recognized, so
Feature extraction is carried out simultaneously using multiple positions of the multiple location Feature Selection Model to birds target afterwards, final basis extracts
Feature identified after, obtain recognition result, use the implementation method that settles at one go, solution in positioning and feature extraction phases
It has determined the low technical problem of existing image-recognizing method recognition efficiency.
The above are a kind of embodiments one of birds image-recognizing method provided by the embodiments of the present application, and the following are the application realities
A kind of embodiment two of birds image-recognizing method of example offer is provided.
Referring to Fig. 2, a kind of flow diagram of the embodiment two of birds image-recognizing method in the embodiment of the present application, packet
It includes:
Step 201, acquisition include the images to be recognized of birds target.
It should be noted that in order to which the description of subsequent step is clear in the present embodiment, for example, in such as the present embodiment
Images to be recognized have a wild goose.
Images to be recognized is normalized in step 202.
It should be noted that the picture size of different images to be recognized is different, and it is subsequently used for carrying out classification
Corresponding activation mapping is handled the image of a dimensions, it is therefore desirable to treat knowledge according to the size of unified specification
Other image is normalized.
It should be noted that normalized specifically includes: according to prediction picture size, images to be recognized is cut,
So that images to be recognized and prediction picture etc. are big；The corresponding pixel value of each pixel in images to be recognized after cutting is carried out
Scaled down.
It is understood that the size for the images to be recognized that such as step 201 is got is 48*52, and classification activation is reflected
The size that can be handled when penetrating is 46*46, thus needs the images to be recognized by 48*52 size to be cropped to 46*46 big.
For traditional pixel value between 0-255, this kind of value data is larger in the present embodiment, when calculating, computational efficiency
It is low, calculating speed is slow, therefore the corresponding pixel value of each pixel is subjected to scaled down in the present embodiment.In the present embodiment
Diminution ratio according toIt carries out.
Step 203 carries out classification activation mapping to images to be recognized, obtains the corresponding thermodynamic chart of images to be recognized.
It should be noted that carrying out classification activation to images to be recognized after images to be recognized is normalized
Mapping, obtains the corresponding thermodynamic chart of images to be recognized.It is understood that as shown in figure 4, classification activation mapping can be
Class Activation Mapping (referred to as CAM) method, just as the image that thermal imaging system generates, CAM can be generated therewith
Similar image, CAM can be shown its decision-making foundation in the form of thermodynamic chart, the target area in original image obtained in a manner of visual
Domain.Therefore the CAM regional location positioning for being used for bird in birds image is very effective method.
Step 204, the pixel average that thermodynamic chart is calculated according to all pixels value in thermodynamic chart.
In the present embodiment, after obtaining thermodynamic chart, the pixel for calculating thermodynamic chart according to all pixels value in thermodynamic chart is average
Value, for example, the pixel average of thermodynamic chart obtained in the present embodiment is 180.
All pixels point in thermodynamic chart is carried out binary conversion treatment according to pixel average by step 205, and by pixel value
For 1 the region that surrounds of pixel as the region where birds target.
It can be with it should be noted that all pixels point in thermodynamic chart is carried out binary conversion treatment according to pixel average
It is that the pixel less than pixel value average value is set to 0, the pixel greater than pixel average is set to 1, is then 1 by pixel value
The region that surrounds of pixel as the region where birds target.
Step 206 carries out convolution operation to the region where birds target according to preset convolutional neural networks model, and will
Positioned at different convolutional layers convolution results successively up-sampled, down-sampling, weighting be multiplied after, obtain corresponding sampled result,
Using a sampled result as a genius loci.
It should be noted that convolutional neural networks model be by convolutional neural networks it is trained after obtain, to convolution
Neural network model in training, choose it is a certain number of include birds image training picture, treat trained icon and carry out
Then the training picture for being added to label is input in convolutional neural networks, until convolutional Neural net by the addition of position label
When training result obtained in network is similar to the birds feature in picture, correspondence obtains convolutional neural networks model.
What the different parts for the birds that different layers is carried out, therefore in utilization convolutional neural networks model to birds mesh
After region where mark carries out feature extraction, the convolution results of different layers correspond to the genius loci of different parts, for example, head,
Foot, tail portion and body.In order to enable recognition result is more acurrate, each genius loci is according to its complexity or important
Degree carries out weight imparting, therefore in feature extraction, is weighted multiplication processing to convolution results.It is by the layer that weighting, which is multiplied,
Convolution results be multiplied with the weighted value of this layer, specifically may refer to Fig. 5 process carry out.
It is understood that the up-sampling of image, down-sampling belong to the prior art, details are not described herein.
Preset convolutional neural networks model in the present embodiment is as shown in Figure 6.
Step 207 is obtained for verifying the verifying genius loci for taking each genius loci.
It should be noted that need to carry out each genius loci corresponding identification verifying after obtaining each genius loci,
Therefore before carrying out identification verifying to each genius loci, the verifying genius loci of the corresponding verifying birds of each genius loci is obtained.
Such as the genius loci of the birds target of images to be recognized has head feature, foot's feature, tail feature and physical trait respectively,
Find verifying birds image unique characteristics are divided according to above-mentioned position according to above-mentioned position after, be verified position
Feature.
Step 208, according to classifier, corresponding genius loci is identified using verifying genius loci, is corresponded to
Similarity score.
It should be noted that respectively by head feature, foot's feature, tail feature and the physical trait in images to be recognized
Classify with corresponding verifying genius loci, obtains head feature, foot's feature, tail feature and physical trait respectively
Corresponding similarity score.
Step 209, to all similarity score averageds, obtain similarity score average value.
Step 210 judges to identify whether score average is greater than preset threshold, if so then execute step 211, if otherwise holding
Row step 212.
It should be noted that if identification score average at this time is greater than preset threshold, then it is assumed that be identified for identification
Birds in the verifying birds and images to be recognized of image be it is more similar, then using at this time verifying birds bird kind as
The recognition result of birds target.
Verifying genius loci is made the bird kind of corresponding verifying birds as the recognition result of birds target by step 211.
Step 212, then to executing subsequent step after re-starting feature extraction to the region where birds target.
In the present embodiment, the overall space in the region where entire birds target is directly positioned in images to be recognized, so
Feature extraction is carried out simultaneously using multiple positions of the multiple location Feature Selection Model to birds target afterwards, final basis extracts
Feature identified after, obtain recognition result, use the implementation method that settles at one go, solution in positioning and feature extraction phases
It has determined the low technical problem of existing image-recognizing method recognition efficiency.
The above are a kind of embodiments two of birds image-recognizing method provided by the embodiments of the present application, and the following are the application realities
A kind of embodiment of birds pattern recognition device of example offer is provided.
Referring to Fig. 3, a kind of structural schematic diagram of the embodiment of birds pattern recognition device in the embodiment of the present application, packet
It includes:
Acquiring unit 301, for obtain include birds target images to be recognized；
Positioning unit 302 after carrying out local area localization to images to be recognized, is obtained for being based on preset location algorithm
Region where birds target；
Feature extraction unit 303, for being carried out to the region where birds target special according to multiple location Feature Selection Model
Sign is extracted, and multiple genius locis of birds target are obtained；
First recognition unit 304 is obtained for being identified using classifier and verifying genius loci to each genius loci
To the corresponding similarity score of each genius loci, verifying has one-to-one relationship between genius loci and genius loci；
Second recognition unit 305, for calculating the identification of birds target in images to be recognized according to all similarity scores
As a result.
In the present embodiment, the overall space in the region where entire birds target is directly positioned in images to be recognized, so
Feature extraction is carried out simultaneously using multiple positions of the multiple location Feature Selection Model to birds target afterwards, final basis extracts
Feature identified after, obtain recognition result, use the implementation method that settles at one go, solution in positioning and feature extraction phases
It has determined the low technical problem of existing image-recognizing method recognition efficiency.
The embodiment of the present application also provides a kind of birds image recognition apparatus, including processor and memory；Memory
Processor is transferred to for storing program code, and by program code；Processor is used for real according to the instruction execution of program code
Apply example one or any birds image-recognizing method of embodiment two.
The embodiment of the present application also provides a kind of storage medium, storage medium is used for storing program code, program code
In any birds image-recognizing method of execution embodiment one or embodiment two.
It is apparent to those skilled in the art that for convenience and simplicity of description, foregoing description wait pacify
Electricity grid network is filled, the specific work process of device and unit can refer to corresponding processes in the foregoing method embodiment, herein not
It repeats again.
In several embodiments provided herein, it should be understood that disclosed device and method can pass through it
Its mode is realized.For example, the apparatus embodiments described above are merely exemplary, for example, the division of the unit, only
Only a kind of logical function partition, there may be another division manner in actual implementation, such as multiple units or components can be tied
Another electricity grid network to be installed is closed or is desirably integrated into, or some features can be ignored or not executed.Another point is shown
The mutual coupling, direct-coupling or communication connection shown or discussed can be through some interfaces, between device or unit
Coupling or communication connection are connect, can be electrical property, mechanical or other forms.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple
In network unit.It can select some or all of unit therein according to the actual needs to realize the mesh of this embodiment scheme
's.
It, can also be in addition, each functional unit in each embodiment of the application can integrate in one processing unit
It is that each unit physically exists alone, can also be integrated in one unit with two or more units.Above-mentioned integrated list
Member both can take the form of hardware realization, can also realize in the form of software functional units.
If the integrated unit is realized in the form of SFU software functional unit and sells or use as independent product
When, it can store in a computer readable storage medium.Based on this understanding, the technical solution of the application is substantially
The all or part of the part that contributes to existing technology or the technical solution can be in the form of software products in other words
It embodies, which is stored in a storage medium, including some instructions are used so that a computer
Equipment (can be personal computer, server or the network equipment etc.) executes the complete of each embodiment the method for the application
Portion or part steps.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory (ROM, Read-
OnlyMemory), random access memory (RAM, RandomAccessMemory), magnetic or disk etc. are various can store
The medium of program code.
The above, above embodiments are only to illustrate the technical solution of the application, rather than its limitations；Although referring to before
Embodiment is stated the application is described in detail, those skilled in the art should understand that: it still can be to preceding
Technical solution documented by each embodiment is stated to modify or equivalent replacement of some of the technical features；And these
It modifies or replaces, the spirit and scope of each embodiment technical solution of the application that it does not separate the essence of the corresponding technical solution.
Claims (10)
1. a kind of birds image-recognizing method characterized by comprising
Acquisition includes the images to be recognized of birds target；
Based on preset location algorithm, after carrying out local area localization to the images to be recognized, the birds target place is obtained
Region；
According to multiple location Feature Selection Model, feature extraction is carried out to the region where the birds target, obtains the birds
Multiple genius locis of target；
Each genius loci is identified using classifier and verifying genius loci, it is corresponding to obtain each genius loci
Similarity score, between the verifying genius loci and the genius loci have one-to-one relationship；
The recognition result of birds target described in images to be recognized is calculated according to all similarity scores.
2. birds image-recognizing method according to claim 1, which is characterized in that it is described based on preset location algorithm, it is right
After the images to be recognized carries out local area localization, the region where obtaining the birds target is specifically included:
Classification activation mapping is carried out to the images to be recognized, obtains the corresponding thermodynamic chart of the images to be recognized；
The pixel average of the thermodynamic chart is calculated according to all pixels value in the thermodynamic chart；
The all pixels point in the thermodynamic chart is subjected to binary conversion treatment according to the pixel average, and is 1 by pixel value
The region that surrounds of pixel as the region where the birds target.
3. birds image-recognizing method according to claim 1, which is characterized in that the multiple location Feature Selection Model packet
Convolutional neural networks model is included,
It is then described according to multiple location Feature Selection Model, feature extraction is carried out to the region where the birds target, obtains institute
The multiple genius locis for stating birds target specifically include:
Convolution operation is carried out to the region where the birds target according to the preset convolutional neural networks model, and will be located at
After the convolution results of different convolutional layers are successively up-sampled, down-sampling, weighting are multiplied, corresponding sampled result is obtained, by one
A sampled result is as a genius loci.
4. birds image-recognizing method according to claim 1, which is characterized in that described to be commented according to all similarities
The recognition result for calculating birds target described in images to be recognized is divided to specifically include:
To all similarity score averageds, similarity score average value is obtained, judges that the identification score is average
Whether value is greater than preset threshold, if the verifying genius loci is then made the bird kind of corresponding verifying birds as the bird
Classification target recognition result, if it is not, then to the region where the birds target is re-started execute after feature extraction it is subsequent
Step.
5. birds image-recognizing method according to claim 1, which is characterized in that described to utilize classifier and verifying position
Feature identifies each genius loci, obtains the corresponding similarity score of each genius loci, the proof department
It is specifically included between position feature and the genius loci with one-to-one relationship:
It obtains for verifying the verifying genius loci for taking each genius loci；
According to the classifier, the corresponding genius loci is identified using the verifying genius loci, is corresponded to
Similarity score.
6. birds image-recognizing method according to claim 1, which is characterized in that it is described based on preset location algorithm, it is right
After the images to be recognized carries out local area localization, before obtaining the region where the birds target further include:
The images to be recognized is normalized.
7. birds image-recognizing method according to claim 6, which is characterized in that the normalized specifically includes:
According to prediction picture size, the images to be recognized is cut, so that the images to be recognized and the prediction picture
Deng big；
Scaled down is carried out to the corresponding pixel value of each pixel in the images to be recognized after cutting.
8. a kind of birds pattern recognition device characterized by comprising
Acquiring unit, for obtain include birds target images to be recognized；
Positioning unit after carrying out local area localization to the images to be recognized, obtains described for being based on preset location algorithm
Region where birds target；
Feature extraction unit, for carrying out feature to the region where the birds target according to multiple location Feature Selection Model
It extracts, obtains multiple genius locis of the birds target；
First recognition unit is obtained for being identified using classifier and verifying genius loci to each genius loci
The corresponding similarity score of each genius loci, has between the verifying genius loci and the genius loci and corresponds
Relationship；
Second recognition unit, for calculating the identification of birds target described in images to be recognized according to all similarity scores
As a result.
9. a kind of birds image recognition apparatus characterized by comprising processor and memory；
Said program code is transferred to the processor for storing program code by the memory；
The processor is schemed for the birds according to instruction execution any one of claims 1 to 7 of said program code
As recognition methods.
10. a kind of storage medium, which is characterized in that for storing program code, said program code is used for the storage medium
Birds image-recognizing method described in any one of perform claim requirement 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910775414.4A CN110472609A (en) | 2019-08-21 | 2019-08-21 | A kind of birds image-recognizing method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910775414.4A CN110472609A (en) | 2019-08-21 | 2019-08-21 | A kind of birds image-recognizing method, device, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110472609A true CN110472609A (en) | 2019-11-19 |
Family
ID=68513445
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910775414.4A Pending CN110472609A (en) | 2019-08-21 | 2019-08-21 | A kind of birds image-recognizing method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110472609A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110991502A (en) * | 2019-11-21 | 2020-04-10 | 北京航空航天大学 | Airspace security situation assessment method based on category activation mapping technology |
| CN111144378A (en) * | 2019-12-30 | 2020-05-12 | 众安在线财产保险股份有限公司 | Target object identification method and device |
| CN112036280A (en) * | 2020-08-24 | 2020-12-04 | 方海涛 | Waterfowl population dynamic monitoring method, device and equipment |
| CN112715427A (en) * | 2020-12-30 | 2021-04-30 | 青海保绿丰生态农林科技有限公司 | Automatic monitoring system for rules of hawk-leading nesting, number of spawning, brooding and predation habits |
| CN112926558A (en) * | 2021-05-12 | 2021-06-08 | 广州朗国电子科技有限公司 | Animal identification method and device |
| CN114063641A (en) * | 2021-10-19 | 2022-02-18 | 深圳市优必选科技股份有限公司 | Robot patrol method, patrol robot and computer readable storage medium |
| CN115690448A (en) * | 2022-11-09 | 2023-02-03 | 广东省科学院动物研究所 | AI-based bird species identification method and device |
| CN115761329A (en) * | 2022-11-14 | 2023-03-07 | 广东省科学院动物研究所 | Bird identification method and device |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108399406A (en) * | 2018-01-15 | 2018-08-14 | 中山大学 | The method and system of Weakly supervised conspicuousness object detection based on deep learning |
| CN109558810A (en) * | 2018-11-12 | 2019-04-02 | 北京工业大学 | Divided based on position and merges target person recognition methods |
-
2019
- 2019-08-21 CN CN201910775414.4A patent/CN110472609A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108399406A (en) * | 2018-01-15 | 2018-08-14 | 中山大学 | The method and system of Weakly supervised conspicuousness object detection based on deep learning |
| CN109558810A (en) * | 2018-11-12 | 2019-04-02 | 北京工业大学 | Divided based on position and merges target person recognition methods |
Non-Patent Citations (1)
| Title |
|---|
| BOLEI ZHOU,ET.AL: "Learning Deep Features for Discriminative Localization", 《IN PROCEEDINGS OF THE IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110991502A (en) * | 2019-11-21 | 2020-04-10 | 北京航空航天大学 | Airspace security situation assessment method based on category activation mapping technology |
| CN111144378A (en) * | 2019-12-30 | 2020-05-12 | 众安在线财产保险股份有限公司 | Target object identification method and device |
| CN111144378B (en) * | 2019-12-30 | 2023-10-31 | 众安在线财产保险股份有限公司 | Target object identification method and device |
| CN112036280A (en) * | 2020-08-24 | 2020-12-04 | 方海涛 | Waterfowl population dynamic monitoring method, device and equipment |
| CN112715427A (en) * | 2020-12-30 | 2021-04-30 | 青海保绿丰生态农林科技有限公司 | Automatic monitoring system for rules of hawk-leading nesting, number of spawning, brooding and predation habits |
| CN112926558A (en) * | 2021-05-12 | 2021-06-08 | 广州朗国电子科技有限公司 | Animal identification method and device |
| CN112926558B (en) * | 2021-05-12 | 2021-10-01 | 广州朗国电子科技股份有限公司 | Animal identification method and device |
| CN114063641A (en) * | 2021-10-19 | 2022-02-18 | 深圳市优必选科技股份有限公司 | Robot patrol method, patrol robot and computer readable storage medium |
| CN114063641B (en) * | 2021-10-19 | 2024-04-16 | 深圳市优必选科技股份有限公司 | Robot patrol method, patrol robot and computer readable storage medium |
| CN115690448A (en) * | 2022-11-09 | 2023-02-03 | 广东省科学院动物研究所 | AI-based bird species identification method and device |
| CN115761329A (en) * | 2022-11-14 | 2023-03-07 | 广东省科学院动物研究所 | Bird identification method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110472609A (en) | A kind of birds image-recognizing method, device, equipment and storage medium | |
| CN110188635B (en) | Plant disease and insect pest identification method based on attention mechanism and multi-level convolution characteristics | |
| EP3690721A1 (en) | Method for recognizing face using multiple patch combination based on deep neural network | |
| US8379994B2 (en) | Digital image analysis utilizing multiple human labels | |
| Gao et al. | Reading scene text with attention convolutional sequence modeling | |
| Wu et al. | Rapid target detection in high resolution remote sensing images using YOLO model | |
| CN103927387B (en) | Image indexing system and its correlation technique and device | |
| CN109034210A (en) | Object detection method based on super Fusion Features Yu multi-Scale Pyramid network | |
| Yan et al. | Weakly-and semi-supervised object detection with expectation-maximization algorithm | |
| Ling et al. | An accurate and real-time self-blast glass insulator location method based on faster R-CNN and U-net with aerial images | |
| CN110163069A (en) | Method for detecting lane lines for assisting driving | |
| Wang et al. | Study on the method of transmission line foreign body detection based on deep learning | |
| CN107239759A (en) | A kind of Hi-spatial resolution remote sensing image transfer learning method based on depth characteristic | |
| Fang et al. | Camouflage people detection via strong semantic dilation network | |
| Masita et al. | Pedestrian detection using R-CNN object detector | |
| CN116071701A (en) | YOLOv5 pedestrian detection method based on attention mechanism and GSConv | |
| US11507784B2 (en) | Device and method of digital image content recognition, training of the same | |
| Hussain et al. | A survey of traffic sign recognition systems based on convolutional neural networks | |
| Yu et al. | Pedestrian detection based on improved Faster RCNN algorithm | |
| JP7180590B2 (en) | Object Feature Amount Extraction Apparatus, Object Feature Amount Extraction Method, and Object Feature Amount Extraction Program | |
| Tianhang et al. | A new skeleton based flying bird detection method for low-altitude air traffic management | |
| CN117315499A (en) | Satellite remote sensing image target detection method and system | |
| Symeonidis et al. | AUTH-Persons: A dataset for detecting humans in crowds from aerial views | |
| Ciccone et al. | Methodology for image analysis in airborne search and rescue operations | |
| Cheng et al. | Iterative localization refinement in convolutional neural networks for improved object detection |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191119 |
|
| RJ01 | Rejection of invention patent application after publication |