CN109271972A - Intelligent image identifying system and method based on natural language understanding and image graphics - Google Patents
Intelligent image identifying system and method based on natural language understanding and image graphics Download PDFInfo
- Publication number
- CN109271972A CN109271972A CN201811307019.5A CN201811307019A CN109271972A CN 109271972 A CN109271972 A CN 109271972A CN 201811307019 A CN201811307019 A CN 201811307019A CN 109271972 A CN109271972 A CN 109271972A
- Authority
- CN
- China
- Prior art keywords
- vector
- local
- image
- similarity
- feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
-
- 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
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
- G06V10/507—Summing image-intensity values; Histogram projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
Landscapes
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Oral & Maxillofacial Surgery (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Image Analysis (AREA)
Abstract
The intelligent image identifying system based on natural language understanding and image graphics that the invention discloses a kind of, change module, local feature vectors module including Fourier, after its particular persons Characteristic Contrast analysis with system typing, if similarity reaches the people's facial information that will capture high-resolution multi-angle after preset value, divide high-definition picture to N block again, then with Gabor kernel function and each piece of image convolution, obtain N number of feature, then these features are together in series, form local Gabor characteristic vector；LDA linear discriminant analysis module carries out dimension-reduction treatment to the N number of local feature vectors of an obtained global characteristics vector sum using LDA linear discriminant analysis, has obtained global classification device and local component's classifier.The present invention obtains high-definition picture again in the case where comparing success conditions and then identifies, realizes precise alignment.
Description
Technical field
The present invention relates to a kind of pattern recognition technique, it is based particularly on natural language understanding and facial image is core
Pattern recognition system and method.
Background technique
Image recognition technology is a key areas of artificial intelligence.It, which refers to, carries out Object identifying to image, with identification
The target of various different modes and technology to picture.
Image recognition technology may be based on the main feature of image.Each image has its feature, such as word
Female A has a point, and P has a circle and there is a acute angle etc. at the center of Y.Eye movement studies have shown that sight is always concentrated when to image recognition
In the main feature of image, that is, the place of image outline curvature maximum or contour direction sudden change is concentrated on, these
The information content in place is maximum.And the scanning route of eyes is also always successively gone in another feature from a feature.Thus
As it can be seen that perceptual mechanism must exclude the redundant information of input in image recognition processes, crucial information is extracted out.Meanwhile big
There must be the mechanism for being responsible for integrating information in brain, it can reflect the finish message obtained stage by stage at a complete consciousness
As.
The research of Digital Image Processing and identification starts from nineteen sixty-five.Digital picture has storage compared with analog image,
Transmission, which facilitates, to be not easy to be distorted in compressible, transmission process, handles and the huge advantages such as facilitate, these are all the hair of image recognition technology
Exhibition provides powerful power.
Object identification generally refers to belong to advanced computer to the perception and understanding of the object and environment of three-dimensional world
Vision scope.It is the research direction of the subjects such as combination artificial intelligence, systematics based on Digital Image Processing and identification,
Its research achievement is widely used on various industry and sniffing robot.
There are many top enterprises such as Google, Bing, Ebay, Amazon etc. marching this field in the world.
Google continuing investment has purchased tri- company of Plink, Pixazza, Like, and accumulating sum is more than 100,000,000 2 thousand ten thousand U.S. dollars；Apple
Fruit has then purchased face visual search Polar Rose；Microsoft also widelys popularize image similarity on Bing.And at the beginning of 11 months
The picture shopping search that group of Alibaba has invested view-based access control model, which is washed in a pan, to be searched.General enterprises also develop toward this side, such as south
Capital, which is gently searched, to be provided for client to scheme to search the service of figure
Present image recognition technology is mainly used in:
Remote sensing images identification: air remote sensing and satellite remote sensing images are usually processed with image recognition technology to extract
Useful information.The technology is currently used primarily in topographic and geologic and detects, the resource investigations such as forest, water conservancy, ocean, agricultural, disaster
Prediction, environmental pollution monitoring, meteorological satellite cloud pictures processing and the identification of ground military target etc..
The fields such as military affairs, police criminal detection: image recognition technology is very extensive in military, in terms of police criminal detection application, such as
Scouting, guidance and the warning system of military target；The control of automatic fire extinguisher and counter camouflage；The scene photograph of public security department refers to
The processing and identification of line, original handwriting, seal, portrait etc.；Reparation and management of history text and photo archive etc..
Biomedical Image identification: image recognition is very widely used in modern medicine, it has intuitive, noninvasive
The features such as wound, safe ready.Extensively by image recognition technology, such as CT (Computed in clinical diagnosis and pathological study
Tomography) technology etc..
Field of machine vision: as the important sense organ of intelligent robot, machine vision is substantially carried out the reason of 3D rendering
Solution and identification, which is also one of the heat subject studied at present.The application field of machine vision is also very extensive, such as with
In military surveillance, the autonomous robot of hazardous environment, postal, hospital and home services intelligent robot.Furthermore machine vision
It can also be used in the workpiece identification and positioning in industrial production, space robotics' is automatically brought into operation.
Communication field: including image transmitting, video telephone, video conference etc. in communication applications.
Summary of the invention
1, goal of the invention.
The invention proposes a kind of based on the intelligent image identifying system that natural language understanding and image graphics are core and
Method solves the safety measure for doing elevator passenger by the threat of non-equipment fault.Intelligent recognition algorithm has been used to identify
Histogram feature, color characteristic, template characteristic and structure feature of people etc. realize acquisition, judge face, limb action and special
The function of article.
2, the technical solution adopted in the present invention.
The intelligent image identifying system based on natural language understanding and image graphics that the invention proposes a kind of, including in Fu
Leaf changes module, does Fourier's variation to it first, retains the coefficient of the real and imaginary parts of low frequency part, as global Fourier
Feature vector；
Local feature vectors module, after its particular persons Characteristic Contrast analysis with system typing, if similarity reaches
The people's facial information of high-resolution multi-angle will be captured after preset value, then divides high-definition picture to N block, then uses Gabor
Kernel function and each piece of image convolution, obtain N number of feature, then these features are together in series, formed local Gabor characteristic to
Amount；
LDA linear discriminant analysis module, it is N number of to an obtained global characteristics vector sum using LDA linear discriminant analysis
Local feature vectors carry out dimension-reduction treatment, have obtained global classification device and local component's classifier；By N number of partial vector classifier
It is weighted summation, obtains local classifiers；Global classification device is also weighted summation with local classifiers, and concurrent integration obtains
To global classification device；
Module is normalized, the two images for treating comparison do the same above processing, then compare their corresponding N respectively
The pixel degree of+1 vector, common normalized crosscorrelation method calculate the similarity of character pair vector.
Further, according to integrated study theory, the similarity that the above global classification device and local classifiers are exported into
Row weighted sum, as final similarity.
Further, if similarity will capture the people's facial information of high-resolution multi-angle after reaching 60%.
The intelligent image recognition methods based on natural language understanding and image graphics that the invention proposes a kind of, including in Fu
Leaf conversion step does Fourier's variation to it first, retains the coefficient of the real and imaginary parts of low frequency part, as global Fourier
Feature vector；
Local feature vectors step, after its particular persons Characteristic Contrast analysis with system typing, if similarity reaches
The people's facial information of high-resolution multi-angle will be captured after 60%, then divides high-definition picture to N block, then uses Gabor core
Function and each piece of image convolution, obtain N number of feature, then these features are together in series, formed local Gabor characteristic to
Amount；
LDA linear discriminant analysis step, it is N number of to an obtained global characteristics vector sum using LDA linear discriminant analysis
Local feature vectors carry out dimension-reduction treatment, have obtained global classification device and local component's classifier；By N number of partial vector classifier
It is weighted summation, obtains local classifiers；Global classification device is also weighted summation with local classifiers, and concurrent integration obtains
To global classification device；
Normalization step, the two images for treating comparison do the same above processing, then compare their corresponding N respectively
The pixel degree of+1 vector, common normalized crosscorrelation method calculate the similarity of character pair vector.
Further, according to integrated study theory, the similarity that the above global classification device and local classifiers are exported into
Row weighted sum, as final similarity.
Further, if similarity will capture the people's facial information of high-resolution multi-angle after reaching 60%.
3, technical effect caused by the present invention.
(1) present invention makes two bites at a cherry identification process, reduces technology consumption.
Low-resolution image identification, whole comparison convenient to carry out are directed to using global characteristics algorithm；Comparing successfully item
High-definition picture is obtained under part again and then is identified, realizes precise alignment.
(2) two algorithm of the present invention combines
Face recognition and bio-identification are combined, the function for comparing personal information and identifying special behavior (article) is completed
Can, create a kind of more convenient and smart safety method realization elevator safety.
Detailed description of the invention
Fig. 1 is flow chart of the present invention.
Fig. 2 is network topological diagram.
Specific embodiment
Embodiment
As shown in figure Fig. 1-2, the present invention is realized using face recognition comparison algorithm and limbs recognizer substantially at present.
On the one hand face recognition algorithm is used, the present invention is mainly used to be known based on the face of eigenface method and global characteristics
Not, after image-capturing apparatus has acquired low-resolution image or video flowing, Fourier's variation is done to it first, retains low frequency portion
The coefficient for the real and imaginary parts divided, as global Fourier's feature vector.By the particular persons Characteristic Contrast of itself and system typing
After analysis, if similarity reach 60% (or more) after will capture the people's facial information of high-resolution multi-angle, then by high score
Resolution image divides N block, then with Gabor kernel function and each piece of image convolution, N number of feature is obtained, then by these feature strings
Connection gets up, and forms local Gabor characteristic vector.Then, using LDA linear discriminant analysis to an obtained global characteristics vector
Dimension-reduction treatment is carried out with N number of local feature vectors, has obtained global classification device and local component's classifier.By N number of partial vector
Classifier is weighted summation, obtains local classifiers.Global classification device is also weighted summation with local classifiers, parallel to collect
At obtaining global classification device.The two images for treating comparison do the same above processing, then compare their corresponding N+ respectively
The pixel degree of 1 vector, common normalized crosscorrelation (normalized cross-correlation, abbreviation NCC) method
To calculate the similarity of character pair vector.According to integrated study theory, the above global classification device and local classifiers are exported
Similarity be weighted summation, as final similarity.
On the other hand biological limbs recognizer can be multi-thread using large size multiple views system Panoptic Studio in equipment
Journey carries out the limbs identification depth information calculating based on chessboard tree after reading each position of limbs, carries out data by Kinect
The set of a coordinate comprising movement point of interest can be generated in the method for acquisition, thus instead of traditional matching template.And
In learning process, constantly there are new branch and node to generate, so as to identify to more movements.Select artis
Position is simultaneously compared and analyzed with existing node form in visual database, the highest form of similarity is found out, to identify limb
Whether body movement is dangerous.The offensive weapons such as controlled knife, firearms can be precisely identified on this basis.
By two algorithms in conjunction with come detection and tracking face, identification human action whether there is risk " to make electricity to reach
The basic goal of the terraced passenger not threat by non-equipment fault ".
Claims (6)
1. a kind of intelligent image identifying system based on natural language understanding and image graphics, it is characterised in that: including Fourier
Change module, does Fourier's variation to it first, retain the coefficient of the real and imaginary parts of low frequency part, it is special as global Fourier
Levy vector；
Local feature vectors module, after its particular persons Characteristic Contrast analysis with system typing, if similarity reaches default
The people's facial information of high-resolution multi-angle will be captured after value, then divides high-definition picture to N block, then uses Gabor core letter
Several and each piece of image convolution, obtains N number of feature, then these features is together in series, and forms local Gabor characteristic vector；
LDA linear discriminant analysis module, using LDA linear discriminant analysis to an obtained N number of part of global characteristics vector sum
Feature vector carries out dimension-reduction treatment, has obtained global classification device and local component's classifier；N number of partial vector classifier is carried out
Weighted sum obtains local classifiers；Global classification device is also weighted summation with local classifiers, and concurrent integration obtains complete
Office's classifier；
Module is normalized, the two images for treating comparison do the same above processing, then compare them respectively corresponding N+1
The pixel degree of vector, common normalized crosscorrelation method calculate the similarity of character pair vector.
2. the intelligent image identifying system according to claim 1 based on natural language understanding and image graphics, feature
It is: according to integrated study theory, the similarity that the above global classification device and local classifiers export is weighted summation, is made
For final similarity.
3. the intelligent image identifying system according to claim 1 based on natural language understanding and image graphics, feature
It is: if similarity will capture the people's facial information of high-resolution multi-angle after reaching 60%.
4. a kind of intelligent image recognition methods based on natural language understanding and image graphics, it is characterised in that: including Fourier
Conversion step does Fourier's variation to it first, retains the coefficient of the real and imaginary parts of low frequency part, special as global Fourier
Levy vector；
Local feature vectors step, after its particular persons Characteristic Contrast analysis with system typing, if similarity reaches 60%
The people's facial information of high-resolution multi-angle will be captured afterwards, then divides high-definition picture to N block, then uses Gabor kernel function
With each piece of image convolution, N number of feature is obtained, then these features are together in series, forms local Gabor characteristic vector；
LDA linear discriminant analysis step, using LDA linear discriminant analysis to an obtained N number of part of global characteristics vector sum
Feature vector carries out dimension-reduction treatment, has obtained global classification device and local component's classifier；N number of partial vector classifier is carried out
Weighted sum obtains local classifiers；Global classification device is also weighted summation with local classifiers, and concurrent integration obtains complete
Office's classifier；
Normalization step, the two images for treating comparison do the same above processing, then compare them respectively corresponding N+1
The pixel degree of vector, common normalized crosscorrelation method calculate the similarity of character pair vector.
5. the intelligent image recognition methods according to claim 4 based on natural language understanding and image graphics, feature
It is: according to integrated study theory, the similarity that the above global classification device and local classifiers export is weighted summation, is made
For final similarity.
6. the intelligent image recognition methods according to claim 4 based on natural language understanding and image graphics, feature
It is: if similarity will capture the people's facial information of high-resolution multi-angle after reaching 60%.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811307019.5A CN109271972A (en) | 2018-11-05 | 2018-11-05 | Intelligent image identifying system and method based on natural language understanding and image graphics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811307019.5A CN109271972A (en) | 2018-11-05 | 2018-11-05 | Intelligent image identifying system and method based on natural language understanding and image graphics |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109271972A true CN109271972A (en) | 2019-01-25 |
Family
ID=65191996
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811307019.5A Pending CN109271972A (en) | 2018-11-05 | 2018-11-05 | Intelligent image identifying system and method based on natural language understanding and image graphics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109271972A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110992257A (en) * | 2019-12-20 | 2020-04-10 | 北京航天泰坦科技股份有限公司 | Remote sensing image sensitive information automatic shielding method and device based on deep learning |
| WO2022087778A1 (en) * | 2020-10-26 | 2022-05-05 | 深圳大学 | Low-resolution image recognition method based on multi-layer coupled mapping |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105117688A (en) * | 2015-07-29 | 2015-12-02 | 重庆电子工程职业学院 | Face identification method based on texture feature fusion and SVM |
-
2018
- 2018-11-05 CN CN201811307019.5A patent/CN109271972A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105117688A (en) * | 2015-07-29 | 2015-12-02 | 重庆电子工程职业学院 | Face identification method based on texture feature fusion and SVM |
Non-Patent Citations (1)
| Title |
|---|
| MISSLIBRA网友: "基于全局和局部特征的人脸识别", 《CSDN博客》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110992257A (en) * | 2019-12-20 | 2020-04-10 | 北京航天泰坦科技股份有限公司 | Remote sensing image sensitive information automatic shielding method and device based on deep learning |
| WO2022087778A1 (en) * | 2020-10-26 | 2022-05-05 | 深圳大学 | Low-resolution image recognition method based on multi-layer coupled mapping |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Panetta et al. | A comprehensive database for benchmarking imaging systems | |
| Ou et al. | Automatic facial expression recognition using Gabor filter and expression analysis | |
| CN112766160A (en) | Face replacement method based on multi-stage attribute encoder and attention mechanism | |
| CN107862300A (en) | A kind of descending humanized recognition methods of monitoring scene based on convolutional neural networks | |
| GB2608975A (en) | Person identification across multiple captured images | |
| CN110462633A (en) | A kind of method, apparatus and electronic equipment of recognition of face | |
| CN109522853A (en) | Face datection and searching method towards monitor video | |
| CN112541433B (en) | Two-stage human eye pupil accurate positioning method based on attention mechanism | |
| CN106886745A (en) | A kind of unmanned plane reconnaissance method based on the generation of real-time online map | |
| Liu et al. | Gesture Recognition for UAV-based Rescue Operation based on Deep Learning. | |
| CN109271972A (en) | Intelligent image identifying system and method based on natural language understanding and image graphics | |
| Miao et al. | Abnormal behavior learning based on edge computing toward a crowd monitoring system | |
| CN114036553A (en) | K-anonymity-combined pedestrian identity privacy protection method | |
| Ikram et al. | A review of the prevalent ICT techniques used for COVID-19 SOP violation detection | |
| CN111325252A (en) | Image processing method, apparatus, device, and medium | |
| CN110135362A (en) | A kind of fast face recognition method based under infrared camera | |
| CN115862128A (en) | Human body skeleton-based customer abnormal behavior identification method | |
| Přihodová et al. | Gender recognition using thermal images from UAV | |
| CN103955708B (en) | Face photo library fast-reduction method for face synthesis portrait recognition | |
| Vázquez et al. | Real time face identification using a neural network approach | |
| Lin et al. | Face detection algorithm based on multi-orientation gabor filters and feature fusion | |
| Nan et al. | Safety Helmet Detection Dynamic Model Based on the Critical Area Attention Mechanism | |
| Alemran et al. | Novel hybrid ear recognition framework in passive human identification | |
| Mahmoud et al. | A Framework for Mask-Wearing Recognition in Complex Scenes for Different Face Sizes. | |
| Liu et al. | Research on Extending Person Re-identification Datasets Based on Generative Adversarial Network |
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: 20190125 |
|
| RJ01 | Rejection of invention patent application after publication |