CN113705443A - Palm print image identification method comprehensively utilizing knowledge graph and depth residual error network - Google Patents
Palm print image identification method comprehensively utilizing knowledge graph and depth residual error network Download PDFInfo
- Publication number
- CN113705443A CN113705443A CN202110995499.4A CN202110995499A CN113705443A CN 113705443 A CN113705443 A CN 113705443A CN 202110995499 A CN202110995499 A CN 202110995499A CN 113705443 A CN113705443 A CN 113705443A
- Authority
- CN
- China
- Prior art keywords
- palm print
- residual error
- error network
- data
- knowledge graph
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Abstract
The invention relates to the technical field of biological feature recognition, and discloses a palm print image recognition method comprehensively utilizing a knowledge graph and a depth residual error network, which comprises the following steps: the original image is randomly cropped into images of different sizes by using a random cropping method, so that the data set is more diverse, and then normalization is used for normalizing the data. The data are divided into two groups according to the information related to the gender by using a knowledge graph, different groups of data are trained by using a deep residual error network (ResNet) to obtain different groups of model parameters, and the model parameters are grouped and stored in a database according to the training. In the identification stage, the characteristics of the palm print to be detected are extracted, the constructed knowledge graph is utilized to inquire the entity data of the palm print and return corresponding gender information, the classification of the palm print is assisted, and then the palm print is classified into a proper category by utilizing a residual error network, so that the identification of the identity is realized.
Description
Technical Field
The invention relates to the technical field of biological feature recognition, in particular to a palm print image recognition method comprehensively utilizing a knowledge graph and a depth residual error network.
Background
In a highly information-based society, effective identity authentication is needed in various occasions such as financial payment, judicial evidence collection, ticket checking when entering a station and the like, and the development of a biological feature recognition technology provides a reliable solution for identity authentication. The biometric identification technology is widely researched and applied by utilizing inherent physiological characteristics of a human body, such as voice, human face, fingerprint, palm print, iris and the like, through a computer to identify personal identity, because the biometric identification technology does not need to be memorized, is difficult to copy, and has the advantages of high precision and high efficiency. The biological characteristics which are successfully applied at present mainly comprise fingerprints, irises, veins, human faces and DNA, and compared with other biological characteristic identification technologies, the palm print identification technology has a plurality of unique advantages: the accuracy is high, and the texture features with high identification degrees such as a main line, a fold, a spinal tip and a bifurcation point are provided; the palm print image is convenient to acquire, can be easily acquired on equipment with lower resolution, and is stable and reliable in palm print identification, and even if the palm print drops off due to special reasons, the new texture still keeps the original unchanged structure. Therefore, the palm print feature identification has the advantages of rich texture features, easy acceptance by users, high safety and stability and the like. The traditional palm print identification method cannot deal with the complicated application requirements, such as the palm print identification problem in a control environment or an open set. With the development of deep learning, the learning mode and the transfer learning based on the convolutional neural network can better solve the problems, various deep neural networks are widely used in the field of biological feature recognition, and the recognition accuracy is greatly improved. Liu attempts to extract deep palm print features using AlexNet. Sun et al use the CNN-F network for palm print recognition. Zhang utilizes the Incepotion-ResNet-v 1 network model to extract the palm print features. Still another scholars try to use the CNN to perform palm positioning and ROI extraction, and good effects are obtained. In recent years, a large amount of training data and a complex deep neural network are used for extracting the shallow and deep features of the palm print, but most algorithms neglect the guiding effect of prior information on palm print recognition, so that the invention provides a depth residual error network model combined with a knowledge graph for palm print recognition, and experiments show that the prior knowledge is introduced to guide the depth network model to obtain higher precision than that of the sole palm print recognition by using the depth network model. Therefore, the palm print image identification method comprehensively utilizing the knowledge graph and the depth residual error network is provided.
Disclosure of Invention
The invention aims to provide a palm print image identification method by comprehensively utilizing a knowledge graph and a depth residual error network, and solves the problems brought forward by the background.
In order to achieve the purpose, the invention provides the following technical scheme: the palm print image identification method comprehensively utilizing the knowledge graph and the depth residual error network comprises the following steps of:
s1: pretreatment:
randomly cutting an original image into images with different sizes by using a random cutting method, so that the data set has more diversity, and then normalizing the data by using normalization; after the data standardization of the original data, all indexes are in the same order of magnitude, and the method is suitable for comprehensive comparison and evaluation and comprises the following specific steps:
s2: construction:
constructing a palmprint knowledge graph, and introducing gender prior knowledge;
s3: training:
firstly, constructing a Resnet-18 residual error network, wherein the structure diagram of the Resnet-18 residual error network is shown in FIG. 2, after construction is completed, transmitting a palm print data sample into the network to extract characteristics, obtaining learning characteristics from a shallow layer L to a deep layer L, then obtaining the gradient of a reverse process by utilizing a chain rule so as to update network parameters, and finally training the palm print data by adopting the deep residual error network ResNet-18;
s4: identity recognition:
and (4) utilizing the trained ResNet-18 to fuse prior knowledge to recognize and classify the palm print images to be recognized, thereby realizing identity recognition.
Preferably, the random cropping method in step S1 is randomresize crop and centrercrop crop.
Preferably, x in step S1 is sample data, u is the mean of all sample data, σ is the standard deviation of all sample data, and x*The data after normalization processing.
Preferably, the step S2 is executed as follows: firstly, extracting palm print entities, extracting association relations among the entities from a palm print database, associating the entities through the relations to form a reticular knowledge structure, collecting gender attribute information of specific entities from different information sources, establishing semantic association, and finally constructing a knowledge graph by using a confusion matrix of CNN prediction results and combining information of nodes and edges.
Preferably, each category is a node of the image knowledge spectrum.
Preferably, the learning characteristic in the step S3 is
Preferably, the grid parameter in the step of S3 is
Preferably, the step S4 is executed as follows: in the identification stage, firstly, the knowledge spectrum is used for classifying the palm print images to be detected into a large group according to the prior knowledge of the sex associated information, then the palm print images are classified into smaller groups G step by step, then the palm prints to be identified are classified through the residual error network, and the knowledge spectrum is used for assisting in correcting the classification result.
The invention provides a palm print image identification method comprehensively utilizing a knowledge graph and a depth residual error network. The palm print image identification method comprehensively utilizing the knowledge graph and the depth residual error network has the following beneficial effects:
(1) the invention introduces knowledge graph concept, extracts the relation among the palm print entity, the sex factor of the palm print entity, the palm print entity and the sex attribute through the step of knowledge extraction, forms a high-quality palm print knowledge base and enriches the knowledge base through the step of knowledge fusion, divides the original sample into two groups by using the sex information in the training stage, subdivides the samples of different groups, respectively trains different model parameters, and optimizes the model according to the recognition condition of the palm print image for testing; in the authentication stage, knowledge reasoning is carried out on the palm print to be identified by using gender information, the palm print to be identified is divided into corresponding groups, and then the group of model parameters is started to send the palm print image to be detected into a deep residual error network for classification, so that identity authentication is realized;
(2) the method for identifying the palm print images comprehensively utilizing the knowledge graph and the depth residual error network can accurately group the palm prints to be identified, effectively reduce the required quantity of training samples, obtain higher identification accuracy and higher identification speed, and can meet the real-time requirement of large-scale palm print identification facing identity judgment;
(3) the palm print image recognition method comprehensively utilizing the knowledge graph and the depth Residual error network comprises the following steps of carrying out information association 2 steps by adding data grouping and introducing prior knowledge, dividing data into two groups in a registration stage according to palm print gender information, training different groups of data by utilizing the depth Residual error network (ResNet) to respectively obtain different groups of model parameters, and grouping and storing the model parameters into a database according to training; in the identification stage, the characteristics of the palmprint to be detected are extracted, the constructed knowledge graph is used for assisting in classifying the palmprint to be detected, and the palmprint is classified into a proper class to realize the identification of the identity.
Drawings
FIG. 1 is a schematic diagram of the general concept of the present invention;
FIG. 2 is a schematic diagram of a residual error network model according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments.
Examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are illustrative and intended to be illustrative of the invention and are not to be construed as limiting the invention.
As shown in fig. 1-2, the present invention provides a technical solution: the palm print image identification method comprehensively utilizing the knowledge graph and the depth residual error network comprises the following steps of:
s1: pretreatment:
random resize crop and centrercrop crop original images into images with different sizes, so that a data set has more diversity, then normalization is used for standardizing data, dimension influence among indexes can be eliminated by normalizing data, comparability among data indexes can be solved, after data standardization is carried out on the original data, all the indexes are in the same order of magnitude and are suitable for comprehensive comparison evaluation, and the method specifically comprises the following steps:
s2: construction:
constructing a palmprint knowledge graph, introducing gender prior knowledge, and performing the following steps: extracting association categories, wherein each category is a node of an image knowledge graph, establishing semantic association according to different conditions, and finally constructing the knowledge graph by using a confusion matrix of a CNN prediction result and combining the information of the nodes and edges;
s3: training:
firstly, constructing a Resnet-18 residual error network, wherein the structure diagram of the Resnet-18 residual error network is shown in figure 2, after construction is finished, a palm print data sample is sent into the network to extract characteristics, learning characteristics from a shallow layer L to a deep layer L are obtained, and the learning characteristics are
Then, by using the chain rule, the gradient of the reverse process can be obtained to update the network parameter, and the network parameter is
Finally, training the palm print data by adopting a deep residual error network ResNet-18;
s4: identity recognition:
the trained ResNet-18 is used for fusing prior knowledge to recognize and classify the palm print images to be recognized, so as to realize identity recognition, and the execution process is as follows: in the identification stage, the palm print images to be detected are firstly classified into a large group by using the prior knowledge of the knowledge graph, then are gradually classified into smaller groups G, then the palm prints to be identified are classified through a residual error network, and the knowledge graph is used for assisting in correcting the classification result.
The invention has the beneficial effects that: the invention achieved by the invention introduces a knowledge graph concept, extracts a palm print entity, sex elements of the palm print entity, and the relationship between the palm print entity and sex attributes through a knowledge extraction step, forms a high-quality palm print knowledge base and enriches the knowledge base through a knowledge fusion step, divides original samples into two groups by using sex information in a training stage, subdivides samples of different groups, respectively trains different model parameters, and optimizes a model according to a palm print image identification condition for testing; in the authentication stage, the gender information is used for carrying out knowledge reasoning on the palmprint to be identified, the palmprint to be identified is classified into corresponding groups, then the group of model parameters is started to send the palm print image to be detected into a depth residual error network for classification, thereby realizing the identity authentication, the invention can accurately group the palmprints to be recognized, effectively reduce the required quantity of training samples, obtain higher recognition accuracy, the identification speed is high, and the real-time requirement of large-scale palm print identification facing identity discrimination can be met, the invention carries out information association 2 steps by adding data grouping and introducing prior knowledge, the data is divided into two groups in the registration stage according to the palm print gender information, different groups of data are trained by utilizing a Deep Residual network (ResNet) to respectively obtain different groups of model parameters, and then the model parameters are grouped and stored in a database according to the training; in the identification stage, the characteristics of the palmprint to be detected are extracted, the constructed knowledge graph is used for assisting in classifying the palmprint to be detected, and the palmprint is classified into a proper class to realize the identification of the identity.
When the palm print image recognition method comprehensively utilizing the knowledge graph and the depth residual error network is used, firstly, data is enhanced, the original image is randomly cut by utilizing random Resizedcrop and centrercrop for enhancing the data, so that the data set has more diversity, after the data is enhanced, the data is normalized, the normalization of the data needs to be carried out, the normalization needs to be carried out, the size influence between indexes can be eliminated, meanwhile, the comparability between data indexes can be solved, secondly, the association category extraction needs to be carried out, the association category is extracted from a Convolutional Neural Network (CNN), then, the association is established according to different conditions, the confusion matrix of CNN prediction results is used, the knowledge graph is constructed by combining the information of nodes and edges, secondly, the Resnet-18 residual error network structure diagram is constructed, and the palm print data sample is sent to a network for extracting characteristics, and (3) obtaining learning characteristics from a shallow layer L to a deep layer L, obtaining gradient of a reverse process by using a chain rule so as to update grid parameters, grouping palm print images to be detected by using prior knowledge of a knowledge graph, classifying the palm prints to be identified by using a residual error network, and correcting a classification result by using the knowledge graph in an auxiliary manner.
In conclusion, the palm prints to be recognized can be accurately grouped, the required quantity of training samples is effectively reduced, the obtained recognition accuracy is high, the recognition speed is high, and the real-time requirement of large-scale palm print recognition facing identity judgment can be met.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising a reference structure" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (8)
1. The palm print image identification method comprehensively utilizing the knowledge graph and the depth residual error network is characterized by comprising the following steps of: the method comprises the following steps:
s1: pretreatment:
randomly cutting an original image into images with different sizes by using a random cutting method, so that the data set has more diversity, and then normalizing the data by using normalization; after the data standardization of the original data, all indexes are in the same order of magnitude, and the method is suitable for comprehensive comparison and evaluation and comprises the following specific steps:
s2: construction:
constructing a palmprint knowledge graph, and introducing gender prior knowledge;
s3: training:
firstly, constructing a Resnet-18 residual error network, wherein the structure diagram of the Resnet-18 residual error network is shown in FIG. 2, after construction is completed, transmitting a palm print data sample into the network to extract characteristics, obtaining learning characteristics from a shallow layer L to a deep layer L, then obtaining the gradient of a reverse process by utilizing a chain rule so as to update network parameters, and finally training the palm print data by adopting the deep residual error network ResNet-18;
s4: identity recognition:
and (4) utilizing the trained ResNet-18 to fuse prior knowledge to recognize and classify the palm print images to be recognized, thereby realizing identity recognition.
2. The palm print image recognition method comprehensively utilizing a knowledge graph and a depth residual error network according to claim 1, characterized in that: the random trimming method in the step S1 is randomresize crop and centrecrop.
3. The palm print image recognition method comprehensively utilizing a knowledge graph and a depth residual error network according to claim 1, characterized in that: in step S1, x is sample data, u is the mean of all sample data, σ is the standard deviation of all sample data, and x*The data after normalization processing.
4. The palm print image recognition method comprehensively utilizing a knowledge graph and a depth residual error network according to claim 1, characterized in that: the execution process of the step S2 is: firstly extracting the correlation category, then establishing semantic correlation according to different conditions, and finally constructing a knowledge spectrum by using a confusion matrix of a CNN prediction result and combining the information of nodes and edges.
5. The palm print image recognition method comprehensively utilizing a knowledge graph and a depth residual error network according to claim 4, characterized in that: each category is a node of the image knowledge spectrum.
6. The palm print image recognition method comprehensively utilizing a knowledge graph and a depth residual error network according to claim 1, characterized in that: the learning characteristic in the step of S3 is
7. The palm print image recognition method comprehensively utilizing a knowledge graph and a depth residual error network according to claim 1, characterized in that: the grid parameters in the step of S3 are
8. The palm print image recognition method comprehensively utilizing a knowledge graph and a depth residual error network according to claim 1, characterized in that: the execution process of the step S4 is: in the identification stage, firstly, the knowledge spectrum is used for classifying the palm print images to be detected into a large group according to the prior knowledge of the sex associated information, then the palm print images are classified into smaller groups G step by step, then the palm prints to be identified are classified through the residual error network, and the knowledge spectrum is used for assisting in correcting the classification result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110995499.4A CN113705443A (en) | 2021-08-27 | 2021-08-27 | Palm print image identification method comprehensively utilizing knowledge graph and depth residual error network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110995499.4A CN113705443A (en) | 2021-08-27 | 2021-08-27 | Palm print image identification method comprehensively utilizing knowledge graph and depth residual error network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113705443A true CN113705443A (en) | 2021-11-26 |
Family
ID=78655973
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110995499.4A Pending CN113705443A (en) | 2021-08-27 | 2021-08-27 | Palm print image identification method comprehensively utilizing knowledge graph and depth residual error network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113705443A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116248680A (en) * | 2023-05-11 | 2023-06-09 | 湖南工商大学 | De novo peptide sequencing method, de novo peptide sequencing device and related equipment |
-
2021
- 2021-08-27 CN CN202110995499.4A patent/CN113705443A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116248680A (en) * | 2023-05-11 | 2023-06-09 | 湖南工商大学 | De novo peptide sequencing method, de novo peptide sequencing device and related equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP2360619A1 (en) | Fast fingerprint searching method and fast fingerprint searching system | |
| Shehu et al. | Detailed identification of fingerprints using convolutional neural networks | |
| Qin et al. | Finger-vein quality assessment by representation learning from binary images | |
| Maltoni et al. | Automated fingerprint identification systems: from fingerprints to fingermarks | |
| CN109145704B (en) | Face portrait recognition method based on face attributes | |
| Arora et al. | Palmhashnet: Palmprint hashing network for indexing large databases to boost identification | |
| Saleem et al. | Hybrid Trainable System for Writer Identification of Arabic Handwriting. | |
| JP2014518579A (en) | Demographic analysis method and system based on multimodal information | |
| CN113705443A (en) | Palm print image identification method comprehensively utilizing knowledge graph and depth residual error network | |
| TWI325568B (en) | A method for face varification | |
| Yuan et al. | A real time fingerprint recognition system based on novel fingerprint matching strategy | |
| Lawal et al. | Face-based gender recognition analysis for Nigerians using CNN | |
| CN115795394A (en) | Biological feature fusion identity recognition method for hierarchical multi-modal and advanced incremental learning | |
| Guarnera et al. | Forensic analysis of handwritten documents with GRAPHJ | |
| Meuwly | Forensic Use of Fingerprints and Fingermarks. | |
| Aboalhsan et al. | Deep learning technique for fingerprint recognition | |
| Srivastava et al. | Three-layer multimodal biometric fusion using SIFT and SURF descriptors for improved accuracy of authentication of human identity | |
| Cenys et al. | Genetic algorithm based palm recognition method for biometric authentication systems | |
| Zhang | Children’s face recognition based on convolutional neural network | |
| Sun et al. | Texture-guided multiscale feature learning network for palmprint image quality assessment | |
| Kumari et al. | Factors affecting the accuracy of automatic signature verification | |
| Li et al. | Multimodal Biometric Fusion Algorithm Based on Ranking Partition Collision Theory | |
| Saraswal et al. | Analysis and Recognition of Handwriting Patterns for Personality Trait Prediction Using Unsupervised Machine Learning Approach | |
| Sabelli et al. | Predictive Modeling Toward the Design of a Forensic Decision Support System Using Cheiloscopy for Identification from Lip Prints | |
| Gheni et al. | Uyghur Language Off-line Signature Verification Based On The Modified Directional Features |
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 |