CN110334667B - Vein recognition method and system with scale rotation invariance based on IRCNN and MTCNN - Google Patents

Vein recognition method and system with scale rotation invariance based on IRCNN and MTCNN Download PDF

Info

Publication number
CN110334667B
CN110334667B CN201910620017.XA CN201910620017A CN110334667B CN 110334667 B CN110334667 B CN 110334667B CN 201910620017 A CN201910620017 A CN 201910620017A CN 110334667 B CN110334667 B CN 110334667B
Authority
CN
China
Prior art keywords
preset
size
feature
mtcnn
multiplied
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.)
Active
Application number
CN201910620017.XA
Other languages
Chinese (zh)
Other versions
CN110334667A (en
Inventor
胡可
黄国恒
周艳妹
彭涛
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangdong University of Technology
Original Assignee
Guangdong University of Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Guangdong University of Technology filed Critical Guangdong University of Technology
Priority to CN201910620017.XA priority Critical patent/CN110334667B/en
Publication of CN110334667A publication Critical patent/CN110334667A/en
Application granted granted Critical
Publication of CN110334667B publication Critical patent/CN110334667B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • 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
    • G06T5/73—
    • 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
    • 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
    • G06V40/161—Detection; Localisation; Normalisation
    • G06V40/165—Detection; Localisation; Normalisation using facial parts and geometric relationships
    • 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
    • G06V40/14—Vascular patterns
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Abstract

The vein identification method with scale rotation invariance based on IRCNN and MTCNN comprises the following steps: inputting the target vein image into a preset IRCNN model for sharpening treatment to obtain a sharpened image; inputting the sharpened image into a preset MTCNN model, and outputting a feature map; the preset MTCNN model is a model which is constructed based on preset parameters; and comparing the relative positions of the feature points in the feature map relative to the boxes of the feature map with preset relative positions so as to realize identification. The method comprises the steps of carrying out image enhancement by using an IRCNN model, making features obvious, and extracting the features by using an MTCNN model to obtain a feature map. Because the MTCNN model has rotation invariance and scale invariance, the extracted features are more accurate, the obtained feature map is more accurate, and the accuracy of identity recognition can be further improved. The application also provides a vein recognition system, a vein recognition device and a vein recognition computer readable storage medium with scale rotation invariance based on IRCNN and MTCNN, which have the beneficial effects.

Description

Vein recognition method and system with scale rotation invariance based on IRCNN and MTCNN
Technical Field
The present application relates to the field of computer vision, and in particular, to a vein recognition method, system, device, and computer readable storage medium with scale rotation invariance based on IRCNN and MTCNN.
Background
At present, when vein images are used for identification, vein direction characteristics are extracted through a direction filter, namely, the images built into a coordinate system are subjected to convolution operation by adopting Radon transformation, so that vein image characteristics are extracted. However, the method is not accurate enough to extract features due to the lack of rotation invariance and scale invariance, thereby leading to inaccurate identification.
Therefore, how to improve the accuracy of identification is a technical problem that needs to be solved by those skilled in the art.
Disclosure of Invention
The purpose of the application is to provide a vein recognition method, a vein recognition system, vein recognition equipment and a vein recognition computer readable storage medium with scale rotation invariance based on IRCNN and MTCNN, which can improve the accuracy of identity recognition.
In order to solve the technical problem, the application provides a vein recognition method with scale rotation invariance based on IRCNN and MTCNN, which comprises the following steps:
inputting the target vein image into a preset IRCNN model for sharpening treatment to obtain a sharpened image;
inputting the sharpened image into a preset MTCNN model, and outputting a feature map; the preset MTCNN model is a model which is constructed based on preset parameters;
and comparing the relative positions of the feature points in the feature map relative to the square frame of the feature map with preset relative positions so as to realize identification.
Preferably, the inputting the sharpened image into a preset MTCNN model, outputting a feature map, includes:
respectively constructing P-Net, R-Net and O-Net according to a preset parameter list to obtain the preset MTCNN model;
and inputting the sharpened image into the preset MTCNN model, and outputting the feature map.
Preferably, the inputting the sharpened image into the preset MTCNN model, outputting the feature map, includes:
inputting the sharpened image into the preset MTCNN model, and sequentially carrying out convolution, pooling and full connection processing to output the characteristic map.
Preferably, the comparing the relative positions of the feature points in the feature map with respect to the square frame of the feature map with a preset relative position to implement identification includes:
storing the relative positions into a matrix to obtain a target matrix;
comparing the target matrix with a preset target matrix corresponding to the preset relative position in a database to obtain a comparison result;
judging whether the comparison result meets a preset threshold value or not; if yes, the identity identification is successful; if not, the identity identification fails.
The present application also provides a vein recognition system with scale rotation invariance based on IRCNN and MTCNN, comprising:
the sharpening image acquisition module is used for inputting the target vein image into a preset IRCNN model for sharpening treatment to obtain a sharpening image;
the feature map acquisition module is used for inputting the sharpened image into a preset MTCNN model and outputting a feature map; the preset MTCNN model is a model which is constructed based on preset parameters;
and the position comparison module is used for comparing the relative positions of the feature points in the feature map relative to the square frame of the feature map with the preset relative positions so as to realize identification.
Preferably, the feature map acquisition module includes:
an MTCNN model construction unit, configured to construct P-Net, R-Net and O-Net according to a preset parameter list, respectively, to obtain the preset MTCNN model;
and the characteristic diagram acquisition unit is used for inputting the sharpened image into the preset MTCNN model and outputting the characteristic diagram.
Preferably, the feature map acquisition unit includes:
and the characteristic diagram acquisition subunit is used for inputting the sharpened image into the preset MTCNN model, sequentially carrying out convolution, pooling and full connection processing, and outputting the characteristic diagram.
Preferably, the position comparison module includes:
the target matrix acquisition unit is used for storing the relative positions into a matrix to obtain a target matrix;
the comparison result acquisition unit is used for comparing the target matrix with a preset target matrix corresponding to the preset relative position in the database to obtain a comparison result;
the comparison result judging unit is used for judging whether the comparison result meets a preset threshold value or not; if yes, the identity identification is successful; if not, the identity identification fails.
The present application also provides an apparatus comprising:
a memory and a processor; the memory is used for storing a computer program, and the processor is used for realizing the steps of the vein recognition method with scale rotation invariance based on IRCNN and MTCNN when the computer program is executed.
The present application also provides a computer readable storage medium storing a computer program which, when executed by a processor, implements the steps of the IRCNN and MTCNN-based vein recognition method with scale rotation invariance described above.
The vein identification method with scale rotation invariance based on IRCNN and MTCNN comprises the following steps: inputting the target vein image into a preset IRCNN model for sharpening treatment to obtain a sharpened image; inputting the sharpened image into a preset MTCNN model, and outputting a feature map; the preset MTCNN model is a model which is constructed based on preset parameters; and comparing the relative positions of the feature points in the feature map relative to the square frame of the feature map with preset relative positions so as to realize identification.
The method comprises the steps of carrying out image enhancement by using an IRCNN model, making features obvious, and extracting the features by using an MTCNN model to obtain a feature map. Because the MTCNN model has rotation invariance and scale invariance, the extracted features are more accurate, the obtained feature map is more accurate, and the accuracy of identity recognition can be further improved. The application further provides a vein recognition system, a vein recognition device and a vein recognition computer readable storage medium with scale rotation invariance based on IRCNN and MTCNN, which have the beneficial effects and are not repeated here.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required to be used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only embodiments of the present application, and that other drawings may be obtained according to the provided drawings without inventive effort to a person skilled in the art.
Fig. 1 is a flowchart of a vein recognition method with scale rotation invariance based on IRCNN and MTCNN provided in an embodiment of the present application;
fig. 2 is a structural diagram of an IRCNN model provided in an embodiment of the present application;
fig. 3 is a structural diagram of an MTCNN model provided in an embodiment of the present application;
fig. 4 is a block diagram of a vein recognition system with scale rotation invariance based on IRCNN and MTCNN according to an embodiment of the present application.
Detailed Description
The core of the application is to provide a vein recognition method with scale rotation invariance based on IRCNN and MTCNN, which can improve the accuracy of identity recognition. Another core of the present application is to provide a vein recognition system, apparatus, and computer readable storage medium with scale rotation invariance based on IRCNN and MTCNN.
For the purposes of making the objects, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions of the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure.
At present, when vein images are used for identification, vein direction characteristics are extracted through a direction filter, namely, the images built into a coordinate system are subjected to convolution operation by adopting Radon transformation, so that vein image characteristics are extracted. However, the method is not accurate enough to extract features due to the lack of rotation invariance and scale invariance, thereby leading to inaccurate identification. Referring to fig. 1 specifically, fig. 1 is a flowchart of a vein recognition method with scale rotation invariance based on IRCNN and MTCNN, which is provided in an embodiment of the present application, and the vein recognition method with scale rotation invariance based on IRCNN and MTCNN specifically includes:
s101, inputting a target vein image into a preset IRCNN model for sharpening, and obtaining a sharpened image;
the target vein image in this embodiment is not particularly limited, and a vein image is generally obtained by photographing a vein on a finger and is used as the target vein image. The target vein image is input into an IRCNN model (CNN noise reducer), the structure of the IRCNN model is shown in fig. 2, and fig. 2 is a structural diagram of the IRCNN model provided by the embodiment of the application. As can be seen from fig. 2, the IRCNN model consists of seven layers, including three different modules, namely an "expanded convolution+relu (activation function)" module in the first layer, five "expanded convolution+bnorm (batch normalization) +relu" middle layer modules, and an "expanded convolution" module in the last layer. The expansion factors of the (3×3) expansion convolution from the first layer to the last layer are set to 1,2,3,4,3,2,1, that is, expansion convolution 1, expansion convolution 2, expansion convolution 3, expansion convolution 4, expansion convolution 3, expansion convolution 2, expansion convolution 1 in fig. 2, respectively.
S102, inputting a sharpened image into a preset MTCNN model, and outputting a feature map; the preset MTCNN model is a model which is constructed based on preset parameters;
further, inputting the sharpened image into a preset MTCNN model, and outputting a feature map generally includes: respectively constructing P-Net, R-Net and O-Net according to a preset parameter list to obtain a preset MTCNN model; and inputting the sharpened image into a preset MTCNN model, and outputting a feature map. Further, inputting the sharpened image into a preset MTCNN model, and outputting a feature map generally includes: inputting the sharpened image into a preset MTCNN model, and sequentially carrying out convolution, pooling and full connection processing to output a feature map. The description of the above embodiments is as follows:
the original MTCNN model (multi-task cascade convolution network) is a multi-task face detection framework, and uses 3 CNN cascade algorithm structures, wherein the CNN networks are P-Net, R-Net and O-Net. And simultaneously carrying out face detection and face feature point detection, and inputting a face picture to obtain a picture with a face boundary box and five feature points. The characteristic points marked by the original MTCNN model are positioned at the left eye, the right eye, the nose tip and the left mouth corner. The embodiment structurally builds the MTCNN model based on preset parameters, is structurally improved compared with the original MTCNN model, and is used for detecting vein images.
The MTCNN model used in this embodiment is shown in fig. 3, fig. 3 is a structural diagram of an MTCNN model provided in this embodiment, and as can be seen from fig. 3, the MTCNN model is composed of P-Net, R-Net, and O-Net, and the P-Net, R-Net, and O-Net respectively correspond to the dashed boxes in fig. 3. As can be seen from fig. 3, when the sharpened image is input to the MTCNN model, the bounding box is resized to a 12×12-sized picture, and converted to a 12×12×3 structure, thereby generating training data of the P-Net network. The training data is passed through 103 x 3 convolution kernels, 2 x 2 max pooling operation, 10 feature maps of 5 x 5 are generated. Then, 16 3×3 feature maps are generated by 16 3×3×10 convolution kernels. Next, 32 1×1 feature maps are generated by 32 convolution kernels of 3×3×16. Finally, for 32 feature maps of 1×1, 32 feature maps of 1×1 can be generated for classification by using 32 convolution kernels of 1×1×32, namely, the vein recognition result in fig. 3; 32 convolution kernels of 1×1×32, and generating 32 feature maps of 1×1 for regression frame judgment, namely, the bounding box regression in fig. 3; 32 1×1×32 convolution kernels, 32 1×1 feature maps are generated for determination of vein feature points, i.e., vein feature point positions in fig. 3. For R-Net, inputting a picture with the size of 24 multiplied by 24, generating 28 characteristic diagrams with 11 multiplied by 11 after the convolution kernels with the size of 28 3 multiplied by 3 and the maximum pooling with the size of 3 multiplied by 3; generating 48 4 x 4 feature maps through 48 convolution kernels of 3 x 28 and 3 x 3 maximum pooling; after passing through 64 convolution kernels of 2×2×48, 64 feature maps of 3×3 are generated; passing the 3×3×64 feature map through a 4096-sized full link layer; and passing through a full connecting layer with the size of 4096; and then through 128 full link layers. The parameters of the last two full-connection layers are trained by a larger learning rate, so that the model can jump out of local optimum and quickly converge. Finally, the background problem is converted into a full connection layer with the size of 1; converting the regression frame classification problem into a full connection layer with the size of 2, namely a vein recognition result in the figure 3; the position regression problem of the bounding box is converted into a full connection layer with the size of 4, namely the bounding box regression in fig. 3; the contour keypoints are transformed into fully connected layers of size 110, i.e. the venous feature point locations in fig. 3. For O-Net, the input is a picture of size 48 x 3, by 32 convolution kernels of 3 x 3 and 3 x 3 the most converting the feature images into 32 feature images of 23×23 after pooling; the feature map is converted into 64 10 multiplied by 10 after being subjected to the maximum pooling of 64 convolution kernels of 3 multiplied by 32 and 3 multiplied by 3; the feature map is converted into 64 4 multiplied by 4 after being subjected to the maximum pooling of 64 convolution kernels of 3 multiplied by 64; converting into 128 characteristic diagrams of 3×3 through 128 convolution kernels of 2×2×64; passing it through a 4096 sized fully-connected layer; and passing through a full connecting layer with the size of 4096; then passing through 257 full link layers; finally, generating a background with the size of 1; regression frame classification characteristics with the size of 2, namely vein recognition results in fig. 3; regression features of regression box positions of size 4, i.e., bounding box regression in FIG. 3; the vein feature point location regression feature, i.e., the vein feature point location in fig. 3, is of size 110.
S103, comparing the relative positions of the feature points in the feature map relative to the boxes of the feature map with preset relative positions so as to realize identification.
Further, the comparing the relative positions of the feature points in the feature map with the preset relative positions to realize the identification generally includes: storing the relative positions into a matrix to obtain a target matrix; comparing the target matrix with a preset target matrix corresponding to a preset relative position in a database to obtain a comparison result; judging whether the comparison result meets a preset threshold value or not; if yes, the identity identification is successful; if not, the identity identification fails. The number of feature points and the preset threshold are not particularly limited, and should be set by those skilled in the art according to actual situations. From the results of the experimental stage, the number of feature points may be 55.
The method provided by the application firstly uses the IRCNN model to enhance the image, the characteristics become obvious, and then uses the MTCNN model to extract the characteristics, so as to obtain the characteristic diagram. Because the MTCNN model has rotation invariance and scale invariance, the extracted features are more accurate, the obtained feature map is more accurate, and the accuracy of identity recognition can be further improved. In addition, the method is faster than prior art identification.
The following describes a vein recognition system, a device and a computer readable storage medium with scale rotation invariance based on IRCNN and MTCNN, where the vein recognition system, the device and the computer readable storage medium with scale rotation invariance based on IRCNN and MTCNN described below and the vein recognition method with scale rotation invariance based on IRCNN and MTCNN described above may be referred to correspondingly.
Referring to fig. 4, fig. 4 is a block diagram of a vein recognition system with scale rotation invariance based on IRCNN and MTCNN according to an embodiment of the present application; the vein recognition system with scale rotation invariance based on IRCNN and MTCNN comprises:
the sharpened image obtaining module 401 is configured to input a target vein image into a preset IRCNN model for sharpening, so as to obtain a sharpened image;
the feature map obtaining module 402 is configured to input the sharpened image into a preset MTCNN model, and output a feature map; the preset MTCNN model is a model which is constructed based on preset parameters;
the position comparison module 403 is configured to compare the relative position of the feature point in the feature map with the relative position of the frame of the feature map with a preset relative position to implement identity recognition.
Based on the above embodiment, the feature map obtaining module 402 in this embodiment includes:
the MTCNN model construction unit is used for respectively constructing the P-Net, the R-Net and the O-Net according to the preset parameter list to obtain a preset MTCNN model;
the feature map acquisition unit is used for inputting the sharpened image into a preset MTCNN model and outputting a feature map.
Based on the above embodiment, the feature map acquiring unit in this embodiment includes:
the characteristic map acquisition subunit is used for inputting the sharpened image into a preset MTCNN model, and sequentially carrying out convolution, pooling and full connection processing to output a characteristic map.
Based on the above embodiment, the position comparison module 403 in this embodiment includes:
the target matrix acquisition unit is used for storing the relative positions into the matrix to obtain a target matrix;
the comparison result acquisition unit is used for comparing the target matrix with a preset target matrix corresponding to a preset relative position in the database to obtain a comparison result;
the comparison result judging unit is used for judging whether the comparison result meets a preset threshold value or not; if yes, the identity identification is successful; if not, the identity identification fails.
The present application also provides an apparatus comprising: a memory and a processor; the memory is used for storing a computer program, and the processor is used for realizing the steps of the vein recognition method with scale rotation invariance based on IRCNN and MTCNN in any embodiment when the computer program is executed.
The present application also provides a computer readable storage medium storing a computer program which when executed by a processor implements the steps of the IRCNN and MTCNN based vein recognition method with scale rotation invariance of any of the embodiments described above.
The computer readable storage medium may include: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
In the description, each embodiment is described in a progressive manner, and each embodiment is mainly described by the differences from other embodiments, so that the same similar parts among the embodiments are mutually referred. The system provided by the embodiment is relatively simple to describe as it corresponds to the method provided by the embodiment, and the relevant points are referred to in the description of the method section.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative elements and steps are described above generally in terms of functionality in order to clearly illustrate the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. The software modules may be disposed in Random Access Memory (RAM), memory, read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
The vein recognition method, system, device and computer readable storage medium with scale rotation invariance based on IRCNN and MTCNN provided by the application are described above in detail. Specific examples are set forth herein to illustrate the principles and embodiments of the present application, and the description of the examples above is only intended to assist in understanding the methods of the present application and their core ideas. It should be noted that it would be obvious to those skilled in the art that various improvements and modifications can be made to the present application without departing from the principles of the present application, and such improvements and modifications fall within the scope of the claims of the present application.

Claims (6)

1. A vein recognition method with scale rotation invariance based on IRCNN and MTCNN, comprising:
inputting the target vein image into a preset IRCNN model for sharpening treatment to obtain a sharpened image; the preset IRCNN model consists of seven layers of expansion convolution, and comprises three different modules, namely an expansion convolution and ReLU activation function module in a first layer, five expansion convolution and BNarm batch normalization and ReLU activation function middle layer modules, and an expansion convolution module in a last layer;
respectively constructing P-Net, R-Net and O-Net according to a preset parameter list to obtain a preset MTCNN model;
inputting the sharpened image into the preset MTCNN model, and outputting a feature map; for the P-Net, when the sharpened image is input into the preset MTCNN model, a bounding box is adjusted to be a picture with the size of 12 multiplied by 12, and the picture is converted into a structure with the size of 12 multiplied by 3, so that training data of the P-Net are generated; the training data is passed through 103 x 3 convolution kernels, 2×2 max pooling operations, generating 10 5×5 feature maps; then generating 16 3×3 feature maps by 16 3×3×10 convolution kernels; then generating 32 1×1 feature maps by 32 convolution kernels of 3×3×16; finally, through 32 convolution kernels of 1×1×32, 32 feature maps of 1×1 are generated for classification; generating 32 1×1 feature maps for regression frame judgment by using 32 1×1×32 convolution kernels, and generating 32 1×1 feature maps for vein feature point judgment by using 32 1×1×32 convolution kernels; inputting a picture with the size of 24 multiplied by 24 for the R-Net, and generating 28 11 multiplied by 11 characteristic graphs after 28 convolution kernels with the size of 3 multiplied by 3 and the maximum pooling of 3 multiplied by 3; generating 48 4 x 4 feature maps through 48 convolution kernels of 3 x 28 and 3 x 3 maximum pooling; generating 64 3×3 feature maps after passing through 64 2×2×48 convolution kernels; passing the 3×3×64 feature map through a 4096-sized full link layer; and passing through a full connecting layer with the size of 4096; then passing through a full connection layer with 128 sizes; finally, the background problem is converted into a full connection layer with the size of 1; converting the regression frame classification problem into a full-connection layer with the size of 2, namely a vein recognition result; the position regression problem of the boundary frame is converted into a full connection layer with the size of 4, namely the boundary frame regression; converting the outline key points into full-connection layers with the size of 110, namely vein feature point positions; for the O-Net, the input is a picture of size 48X 3, by 32 convolution kernels of 3 x 3 and 3 x 3 the most converting the feature images into 32 feature images of 23×23 after pooling; the feature map is converted into 64 10 multiplied by 10 after being subjected to the maximum pooling of 64 convolution kernels of 3 multiplied by 32 and 3 multiplied by 3; the feature map is converted into 64 4 multiplied by 4 after being subjected to the maximum pooling of 64 convolution kernels of 3 multiplied by 64; converting into 128 characteristic diagrams of 3×3 through 128 convolution kernels of 2×2×64; passing the 3×3×128 feature map through a 4096-sized full link layer; and passing through a full connecting layer with the size of 4096; then passing through 257 full-connection layers; finally, generating a background with the size of 1; regression frame classification characteristics with the size of 2, namely vein recognition results; regression features of the regression frame position with the size of 4, namely, the regression of the boundary frame; regression features of vein feature points with the size of 110, namely the positions of the vein feature points;
and comparing the relative positions of the feature points in the feature map relative to the square frame of the feature map with preset relative positions so as to realize identification.
2. The vein recognition method with scale rotation invariance based on IRCNN and MTCNN according to claim 1, wherein the comparing the relative position of the feature point in the feature map with respect to the box of the feature map with a preset relative position to realize the identification comprises:
storing the relative positions into a matrix to obtain a target matrix;
comparing the target matrix with a preset target matrix corresponding to the preset relative position in a database to obtain a comparison result;
judging whether the comparison result meets a preset threshold value or not; if yes, the identity identification is successful; if not, the identity identification fails.
3. An IRCNN and MTCNN-based vein recognition system with scale rotation invariance, comprising:
the sharpening image acquisition module is used for inputting the target vein image into a preset IRCNN model for sharpening treatment to obtain a sharpening image; the preset IRCNN model consists of seven layers of expansion convolution, and comprises three different modules, namely an expansion convolution and ReLU activation function module in a first layer, five expansion convolution and BNarm batch normalization and ReLU activation function middle layer modules, and an expansion convolution module in a last layer;
the feature map acquisition module comprises:
the MTCNN model construction unit is used for respectively constructing the P-Net, the R-Net and the O-Net according to the preset parameter list to obtain a preset MTCNN model;
the feature map obtaining unit is used for inputting the sharpened image into the preset MTCNN model and outputting a feature map; for the P-Net, when the sharpened image is input into the preset MTCNN model, a bounding box is adjusted to be a picture with the size of 12 multiplied by 12, and the picture is converted into a structure with the size of 12 multiplied by 3, so that training data of the P-Net are generated; the training data is passed through 103 x 3 convolution kernels, 2×2 max pooling operations, generating 10 5×5 feature maps; then generating 16 3×3 feature maps by 16 3×3×10 convolution kernels; then generating 32 1×1 feature maps by 32 convolution kernels of 3×3×16; finally, through 32 convolution kernels of 1×1×32, 32 feature maps of 1×1 are generated for classification; generating 32 1×1 feature maps for regression frame judgment by using 32 1×1×32 convolution kernels, and generating 32 1×1 feature maps for vein feature point judgment by using 32 1×1×32 convolution kernels; inputting a picture with the size of 24 multiplied by 24 for the R-Net, and generating 28 11 multiplied by 11 characteristic graphs after 28 convolution kernels with the size of 3 multiplied by 3 and the maximum pooling of 3 multiplied by 3; generating 48 4 x 4 feature maps through 48 convolution kernels of 3 x 28 and 3 x 3 maximum pooling; generating 64 3×3 feature maps after passing through 64 2×2×48 convolution kernels; passing the 3×3×64 feature map through a 4096-sized full link layer; and passing through a full connecting layer with the size of 4096; then passing through a full connection layer with 128 sizes; finally, the background problem is converted into a full connection layer with the size of 1; converting the regression frame classification problem into a full-connection layer with the size of 2, namely a vein recognition result; the position regression problem of the boundary frame is converted into a full connection layer with the size of 4, namely the boundary frame regression; converting the outline key points into full-connection layers with the size of 110, namely vein feature point positions; for the O-Net, the input is a picture of size 48X 3, by 32 convolution kernels of 3 x 3 and 3 x 3 the most converting the feature images into 32 feature images of 23×23 after pooling; the feature map is converted into 64 10 multiplied by 10 after being subjected to the maximum pooling of 64 convolution kernels of 3 multiplied by 32 and 3 multiplied by 3; the feature map is converted into 64 4 multiplied by 4 after being subjected to the maximum pooling of 64 convolution kernels of 3 multiplied by 64; converting into 128 characteristic diagrams of 3×3 through 128 convolution kernels of 2×2×64; passing the 3×3×128 feature map through a 4096-sized full link layer; and passing through a full connecting layer with the size of 4096; then passing through 257 full-connection layers; finally, generating a background with the size of 1; regression frame classification characteristics with the size of 2, namely vein recognition results; regression features of the regression frame position with the size of 4, namely, the regression of the boundary frame; regression features of vein feature points with the size of 110, namely the positions of the vein feature points;
and the position comparison module is used for comparing the relative positions of the feature points in the feature map relative to the square frame of the feature map with the preset relative positions so as to realize identification.
4. The IRCNN and MTCNN based vein recognition system having scale rotation invariance according to claim 3, wherein the location comparison module comprises:
the target matrix acquisition unit is used for storing the relative positions into a matrix to obtain a target matrix;
the comparison result acquisition unit is used for comparing the target matrix with a preset target matrix corresponding to the preset relative position in the database to obtain a comparison result;
the comparison result judging unit is used for judging whether the comparison result meets a preset threshold value or not; if yes, the identity identification is successful; if not, the identity identification fails.
5. An electronic device, comprising:
a memory and a processor; wherein the memory is for storing a computer program, the processor being for implementing the steps of the IRCNN and MTCNN based vein recognition method with scale rotation invariance according to any of claims 1 to 2 when the computer program is executed.
6. A computer-readable storage medium, characterized in that it stores a computer program which, when executed by a processor, implements the steps of the IRCNN and MTCNN-based vein recognition method with scale rotation invariance according to any of claims 1 to 2.
CN201910620017.XA 2019-07-10 2019-07-10 Vein recognition method and system with scale rotation invariance based on IRCNN and MTCNN Active CN110334667B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910620017.XA CN110334667B (en) 2019-07-10 2019-07-10 Vein recognition method and system with scale rotation invariance based on IRCNN and MTCNN

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910620017.XA CN110334667B (en) 2019-07-10 2019-07-10 Vein recognition method and system with scale rotation invariance based on IRCNN and MTCNN

Publications (2)

Publication Number Publication Date
CN110334667A CN110334667A (en) 2019-10-15
CN110334667B true CN110334667B (en) 2023-07-25

Family

ID=68146181

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910620017.XA Active CN110334667B (en) 2019-07-10 2019-07-10 Vein recognition method and system with scale rotation invariance based on IRCNN and MTCNN

Country Status (1)

Country Link
CN (1) CN110334667B (en)

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102222229B (en) * 2011-07-28 2015-12-02 陈庆武 Method for preprocessing finger vein images
CN106295622A (en) * 2016-08-30 2017-01-04 哈尔滨理工大学 A kind of recognition methods based on hand back vein
CN109711233A (en) * 2017-10-26 2019-05-03 北京航天长峰科技工业集团有限公司 A kind of comparison certificate photo face identification method based on deep learning
CN108009520B (en) * 2017-12-21 2020-09-01 西安格威西联科技有限公司 Finger vein identification method and system based on convolution variational self-encoder network
CN109086665A (en) * 2018-06-27 2018-12-25 相舆科技（上海）有限公司 A kind of Management System for Clients Information based on recognition of face
CN109948573B (en) * 2019-03-27 2021-08-17 厦门大学 Noise robust face recognition method based on cascade deep convolutional neural network

Also Published As

Publication number Publication date
CN110334667A (en) 2019-10-15

Similar Documents

Publication Publication Date Title
CN108875732B (en) Model training and instance segmentation method, device and system and storage medium
CN109409222B (en) Multi-view facial expression recognition method based on mobile terminal
EP3916627A1 (en) Living body detection method based on facial recognition, and electronic device and storage medium
EP3885965B1 (en) Image recognition method based on micro facial expressions, apparatus and related device
CN110659582A (en) Image conversion model training method, heterogeneous face recognition method, device and equipment
CN105320945A (en) Image classification method and apparatus
CN105144239A (en) Image processing device, program, and image processing method
CN107784288A (en) A kind of iteration positioning formula method for detecting human face based on deep neural network
CN110619316A (en) Human body key point detection method and device and electronic equipment
CN110245621B (en) Face recognition device, image processing method, feature extraction model, and storage medium
CN112419202B (en) Automatic wild animal image recognition system based on big data and deep learning
CN110705353A (en) Method and device for identifying face to be shielded based on attention mechanism
CN111783761A (en) Certificate text detection method and device and electronic equipment
CN112836653A (en) Face privacy method, device and apparatus and computer storage medium
CN109635755A (en) Face extraction method, apparatus and storage medium
EP3671635B1 (en) Curvilinear object segmentation with noise priors
CN111080592B (en) Rib extraction method and device based on deep learning
CN113221771A (en) Living body face recognition method, living body face recognition device, living body face recognition equipment, storage medium and program product
CN111144407A (en) Target detection method, system, device and readable storage medium
CN114663726A (en) Training method of target type detection model, target detection method and electronic equipment
CN111985488B (en) Target detection segmentation method and system based on offline Gaussian model
CN111353325A (en) Key point detection model training method and device
CN111881803A (en) Livestock face recognition method based on improved YOLOv3
CN111353577B (en) Multi-task-based cascade combination model optimization method and device and terminal equipment
CN110334667B (en) Vein recognition method and system with scale rotation invariance based on IRCNN and MTCNN

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant