CN111368715B - Fingerprint anti-counterfeiting method, device and equipment - Google Patents

Fingerprint anti-counterfeiting method, device and equipment Download PDF

Info

Publication number
CN111368715B
CN111368715B CN202010136601.0A CN202010136601A CN111368715B CN 111368715 B CN111368715 B CN 111368715B CN 202010136601 A CN202010136601 A CN 202010136601A CN 111368715 B CN111368715 B CN 111368715B
Authority
CN
China
Prior art keywords
roi
fingerprint
network
maps
optimization selection
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
CN202010136601.0A
Other languages
Chinese (zh)
Other versions
CN111368715A (en
Inventor
陈书楷
徐志通
杨奇
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.)
Xiamen Entropy Technology Co ltd
Original Assignee
Xiamen Entropy Technology Co ltd
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 Xiamen Entropy Technology Co ltd filed Critical Xiamen Entropy Technology Co ltd
Priority to CN202010136601.0A priority Critical patent/CN111368715B/en
Publication of CN111368715A publication Critical patent/CN111368715A/en
Application granted granted Critical
Publication of CN111368715B publication Critical patent/CN111368715B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

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
    • G06V40/12—Fingerprints or palmprints
    • G06V40/1347—Preprocessing; Feature extraction
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]

Abstract

The application discloses a fingerprint anti-counterfeiting method, a fingerprint anti-counterfeiting device and fingerprint anti-counterfeiting equipment, wherein a plurality of first ROI (region of interest) images are generated by inputting acquired fingerprint images into an ROI recommendation network, and the first ROI images are sorted in a descending order according to the number of fingerprint pixel points in the first ROI images; inputting the sequenced first ROI images into an ROI optimization selection network for optimization selection, and feeding back an optimization selection result to the ROI recommendation network to enable the ROI recommendation network to output the first ROI images after optimization selection; inputting the contour map of the fingerprint image and the first ROI maps of the preset number after the optimized selection into a preset convolution neural network together for feature extraction, fusing the extracted features, predicting based on the fused features, and outputting a predicted value; the authenticity of the fingerprint in the fingerprint image is judged according to the predicted value, and the technical problems that the existing fingerprint anti-counterfeiting method is low in robustness, low in authenticity fingerprint identification rate and incapable of meeting practical requirements are solved.

Description

Fingerprint anti-counterfeiting method, device and equipment
Technical Field
The present application relates to the field of fingerprint anti-counterfeiting technologies, and in particular, to a fingerprint anti-counterfeiting method, apparatus and device.
Background
The fingerprint anti-counterfeiting method aims to effectively identify the true fingerprint and the false fingerprint, so that an authorized person can enjoy the corresponding authority, meanwhile, false fingerprint access is effectively denied, and identity security is improved. The existing fingerprint anti-counterfeiting method mainly adopts hardware equipment such as OCT imaging and the like to carry out fingerprint authenticity auxiliary identification, obtains rich information of a skin structure such as endocrine glands and capillary blood flow through OCT imaging by counting the relation between the number of sweat pores of a finger and a threshold value and is used for assisting in identifying the authenticity of the fingerprint, and the method needs expensive hardware equipment; when the fingerprint data is too small or too biased, the existing fingerprint anti-counterfeiting method has poor robustness and the problems of low true and false fingerprint identification rate and incapability of meeting practical requirements.
Disclosure of Invention
The application provides a fingerprint anti-counterfeiting method, a fingerprint anti-counterfeiting device and fingerprint anti-counterfeiting equipment, which are used for solving the technical problems that the existing fingerprint anti-counterfeiting method is low in robustness, low in true and false fingerprint identification rate and incapable of meeting practical requirements.
In view of the above, a first aspect of the present application provides a fingerprint anti-counterfeiting method, including:
acquiring a fingerprint image;
inputting the fingerprint image into an ROI recommendation network, enabling the ROI recommendation network to process the fingerprint image, generating a plurality of first ROI images, and sequencing the first ROI images in a descending order according to the number of fingerprint pixel points in the first ROI images, wherein the fingerprint pixel points are pixel points with pixel values smaller than a pixel threshold value;
inputting the first ROI images after the descending sorting into an ROI optimization selection network, enabling the ROI optimization selection network to perform optimization selection on the first ROI images after the descending sorting, and feeding back an optimization selection result to the ROI recommendation network, so that the ROI recommendation network outputs the first ROI images after the optimization selection;
inputting the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after the optimized selection into a preset convolutional neural network together, so that the preset convolutional neural network extracts the features of the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after the optimized selection, fuses the extracted features, predicts based on the fused features, and outputs a predicted value;
and judging the authenticity of the fingerprint in the fingerprint image according to the predicted value.
Preferably, the ROI optimization selection network performs optimization selection on the first ROI map sorted in the descending order, and feeds back an optimization selection result to the ROI recommendation network, so that the ROI recommendation network outputs the first ROI map after optimization selection, including:
the ROI optimization selection network performs feature extraction on each first ROI image after descending sorting to obtain a plurality of ROI feature images;
the ROI optimization selection network performs feature matching on two adjacent ROI feature maps and calculates an IOU value;
when the IOU value is larger than a preset threshold value, the ROI optimization selection network deletes the first ROI map corresponding to the next ROI feature map in the two adjacent ROI feature maps, and feeds back the position of the deleted first ROI map to the ROI recommendation network, so that the ROI recommendation network deletes the first ROI map corresponding to the position of the deleted first ROI feature map, and outputs the first ROI map after optimization selection.
Preferably, the loss function of the preset convolutional neural network is:
L total ＝ξ·L rank +τ·L focal ；
wherein xi and tau are L respectively rank 、L focal Weight parameter of L rank As a function of rank loss, L focal Is the focal loss function.
Preferably, the inputting the contour map of the fingerprint image and the first ROI map of the first ROI map after the optimization selection together into a preset convolutional neural network includes:
when more than two region recommended sizes are set in the ROI recommended network, the contour map of the fingerprint image and the first ROI maps with the preset number after optimized selection are input into a preset convolutional neural network together after the first ROI maps are reshaped to a preset size.
Preferably, the inputting the contour map of the fingerprint image and the first ROI map of the first ROI map after the optimization selection together into a preset convolutional neural network further comprises:
acquiring a fingerprint image to be trained;
inputting the fingerprint image to be trained into the ROI recommendation network, enabling the ROI recommendation network to process the fingerprint image to be trained, generating a plurality of second ROI maps, and sequencing the second ROI maps in a descending order according to the number of the fingerprint pixel points in the second ROI maps;
inputting the second ROI images after the descending sorting into the ROI optimization selection network, enabling the ROI optimization selection network to perform optimization selection on the second ROI images after the descending sorting, and feeding back an optimization selection result to the ROI recommendation network, so that the ROI recommendation network outputs the second ROI images after the optimization selection;
inputting the contour map of the fingerprint image to be trained and the second ROI maps with the preset number in front of the second ROI maps after optimization selection into a convolutional neural network together, and training the convolutional neural network;
and when the convolutional neural network reaches a convergence condition, obtaining the trained convolutional neural network, and taking the trained convolutional neural network as the preset convolutional neural network.
Preferably, the determining the authenticity of the fingerprint in the fingerprint image according to the predicted value includes:
when the predicted value is larger than 0.5, judging that the fingerprint in the fingerprint image is a true fingerprint;
and when the predicted value is less than or equal to 0.5, judging that the fingerprint in the fingerprint image is a false fingerprint.
This application second aspect provides a fingerprint anti-counterfeiting device, includes:
the first image acquisition module is used for acquiring a fingerprint image;
the first sequencing module is used for inputting the fingerprint image into an ROI recommendation network, so that the ROI recommendation network processes the fingerprint image to generate a plurality of first ROI maps, and the first ROI maps are subjected to descending sequencing according to the number of fingerprint pixel points in the first ROI maps, wherein the fingerprint pixel points are pixel points with pixel values smaller than a pixel threshold value;
the first optimization selection module is used for inputting the first ROI maps after the descending sorting into an ROI optimization selection network, so that the ROI optimization selection network performs optimization selection on the first ROI maps after the descending sorting and feeds an optimization selection result back to the ROI recommendation network, and the ROI recommendation network outputs the first ROI maps after the optimization selection;
the prediction module is used for inputting the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after the optimized selection into a preset convolutional neural network together, so that the preset convolutional neural network extracts the features of the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after the optimized selection, fuses the extracted features, predicts based on the fused features and outputs a predicted value;
and the judging module is used for judging the authenticity of the fingerprint in the fingerprint image according to the predicted value.
Preferably, the method further comprises the following steps:
the second image acquisition module is used for acquiring a fingerprint image to be trained;
the second sequencing module is used for inputting the fingerprint images to be trained into the ROI recommendation network, so that the ROI recommendation network processes the fingerprint images to be trained to generate a plurality of second ROI maps, and the second ROI maps are sequenced in a descending order according to the number of the fingerprint pixel points in the second ROI maps;
the second optimization selection module is used for inputting the second ROI images after the descending sorting into the ROI optimization selection network, so that the ROI optimization selection network performs optimization selection on the second ROI images after the descending sorting, and feeds an optimization selection result back to the ROI recommendation network, so that the ROI recommendation network outputs the second ROI images after the optimization selection;
the training module is used for inputting the contour map of the fingerprint image to be trained and the second ROI maps of the preset number after the optimal selection into a convolutional neural network together to train the convolutional neural network;
and the convergence module is used for obtaining the trained convolutional neural network when the convolutional neural network reaches a convergence condition, and taking the trained convolutional neural network as the preset convolutional neural network.
Preferably, the determining module is specifically configured to:
when the predicted value is larger than 0.5, judging that the fingerprint in the fingerprint image is a true fingerprint;
and when the predicted value is less than or equal to 0.5, judging the fingerprint in the fingerprint image to be a false fingerprint.
A third aspect of the application provides a fingerprint anti-counterfeiting device, which comprises a processor and a memory;
the memory is used for storing program codes and transmitting the program codes to the processor;
the processor is configured to execute the fingerprint anti-counterfeiting method according to any one of the first aspect according to instructions in the program code.
According to the technical scheme, the method has the following advantages:
the application provides a fingerprint anti-counterfeiting method, which comprises the following steps: acquiring a fingerprint image; inputting the fingerprint image into an ROI recommendation network, enabling the ROI recommendation network to process the fingerprint image, generating a plurality of first ROI images, and sequencing the first ROI images in a descending order according to the number of fingerprint pixel points in the first ROI images, wherein the fingerprint pixel points are pixel points with pixel values smaller than a pixel threshold value; inputting the first ROI images after the descending sorting into an ROI optimization selection network, enabling the ROI optimization selection network to perform optimization selection on the first ROI images after the descending sorting, and feeding back an optimization selection result to the ROI recommendation network, so that the ROI recommendation network outputs the first ROI images after the optimization selection; inputting the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after the optimized selection into a preset convolution neural network together, so that the preset convolution neural network extracts the features of the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after the optimized selection, fuses the extracted features, predicts based on the fused features, and outputs a predicted value; and judging the authenticity of the fingerprint in the fingerprint image according to the predicted value.
According to the fingerprint anti-counterfeiting method, ROI recommendation is carried out on the acquired fingerprint images by adopting an ROI recommendation network, and the first ROI images are sorted in a descending order according to the number of pixel points belonging to the fingerprint in the generated multiple first ROI images; the ROI optimization selection network is adopted to carry out optimization selection on the sequenced first ROI maps, the first ROI maps with more fingerprint details are reserved to the maximum extent, and the method is beneficial to presetting a convolution neural network to extract more fingerprint characteristic information, so that the accuracy of identifying true and false fingerprints is improved, the first ROI maps with less fingerprint information are screened, the speed of identifying the true and false fingerprints can be improved, and the practicability is improved; the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after optimized selection are input into the preset convolutional neural network together, so that the preset convolutional neural network can extract various channel characteristics, the characteristic representation is enhanced by fusing the extracted various channel characteristics, the generalization capability and robustness of the model are improved, the anti-counterfeiting prediction accuracy of the fingerprint is improved, and the technical problems that the existing fingerprint anti-counterfeiting method is low in robustness, low in true and false fingerprint identification rate and incapable of meeting practical requirements are solved.
Drawings
Fig. 1 is a schematic flowchart of a fingerprint anti-counterfeiting method provided in an embodiment of the present application;
fig. 2 is another schematic flow chart of a fingerprint anti-counterfeiting method according to an embodiment of the present disclosure;
fig. 3 is a schematic structural diagram of a fingerprint anti-counterfeiting device according to an embodiment of the present application;
FIG. 4 is a schematic diagram illustrating ROI map recommendation generation provided in an embodiment of the present application;
fig. 5 is a schematic structural diagram of four-channel data obtained by channel splicing the contour map provided in the embodiment of the present application and the ROI map of top 3.
Detailed Description
In order to make the technical solutions of the present application better understood, the technical solutions in 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 obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
For easy understanding, referring to fig. 1, an embodiment of a fingerprint anti-counterfeiting method provided in the present application includes:
step 101, acquiring a fingerprint image.
It should be noted that the fingerprint image may be acquired by a fingerprint acquisition device.
102, inputting the fingerprint image into an ROI recommendation network, enabling the ROI recommendation network to process the fingerprint image, generating a plurality of first ROI images, and sequencing the first ROI images in a descending order according to the number of fingerprint pixel points in the first ROI images.
It should be noted that, because the sizes of the fingerprint images acquired by different fingerprint acquisition devices may be different, if a reshaping operation is simply performed, the fingerprint contour and the fingerprint texture minutiae may be damaged, which affects the accuracy of fingerprint anti-counterfeiting prediction, and therefore, in the embodiment of the present application, before the preset convolutional neural network is used to extract the features of the fingerprint image, an ROI (Region of interest) minutiae map of the fingerprint is acquired first, so as to ensure that the fingerprint texture features are not damaged. In the embodiment of the application, the ROI recommendation Network is adopted to perform regional recommendation on the fingerprint image to generate a plurality of first ROI maps, and the existing regional recommendation Network can be adopted as the ROI recommendation Network, for example, an RPN (Region proxy Network); in the ROI recommending process, the ROI recommending network conducts descending sequencing on the first ROI images according to the number of fingerprint pixel points in the first ROI images to obtain the sequenced first ROI images, wherein the fingerprint pixel points are pixel points with pixel values smaller than a pixel threshold value, and the pixel points with the pixel values smaller than 50 in the ROI images are found through experiments.
And 103, inputting the first ROI images after the descending sorting into an ROI optimization selection network, so that the ROI optimization selection network performs optimization selection on the first ROI images after the descending sorting, and feeding back an optimization selection result to the ROI recommendation network, so that the ROI recommendation network outputs the first ROI images after the optimization selection.
It should be noted that, in order to retain the detail features of the fingerprint image as much as possible, and to discard redundant feature information and improve the data processing speed, in the embodiment of the present application, an ROI optimization selection network is used to perform optimization selection on the sorted first ROI map.
And 104, inputting the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after the optimized selection into a preset convolution neural network together, so that the preset convolution neural network extracts the features of the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after the optimized selection, fuses the extracted features, predicts based on the fused features, and outputs a predicted value.
It should be noted that the first ROI maps of the first ROI map after the optimization selection are selected to be used as input data, that is, the first ROI map retaining the most fingerprint detail information is used as the input of the network, so that the preset convolutional neural network can extract more fingerprint features, it is found through experiments that the first ROI maps of the first 3 ROI maps after the optimization selection can substantially contain more than 90% of the area of the fingerprint, in order to improve the data processing speed, in the embodiment of the present application, the first ROI maps of the first 3 ROI maps after the optimization selection and the outline maps of the fingerprint images are preferably used as input data, wherein the fingerprint images acquired by the fingerprint acquisition device are rectangular, there is height > width, the positions (upper or lower) of the fingerprint images of the fingerprint areas in the fingerprint images are determined by counting the fingerprint pixel values in the fingerprint images, then according to the dimensions of the fingerprint images and the dimensions of the fingerprint images, the side length is square with the centroid as the center, the obtained fingerprint images are square images, and when the fingerprint images are cut off the upper or lower side of the centroid of the square images, the fingerprint images, and the side edge of the fingerprint images are obtained according to the side edge of the centroid positions of the fingerprint images.
In the embodiment of the application, a preset convolutional neural network is adopted to extract the features of the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after optimization selection, the extracted features are fused to enhance feature representation, prediction is carried out based on the fused features, and a predicted value is output and is a probability value that the fingerprint in the fingerprint image is a true fingerprint.
And 105, judging the authenticity of the fingerprint in the fingerprint image according to the predicted value.
The authenticity of the fingerprint in the fingerprint image can be judged according to the size of the predicted value.
According to the fingerprint anti-counterfeiting method in the embodiment of the application, ROI recommendation is carried out on the acquired fingerprint images by adopting an ROI recommendation network, and the ROI images are sorted in a descending order according to the number of pixel points belonging to fingerprints in the plurality of generated ROI images; the ROI optimization selection network is adopted to carry out optimization selection on the sequenced first ROI images, so that the first ROI images with more fingerprint details are reserved to the maximum extent, the preset convolution neural network is facilitated to extract more fingerprint characteristic information, the accuracy rate of true and false fingerprint identification is improved, the first ROI images with less fingerprint information are screened, the speed of true and false fingerprint identification can be improved, and the practicability is improved; the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after optimized selection are input into the preset convolutional neural network together, so that the preset convolutional neural network can extract various channel features, the feature representation is enhanced by fusing the extracted various channel features, the generalization capability and robustness of the model are improved, the fingerprint anti-counterfeiting prediction accuracy is improved, and the technical problems that the existing fingerprint anti-counterfeiting method is low in robustness, low in true and false fingerprint identification rate and incapable of meeting practical requirements are solved.
For easy understanding, referring to fig. 2, fig. 4 and fig. 5, another embodiment of a fingerprint anti-counterfeiting method provided by the present application includes:
step 201, obtaining a fingerprint image to be trained.
It should be noted that the fingerprint image to be trained may pass through the fingerprint acquisition device or may be acquired from the fingerprint database.
Step 202, inputting the fingerprint image to be trained into an ROI recommendation network, enabling the ROI recommendation network to process the fingerprint image to be trained, generating a plurality of second ROI maps, and sequencing the second ROI maps in a descending order according to the number of fingerprint pixel points in the second ROI maps.
It should be noted that, because the fingerprint image features are less, if the fingerprint areas included in the fingerprint image to be trained are more, the more fingerprint feature information can be included, assuming that the size of the fingerprint image to be trained is w × h, the ROI recommendation network is used to perform area recommendation on the fingerprint image to be trained, that is, random search cropping is performed on the fingerprint image to be trained by using anchors to generate a plurality of second ROI maps, in the embodiment of the present application, the ROI recommendation network sets two area recommendation sizes, which are s, respectively 1 ：
Figure BDA0002397545100000081
And s 2 :122, and a magnetic field generator, 122, i.e. based on>
Figure BDA0002397545100000082
The search step sizes are respectively set to
Figure BDA0002397545100000083
Generating second ROI maps As shown in FIG. 4, assume that a total of n second ROI maps are generated, which are +>
Figure BDA0002397545100000087
Represents all second ROI maps and->
Figure BDA0002397545100000086
In the ROI recommending process, the ROI recommending network performs descending sequencing on the second ROI images according to the number of fingerprint pixel points in each second ROI image, and experiments are performedFinding that pixel points with pixel values less than 50 in the second ROI image are fingerprint pixel points, and selecting any second ROI image based on the fingerprint pixel points>
Figure BDA0002397545100000088
Let the sorting function be μ, let->
Figure BDA0002397545100000085
Sorting the second ROI map sorted in descending order, then->
Figure BDA0002397545100000084
And 203, inputting the second ROI images after the descending sorting into an ROI optimization selection network, so that the ROI optimization selection network performs optimization selection on the second ROI images after the descending sorting, and feeding back an optimization selection result to the ROI recommendation network, so that the ROI recommendation network outputs the second ROI images after the optimization selection.
It should be noted that after the first round of second ROI map ordering is obtained through the ROI recommendation network, all the ordered second ROI maps are input to the ROI optimization selection network, and the ROI optimization selection network performs feature extraction on each second ROI map ordered in a descending order to obtain a plurality of ROI feature maps;
the ROI optimization selection network performs feature matching on any two adjacent ROI feature maps according to the descending order and calculates the IOU value of the two adjacent ROI feature maps;
when the IOU value is larger than a preset threshold value, the ROI optimization selection network deletes a second ROI (region of interest) graph corresponding to the next ROI feature graph in two adjacent ROI feature graphs, the subsequent ROI feature graphs move forward by one bit in sequence, the ROI optimization selection network feeds the position of the deleted second ROI graph back to the ROI recommendation network, the ROI recommendation network deletes the second ROI graph corresponding to the position of the deleted second ROI feature graph, the subsequent ROI graphs move forward by one bit in sequence, and for example, the situation that the second ROI graph r after sequencing is supposed to be calculated is that 1 And r 2 If the IOU value exceeds a preset threshold, the ROI optimization selection network will delete r 2 And feeding back the result to the ROI recommendation network to execute the same operation and delete r 2 The final optimized selected second ROI map may be
Figure BDA0002397545100000091
Let δ be the ordering result after the optimized selection mechanism, λ i Is the i-th second ROI map after sorting, then delta = { lambda = { (λ) 1 ,λ 2 ,…,λ t T is the number of final secondary ROI maps. The preset threshold in this embodiment of the present application is preferably set to 0.8, and when the IOU value is greater than 0.8, that is, the data overlap amount exceeds 80%, the subsequently learned features are similar, and in order to avoid feature redundancy and retain as much fingerprint information as possible, the second ROI map of the next bit is deleted.
And step 204, inputting the contour map of the fingerprint image to be trained and the second ROI maps with the preset number in front of the second ROI maps after optimization selection into the convolutional neural network together, and training the convolutional neural network.
It should be noted that, in the experimental process, by observing the characteristics of the fingerprint image, the 3 regions are adopted to basically contain more than 90% of the fingerprint, so that in order not to influence the depth of network training, the top3 second ROI map of the second ROI map after optimization selection, that is, δ = { λ = is selected in the present application 1 ,λ 2 ,λ 3 The method can better ensure that the fingerprint image detail areas are not lost when the network is trained, improve the problem of inaccurate clipping due to the mass center clipping of the ROI image, and simultaneously can better improve the data processing speed; and feature fusion is carried out by combining the profile map features of the fingerprint image to be trained, so that the fingerprint anti-counterfeiting model with strong robustness on small fingerprints, partial fingerprints and bright fingerprints can be trained.
In the embodiment of the application, an existing Feature Pyramid Network (FPN) is improved, a fingerprint attention mechanism is added to the existing FPN network, and the improved network is named as FPN5_ HOA. Because two region recommended sizes are set in the ROI recommended network, sizes of the obtained second ROI maps are different, in this embodiment of the present application, the contour map of the fingerprint image to be trained and the first 3 second ROI maps of the second ROI map after optimization selection are reshaped to a preset size, the preset size is preferably 112 × 112, and then the reshaped contour map of the fingerprint image to be trained and the second ROI map of top3 are channel-spliced to obtain four-channel data as input data of the fpn5_ HOA network, which may refer to fig. 5; the fpn5_ HOA network performs a series of convolution, pooling and activation operations on the input four-channel data to perform feature extraction, three groups of feature vectors are obtained, concat fusion is performed on the three groups of feature vectors in channel dimensions, namely the three groups of feature vectors are vertically spliced to obtain spliced feature vectors, and finally the probability values of true and false fingerprints are output through a linear layer and a softmax layer.
It should be noted that the fpn5_ HOA network in the embodiment of the present application uses a loss function combining rank loss and focal loss, and uses rank loss to perform a sorting penalty on the ROI of the fingerprint, that is, the aforementioned fingerprint attention mechanism uses focal loss to deal with the problem of unbalanced training data samples, where the expression L of rank loss is used to represent that the ROI of the fingerprint is unbalanced rank Comprises the following steps:
Figure BDA0002397545100000101
and when the sequence of the second round of second ROI maps after sequencing is consistent with the sequence of the first round of second ROI maps after sequencing, performing no sequencing loss penalty, and if the sequence is inconsistent, performing loss penalty on the first round of second ROI maps by using an f (x) function.
Expression L of Focal loss focal Comprises the following steps:
Figure BDA0002397545100000102
wherein y is the real label of the fingerprint image to be trained, y =1 represents the real fingerprint, y =0 represents the false fingerprint, y' is the predicted value of the network, the parameter α is used for balancing the category of the real and false fingerprints in the fingerprint image to be trained, and is preferably set to 0.25, and the parameter γ is used for controlling the weight reduction rate of the simple samples, reducing the loss of the samples which are easily classified, and making the fpn5_ HOA network focus more on the difficult samples, and is preferably set to 3.
The overall loss function for the fpn5_ HOA network is:
L total ＝ξ·L rank +τ·L focal ；
wherein xi and tau are L respectively rank 、L focal Preferably, ξ and τ in the embodiment of the present application are set to 1.
And step 205, when the convolutional neural network reaches the convergence condition, obtaining the trained convolutional neural network, and taking the trained convolutional neural network as a preset convolutional neural network.
It should be noted that the convergence condition may be that a preset iteration number is reached, or that a training error is lower than a preset error, when the fpn5_ HOA network reaches the convergence condition, a trained fpn5_ HOA network is obtained, and the trained fpn5_ HOA network is used as a preset convolutional neural network.
Step 206, acquiring a fingerprint image.
It should be noted that the fingerprint image may be acquired by a fingerprint acquisition device.
And step 207, inputting the fingerprint image into an ROI recommendation network, so that the ROI recommendation network processes the fingerprint image to generate a plurality of first ROI maps, and sequencing the first ROI maps in a descending order according to the number of fingerprint pixel points in the first ROI maps.
It should be noted that step 207 is the same as step 202, and is not described herein again.
And 208, inputting the first ROI images after the descending order into an ROI optimization selection network, so that the ROI optimization selection network performs optimization selection on the first ROI images after the descending order, and feeding back an optimization selection result to the ROI recommendation network, so that the ROI recommendation network outputs the first ROI images after the optimization selection.
It should be noted that, in order to retain the detail features of the fingerprint image as much as possible, and to discard redundant feature information and improve the data processing speed, in the embodiment of the present application, an ROI optimization selection network is used to perform optimization selection on the sorted first ROI map, specifically: inputting the first ROI maps after descending sorting into an ROI optimization selection network, and performing feature extraction on each first ROI map after descending sorting by the ROI optimization selection network to obtain a plurality of ROI feature maps;
the ROI optimization selection network performs feature matching on two adjacent ROI feature maps and calculates an IOU value;
and when the IOU value is larger than a preset threshold value, the ROI optimization selection network deletes the first ROI image corresponding to the next ROI feature image in the two adjacent ROI feature images, and feeds back the position of the deleted first ROI image to the ROI recommendation network, so that the ROI recommendation network deletes the first ROI image corresponding to the position of the deleted first ROI feature image, and outputs the first ROI image after optimization selection.
And 209, inputting the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after the optimized selection into a preset convolution neural network together, so that the preset convolution neural network extracts the features of the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after the optimized selection, fuses the extracted features, predicts based on the fused features, and outputs a predicted value.
It should be noted that, because two region recommended sizes are set in the ROI recommendation network in step 207, in the case that the first ROI maps after the optimized selection may have different sizes, the first ROI maps of the first 3 of the first ROI maps after the fingerprint image and the contour map after the optimized selection need to be reshaped to a preset size and then input to the preset convolutional neural network together, where the preset size is consistent with the preset size reshaped during training. The method includes the steps that first ROI maps with the preset number in front of the first ROI maps after optimization selection are selected to serve as input data, namely the first ROI maps retaining most fingerprint detail information serve as input of a network, so that more fingerprint features can be extracted by the preset convolutional neural network, experiments show that the first 3 ROI maps of the first ROI maps after optimization selection basically contain more than 90% of regions of fingerprints, in order to improve data processing speed, the first ROI maps of the first 3 ROI maps after optimization selection and outline maps of fingerprint images are preferably adopted as the input data in the embodiment of the application, and the outline maps of the fingerprint images are obtained by feature extraction of the fingerprint images.
In the embodiment of the application, a preset convolution neural network is adopted to extract the characteristics of the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after optimized selection, the extracted characteristics are fused to enhance characteristic representation, prediction is carried out based on the fused characteristics, and a predicted value is output and is the probability value that the fingerprint in the fingerprint image is a true fingerprint; the fingerprint anti-counterfeiting method in the embodiment of the application can extract the ROI images with the same size from the fingerprint images with different sizes collected by any fingerprint collecting equipment through the ROI recommendation network, and compared with direct remolding operation, the fingerprint anti-counterfeiting method can better keep the detail texture characteristics of the fingerprint images, can effectively process small fingerprints and partial fingerprints, and can realize the anti-counterfeiting function of the fingerprint images with different sizes collected by different fingerprint collecting equipment.
And step 210, judging the authenticity of the fingerprint in the fingerprint image according to the predicted value.
It should be noted that, when the predicted value is greater than 0.5, the fingerprint in the fingerprint image is determined to be a true fingerprint; and when the predicted value is less than or equal to 0.5, judging the fingerprint in the fingerprint image to be a false fingerprint.
For easy understanding, referring to fig. 3, an embodiment of a fingerprint anti-counterfeiting device provided by the present application includes:
the first image acquisition module is used for acquiring a fingerprint image.
The first sequencing module is used for inputting the fingerprint images into the ROI recommendation network, so that the ROI recommendation network processes the fingerprint images to generate a plurality of first ROI images, the first ROI images are subjected to descending sequencing according to the number of fingerprint pixel points in the first ROI images, and the fingerprint pixel points are pixel points with pixel values smaller than a pixel threshold value.
And the first optimization selection module is used for inputting the first ROI maps after the descending sorting into an ROI optimization selection network, so that the ROI optimization selection network performs optimization selection on the first ROI maps after the descending sorting and feeds an optimization selection result back to the ROI recommendation network, and the ROI recommendation network outputs the first ROI maps after the optimization selection.
And the prediction module is used for inputting the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after the optimization selection into a preset convolution neural network together, so that the preset convolution neural network extracts the features of the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after the optimization selection, fuses the extracted features, predicts based on the fused features and outputs a predicted value.
And the judging module is used for judging the authenticity of the fingerprint in the fingerprint image according to the predicted value.
Further, the method also comprises the following steps:
and the second image acquisition module is used for acquiring the fingerprint image to be trained.
And the second sequencing module is used for inputting the fingerprint images to be trained into the ROI recommendation network, so that the ROI recommendation network processes the fingerprint images to be trained to generate a plurality of second ROI images, and the second ROI images are subjected to descending sequencing according to the number of fingerprint pixel points in the second ROI images.
And the second optimization selection module is used for inputting the second ROI images after the descending sorting into the ROI optimization selection network, so that the ROI optimization selection network performs optimization selection on the second ROI images after the descending sorting, and feeds back the optimization selection result to the ROI recommendation network, so that the ROI recommendation network outputs the second ROI images after the optimization selection.
And the training module is used for inputting the contour map of the fingerprint image to be trained and the second ROI maps with the preset number in front of the second ROI maps after the optimized selection into the convolutional neural network together to train the convolutional neural network.
And the convergence module is used for obtaining the trained convolutional neural network when the convolutional neural network reaches a convergence condition, and taking the trained convolutional neural network as a preset convolutional neural network.
Further, the judging module is specifically configured to:
when the predicted value is larger than 0.5, judging the fingerprint in the fingerprint image to be a true fingerprint;
and when the predicted value is less than or equal to 0.5, judging the fingerprint in the fingerprint image to be a false fingerprint.
The embodiment of the application also provides fingerprint anti-counterfeiting equipment, which comprises a processor and a memory;
the memory is used for storing the program codes and transmitting the program codes to the processor;
the processor is used for executing the fingerprint anti-counterfeiting method in the embodiment of the fingerprint anti-counterfeiting method according to the instructions in the program code.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application, or portions or all or portions of the technical solutions that contribute to the prior art, may be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for executing all or part of the steps of the methods described in the embodiments of the present application through a computer device (which may be a personal computer, a server, or a network device). And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
The above embodiments are only used to illustrate the technical solutions of the present application, and not to limit the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present application.

Claims (9)

1. A fingerprint anti-counterfeiting method is characterized by comprising the following steps:
acquiring a fingerprint image;
inputting the fingerprint image into an ROI recommendation network, enabling the ROI recommendation network to process the fingerprint image, generating a plurality of first ROI images, and sequencing the first ROI images in a descending order according to the number of fingerprint pixel points in the first ROI images, wherein the fingerprint pixel points are pixel points with pixel values smaller than a pixel threshold value;
inputting the first ROI graphs after the descending sorting into an ROI optimization selection network, so that the ROI optimization selection network performs optimization selection on the first ROI graphs after the descending sorting, and feeding back an optimization selection result to the ROI recommendation network, so that the ROI recommendation network outputs the first ROI graphs after the optimization selection, wherein the method comprises the following steps:
the ROI optimization selection network performs feature extraction on each first ROI image after descending sorting to obtain a plurality of ROI feature images;
the ROI optimization selection network performs feature matching on two adjacent ROI feature maps and calculates an IOU value;
when the IOU value is larger than a preset threshold value, the ROI optimization selection network deletes the first ROI map corresponding to the next ROI feature map in two adjacent ROI feature maps, and feeds back the deleted position of the first ROI map to the ROI recommendation network, so that the ROI recommendation network deletes the first ROI map corresponding to the deleted position of the first ROI map, and outputs the first ROI map after optimization selection;
inputting the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after the optimized selection into a preset convolutional neural network together, so that the preset convolutional neural network extracts the features of the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after the optimized selection, fuses the extracted features, predicts based on the fused features, and outputs a predicted value;
and judging the authenticity of the fingerprint in the fingerprint image according to the predicted value.
2. The fingerprint anti-counterfeiting method according to claim 1, wherein the loss function of the preset convolutional neural network is as follows:
L total ＝ξ·L rank +τ·L focal ；
wherein xi and tau are L respectively rank 、L focal Weight parameter of L rank As a function of rank loss, L focal Is the focal loss function.
3. The fingerprint anti-counterfeiting method according to claim 1, wherein the step of inputting the profile map of the fingerprint image and the first ROI map of the first ROI map after optimization selection into a preset convolutional neural network together comprises the following steps:
when more than two region recommended sizes are set in the ROI recommended network, the contour map of the fingerprint image and the first ROI maps with the preset number after optimized selection are input into a preset convolutional neural network together after the first ROI maps are reshaped to a preset size.
4. The fingerprint anti-counterfeiting method according to claim 1, wherein the inputting the profile map of the fingerprint image and the first ROI map of the first ROI map after the optimization selection into a preset convolutional neural network together comprises:
acquiring a fingerprint image to be trained;
inputting the fingerprint image to be trained into the ROI recommendation network, enabling the ROI recommendation network to process the fingerprint image to be trained, generating a plurality of second ROI graphs, and sequencing the second ROI graphs in a descending order according to the number of the fingerprint pixel points in the second ROI graphs;
inputting the second ROI images after the descending sorting into the ROI optimization selection network, enabling the ROI optimization selection network to perform optimization selection on the second ROI images after the descending sorting, and feeding back an optimization selection result to the ROI recommendation network, so that the ROI recommendation network outputs the second ROI images after the optimization selection;
inputting the contour map of the fingerprint image to be trained and the second ROI maps with the preset number in front of the second ROI maps after optimization selection into a convolutional neural network together, and training the convolutional neural network;
and when the convolutional neural network reaches a convergence condition, obtaining the trained convolutional neural network, and taking the trained convolutional neural network as the preset convolutional neural network.
5. The fingerprint anti-counterfeiting method according to claim 1, wherein the judging whether the fingerprint in the fingerprint image is true or false according to the predicted value comprises the following steps:
when the predicted value is larger than 0.5, judging that the fingerprint in the fingerprint image is a true fingerprint;
and when the predicted value is less than or equal to 0.5, judging that the fingerprint in the fingerprint image is a false fingerprint.
6. A fingerprint security device, comprising:
the first image acquisition module is used for acquiring a fingerprint image;
the first sequencing module is used for inputting the fingerprint image into an ROI recommendation network, so that the ROI recommendation network processes the fingerprint image to generate a plurality of first ROI maps, and the first ROI maps are subjected to descending sequencing according to the number of fingerprint pixel points in the first ROI maps, wherein the fingerprint pixel points are pixel points with pixel values smaller than a pixel threshold value;
the first optimization selection module is configured to input the first ROI map after the descending order to an ROI optimization selection network, so that the ROI optimization selection network performs optimization selection on the first ROI map after the descending order, and feeds back an optimization selection result to the ROI recommendation network, so that the ROI recommendation network outputs the first ROI map after the optimization selection, and the first optimization selection module includes:
the ROI optimization selection network extracts the features of the first ROI pictures after the descending sorting to obtain a plurality of ROI feature pictures;
the ROI optimization selection network performs feature matching on two adjacent ROI feature maps and calculates an IOU value;
when the IOU value is larger than a preset threshold value, the ROI optimization selection network deletes the first ROI map corresponding to the next ROI feature map in two adjacent ROI feature maps, and feeds back the deleted position of the first ROI map to the ROI recommendation network, so that the ROI recommendation network deletes the first ROI map corresponding to the deleted position of the first ROI map, and outputs the first ROI map after optimization selection;
the prediction module is used for inputting the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after the optimized selection into a preset convolutional neural network together, so that the preset convolutional neural network extracts the features of the contour map of the fingerprint image and the first ROI maps with the preset number in front of the first ROI maps after the optimized selection, fuses the extracted features, predicts based on the fused features and outputs a predicted value;
and the judging module is used for judging the authenticity of the fingerprint in the fingerprint image according to the predicted value.
7. The fingerprint anti-counterfeiting device according to claim 6, further comprising:
the second image acquisition module is used for acquiring a fingerprint image to be trained;
the second sequencing module is used for inputting the fingerprint images to be trained into the ROI recommendation network, so that the ROI recommendation network processes the fingerprint images to be trained to generate a plurality of second ROI maps, and the second ROI maps are sequenced in a descending order according to the number of the fingerprint pixel points in the second ROI maps;
the second optimization selection module is used for inputting the second ROI images after the descending sorting into the ROI optimization selection network, so that the ROI optimization selection network performs optimization selection on the second ROI images after the descending sorting, and feeds an optimization selection result back to the ROI recommendation network, so that the ROI recommendation network outputs the second ROI images after the optimization selection;
the training module is used for inputting the contour map of the fingerprint image to be trained and the second ROI maps with the preset number in front of the second ROI maps after optimization selection into a convolutional neural network together to train the convolutional neural network;
and the convergence module is used for obtaining the trained convolutional neural network when the convolutional neural network reaches a convergence condition, and taking the trained convolutional neural network as the preset convolutional neural network.
8. The fingerprint anti-counterfeiting device according to claim 6, wherein the judging module is specifically configured to:
when the predicted value is larger than 0.5, judging that the fingerprint in the fingerprint image is a true fingerprint;
and when the predicted value is less than or equal to 0.5, judging the fingerprint in the fingerprint image to be a false fingerprint.
9. A fingerprint anti-counterfeiting device, characterized in that the device comprises a processor and a memory;
the memory is used for storing program codes and transmitting the program codes to the processor;
the processor is used for executing the fingerprint anti-counterfeiting method according to instructions in the program code, wherein the fingerprint anti-counterfeiting method comprises the steps of 1-5.
CN202010136601.0A 2020-03-02 2020-03-02 Fingerprint anti-counterfeiting method, device and equipment Active CN111368715B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010136601.0A CN111368715B (en) 2020-03-02 2020-03-02 Fingerprint anti-counterfeiting method, device and equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010136601.0A CN111368715B (en) 2020-03-02 2020-03-02 Fingerprint anti-counterfeiting method, device and equipment

Publications (2)

Publication Number Publication Date
CN111368715A CN111368715A (en) 2020-07-03
CN111368715B true CN111368715B (en) 2023-04-07

Family

ID=71208505

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010136601.0A Active CN111368715B (en) 2020-03-02 2020-03-02 Fingerprint anti-counterfeiting method, device and equipment

Country Status (1)

Country Link
CN (1) CN111368715B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20160037305A (en) * 2014-09-26 2016-04-06 창신정보통신(주) Method for User Authentication using Fingerprint Recognition
CN107077615A (en) * 2017-01-12 2017-08-18 厦门中控生物识别信息技术有限公司 Fingerprint method for anti-counterfeit and equipment
CN108932507A (en) * 2018-08-06 2018-12-04 深圳大学 A kind of automatic anti-fake method and system based on OCT fingerprint image
CN109063572A (en) * 2018-07-04 2018-12-21 南京信息工程大学 It is a kind of based on multiple dimensioned and multireel lamination Fusion Features fingerprint activity test methods

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20160037305A (en) * 2014-09-26 2016-04-06 창신정보통신(주) Method for User Authentication using Fingerprint Recognition
CN107077615A (en) * 2017-01-12 2017-08-18 厦门中控生物识别信息技术有限公司 Fingerprint method for anti-counterfeit and equipment
CN109063572A (en) * 2018-07-04 2018-12-21 南京信息工程大学 It is a kind of based on multiple dimensioned and multireel lamination Fusion Features fingerprint activity test methods
CN108932507A (en) * 2018-08-06 2018-12-04 深圳大学 A kind of automatic anti-fake method and system based on OCT fingerprint image

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
Fingerprint ROI segmentation based on deep learning;Branka Stojanovic et al.;《24th Telecommunications forum TELFOR 2016》;20161123;全文 *
基于韦伯二值感知特征的指纹活性检测;吕锐 等;《应用科学学报》;20160930;第34卷(第5期);全文 *
轻量化多特征融合的指纹分类算法研究;甘俊英等;《信号处理》;20190525;第35卷(第05期);全文 *

Also Published As

Publication number Publication date
CN111368715A (en) 2020-07-03

Similar Documents

Publication Publication Date Title
Tudor Ionescu et al. How hard can it be? Estimating the difficulty of visual search in an image
WO2021212659A1 (en) Video data processing method and apparatus, and computer device and storage medium
KR102554724B1 (en) Method for identifying an object in an image and mobile device for practicing the method
CN112801057B (en) Image processing method, image processing device, computer equipment and storage medium
CN104580143A (en) Security authentication method based on gesture recognition, terminal, server and system
Yang et al. Personal authentication using finger vein pattern and finger-dorsa texture fusion
CN113408574B (en) License plate classification method, license plate classification device and computer readable storage medium
CN111914665A (en) Face shielding detection method, device, equipment and storage medium
CN110555439A (en) identification recognition method, training method and device of model thereof and electronic system
CN112966574A (en) Human body three-dimensional key point prediction method and device and electronic equipment
CN111444817B (en) Character image recognition method and device, electronic equipment and storage medium
CN113191216A (en) Multi-person real-time action recognition method and system based on gesture recognition and C3D network
CN110348366B (en) Automatic optimal face searching method and device
CN113936309A (en) Facial block-based expression recognition method
CN114359962A (en) Posture recognition method and device and readable storage medium
CN107480628B (en) Face recognition method and device
CN113673308A (en) Object identification method, device and electronic system
CN111368715B (en) Fingerprint anti-counterfeiting method, device and equipment
CN113011307A (en) Face recognition identity authentication method based on deep residual error network
CN112749605A (en) Identity recognition method, system and equipment
KR100696251B1 (en) Method and apparatus for setting of comparison area and generating of user authentication information for iris recognition
CN112101479B (en) Hair style identification method and device
CN110852203B (en) Multi-factor suspicious person identification method based on video feature learning
EP3702958A1 (en) Method for verifying the identity of a user by identifying an object within an image that has a biometric characteristic of the user and separating a portion of the image comprising the biometric characteristic from other portions of the image
CN111382703A (en) Finger vein identification method based on secondary screening and score fusion

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
CB02 Change of applicant information
CB02 Change of applicant information

Address after: Room 1301, No.132, Fengqi Road, phase III, software park, Xiamen City, Fujian Province

Applicant after: Xiamen Entropy Technology Co.,Ltd.

Address before: 361000, Xiamen three software park, Fujian Province, 8 North Street, room 2001

Applicant before: XIAMEN ZKTECO INFORMATION TECHNOLOGY Co.,Ltd.

GR01 Patent grant
GR01 Patent grant