CN114863130B - Finger vein biological feature identification method, system and matching and identification method - Google Patents
Finger vein biological feature identification method, system and matching and identification method Download PDFInfo
- Publication number
- CN114863130B CN114863130B CN202210533024.8A CN202210533024A CN114863130B CN 114863130 B CN114863130 B CN 114863130B CN 202210533024 A CN202210533024 A CN 202210533024A CN 114863130 B CN114863130 B CN 114863130B
- Authority
- CN
- China
- Prior art keywords
- finger vein
- image
- gabor
- formula
- cgdbp
- 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
Links
- 210000003462 vein Anatomy 0.000 title claims abstract description 197
- 238000000034 method Methods 0.000 title claims abstract description 78
- 230000004044 response Effects 0.000 claims abstract description 31
- 238000001914 filtration Methods 0.000 claims abstract description 16
- 230000000903 blocking effect Effects 0.000 claims abstract description 6
- 238000012360 testing method Methods 0.000 claims description 18
- 230000008569 process Effects 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 13
- 239000000203 mixture Substances 0.000 claims description 10
- 238000010586 diagram Methods 0.000 claims description 9
- 238000007781 pre-processing Methods 0.000 claims description 9
- 125000004122 cyclic group Chemical group 0.000 claims description 8
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 claims description 5
- 230000002860 competitive effect Effects 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 4
- 230000004931 aggregating effect Effects 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 abstract description 22
- 238000013519 translation Methods 0.000 abstract description 15
- 238000005286 illumination Methods 0.000 abstract description 10
- 238000012545 processing Methods 0.000 abstract description 4
- 238000000605 extraction Methods 0.000 description 16
- 230000006870 function Effects 0.000 description 11
- 230000000694 effects Effects 0.000 description 10
- 238000013527 convolutional neural network Methods 0.000 description 6
- 230000008859 change Effects 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 5
- 238000009826 distribution Methods 0.000 description 5
- 239000000284 extract Substances 0.000 description 4
- 238000005452 bending Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000002679 ablation Methods 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000036961 partial effect Effects 0.000 description 1
- 238000002203 pretreatment Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
- G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
- G06V10/451—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters with interaction between the filter responses, e.g. cortical complex cells
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/467—Encoded features or binary features, e.g. local binary patterns [LBP]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/54—Extraction of image or video features relating to texture
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
- G06V10/7515—Shifting the patterns to accommodate for positional errors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/761—Proximity, similarity or dissimilarity measures
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Biodiversity & Conservation Biology (AREA)
- Vascular Medicine (AREA)
- Human Computer Interaction (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
- Collating Specific Patterns (AREA)
Abstract
The invention belongs to the field of medical image processing, and particularly relates to a finger vein biological feature recognition method, a finger vein biological feature recognition system and a finger vein biological feature matching and recognition method. The invention adopts a competition Gabor direction binary statistical characteristic histogram for extracting the finger vein structural characteristics with discrimination. Firstly, a multidirectional Gabor filter is utilized, and a maximum filter response value index is obtained as a dominant direction to obtain a rotation invariant feature. And secondly, comparing the sequence difference relation of adjacent three directions according to the filtering value of each pixel point on the finger vein image, and constructing a Competition Gabor Direction Binary Pattern (CGDBP) with high discrimination. Finally, the finger vein CGDBP characteristics are extracted in a blocking mode, discrete characteristic codes are aggregated into a histogram representation, a combined characteristic histogram HCGDBS is constructed, and translation of an image is overcome. A large number of experiments are carried out on four widely used finger vein databases, and the results show that the method can effectively improve the finger vein recognition performance and is more robust to illumination, translation, noise and small-range rotation.
Description
Technical Field
The invention belongs to the field of medical image processing, and particularly relates to an anti-interference finger vein biological feature identification method and system, and a corresponding matching and identification method.
Background
The finger vein belongs to the internal characteristics, is difficult to forge and has high safety. Compared with other biological characteristics, such as fingerprints, palmprints, faces and the like, the vein characteristics are living body identification and are not easy to abrade, so that the finger vein identification becomes a current research hot spot.
The process of finger vein identification typically involves 3 steps. (1) Preprocessing, including extraction of a region of interest (ROI), image enhancement; (2) extracting features, namely extracting features of finger veins; (3) And matching and identifying, namely matching the feature vector of the test sample with the feature vector of the training sample, and classifying and identifying the features. Among these, finger vein feature extraction is the most important step, which significantly affects the performance of recognition.
The traditional method for extracting the vein texture change and the direction characteristics is easy to be influenced by illumination, translation, noise and rotation, and is difficult to directly extract reliable structural characteristics from an original image.
Disclosure of Invention
Aiming at the technical problems that the prior method for taking the vein texture change and direction characteristics is easy to be influenced by illumination, translation, noise and rotation, and reliable structural characteristics are difficult to be directly extracted from an original image, the invention provides an anti-interference vein biological characteristic identification method and system, and a corresponding matching and identification method.
The invention is realized by adopting the following technical scheme: an anti-interference finger vein biological feature identification method comprises the following steps:
step one, extracting the dominant direction of each pixel point on a finger vein image based on a multidirectional Gabor filter;
step two, referring to the dominant direction, sequentially encoding the sequential intensity difference relation of each pixel point in adjacent directions, and constructing a competition Gabor direction binary pattern CGDBP;
and thirdly, extracting CGDBP characteristics on the finger vein image in a blocking way by utilizing the competitive Gabor direction binary pattern CGDBP, aggregating discrete CGDBP characteristic codes into a histogram representation, and constructing a combined characteristic histogram HCGDBS.
As a further improvement of the above solution, in step one, the model of the Gabor filter is:
wherein x and y respectively refer to two-dimensional coordinate values of vein images, θ is Gabor function direction in radian units, μ is radial frequency in radian units, σ is Gaussian function standard deviation,
further, the Gabor filter adopts a linear filter with adjustable direction.
Further, the finger vein image is filtered using eight Gabor filters of direction j pi/8 for the finger vein image, j=0, 1,..7, assuming G j Is the real part of the function when the Gabor function direction θ is jpi/8, and the convolution of the real part and the finger vein image is as follows:
g j (x,y)＝G j *I(x,y) (3)
wherein g j (x, y) refers to venous images I (x, y) and G j Is a convolution operation, I (x, y) is an image at the (x, y) position in the finger vein image;
and extracting the dominant direction in the direction with the maximum filter response value on the finger vein image.
Further, in the second step, the first step,
(a) Preprocessing the finger vein image;
(b) Eight filter response values of a certain pixel point on the finger vein image and a Gabor filter are obtained,
(c) Circularly shifting the filtering results in eight directions according to the maximum value of the filtering response;
(d) Grouping, namely dividing filtering values of each direction and two adjacent directions into a group;
(e) Sorting the filter values in each group;
(f) Calculating the sequential intensity difference relation between each direction of each pixel point and the adjacent direction;
(g) And calculating the CGDBP value corresponding to a certain pixel point of the finger vein image.
Still further, the dominant direction is taken as an index value C (x, y) corresponding to the maximum filter response:
second, the rotation sequence is cycled until the point indexed by C (x, y) is at the first position:
g’ 0 (x,y),g’ 1 (x,y),...,g’ t (x,y),...,g’ 7 (x,y)：＝g C (x,y),...,g 7 (x,y),g 0 (x,y),...,g C-1 (x,y) (5)
wherein the symbol ": = "execute element assignment;
then, based on the filter response values of the Gabor filter in eight directions, the points in the cyclic sequence are uniformly distributed on the circle,
there are therefore 8 packets, namely:
wherein,is a vector whose element composition is the filtered value of the pixel point in the j direction,and phi (j) is the index of the right and left adjacent directions of j, N θ ＝8；
Next, the three filtered values in each group are ordered to obtain:
wherein,is a vector whose element composition is +.>Descending order arrangement;
finally, coding according to the relation of the intensity sequence difference among the filtering values in three directions in each group according to a formula (10) to form a competing Gabor direction binary mode CGDBP:
as a further improvement of the above solution, in step three, the finger vein image is divided into a plurality of sub-images, and a sub-histogram is constructed from each sub-image; the constructed sub-histograms are then normalized and concatenated together.
Further, first, the finger vein image is divided into small cells; then, each cell is formed into a block, CGDBP of all cells in the block is calculated, and the size of the cell is used as a moving step.
The invention also provides an anti-interference finger vein biological feature recognition system, which applies the random anti-interference finger vein biological feature recognition method, and the working process of the finger vein biological feature recognition system is as follows:
1. receiving finger vein images, wherein the size of the finger vein images is H multiplied by W, H is high, and W is wide;
2. constructing eight Gabor filters according to formula (2), containing eight directions;
wherein, formula (2) is:
x and y respectively refer to two-dimensional coordinate values of the vein image, θ is the Gabor function direction in radian units, μ is the radial frequency in radian units, σ is the standard deviation of the Gaussian function,
3. linear filtering is carried out on the finger vein image according to the formula (3), the formula (4) and the formula (5), so as to obtain filter response values in eight directions and a cyclic sampling sequence in a reference dominant direction;
wherein, formula (3) is: g j (x,y)＝G j *I(x,y)
Wherein g j (x, y) refers to venous images I (x, y) and G j Filter response value G of (2) j Is the real part of the function when the Gabor function direction θ is jpi/8, is the convolution operation, and I (x, y) is the image at the (x, y) position in the finger vein image;
the formula (4) is:
c (x, y) is the index value corresponding to the maximum filter response;
the formula (5) is:
g’ 0 (x,y),g’ 1 (x,y),...,g’ t (x,y),...,g’ 7 (x,y)：＝g C (x,y),...,g 7 (x,y),g 0 (x,y),...,g C-1 (x,y)
sign ": = "execute element assignment;
4. based on the cyclic sampling sequence, obtaining a characteristic diagram T according to a formula (6) -a formula (11), wherein the dimension of the characteristic diagram T is H multiplied by W;
wherein, formula (6) is:
the formula (7) is:
equation (8) is: phi (j) =mod (mod (j, N) θ )+1,N θ )
Equation (9) is:
the formula (10) is:
the formula (11) is:
is a vector whose element composition is the filtered value of the pixel point in the j direction,/->And phi (j) is the index of the right and left adjacent directions of j, N θ ＝8；/>Is a vector whose element composition is +.>Descending order arrangement;
5. dividing a characteristic image T into small cells, and then forming a block by every few cells, and marking the block as n image blocks;
6. constructing competition for each area separatelyGabor direction binary statistics feature
7. All V are sequentially arranged i String into feature vector v= [ V 1 ,V 2 ,...,V n ]And obtaining the binary statistical feature vector V of the competing Gabor direction.
The invention also provides a matching and identifying method for measuring the similarity between two finger vein images, which is characterized in that the similarity is measured by image features, and the image features of one finger vein image are obtained by adopting the random anti-interference finger vein biological feature identifying method, and the matching and identifying method comprises the following steps:
similarity between image features of two finger vein images is measured using normalized correlation coefficient NCC, a is a feature vector corresponding to a training set finger vein, B is a feature vector corresponding to a test set finger vein, a= (a) 1 ,a 2 ,...,a n ),B＝(b 1 ,b 2 ,...,b n )。
Wherein mu A (μ B ) Is the mean value of the feature vector A (B), sigma A (σ B ) Is the standard deviation of A (B), l is the length value of A or B, and the value of NCC is between-1 and 1;
if the NCC is close to 1, this means that the two finger vein images may be identical; otherwise, they are considered to be different.
The competitive Gabor direction binary statistical feature histogram of the present invention is an effective image representation. The discrimination characteristic of each pixel point of the finger vein image is obtained by calculating the sequential filtering difference relation of adjacent three directions under the multidirectional Gabor filter by referring to the dominant direction. The HCGDBS is untrained, is well designed, robust to illumination, translation, noise and rotation, and verifies the effectiveness of the proposed HCGDBS method using four widely used finger vein image databases.
Drawings
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention. In the drawings:
FIG. 1 is a flow chart of a method for identifying biological characteristics of a finger vein according to the present invention.
Fig. 2 is a schematic diagram of a calculation process of the dominant direction of the finger vein in fig. 1.
Fig. 3 is a schematic diagram of the formation process of CGDBP in fig. 1.
Fig. 4 is a schematic diagram of one of the image pre-processes of finger vein database (PolyU) at hong Kong university.
FIG. 5 is a schematic representation of one of the image pre-processing of the finger vein database (SDUMLA-FV) of the Shandong university MLA laboratory.
FIG. 6 is a schematic representation of one of the image pretreatments of the finger vein library (FV-USM) of the university of Malaysia.
FIG. 7 is a schematic diagram of the pretreatment of one of the images of the Tianjin intelligent signal and image processing emphasis laboratory finger vein library (FV-TJ).
Fig. 8 is a graph of the ROC database of the finger vein of the present invention.
Fig. 9 is a graph of finger vein database matching scores of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
The finger vein biological feature recognition method provided by the invention is used for extracting finger vein structural features with discrimination by the competition Gabor direction binary statistical feature Histogram (HCGDBS). HCGDBS is a histogram representation that explicitly encodes structural information in an image in feature space. Firstly, a multidirectional Gabor filter is utilized, and a maximum filter response value index is obtained as a dominant direction to obtain a rotation invariant feature. And secondly, comparing the sequence difference relation of adjacent three directions according to the filtering value of each pixel point on the finger vein image, and constructing a Competition Gabor Direction Binary Pattern (CGDBP) with high discrimination. Finally, the finger vein CGDBP characteristics are extracted in a blocking mode, discrete characteristic codes are aggregated into a histogram representation, a combined characteristic histogram HCGDBS is constructed, and translation of an image is overcome. A large number of experiments are carried out on four widely used finger vein databases, and the results show that the method can effectively improve the finger vein recognition performance and is more robust to illumination, translation, noise and small-range rotation.
Example 1
Referring to fig. 1, the finger vein biometric feature recognition method of the present invention mainly includes the following steps:
step one, extracting the dominant direction of each pixel point on a finger vein image based on a multidirectional Gabor filter;
step two, referring to the dominant direction, sequentially encoding the sequential intensity difference relation of each pixel point in adjacent directions, and constructing a competition Gabor direction binary pattern CGDBP;
and thirdly, extracting CGDBP characteristics on the finger vein image in a blocking way by utilizing the competitive Gabor direction binary pattern CGDBP, aggregating discrete CGDBP characteristic codes into a histogram representation, and constructing a combined characteristic histogram HCGDBS.
The finger vein biological feature recognition method performs image recognition on the finger vein structure information by explicitly encoding and extracting high-discrimination features. It is proposed to extract the dominant direction of each pixel point of the finger vein based on a multidirectional Gabor filter. And (3) sequentially encoding the adjacent three-direction sequential intensity difference relation of each pixel point by referring to the dominant direction, and constructing a Competition Gabor Direction Binary Pattern (CGDBP). And then, extracting finger vein image features by blocks to combine into a histogram, so as to form a competition Gabor direction binary statistical feature Histogram (HCGDBS). For the test samples, classification and identification are performed by using a Normalized Correlation Coefficient (NCC) according to the one-dimensional finger vein characteristics.
In step one, the design of the Gabor filter is critical. The Gabor filter is used as a linear filter with adjustable direction, the frequency and the direction expression of the Gabor filter are similar to those of a human visual system, and the Gabor filter can well approximate the receptive field function of single cells. Furthermore, gabor filters have good textured two-dimensional spectral characteristics, and the texture varies with the two-dimensional spatial position. Because of these excellent characteristics, gabor filters are widely used for finger vein feature recognition because they enable images to be well characterized. A typical two-dimensional Gabor filter has the following general form:
where μ is the radial frequency in radians, θ is the Gabor function direction in radians, σ is the standard deviation of the Gaussian function,similar to the competing encoding method, the real part of the Gabor filter is used to extract the features of the finger veins. To extract features more accurately, the real part of the Gabor filter is converted into an "inverted" form, so the largest response corresponds to the smallest convolution value. The transformed Gabor filter is defined as:
wherein,N θ ＝8。
the dominant direction in step one is the finger vein dominant direction. Because of the influence of factors such as acquisition environment, illumination change, finger gesture and the like, irrelevant information such as rotation, translation, noise and the like exists in the captured finger vein image, and the recognition performance is influenced. Here, we use Gabor filter to filter the vein image to get the dominant direction of the pixel point, so as to overcome the rotation of the image. First, eight Gabor filters of direction j pi/8 (j=0, 1,..7) are used to filter the finger vein image for the preprocessed image. Suppose G j Is the real part of G in the direction j pi/8 (j=0, 1,., 7), which convolves with the finger vein image is:
g j (x,y)＝G j *I(x,y) (3)
where I refers to venous images, "x" is the convolution operation, g j (x, y) refers to venous images I (x, y) and G j (x, y) represents the position of the pixel in I.
The determination of the dominant direction of the finger veins is critical to our encoding. The Gabor filter is less likely to produce a larger filter response, and it is more discriminative to extract its maximum filter response value index as the dominant direction. By using eight Gabor filters in different directions to convolve the finger vein image, the direction with the largest filter response value is extracted as the dominant direction, and the maximum filter response is used to improve the recognition capability of the computing features and the robustness to noise and rotation.
Wherein C (x, y) is the dominant direction at the pixel point (x, y), which is the index value corresponding to the maximum filter response.
After obtaining the dominant direction, we loop through the rotation sequence until the point indexed by C (x, y) is at the first position:
g’ 0 (x,y),g’ 1 (x,y),...,g’ t (x,y),...,g’ 7 (x,y)：＝g C (x,y),...,g 7 (x,y),g 0 (x,y),...,g C-1 (x,y) (5)
wherein the symbol ": = "perform element assignment (as shown in fig. 2, a schematic diagram of a process of extracting dominant directions of a certain pixel point on a finger vein).
In step two, local Binary Pattern (LBP) and its various variants have been shown to extract the discriminatory power of venous features. However, most of these LBP methods focus on the local intensity differences between the encoding center pixel and its neighbors, and do not fully consider the local intensity order relationship between the pixel point neighboring directions. In order to solve the problem, the invention provides a method for acquiring the dominant direction based on Gabor filtering, then encoding the intensity sequence difference relation of adjacent three directions of each pixel point on the finger vein image, and improving the discrimination of the extracted features.
Referring to fig. 3, we uniformly distribute the points in the cyclic sequence on a circle based on the filter response values of Gabor filtering in eight directions. The cyclic ordering thereof (see formula (5)) according to the dominant direction mentioned above, eight directions are used in the present invention, so there are 8 packets, namely:
φ(j)＝mod(mod(j,N θ )+1,N θ ) (8)
wherein,is a vector whose element composition is the filtered value of the pixel point in the j direction,/->And phi (j) is the index of the right and left adjacent directions of j.
Next, the three filtered values in each group are ordered to obtain:
wherein,is a vector whose element composition is +.>And (5) arranging in a descending order.
Finally, we encode according to equation (10) based on the intensity order difference relationship between the filtered values in three directions in each group:
the method comprises the steps of (a) preprocessing an ROI image of a finger vein, (b) convoluting eight filter response values of a certain pixel point on the finger vein image and a Gabor filter, (c) circularly shifting filter results in eight directions according to a maximum value of the filter response, (d) grouping the filter values in each direction and two adjacent directions into a group, (e) sorting the filter values in each group, (f) calculating a sequential intensity difference relation between each direction of each pixel point and the adjacent direction according to a formula in the figure, and (g) calculating a CGDBP value corresponding to the certain pixel point of the finger vein image according to a formula (10).
In order to improve the recognition capability, the histogram-based description method can well overcome the translation existing in the image to acquire the discrimination characteristics. A feature image is divided into a plurality of sub-images, and a sub-histogram is constructed from each sub-image. The constructed sub-histograms are then normalized and concatenated together. First, the feature image is divided into small cells (m×n), then, each cell is formed into a block (k×l), and CGDBP of all cells in a block is calculated, and the size of the cell is used as a moving step. As can be seen from the above equation (10), the CGDBP has a value range of [0,255], and 4 is used as a moving step within [0,255], so that the feature dimension of each block can be reduced from 255 to 64. HCGDBS feature dimension of one finger vein image (H W) is
Length(HCGDBS)＝((H-K)/M+1)*((W-L)/N+1)*64 (12)
The HCGDBS proposed by the present invention has the following properties. First, the discrimination features are robust to illumination changes by referencing dominant directions to finger vein image rotation and then calculating CGDBP using local sequential difference relationships. And secondly, the HCGDBS histogram extracts the characteristics of the finger vein image through blocking, so that local detail information is obtained, the translation of the image is overcome, and the finger vein recognition performance is improved. Third, HCGDBS enriches local features including local dominant direction, local neighbor three directions and their sequential difference relationships. None of these were explored in Gabor and LBP.
When the finger vein biological characteristic recognition method is applied, the finger vein biological characteristic recognition system can be designed into a corresponding finger vein biological characteristic recognition system, such as an APP software installation package, and the working process of the finger vein biological characteristic recognition system is shown in table 1 no matter what way is adopted.
Table 1 shows the detailed procedure for HCGDBS feature extraction
In the identification experiment, the matching and identification method adopted by the invention is as follows: the similarity between two finger vein images was measured using a Normalized Correlation Coefficient (NCC). A is a feature vector corresponding to the finger vein of the training set, B is a feature vector corresponding to the finger vein of the test set, a= (a) 1 ,a 2 ,...,a n ),B＝(b 1 ,b 2 ,...,b n )。
Wherein mu A (μ B ) Is the mean value of the feature vector A (B), sigma A (σ B ) Is the standard deviation of A (B), l is the length value of A or B, and the value of NCC is between-1 and 1. If the NCC is close to 1, this means that the two finger vein images may be identical; otherwise, they are considered to be different.
Class labels for finger vein images are known in authentication experiments. We evaluate experimental performance by performing true-false matching analysis on four different finger vein databases.
Example 2
Example 2 was experimentally verified for the method of example 1.
To assess the effectiveness of the proposed method, we performed a number of experiments on four public finger vein databases, including the PolyU, SDUMLA-FV, FV-USM and FV-TJ finger vein databases.
A. Finger vein database
Finger vein database (PolyU) at university of hong Kong, finger vein images were collected in duplicate in PolyU dataset with a minimum interval of one month and a maximum interval of more than six months and an average interval of 66.8 days. In our experiments, a first stage database was used, with a total of 156 subjects providing images of the finger veins, 2 fingers per person, and 6 image samples per finger. Since there are differences between different fingers in the same individual, different finger vein images from the same individual belong to different categories, namely 312 (156 subjects×2 fingers) categories, each category having 6 samples. The image is subjected to a series of preprocessing operations, the size of the image being 150 x 96. In our experiments, samples 1,3,5 were selected for training and samples 2,4,6 were used for testing.
To assess the effectiveness of the proposed method, we performed a number of experiments on four public finger vein databases, including the PolyU, SDUMLA-FV, FV-USM and FV-TJ finger vein databases.
B. Finger vein database
As shown in fig. 4, finger vein database (PolyU) of university of hong kong, in the PolyU dataset, finger vein images were collected in two times, with a minimum interval of one month and a maximum interval of more than six months, and an average interval of 66.8 days. In our experiments, a first stage database was used, with a total of 156 subjects providing images of the finger veins, 2 fingers per person, and 6 image samples per finger. Since there are differences between different fingers in the same individual, different finger vein images from the same individual belong to different categories, namely 312 (156 subjects×2 fingers) categories, each category having 6 samples. The image is subjected to a series of preprocessing operations, the size of the image being 150 x 96. In our experiments, samples 1,3,5 were selected for training and samples 2,4,6 were used for testing.
As shown in FIG. 5, the finger vein database (SDUMLA-FV) of the Shandong university MLA laboratory. The SDUMLA-FV database contains images of finger veins obtained from 106 individuals, each providing 6 fingers, each providing 6 image samples. Since finger veins differ between different fingers in the same individual, different finger vein images from the same individual belong to different categories, i.e., 636 (106 subjects×6 fingers) category, 6 samples each. The image is subjected to a series of preprocessing operations, the size of the image being 150 x 96. In our experiments, samples 1,3,5 were selected for training and samples 2,4,6 were used for testing.
As shown in FIG. 6, finger vein library (FV-USM) of university of Malaysia. The FV-USM database contains 4 different fingers provided per person from 123 different subjects, each finger capturing 6 images, since different fingers of the same individual represent different categories, namely 492 (123 subjects x 4 fingers) categories, 6 samples each. The size of the image was 300 x 100, in our experiments samples 1,3,5 were chosen for training and samples 2,4,6 were used for testing.
As shown in FIG. 7, tianjin smart signal and image processing focused laboratory finger vein library (FV-TJ). The FV-TJ database contains images of the finger veins obtained from 64 different fingers, 15 images are acquired for each finger, 64 classes are used, and the sample images in the database are all preprocessed, and the image size is 172×76. In our experiment, the first 5 samples were selected for training and the last 10 samples were used for testing.
C. Parameter selection
In this experiment, we studied the performance of the finger vein recognition method proposed by the present invention at different parameter values, including frequency u and standard deviation σ. The binary statistical characteristic histogram of the competing Gabor direction can achieve the optimal recognition effect in the finger vein recognition method, and the influence of different u and sigma on the recognition effect is analyzed on the PolyU, SDUMLA-FV, FV-USM and FV-TJ databases respectively. Performance was quantified using two indicators, RR and EER. RR refers to the recognition rate of a dataset in the present invention. EER is an abbreviation for equal error rate, which relates to False Acceptance Rate (FAR) and False Rejection Rate (FRR). Eqn. (14) and eqn. (15) show how it calculates FAR and FRR.
When far=frr, the error rate is EER.
First, we evaluate the effect of frequency u in the Gabor filter on the recognition performance and select the best value for the subsequent experiments. The frequency u varies between 0.0085 and 0.0435 with a spacing of 0.005. It can be observed from the table that when u=0.0385, the experimental effect was optimal on the PolyU database and the experimental performance was also the best. When the frequency u parameter is fixed, it can be seen that the influence of the standard deviation sigma on the experimental result is obvious, and by comparing the following eight groups of experiments, when the standard deviation sigma= 12.2998, the identification rate of the PolyU database reaches 99.89%, and the error rate reaches 0.5342%. The SDUMLA-FV database has a recognition rate of 99.11% and an error rate of 0.8386% when u=0.0385, sigma= 13.2998. The SDUMLA-FV database is not particularly sharp in image compared to the PolyU database, so the recognition rate is reduced. The FV-USM database is less affected by parameters, the recognition rate is 99.73%, and the error rate is 0.2710%. The finger vein images in the FV-TJ database are clear, the recognition rate can reach 100%, and the error rate reaches 0%.
TABLE 2 results of the test for parameter u on the venous store
Table 3 test results for parameter σ when u=0.0385 on venous bank
D. Finger vein identification
Finger vein identification is a one-to-many matching process that determines a test finger vein image category label. Typically, a set of finger vein images with known class labels is selected as the training samples, with which the test sample will be compared, and the class label of the training sample with the greatest similarity to the test sample is considered the class label of the test sample.
In the following identification experiments, the settings of the database training samples and the test samples are set forth in the database description. To verify the effectiveness of the proposed method, we compared the spatial domain-based and frequency domain-based finger vein feature extraction methods, such as LBP, global WLD, block WLD, HOG, DBC, DBD, DCGWLD, and RDGT. In addition, we also compared three deep learning-based finger vein recognition methods, including FCN & CNN, DNN, and Gabor & CNN.
Table 4 summarizes the results of comparisons of finger vein recognition performed on PolyU, SDUMLA-FV, FV-USM and FV-TJ databases. The finger vein feature extraction method based on airspace comprises the following steps: LBP, HOG are widely used local descriptors, simple to calculate and robust to light, but susceptible to noise. Global WLD and partitioned WLD are susceptible to illumination, translation, rotation, etc., so their recognition performance is not as good as partitioned WLD. The DBC and the DBD extract local features through a feature learning method, and compared with LBP, WLD and HOG, the performance is obviously improved, but the DBC and the DBD not only comprise the process of extracting the features but also feature learning, and are relatively inferior to the traditional manual local descriptors in time. Compared with other local descriptors, the DCGWLD has obvious improvement on the effect by considering the bending degree of veins, but the acquired vein image is easily influenced by rotation, and the single consideration of the bending degree of veins is insufficient to reflect the discrimination information, so that the optimal effect is not achieved. The finger vein feature extraction method based on the frequency domain comprises the following steps: RDGT is a method of transforming it into space-frequency domain by discrete Gabor transformation, and its feature dimension is low, the calculation amount is small, but the identification effect is not obvious. The finger vein feature extraction method based on deep learning comprises the following steps: FCN & CNN is based on CNN learning finger vein features, FCN recovery vein image solves the loss caused by preprocessing and feature extraction. DNN automatically learns features from an image and directly learns robust feature representations from original pixels of the image independent of manually extracted features. Gabor & CNN self-adaptively learns parameters of the Gabor filter through a convolutional neural network, so that the problem that the parameters of the Gabor filter are difficult to adjust is solved.
Table 4 shows a comparison of different methods on a venous database
From table 4, it can be observed that the proposed HCGDBS is superior to the listed finger vein feature extraction method based on spatial, frequency and deep learning. Therefore, the method not only overcomes the influence of illumination, translation, rotation and the like of the image, but also considers the sequential difference relation of Gabor filter response values under adjacent three directions to obtain a high-discrimination local feature histogram. Compared with other finger vein feature extraction methods, the HCGDBS proposed on the PolyU database has the average error rate reduced by 1.9883 percent compared with the average error rate of other methods. Because the sample image of the PolyU database has translation and rotation, the HCGDBS provided by the invention can well solve the problem, so that the experiment has a good effect. For the SDUMLA-FV database, the method is improved compared with the finger vein feature extraction method based on frequency domain and deep learning, but is improved by 0.0524% compared with the DCGWLD method in the space domain. Because the image of the SDUMLA-FV database is blurred and has serious deformation, the variable curvature Gabor filter of the DCGWLD structure takes the bending degree of the finger vein into consideration, and enriches the line characteristics of the finger vein image. In later experiments, we consider the degree of curvature of the vein and the convexity of the vein curve, thereby extracting highly discriminative structural features of the vein. Compared with a space domain feature extraction method, a frequency domain feature extraction method and the like, the FV-USM has the advantages that the error rate is reduced by 1.3167% and 0.2033%, and compared with three deep learning methods, the FV-USM has the advantage that the error rate is reduced by 0.6590%. The method provided by the invention achieves a better experimental effect in FV-USM. The FV-TJ database has clear images, the extracted features have high identification rate, compared with other methods compared with the FV-TJ database, the FV-TJ database has obvious experimental effect, the identification rate reaches 100%, and the error rate is reduced to 0%. In summary, on the PolyU, SDUMLA-FV, FV-TJ and FV-USM databases, the method provided by the invention has strong robustness to noise of images, has good robustness to images with translation and rotation, can well overcome differences among similar samples, and extracts high-discrimination features. Fig. 8 shows ROC curves for each method based on manual features. The result shows that the method has better performance on finger vein image feature extraction.
Ablation experiments of HCGDBS
To further elucidate the effectiveness of HCGDBS, the proposed CGDBP was verified to have a high discrimination, we performed the following experiments: (1) Sequentially encoding eight-direction response values after determining the dominant direction according to the magnitude relation of two adjacent directions to obtain an eight-bit binary sequence which is recorded as CGDBP 1 . (2) Sequentially encoding eight-direction response values after determining the dominant direction according to the magnitude relation of adjacent three-direction difference values to obtain an eight-bit binary sequence, and marking the eight-bit binary sequence as CGDBP 2 . (3) Sequentially encoding eight-direction response values after determining the dominant direction according to the magnitude relation of adjacent three-direction descending order difference values to obtain an eight-bit binary sequence, and marking the eight-bit binary sequence as CGDBP 3 (HCGDBS). In the three above representations, three local binary statistical feature histograms are formed. In this study, we performed experiments in four different databases described above in order to better verify the effectiveness of the proposed method.
From the comparison, we can derive the following observations. CGDBP on four finger vein databases 3 (HCGDBS) has better performance than CGDBP 1 And CGDBP 2 . This result shows that HCGDBS improves the discrimination of extracted features by encoding the magnitude relationship of adjacent three-way sequential differences. CGDBP on PolyU, FV-USM and FV-TJ databases 2 Ratio CGDBP 1 Has improved identification rate of SDUMLA-FV numberCGDBP on database 2 Ratio CGDBP 1 The error rate of (a) is reduced by 0.1612%, which means that the extraction of the three-direction difference features has more discrimination than the extraction of the adjacent two-direction features.
Table 5 three local descriptor comparisons
F. Finger vein authentication
In finger vein authentication, class labels of finger vein images are known. Each finger vein image is matched to all other finger vein images in the same database. If both finger vein images are from the same finger, then the match is referred to as a true match; otherwise, the match is considered a false match. That is, matching between similar samples is called true matching (genuine match), and matching between dissimilar samples is called false matching (imposter match). For the PolyU finger vein database, there were 1872 samples, 936 (312×3) training samples, and the rest were test samples. Thus, the total number of matches is 876096, with 2808 being true matches and 873288 being false matches. The SDUMLA-FV database contains 3640464 matches, including 5724 true matches and 3634740 false matches, respectively. In the FV-USM finger vein database, the total number of matches is 2178576, and the true and false matches are 4428 and 2174148, respectively. The FV-TJ database contains 204800 matches, including 3200 true matches and 201600 false matches, respectively.
We calculate the true-false matching score at the same time using the proposed method. The quality of the algorithm performance can be qualitatively measured through the distribution condition of the true and false matching scores. FIGS. 9 (a) - (d) show the distribution of true and false matching scores over PolyU, SDUMLA-FV, FV-USM and FV-TJ databases, respectively. From these distributions we can see that the true and false match scores have a highly independent distribution over the four finger vein databases. Wherein the true match scores are distributed primarily centrally in areas greater than 0.8 and the false match scores are distributed primarily centrally in areas less than 0.8. The HCGDBS features are shown to have a strong discrimination of finger vein images. As can be seen from fig. 9, the distribution of true and false match scores is not as independent on the poly u, SDUMLA-FV and FV-USM databases as on the FV-TJ databases, with partial crossover. The main reason for this is that the images in the three databases have serious problems of deformation, blurring, scale change, etc., which cause the features extracted by the Gabor filter to change. In the next study we considered the use of multi-scale multi-direction and multi-curvature to enhance the robustness of HCGDBS features to shape and scale variations.
In summary, the proposed competitive Gabor direction binary statistical feature histogram is an effective image representation. The discrimination characteristic of each pixel point of the finger vein image is obtained by calculating the sequential filtering difference relation of adjacent three directions under the multidirectional Gabor filter by referring to the dominant direction. The HCGDBS is untrained, is well designed, robust to illumination, translation, noise and rotation, and verifies the effectiveness of the proposed HCGDBS method using four widely used finger vein image databases.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.
Claims (8)
1. An anti-interference finger vein biological feature recognition method is characterized by comprising the following steps:
step one, extracting the dominant direction of each pixel point on a finger vein image based on a multidirectional Gabor filter;
step two, referring to the dominant direction, sequentially encoding the sequential intensity difference relation of each pixel point in adjacent directions, and constructing a competition Gabor direction binary pattern CGDBP; the process is as follows:
(a) Preprocessing the finger vein image;
(b) Eight filter response values of a certain pixel point on the finger vein image and a Gabor filter are obtained,
(c) Circularly shifting the filtering results in eight directions according to the maximum value of the filtering response;
(d) Grouping, namely dividing filtering values of each direction and two adjacent directions into a group;
(e) Sorting the filter values in each group;
(f) Calculating the sequential intensity difference relation between each direction of each pixel point and the adjacent direction;
(g) Calculating a CGDBP value corresponding to a certain pixel point of the finger vein image;
taking the dominant direction as an index value C (x, y) corresponding to the maximum filter response:
wherein x and y refer to two-dimensional coordinate values of the vein image, respectively, and then, cyclically rotating the sequence until a point indexed by C (x, y) is located at a first position:
g’ 0 (x,y),g’ 1 (x,y),...,g’ t (x,y),...,g’ 7 (x,y)：＝g C (x,y),...,g 7 (x,y),g 0 (x,y),...,g C-1 (x,y) (5)
wherein the symbol ": = "execute element assignment;
then, based on the filter response values of the Gabor filter in eight directions, the points in the cyclic sequence are uniformly distributed on the circle,
there are therefore 8 packets, namely:
φ(j)＝mod(mod(j,N θ )+1,N θ ) (8)
wherein,is a vector whose element composition is the filtered value of the pixel point in the j direction,/->And phi (j) is the index of the right and left adjacent directions of j, N θ ＝8；
Next, the three filtered values in each group are ordered to obtain:
wherein,is a vector whose element composition is +.>Descending order arrangement;
finally, coding according to the relation of the intensity sequence difference among the filtering values in three directions in each group according to a formula (10) to form a competing Gabor direction binary mode CGDBP:
and thirdly, extracting CGDBP characteristics on the finger vein image in a blocking way by utilizing the competitive Gabor direction binary pattern CGDBP, aggregating discrete CGDBP characteristic codes into a histogram representation, and constructing a combined characteristic histogram HCGDBS.
2. The method for identifying anti-interference finger vein biometric features according to claim 1, wherein: in the first step, the Gabor filter is modeled as follows:
wherein x and y respectively refer to two-dimensional coordinate values of vein images, θ is Gabor function direction in radian units, μ is radial frequency in radian units, σ is Gaussian function standard deviation,
3. the method for identifying anti-interference finger vein biometric features according to claim 2, wherein: the Gabor filter adopts a linear filter with adjustable direction.
4. The method for identifying anti-interference finger vein biometric features according to claim 2, wherein:
the finger vein image was filtered using eight Gabor filters with a direction of jpi/8, j=0, 1,..7, assuming G j Is the real part of the function when the Gabor function direction θ is jpi/8, and the convolution of the real part and the finger vein image is as follows:
g j (x,y)＝G j *I(x,y) (3)
wherein g j (x, y) refers to venous images I (x, y) and G j Is a convolution operation, I (x, y) is an image at the (x, y) position in the finger vein image;
and extracting the dominant direction in the direction with the maximum filter response value on the finger vein image.
5. The method for identifying anti-interference finger vein biometric features according to claim 1, wherein: in the third step of the process, the process is carried out,
dividing the finger vein image into a plurality of sub-images, and constructing a sub-histogram from each sub-image;
the constructed sub-histograms are then normalized and concatenated together.
6. The method for identifying anti-interference finger vein biometric features according to claim 5, wherein:
firstly, dividing the finger vein image into small cells;
then, each cell is formed into a block, CGDBP of all cells in the block is calculated, and the size of the cell is used as a moving step.
7. An anti-interference finger vein biometric identification system, which applies the anti-interference finger vein biometric identification method according to any one of claims 1 to 6, characterized in that the working process of the finger vein biometric identification system is as follows:
1. receiving finger vein images, wherein the size of the finger vein images is H multiplied by W, H is high, and W is wide;
2. constructing eight Gabor filters according to formula (2), containing eight directions;
wherein, formula (2) is:
x and y respectively refer to two-dimensional coordinate values of the vein image, θ is the Gabor function direction in radian units, μ is the radial frequency in radian units, σ is the standard deviation of the Gaussian function,
3. linear filtering is carried out on the finger vein image according to the formula (3), the formula (4) and the formula (5), so as to obtain filter response values in eight directions and a cyclic sampling sequence in a reference dominant direction;
wherein, formula (3) is: g j (x,y)＝G j *I(x,y)
Wherein g j (x, y) refers to vein image I (x, y)) And G j Filter response value G of (2) j Is the real part of the function when the Gabor function direction θ is jpi/8, is the convolution operation, and I (x, y) is the image at the (x, y) position in the finger vein image;
the formula (4) is:
c (x, y) is the index value corresponding to the maximum filter response;
the formula (5) is:
g’ 0 (x,y),g’ 1 (x,y),...,g’ t (x,y),...,g’ 7 (x,y)：＝g C (x,y),...,g 7 (x,y),g 0 (x,y),...,g C-1 (x,y)
sign ": = "execute element assignment;
4. based on the cyclic sampling sequence, obtaining a characteristic diagram T according to a formula (6) -a formula (11), wherein the dimension of the characteristic diagram T is H multiplied by W;
wherein, formula (6) is:
the formula (7) is:
equation (8) is: phi (j) =mod (mod (j, N) θ )+1,N θ )
Equation (9) is:
the formula (10) is:
the formula (11) is:
is a vector whose element composition is the filtered value of the pixel point in the j direction,/->And phi (j) is the index of the right and left adjacent directions of j, N θ ＝8；/>Is a vector whose element composition is +.>Descending order arrangement;
5. dividing a characteristic image T into small cells, and then forming a block by every few cells, and marking the block as n image blocks;
6. respectively constructing competition Gabor direction binary statistical characteristics V of each area i :(i＝1,2,...,n)；
7. All V are sequentially arranged i String into feature vector v= [ V 1 ,V 2 ,...,V n ]And obtaining the binary statistical feature vector V of the competing Gabor direction.
8. A matching and identifying method for measuring similarity between two finger vein images, wherein the similarity is measured by image features, and wherein the image features of one of the finger vein images are obtained by an anti-interference finger vein biometric identification method according to any one of claims 1 to 6, the matching and identifying method comprising:
the similarity between the image features of the two finger vein images is measured using the normalized correlation coefficient NCC, a being the feature vector corresponding to the finger vein of the training set, B being the feature vector corresponding to the finger vein of the test set,
A＝(a 1 ,a 2 ,...,a n ),B＝(b 1 ,b 2 ,...,b n )；
wherein mu A (μ B ) Is the mean value of the feature vector A (B), sigma A (σ B ) Is the standard deviation of A (B), l is the length value of A or B, and the value of NCC is between-1 and 1;
if the NCC is close to 1, this means that the two finger vein images may be identical; otherwise, they are considered to be different.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210533024.8A CN114863130B (en) | 2022-05-11 | 2022-05-11 | Finger vein biological feature identification method, system and matching and identification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210533024.8A CN114863130B (en) | 2022-05-11 | 2022-05-11 | Finger vein biological feature identification method, system and matching and identification method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114863130A CN114863130A (en) | 2022-08-05 |
| CN114863130B true CN114863130B (en) | 2024-03-05 |
Family
ID=82636656
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210533024.8A Active CN114863130B (en) | 2022-05-11 | 2022-05-11 | Finger vein biological feature identification method, system and matching and identification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114863130B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115311691B (en) * | 2022-10-12 | 2023-02-28 | 山东圣点世纪科技有限公司 | Joint identification method based on wrist vein and wrist texture |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111353472A (en) * | 2020-03-30 | 2020-06-30 | 蚌埠学院 | Finger vein identification method based on LBP and Gabor direction Weber local descriptor |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110516087B (en) * | 2019-04-23 | 2021-12-07 | 广州麦仑信息科技有限公司 | Flash search comparison method for large-scale distributed full-palmar venation data |
-
2022
- 2022-05-11 CN CN202210533024.8A patent/CN114863130B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111353472A (en) * | 2020-03-30 | 2020-06-30 | 蚌埠学院 | Finger vein identification method based on LBP and Gabor direction Weber local descriptor |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114863130A (en) | 2022-08-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107292230B (en) | Embedded finger vein identification method based on convolutional neural network and having counterfeit detection capability | |
| Park et al. | Presentation attack detection using a tiny fully convolutional network | |
| Su et al. | A deep learning approach towards pore extraction for high-resolution fingerprint recognition | |
| CN111126240B (en) | Three-channel feature fusion face recognition method | |
| Gupta et al. | Fingerprint indexing schemes–a survey | |
| Pamplona Segundo et al. | Pore-based ridge reconstruction for fingerprint recognition | |
| CN108805027B (en) | Face recognition method under low resolution condition | |
| Song et al. | Fingerprint indexing based on pyramid deep convolutional feature | |
| Akhtar et al. | Face recognition under ageing effect: a comparative analysis | |
| CN114863130B (en) | Finger vein biological feature identification method, system and matching and identification method | |
| CN109523484B (en) | Fractal feature-based finger vein network repair method | |
| Raghavendra et al. | An efficient finger vein indexing scheme based on unsupervised clustering | |
| Anand et al. | Pore-based indexing for fingerprints acquired using high-resolution sensors | |
| CN111209850B (en) | Method for generating applicable multi-device identification finger vein image based on improved cGAN network | |
| Ghoualmi et al. | A multi-objective optimization approach for palm vein feature selection based on the discrete PSO | |
| Grudzien et al. | Face re-identification in thermal infrared spectrum based on ThermalFaceNet neural network | |
| Liu et al. | A novel high-resolution fingerprint representation method | |
| Bindu et al. | Kernel-based scale-invariant feature transform and spherical SVM classifier for face recognition | |
| CN114913610A (en) | Multi-mode identification method based on fingerprints and finger veins | |
| Huang et al. | Axially-enhanced local attention network for finger vein recognition | |
| Gu et al. | Selection of optimal features for iris recognition | |
| Bashar et al. | Texture based vein biometrics for human identification: A comparative study | |
| Sahmoud | Enhancing iris recognition | |
| Lee et al. | Automatic writer verification algorithm for Chinese characters using semi-global features and adaptive classifier | |
| Kuban et al. | A NOVEL MODIFICATION OF SURF ALGORITHM FOR FINGERPRINT MATCHING. |
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 |