KR20220135752A - finger vein recognition apparatus using peak data and method therefor - Google Patents

Abstract

The present disclosure relates to a finger vein detection device using peak data and a method thereof. The device can reduce detection time by performing, without image preprocessing, the detection of knuckle position and vein pattern through peak data detection, and minimize detection errors in relatively dark vein images, as infrared transmittance is low. In order to clearly recognize even thin veins, finger veins are more clearly highlighted through gray scale conversion of the finger image, and a finger vein is detected by extracting brightness values of the image to detect peak data.

Description

Finger vein recognition apparatus using peak data and method therefor

The present disclosure relates to a technique capable of detecting a finger vein by extracting blood vessel pattern features using peak data for finger vein recognition, and more specifically, image preprocessing through finger joint position and vein pattern detection through peak data detection A finger vein detection device and method using peak data that can be performed without a process to reduce detection time and minimize detection errors of relatively dark vein images due to low infrared transmittance.

Biometrics is one of the security methods to identify users based on the unique physical characteristics of an individual. It must satisfy seven conditions, such as unaffected accuracy, user acceptance without objection, and deception safe from arbitrary illegal use. is being used as

Currently, as for biometric recognition methods using biometric features, fingerprint recognition using fingerprint patterns is used in various fields, iris recognition using iris patterns, face recognition using face size and shape, EEG (electroencephalography) based EEG (electroencephalography) based EEG, and vein Vein recognition using patterns, etc. In addition to the recognition methods using the shape and size of body parts, behavioral recognition methods such as human behavior patterns and voices were also studied.

Among biometric methods, fingerprint recognition is excellent in development cost and usability, and is currently being used in various fields. However, fingerprint recognition has a problem in that the recognition rate is lowered due to external factors such as sweat and dust, or when deformation such as a wound occurs, recognition is not performed. In addition, since fingerprints can be copied with gelatin, etc., and forgery and falsification are relatively easier than other body parts, there is a high possibility that this will become a problem in places where personal authentication and security are important. Accordingly, iris recognition has been developed to compensate for the shortcomings of fingerprint recognition, and it is one of the biorecognition methods that are attracting attention. The probability of the iris being the same is about 1 in 1 billion, similar to that of a fingerprint, and there are about 260 unique patterns, so it has the characteristic of being more stable than a fingerprint. However, it was discovered that the iris could be duplicated as a high-resolution photo, making security weak. A vein recognition method that can compensate for the problems and shortcomings of the previous biorecognition method has recently attracted attention and a lot of research is being conducted. The same probability of veins is about 1 in 3.2 billion, which has a relatively low probability compared to fingerprints and irises, and has a characteristic that it is relatively difficult to forge, falsify and duplicate compared to fingerprints and iris, so it is used in places where personal authentication and security are important. can be used efficiently.

On the other hand, finger vein recognition is spotlighted as one of the biorecognition methods for authenticating a user through processes such as feature extraction, classification, and matching after acquiring an image by irradiating a finger with near-infrared light, and when acquiring a vein image, an individual's finger The difference in infrared transmittance according to thickness, ambient noise, and an image acquired in a non-uniform state have many difficulties in separating veins from the background in the image, and there is a problem in that the recognition rate is lowered when the position of the finger is changed.

According to an embodiment of the present disclosure, in order to reliably recognize even thin finger veins, the finger image is more clearly emphasized through gray scale conversion, and the brightness value of the image is extracted to detect peak data and designate it through this pulse can be detected.

For this, peak data can be detected using SWA (Shifted Waveform Analysis) algorithm or DSTW (Down Slope Trace Waveform) algorithm for video images. For clearer detection, blue LED irradiation, outline highlighting, data smoothing, etc. It is possible to derive the result of minimizing the detection error and improving the accuracy of the relatively dark vein image due to the low infrared transmittance.

According to one aspect of the present disclosure, there is provided an image capture unit for generating a plurality of image images by capturing an image of a user's finger photographed using a complex wavelength in real time; a gray scale conversion unit configured to perform gray scale conversion on the plurality of video images according to a gray scale conversion ratio to which a ratio of red among RGB colors is highest; a peak data detector extracting brightness values of the plurality of video images and detecting peak data using the extracted brightness values; And there is provided a finger vein detection device using peak data, characterized in that it comprises a finger vein detection unit for detecting a finger vein using the detected peak data.

In an embodiment, the image capture unit may adjust an image brightness setting value according to an average brightness value of a plurality of image images.

In an embodiment, the apparatus may further include an outline highlighter for emphasizing an outline of a finger by removing a higher frequency or a lower frequency than a preset reference frequency range for the plurality of image images on which the gray scale has been performed.

In an embodiment, a data smoothing performing unit that generates a single output point by calculating the frequency average of the plurality of image images and smoothes the data by filtering data or signals through a low-pass filter using the single output point as a starting point further may include

In an embodiment, the gray scale conversion unit may have a ratio of red, green, and blue in the gray scale conversion ratio of 7:1.5:1.5.

In one embodiment, the image capture unit uses a blue LED to adjust the image brightness setting value, and controls the irradiation angle, illuminance, and time of the blue LED to increase the sharpness of the fingers included in the plurality of image images. can

In an embodiment, the peak data detection unit applies a shifted waveform analysis (SWA) algorithm to the video image by as many as the number of pixels in the horizontal direction in the vertical direction and the number of pixels in the horizontal direction in the vertical direction in the video image. The SWA technique can be repeated.

In an embodiment, the peak data detection unit may calculate a difference between the SWA data and the original video image data after the repetition of the SWA technique is completed, and determine a peak when a preset threshold value is exceeded.

In an embodiment, the finger vein detection unit may reconstruct the video image by converting the pixel value of the peak data to 255 when the pixel value is greater than a preset reference value and fixing the pixel value of the non-peak portion to 0.

In an embodiment, the apparatus may further include a knurl detector configured to detect a knuckle by using the detected peak data.

According to another aspect of the present disclosure, generating a plurality of video images by capturing an image of a user's finger photographed using a complex wavelength in real time; performing gray scale conversion on the plurality of video images according to a gray scale conversion ratio to which a ratio of red is the highest among RGB colors; extracting brightness values of the plurality of video images and detecting peak data using the extracted brightness values; And there is provided a finger vein detection method using peak data, characterized in that it comprises the step of detecting a finger vein using the detected peak data.

In an embodiment, the generating of the video image may include adjusting an image brightness setting value according to an average brightness value of a plurality of video images.

In an embodiment, the method may further include emphasizing the outline of a finger by removing a higher frequency or a lower frequency than a preset reference frequency range for the plurality of image images on which the gray scale has been performed.

In an embodiment, the method may further include generating a single output point by calculating a frequency average of the plurality of video images, and smoothing the data by filtering data or signals through a low-pass filter using the single output point as a starting point. can

In an embodiment, in the performing of the gray scale conversion, a ratio of red, green, and blue in the gray scale conversion ratio may be 7:1.5:1.5.

In an embodiment, the generating of the video image uses a blue LED to adjust the image brightness setting value, and controls the irradiation angle, illuminance, and time of the blue LED to control a finger included in the plurality of video images. can increase the clarity of

In an embodiment, the detecting of the peak data includes applying a shifted waveform analysis (SWA) algorithm to the video image, so that the number of pixels in the horizontal and vertical directions in the video image is The SWA technique can be repeated as many as the number of pixels.

In an embodiment, the detecting of the peak data includes calculating a difference between the SWA data and the original video image data after the repetition of the SWA technique is completed, and determining a peak when a preset threshold value is exceeded. data can be detected.

In one embodiment, in the detecting of the finger vein, if the pixel value of the peak data is greater than a preset reference value, it is converted to 255 and the pixel value of the non-peak portion is fixed to 0 to reconstruct the image image.

In an embodiment, the method may further include detecting a knuckle by using the detected peak data.

Through the finger vein detection device using peak data implemented according to an embodiment of the present disclosure, it is possible to detect a finger vein through less computational resources and time without preprocessing the image image, and SWA (Shifted Waveform Analysis) algorithm or High precision can be derived by detecting peak data using DSTW (Down Slope Trace Waveform) algorithm and detecting finger veins through it. The effect of providing a detection function exists.

1 is a block diagram of a finger vein detection device using peak data according to a first embodiment of the present disclosure.
2 is a block diagram of a finger vein detection device using peak data according to a second embodiment of the present disclosure.
3 is a block diagram of a finger vein detection device using peak data according to a third embodiment of the present disclosure.
4 is a block diagram of a finger vein detection device using peak data according to a fourth embodiment of the present disclosure.
5 is a diagram illustrating a feature point detection algorithm on a finger vein image according to an embodiment of the present disclosure.
Figure 6 is a view showing the process of inverting the image from the original finger vein image, according to an embodiment of the present disclosure.
7 is a diagram illustrating an example of a SWA algorithm according to an embodiment of the present disclosure, and FIG. 8 is a diagram illustrating a method of finding a peak point in the SWA algorithm.
9 is a diagram illustrating a gray scale conversion result image for each gray scale conversion ratio performed according to an embodiment of the present disclosure.
10 is a diagram illustrating a node detection criterion according to an embodiment of the present disclosure, and FIG. 11 is a diagram illustrating an application direction of a SWA technique performed on a video image according to an embodiment of the present disclosure.
12 is a diagram illustrating a characteristic of a peak value in a difference between two signals, according to an embodiment of the present disclosure.
13 is a view showing a performance check result of a feature detection algorithm according to an embodiment of the present disclosure. In FIG. 13 (A) is a result based on the X axis, (B) is a result based on the Y axis, , (C) is the result of combining the results of (A) and (B), and (D) is the original image.
14 is a view showing a result of checking the performance of a feature detection algorithm according to another embodiment of the present disclosure, wherein (A) of FIG. 14 is a result based on the X axis, (B) is a result based on the Y axis, , (C) is the result of combining the results of (A) and (B), and (D) is the original image.
15 is a diagram illustrating a comparison of a feature extraction algorithm according to an embodiment of the present disclosure.
16 is a diagram illustrating a process of performing SWA performed on a test template for SWA algorithm learning according to an embodiment of the present disclosure.
17 is a diagram illustrating a result of performing a SWA technique performed on a video image according to an embodiment of the present disclosure.
18 is a diagram illustrating a knurl detection image obtained by detecting a knuckle by using peak data for a video image according to an embodiment of the present disclosure.
19 is a flowchart illustrating a learning process of the SWA algorithm according to an embodiment of the present disclosure.
20 is a flowchart of a finger vein detection method using peak data according to an embodiment of the present disclosure.

Hereinafter, with reference to the accompanying drawings, embodiments of the present disclosure will be described in detail so that those of ordinary skill in the art to which the present disclosure pertains can easily implement them. However, the present disclosure may be implemented in several different forms and is not limited to the embodiments described herein.

And in order to clearly explain the present disclosure in the drawings, parts irrelevant to the description are omitted, and similar reference numerals are attached to similar parts throughout the specification.

Throughout the specification, when a part "includes" a certain element, it means that other elements may be further included, rather than excluding other elements, unless otherwise stated.

Hereinafter, an apparatus and method for detecting finger veins using peak data according to an embodiment of the present disclosure will be described with reference to the drawings.

1 is a block diagram of a finger vein detection apparatus 1000 using peak data according to a first embodiment of the present disclosure.

According to an embodiment of the present disclosure, the finger vein detection device 1000 using peak data may include an image capture unit 100 , a gray scale conversion unit 200 , a peak data detection unit 300 , and a finger vein detection unit 400 . can

The image capture unit 100 may generate a plurality of image images by capturing an image of a user's finger photographed using a complex wavelength in real time.

In an embodiment of the present disclosure, the image capture unit 100 may photograph a user's finger using a complex wavelength of an infrared LED and a green LED as a light source.

According to the above embodiment, by photographing the user's finger using the complex wavelength of the infrared LED and the green LED as a light source, the outline of the vein inside the finger can be photographed relatively clearly.

According to an embodiment of the present disclosure, by capturing an image of a user's finger in real time, it is possible to generate a plurality of image images in real time without delay in shooting time.

According to an embodiment of the present disclosure, the image capture unit 100 may adjust an image brightness setting value according to an average brightness value of a plurality of image images.

According to the embodiment, auto leveling of automatically adjusting the step value of the video image by using the average brightness value of the LED and the video image may be performed.

According to an embodiment of the present disclosure, the image capture unit 100 uses a blue LED to adjust an image brightness setting value, and controls the irradiation angle, illuminance, and time of the blue LED to display a plurality of image images. You can increase the sharpness of the included fingers.

According to an embodiment of the present disclosure, by using a blue LED among LEDs for providing illumination, the finger veins inside the finger can be more clearly highlighted and photographed.

According to the above embodiment, it is possible to increase the sharpness of the fingers included in the plurality of video images by controlling the irradiation angle, the illuminance, and the time of the blue LED until the sharpness that meets the preset standard is secured.

The gray scale conversion unit 200 may perform gray scale conversion on a plurality of image images according to a gray scale conversion ratio to which a ratio of red is applied the highest among RGB colors.

In general, a vein image obtained by photographing a finger has a dark blood vessel portion and a bright background portion, and there is a problem in that the accuracy of recognizing a finger vein is lowered with such an image. In order to solve this problem, one of the present disclosure According to an embodiment, the gray scale conversion unit 200 may perform gray scale conversion, and in particular, may perform gray scale conversion using a gray scale conversion ratio to which a ratio of red among RGB colors is applied the highest.

When the RGB pixel values of the image are inverted according to the above embodiment, the blood vessel portion is bright and the remaining background and noise portions are converted to be dark, so that the vein image can be reconstructed into a clear image emphasizing the blood vessel.

According to an embodiment of the present disclosure, gray scale conversion may be performed using an image processing function as in Equation 1 below.

[Equation 1]

According to an embodiment of the present disclosure, the gray scale conversion unit 200 may have a ratio of red, green, and blue in a gray scale conversion ratio of 7:1.5:1.5.

According to the above embodiment, the gray scale conversion ratio is conventional, and the red, green, and blue ratios are 1: 1: 1, rather than performing gray scale conversion, the blood vessels are clearer when the red is more emphasized and converted to gray. In the case of gray scale conversion in which the ratio of red, green, and blue is 7:1.5:1.5 as a result of the experiment, the finger veins can be recognized most clearly, and this will be described in more detail with reference to FIG. 5 .

The peak data detector 300 may extract brightness values of a plurality of video images and detect peak data using the extracted brightness values.

According to an embodiment of the present disclosure, the peak data detection unit 300 may apply a shifted waveform analysis (SWA) algorithm to the video image, and the number of pixels in the horizontal and vertical directions to the video image through the SWA algorithm , a peak may be determined by repeating the SWA technique as many as the number of pixels in the horizontal direction in the vertical direction.

According to the above embodiment, by applying the SWA algorithm to the video image, the point at which the slope of the signal changes from a positive number to a negative number can be determined as a peak. By calculation, the peak point of the original signal can be obtained.

By repeatedly performing the SWA technique according to the above embodiment, it is possible to perform separation of blood vessels, backgrounds, and noises that cause a decrease in the recognition rate when extracting finger vein feature points.

According to an embodiment of the present disclosure, a peak is determined by repeating the SWA technique by the number of pixels in the vertical direction in the horizontal direction and the number of pixels in the horizontal direction in the vertical direction of the video image using the template of the learned SWA algorithm. can

According to an embodiment of the present disclosure, the peak data detector 300 calculates a difference between the SWA data and the original video image data after the repetition of the SWA technique is completed, and determines as a peak when a preset threshold value is exceeded. Thus, peak data can be detected.

According to an embodiment of the present disclosure, a Down Slope Trace Waveform (DSTW) algorithm, not a SWA algorithm, may be used to determine a peak using a video image.

The finger vein detection unit 400 may detect a finger vein using the detected peak data.

According to an embodiment of the present disclosure, an image image may be reconstructed so that veins are highlighted using peak data, and the distribution of vein blood vessels in the reconstructed image image may be expressed as a composite vector component of x and y axes. .

According to the above embodiment, the vector component of the x-axis can be derived by repeating the SWA technique by the number of pixels in the vertical direction in the horizontal direction of the video image, and the vector component of the y-axis by repeating the SWA technique by the number of pixels in the horizontal direction in the vertical direction can be derived, and a synthetic vector component can be derived from this.

According to an embodiment of the present disclosure, the finger vein detection unit 400 may reconstruct the video image by converting the pixel value of the peak data to 255 when the pixel value is greater than the reference value and fixing the pixel value of the non-peak portion to 0.

According to an embodiment of the present disclosure, the difference between the SWA data and the original data, which is the result of performing the SWA algorithm, is calculated and set as a reference value. If the pixel value of the peak data is greater than the reference value, it is converted to 255 and the pixel value of the non-peak part is fixed to 0, and the x-axis vector and y-axis vector on a black background can be reconstructed as an image image displayed in white, and finger veins can be detected through a composite vector of the x-axis vector and y-axis vector.

2 is a block diagram of a finger vein detection device using peak data according to a second embodiment of the present disclosure.

Referring to FIG. 2 , the finger vein detection apparatus 1100 using peak data according to a second embodiment of the present disclosure is an outline highlighting unit 500 in the finger vein detection apparatus 1000 using peak data according to the first embodiment. ) may be further included.

According to the second embodiment of the present disclosure, the outline highlighting unit 500 may emphasize the outline of a finger by removing a higher frequency or a lower frequency than a preset reference frequency range for a plurality of image images on which gray scale has been performed. .

According to an embodiment of the present disclosure, a baseline filter may be used to remove a higher frequency or a lower frequency than a preset reference frequency range.

According to the above embodiment, the baseline filter can determine the score with the highest score only with the baseline through an automatic calculation method using a Weighted Least Squares (WLS) criterion algorithm, and by repeatedly performing this, each spectrum is By fitting a baseline and determining which variables are clearly above the baseline, high or low frequencies above a preset reference frequency range can be removed.

3 is a block diagram 1200 of a finger vein detection device using peak data according to a third embodiment of the present disclosure.

Referring to FIG. 3 , the finger vein detection device 1200 using peak data according to a third embodiment of the present disclosure is a data smoothing performing unit 600 to the finger vein detection device 1100 using peak data according to the second embodiment. ) may be further included.

According to the third embodiment of the present disclosure, the data smoothing unit 600 calculates the frequency average of a plurality of video images to generate a single output point, and uses the single output point as a starting point to convert data or signals through a low-pass filter. You can smooth the data by filtering.

According to an embodiment of the present disclosure, a moving average filter may be used to filter data or a signal through a low-pass filter based on a single output point, and the moving average filter is an array of sampled data/signals. It can be a simple low-pass finite impulse response (FIR) filter commonly used to smooth . It can have a simple low-pass filter (LPF) structure to filter out unwanted noisy data/signals.

4 is a block diagram of a finger vein detection device using peak data according to a fourth embodiment of the present disclosure.

Referring to FIG. 4 , the finger vein detection device 1200 using peak data according to a fourth embodiment of the present disclosure includes a node detection unit 700 to the finger vein detection device 1000 using peak data according to the first embodiment. may include more.

According to the fourth embodiment of the present disclosure, the knuckle detection unit 700 may detect a finger knuckle by using the detected peak data.

According to an embodiment of the present disclosure, it is possible to calculate and detect the positions of two knurls of a finger using peak data.

According to the above embodiment, since two knurls exist in each of the index, middle, ring, and little fingers except for the thumb, the human finger can be detected by calculating the positions of the two knurls using peak data.

According to an embodiment of the present disclosure, the video image is divided in half, and the peak having the largest value calculated by calculating the difference between the SWA data and the original video image data on the left side and the peak with the largest value on the right side are selected and detected as nodes. have.

According to an embodiment of the present disclosure, it can be used for the purpose of determining whether or not the target object to be photographed is a finger by detecting the nodes, and reducing the detection time and detecting finger veins by limiting the fingers between the detected nodes. It can also be used to improve accuracy.

5 is a diagram illustrating a feature point detection algorithm on a finger vein image according to an embodiment of the present disclosure.

When acquiring a finger vein image, infrared light is transmitted through the finger, so the background is bright due to the difference in absorbance, and blood vessels appear dark. In addition, since the infrared transmittance varies according to the thickness of an individual's finger, a person with thick fingers has a dark overall image, and a person with a thin finger has a bright image, making it difficult to identify the vein location. The existing detection algorithm performs an image preprocessing process, so the detection time is increased, and there is a problem in that a finger vein detection error occurs in an image acquired in an environment with non-uniform lighting. The vein extraction algorithm developed by the present inventor reduces the detection time and applies the SWA (Shifted Waveform Analysis) algorithm as it is performed without pre-processing, so that even images acquired in an environment with uneven lighting can be detected even when the finger position changes. It has a characteristic that the recognition rate is not lowered by holding the reference point with the knuckle of a finger. The algorithm execution process performs pixel value inversion to brighten blood vessels in the acquired finger vein image, and then calculates the difference between the original image and the SWA value by applying the SWA algorithm. The calculated value can know the node position by applying the node detection algorithm. When the feature detection algorithm is applied, the background and noise are removed, and only the blood vessel features are detected and reconstructed into an image.

Figure 6 is a view showing the process of inverting the image from the original finger vein image, according to an embodiment of the present disclosure.

The acquired vein image shows relatively dark blood vessels and a bright background. After converting the RGB format vein image to gray scale, if the entire image pixel value is inverted, the blood vessels are relatively bright and the background and noise are reconstructed to be dark. 6A is an original 3-Channel Gray Scale vein image obtained by the finger vein detection device 1000, and FIG. 6B is an image converted to 1-Channel Gray Scale. Although there is no change visually, it has the characteristics of reducing memory usage and shortening the execution time in the subsequent execution process. Finally, FIG. 6C is an image reconstructed by inverting all pixel values of the image of FIG. 6B.

7 is a diagram illustrating an example of a SWA algorithm according to an embodiment of the present disclosure, and FIG. 8 is a diagram illustrating a method of finding a peak point in the SWA algorithm.

In FIG. 7, a dotted line indicates a SWA signal, and a solid line indicates an original signal x(n). The SWA signal maintains the original signal at the point where the original signal increases, and at the inflection point, which is the point where the slope of the signal becomes 0, the SWA signal maintains a peak value from the p point to the d point. After that, the SWA signal repeats the process of decreasing like a change from point d to point x until it is equal to or smaller than the original signal. At this time, the characteristics of the signal are classified and detected in the form shown in FIG. 8 . Fig. 8(A) shows finding a peak, and Fig. 8(B) shows shifting from the peak.

9 is a diagram illustrating a gray scale conversion result image for each gray scale conversion ratio performed according to an embodiment of the present disclosure.

Referring to FIG. 9 , a gray scale conversion result image for each gray scale conversion ratio performed according to an embodiment of the present disclosure is shown, and red and green than when gray scale conversion is performed with other gray scale conversion ratios as shown in FIG. 9 . , it can be seen that the venous blood vessels are relatively clearly visible when the gray scale conversion is performed through the gray scale conversion ratio of 7: 1.5: 1.5 in the ratio of blue.

10 is a diagram illustrating a node detection criterion according to an embodiment of the present disclosure, and FIG. 11 is a diagram illustrating an application direction of a SWA technique performed on a video image according to an embodiment of the present disclosure.

The SWA algorithm is performed only once based on Yc as shown in FIG. 10 for the node detection process. In addition, for the feature detection process, as shown in (A) of FIG. 11 , Xn times on the x-axis and Ym times on the y-axis as shown in (B) of FIG. 11 are performed.

Human fingers have two joints on each index, middle, ring, and little finger except for the thumb. It is possible to calculate the position of two nodes through image processing. The SWA signal is obtained by applying the SWA algorithm to (x _0, y _c ), (x _1, y _c ) to (x _n, y _c ), which are the same components as the Yc position of FIG. 10 for the image-converted image. Thereafter, the position of the largest value among the values having the same characteristics as in FIG. 8 is calculated on the left based on the Xc position of FIG. 10 , and the same process is performed on the right side based on the Xc position to calculate the node position.

The original signal calculates the difference from the signal of the SWA algorithm execution result in order to obtain a value necessary for the next process. At this time, the calculation of the difference between the original signal and the SWA signal is performed as many times as the number of repetitions of the node position calculation process.

Thereafter, the calculated value sets a threshold, and the pixel value of the highest point, which is a component having the same characteristics as in FIG. 12 , is converted into 0 (white). 12 is a diagram illustrating a characteristic of a peak value in a difference between two signals, according to an embodiment of the present disclosure. The converted value may reconstruct an image into an x-axis vector and a y-axis vector on a black background. In this case, the blood vessel feature can be detected as a composite vector of an x-axis vector and a y-axis vector.

13 is a view showing a performance check result of a feature detection algorithm according to an embodiment of the present disclosure. In FIG. 13 (A) is a result based on the X axis, (B) is a result based on the Y axis, , (C) is the result of combining the results of (A) and (B), and (D) is the original image.

14 is a view showing a performance check result of a feature detection algorithm according to another embodiment of the present disclosure, wherein (A) of FIG. 14 is a result based on the X-axis, (B) is a result based on the Y-axis, , (C) is the result of combining the results of (A) and (B), and (D) is the original image.

The present inventor conducted an experiment to confirm the accuracy of whether the feature detection algorithm accurately detected the blood vessel pattern as shown in FIGS. 13 and 14 . 13 and 14 show the results of checking the performance of the feature detection algorithm. First, randomly generated lines (start: (10, 10), end: (60, 40), color: 125 (a number between 0 and 255), thickness: 1) are D) and the same results were obtained. The original image consists of 51 pixels, and 100% accuracy was obtained with 51 feature detection results. As a result, the total number of pixels was 399 and the number of feature-detected pixels was 389, showing an accuracy of 97.5%. Accordingly, it is determined that a high reliability result can be obtained when the feature detection is performed with the finger vein image.

15 is a diagram illustrating a comparison of a feature extraction algorithm according to an embodiment of the present disclosure.

15 shows the original vein image and the feature extracted image by applying the algorithm developed by the present inventor and the augmented reality (AR) algorithm. In Fig. 15, column 2 shows the original photo, column 3 shows the result of using the augmented reality (AR) algorithm, and column 4 shows the finger blood vessel pattern obtained by the SWA algorithm. The result of the SWA algorithm has less noise than the result using the reality augmentation algorithm, and shows the characteristics of extracting only venous blood vessel patterns. This means that the performance of FAR (False Acceptance Rate), FRR (False Rejection Rate), and EER (Equal Error Rate), which are the last stage recognition performance indicators of the recognition algorithm, is improved.

The detection result of the developed algorithm is proven as a highly reliable algorithm by showing an accuracy of more than 99% as a result of experimenting with images made up of several lines with different positions, sizes, and colors. In addition, the results obtained by performing the node position detection algorithm are judged to be able to compensate for the problem of a decrease in the recognition rate due to a change in the finger position. If the proposed algorithm is applied to the recognition field using body parts such as wrists and palms in the future, it is expected that it will contribute to reducing the detection accuracy and detection execution time of biometric features.

16 is a diagram illustrating a process of performing SWA performed on a test template for SWA algorithm learning according to an embodiment of the present disclosure.

Referring to FIG. 16 , a process of performing the SWA technique performed on a test template for learning the SWA algorithm according to an embodiment of the present disclosure is illustrated, and the x-axis and y-axis of the test template for learning the SWA algorithm are shown. By moving to and applying the SWA technique, higher accuracy can be obtained by learning the algorithm.

17 is a diagram illustrating a result of performing a SWA technique performed on a video image according to an embodiment of the present disclosure.

Referring to FIG. 17 , the result of performing the SWA technique according to an embodiment of the present disclosure is shown, and the result of applying the SWA technique through the SWA algorithm only on the x-axis to the acquired vein image (video image) and the x-axis as well as the y The results of applying the SWA technique through the SWA algorithm up to the axis are shown.

According to the above embodiment, the vector component of the x-axis can be derived by repeating the SWA technique by the number of pixels in the vertical direction in the horizontal direction (x-axis) of the video image, and SWA by the number of pixels in the horizontal direction in the vertical direction (y-axis) By repeating the technique, the vector component of the y-axis can be derived, and the synthetic vector component can be derived through this, which can increase the recognition rate of the finger veins.

18 is a diagram illustrating a knurl detection image obtained by detecting a knuckle by using peak data for a video image according to an embodiment of the present disclosure.

Referring to FIG. 18 , a node detection image detected according to an embodiment of the present disclosure is shown. According to the embodiment, a peak is derived from an original image image, and the image image among the derived peaks is divided into the middle. A node can be detected by selecting the largest peak value from the left and the largest peak value from the right.

19 is a flowchart illustrating a learning process of the SWA algorithm according to an embodiment of the present disclosure.

Referring to FIG. 19 , the learning process of the SWA algorithm according to an embodiment of the present disclosure is shown, a video image is captured, a grayscale conversion is performed on the captured video image, and the image resizing operation is performed. .

Here, the resizing operation may be performed to shorten the detection time, and this may be selectively applied.

According to the above embodiment, the peak is determined by applying the SWA algorithm to the resized image, and the feature point extraction of the finger vein and reconstruction of the image image can be performed using the derived peak data, and then the feature point expansion operation is performed. can be performed and the learning level can be increased by repeating this process a number of times.

After that, it is possible to learn with the SWA algorithm having a more learned template by saving the image and saving the template through the feature point combining step.

20 is a flowchart of a finger vein detection method using peak data according to an embodiment of the present disclosure.

A plurality of video images are generated by capturing an image of a user's finger photographed using a complex wavelength in real time (S10).

According to an embodiment of the present disclosure, a plurality of video images may be generated by capturing an image of a user's finger photographed using a complex wavelength in real time.

In an embodiment of the present disclosure, a user's finger may be photographed by using a complex wavelength of an infrared LED and a green LED as a light source.

According to the above embodiment, by photographing the user's finger using the complex wavelength of the infrared LED and the green LED as a light source, the outline of the vein inside the finger can be photographed relatively clearly.

According to an embodiment of the present disclosure, by capturing an image of a user's finger in real time, it is possible to generate a plurality of image images in real time without delay in shooting time.

According to an embodiment of the present disclosure, an image brightness setting value may be adjusted according to an average brightness value of a plurality of video images.

According to the embodiment, auto leveling of automatically adjusting the step value of the video image by using the average brightness value of the LED and the video image may be performed.

According to an embodiment of the present disclosure, a blue LED (blue LED) is used to adjust the image brightness setting value, and the brightness of a finger included in a plurality of video images is increased by controlling the irradiation angle, illuminance, and time of the blue LED. can

According to an embodiment of the present disclosure, by using a blue LED among LEDs for providing illumination, the finger veins inside the finger can be more clearly highlighted and photographed.

According to the above embodiment, it is possible to increase the sharpness of the fingers included in the plurality of video images by controlling the irradiation angle, the illuminance, and the time of the blue LED until the sharpness that meets the preset standard is secured.

Gray scale conversion is performed on a plurality of video images according to the gray scale conversion ratio to which the ratio of red among the RGB colors is applied ( S20 ).

According to an embodiment of the present disclosure, gray scale conversion may be performed on a plurality of video images according to a gray scale conversion ratio to which a ratio of red among RGB colors is applied the highest.

According to an embodiment of the present disclosure, in the vein image obtained by photographing the finger, the blood vessel portion is dark and the background portion is bright. For this purpose, gray scale conversion may be performed according to an embodiment of the present disclosure, and in particular, gray scale conversion may be performed using a gray scale conversion ratio to which the ratio of red among RGB colors is applied the highest.

When the RGB pixel values of the image are inverted according to the above embodiment, the blood vessel portion is bright and the remaining background and noise portions are converted to be dark, so that the vein image can be reconstructed into a clear image emphasizing the blood vessel.

According to an embodiment of the present disclosure, gray scale conversion may be performed using an image processing function as in Equation 1 above.

According to an embodiment of the present disclosure, a ratio of red, green, and blue in a gray scale conversion ratio may be 7:1.5:1.5.

According to the above embodiment, the gray scale conversion ratio is conventional, and the red, green, and blue ratios are 1: 1: 1, rather than performing gray scale conversion, the blood vessels are clearer when the red is more emphasized and converted to gray. In the case of gray scale conversion with the ratio of red, green, and blue being 7:1.5:1.5 as a result of the experiment, the finger veins can be recognized most clearly.

According to an embodiment of the present disclosure, the outline of a finger may be emphasized by removing a higher frequency or a lower frequency than a preset reference frequency range from a plurality of gray scaled image images.

According to an embodiment of the present disclosure, a baseline filter may be used to remove a higher frequency or a lower frequency than a preset reference frequency range.

According to the above embodiment, the baseline filter can determine the score with the highest score only with the baseline through an automatic calculation method using a Weighted Least Squares (WLS) criterion algorithm, and by repeatedly performing this, each spectrum is By fitting a baseline and determining which variables are clearly above the baseline, high or low frequencies above a preset reference frequency range can be removed.

According to an embodiment of the present disclosure, a single output point may be generated by calculating a frequency average of a plurality of video images, and data or a signal may be smoothed by filtering data or a signal through a low-pass filter using the single output point as a starting point.

According to an embodiment of the present disclosure, a moving average filter may be used to filter data or signals through a low-pass filter based on a single output point, and the moving average filter is an array of sampled data/signals. It can be a simple low-pass finite impulse response (FIR) filter commonly used to smooth It may have a filter (LPF) structure.

The brightness values of a plurality of video images are extracted, and peak data is detected using the extracted brightness values (S30).

According to an embodiment of the present disclosure, brightness values of a plurality of video images may be extracted, and peak data may be detected using the extracted brightness values.

According to an embodiment of the present disclosure, a Shifted Waveform Analysis (SWA) algorithm may be applied to the video image, and the number of pixels in the horizontal direction and the horizontal direction in the video image through the SWA algorithm A peak may be determined by repeating the SWA technique as many as the number of pixels.

According to the above embodiment, by applying the SWA algorithm to the video image, the point at which the slope of the signal changes from a positive number to a negative number can be determined as a peak. By calculation, the peak point of the original signal can be obtained.

By repeatedly performing the SWA technique according to the above embodiment, it is possible to perform separation of blood vessels, backgrounds, and noises that cause a decrease in the recognition rate when extracting finger vein feature points.

According to an embodiment of the present disclosure, a peak is determined by repeating the SWA technique by the number of pixels in the vertical direction in the horizontal direction and the number of pixels in the horizontal direction in the vertical direction of the video image using the template of the learned SWA algorithm. can

According to an embodiment of the present disclosure, after the repetition of the SWA technique is completed, the difference between the SWA data and the original video image data is calculated, and when a preset threshold value is exceeded, the peak data can be detected by determining it as a peak. .

According to an embodiment of the present disclosure, a Down Slope Trace Waveform (DSTW) algorithm, not a SWA algorithm, may be used to determine a peak using a video image.

Finger veins are detected using the detected peak data (S40).

According to an embodiment of the present disclosure, it is possible to detect a finger vein using the detected peak data.

According to an embodiment of the present disclosure, an image image may be reconstructed so that veins are highlighted using peak data, and the distribution of vein blood vessels in the reconstructed image image may be expressed as a composite vector component of x and y axes. .

According to the above embodiment, the vector component of the x-axis can be derived by repeating the SWA technique by the number of pixels in the vertical direction in the horizontal direction of the video image, and the vector component of the y-axis by repeating the SWA technique by the number of pixels in the horizontal direction in the vertical direction can be derived, and a synthetic vector component can be derived from this.

According to an embodiment of the present disclosure, if the pixel value of the peak data is greater than the reference value, the video image may be reconstructed by converting it to 255 and fixing the pixel value of the non-peak portion to 0.

According to an embodiment of the present disclosure, the difference between the SWA data and the original data, which is the result of performing the SWA algorithm, is calculated and set as a reference value. If the pixel value of the peak data is greater than the reference value, it is converted to 255 and the pixel value of the non-peak part is fixed to 0, and the x-axis vector and y-axis vector on a black background can be reconstructed as an image image displayed in white, and finger veins can be detected through a composite vector of the x-axis vector and y-axis vector.

According to an embodiment of the present disclosure, a knuckle may be detected using the detected peak data.

According to an embodiment of the present disclosure, two joints of a finger using peak data

It can be detected by calculating the position.

According to the above embodiment, since two knurls exist in each of the index, middle, ring, and little fingers except for the thumb, the human finger can be detected by calculating the positions of the two knurls using peak data.

According to an embodiment of the present disclosure, the video image is divided in half, and the peak having the largest value calculated by calculating the difference between the SWA data and the original video image data on the left side and the peak with the largest value on the right side are selected and detected as nodes. have.

According to an embodiment of the present disclosure, it can be used for the purpose of determining whether or not the target object to be photographed is a finger by detecting the nodes, and reducing the detection time and detecting finger veins by limiting the fingers between the detected nodes. It can also be used to improve accuracy.

Embodiments of the present disclosure are not implemented only through the devices and/or methods described above, and have been described in detail with respect to the embodiments of the present disclosure above, but the scope of the present disclosure is not limited thereto, and the following claims Various modifications and improvements by those skilled in the art using the basic concept of the present disclosure as defined in are also within the scope of the present disclosure.

Claims (20)

an image capture unit generating a plurality of image images by capturing an image of a user's finger photographed using a complex wavelength in real time;
a gray scale conversion unit configured to perform gray scale conversion on the plurality of video images according to a gray scale conversion ratio to which a ratio of red among RGB colors is highest;
a peak data detector extracting brightness values of the plurality of video images and detecting peak data using the extracted brightness values; and
Finger vein detection device using peak data, characterized in that it comprises a finger vein detection unit for detecting the finger vein using the detected peak data. The method of claim 1,
The image capture unit finger vein detection device using peak data for adjusting the image brightness setting value according to the average brightness value of a plurality of image images. The method of claim 1,
Finger vein detection device using peak data further comprising an outline highlighter for emphasizing the outline of a finger by removing a high frequency or a low frequency than a preset reference frequency range for the plurality of image images on which the gray scale is performed. 4. The method of claim 3,
Peak data further comprising a data smoothing unit for generating a single output point by calculating the frequency average of the plurality of image images, and smoothing the data by filtering data or signals through a low-pass filter using the single output point as a starting point Finger vein detection device used. The method of claim 1,
The gray scale conversion unit red, green, and blue ratio in the gray scale conversion ratio 7: 1.5: finger vein detection device using peak data, characterized in that 1.5. 3. The method of claim 2,
The image capture unit uses a blue LED to adjust the image brightness setting value, and controls the irradiation angle, illuminance, and time of the blue LED to increase the sharpness of the fingers included in the plurality of image images. Finger vein detection device using data. The method of claim 1,
By applying a shifted waveform analysis (SWA) algorithm to the video image, the peak data detector repeats the SWA technique by the number of pixels in the horizontal direction and the number of pixels in the vertical direction in the video image. Finger vein detection device using peak data. 8. The method of claim 7,
After the repetition of the SWA technique is completed, the peak data detection unit performs SWA data
A finger vein detection device using peak data that calculates the difference between and the original video image data and detects the peak data by determining a peak when it exceeds a preset threshold value. 9. The method of claim 8,
The finger vein detection unit using peak data for reconstructing an image by converting the pixel value of the peak data to 255 when the pixel value is greater than a preset reference value and fixing the pixel value of the non-peak portion to 0. The method of claim 1,
Finger vein detection device using peak data further comprising a joint detection unit for detecting a finger joint using the detected peak data. generating a plurality of video images by capturing an image of a user's finger photographed using a complex wavelength in real time;
performing gray scale conversion on the plurality of video images according to a gray scale conversion ratio to which a ratio of red is the highest among RGB colors;
extracting brightness values of the plurality of video images and detecting peak data using the extracted brightness values; and
Finger vein detection method using peak data, characterized in that it comprises the step of detecting the finger veins using the detected peak data. 12. The method of claim 11,
The step of generating the video image is a finger vein detection method using peak data for adjusting the image brightness setting value according to the average brightness value of a plurality of video images. 12. The method of claim 11,
Finger vein detection method using peak data further comprising the step of emphasizing the outer line of the finger by removing a higher frequency or a lower frequency than a preset reference frequency range for the plurality of image images on which the gray scale is performed. 14. The method of claim 13,
Designation using peak data, further comprising the step of calculating a frequency average of the plurality of video images to generate a single output point, and smoothing the data by filtering the data or signal through a low-pass filter using the single output point as a starting point How to detect a mac. 12. The method of claim 11,
The step of performing the gray scale conversion is a finger vein detection method using peak data, characterized in that the ratio of red, green, and blue in the gray scale conversion ratio is 7: 1.5: 1.5. 13. The method of claim 12,
The step of generating the video image uses a blue LED to adjust the image brightness setting value, and controls the irradiation angle, illuminance, and time of the blue LED to increase the sharpness of the fingers included in the plurality of video images. A method for detecting finger veins using peak data, which is characterized. 12. The method of claim 11,
In the detecting of the peak data, by applying a shifted waveform analysis (SWA) algorithm to the video image, the number of pixels in the vertical direction in the horizontal direction and the number of pixels in the horizontal direction in the vertical direction in the video image are SWA techniques Finger vein detection method using peak data that repeats. 18. The method of claim 17,
In the detecting of the peak data, after the repetition of the SWA technique is completed, the difference between the SWA data and the original video image data is calculated, and when a preset threshold value is exceeded, the peak is determined as a peak to detect the peak data. A method for detecting finger veins using data. 19. The method of claim 18,
In the detecting of the finger veins, if the pixel value of the peak data is greater than a preset reference value, it is converted to 255, and the pixel value of the non-peak part is fixed to 0 to reconstruct the image image. Finger vein detection method using peak data . 12. The method of claim 11,
Finger vein detection method using peak data further comprising the step of detecting a finger joint using the detected peak data.

Images