TW201447625A - Palm vein recognition method using adaptive Gabor filter - Google Patents
Palm vein recognition method using adaptive Gabor filter Download PDFInfo
- Publication number
- TW201447625A TW201447625A TW102120313A TW102120313A TW201447625A TW 201447625 A TW201447625 A TW 201447625A TW 102120313 A TW102120313 A TW 102120313A TW 102120313 A TW102120313 A TW 102120313A TW 201447625 A TW201447625 A TW 201447625A
- Authority
- TW
- Taiwan
- Prior art keywords
- palm vein
- adaptive
- gabor filter
- standard deviation
- image
- Prior art date
Links
Landscapes
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
- Collating Specific Patterns (AREA)
Abstract
Description
本發明係一種手掌靜脈識別的方法，特別是一種具有自適應加伯濾波器的手掌靜脈識別方法。 The invention relates to a method for palm vein recognition, in particular to a palm vein recognition method with an adaptive Gabor filter.
習知技術，身分識別的方法，包括使用密碼，個人識別碼(PIN)，磁條卡，鑰匙，智能卡，但無論所採用的方法，它只能提供有限的安全性。許多生物特徵的識別系統，處理各種人的生理特徵或行為，包括有面部圖像、手的外型、指紋、掌紋、視網膜、手寫簽名、以及步態，以改善個人驗證的安全性。其中，每個系統都有優點和缺點。直接接觸的手指的指紋圖像中，經由感測器提取指紋的情況下，會引起的感測性能下降，特別是在工廠環境中，由於手指容易沾油、水分、灰塵等因素下，優質的指紋是很難取得。此外，虹膜/視網膜掃描儀的使用下，用戶必須將眼睛靠近掃描儀，導致不舒適的感覺、或是侵犯隱私的感覺。再者，手的形狀識別方法，可能會出現，用戶的使用者中，誰患有關節炎或風濕病的病患隱私問題，所以他們很少使用。因此，與其他上述的物理特性相比，手掌靜脈辨識的方法，已被開發來解決如上所述的問題。 Conventional techniques, identity identification methods, including the use of passwords, personal identification numbers (PINs), magnetic stripe cards, keys, smart cards, but regardless of the method used, it can only provide limited security. Many biometric recognition systems that deal with various human physiological characteristics or behaviors, including facial images, hand shapes, fingerprints, palm prints, retinas, handwritten signatures, and gaits to improve the security of personal verification. Among them, each system has advantages and disadvantages. In the fingerprint image of the finger that is in direct contact, the sensory performance is degraded when the fingerprint is extracted via the sensor, especially in a factory environment, because the finger is easily exposed to oil, moisture, dust, etc., high quality Fingerprints are hard to get. In addition, with the use of an iris/retina scanner, the user must bring their eyes close to the scanner, resulting in an uncomfortable feeling or a feeling of privacy violation. Furthermore, the shape recognition method of the hand may occur, and the user of the user who has arthritis or rheumatism has privacy problems, so they are rarely used. Therefore, the method of palm vein recognition has been developed to solve the above problems as compared with other physical properties described above.
本發明提一種自適應加伯濾波器的手掌靜脈識別方法，使用身體的一部分的手掌靜脈是一種理想用於生物特徵的識別方法，用於拍攝的血管圖案時，不容易受到不同的手指或手背的皮膚顏色變化影響。本發明手掌靜脈識別方法，其屬性的唯一性、穩定性、和較強的抗偽造性，是一種的良好的生物識別特點，足以提供身份驗證安全性和可靠性。 The invention provides an adaptive Jiabo filter for palm vein recognition. The palm vein of a part of the body is an ideal identification method for biometrics. When used for photographing blood vessel patterns, it is not easy to be subjected to different fingers or backs of hands. The effect of skin color changes. The method for identifying the palm vein of the present invention has unique properties, stability, and strong anti-forgery property, and is a good biometric feature, which is sufficient to provide identity verification security and reliability.
本發明一實施例，係一種自適應加伯濾波器的手掌靜脈識別方法，其中該手掌靜脈識別方法，包括步驟：(s1)使用一紅外線裝置擷取一手掌靜脈影像的紋路特徵；(s2)將擷取該手掌靜脈影像的紋路特徵，做影像處理；(s3)經由自適應加伯濾波器，擷取影像處理後該手掌靜脈影像的紋路特徵，並將該影像處理後該手掌靜脈影像切割成複數個矩陣區塊，該複數個矩陣區塊系MxM個，其中該自適應加伯濾波器於該複數個矩陣區塊的各自獨立矩陣區塊中，選取一方向角θ、一頻率u、以及一標準差σ(standard deviation)；以及(s4)使用正規化漢明距離(normalized Hamming dstance)，作該手掌靜脈影像的該複數個矩陣區塊匹配比對；其中該自適應加伯濾波器(Gabor filter)，具有頻域、以及空間域的高速識別能力，係G σ,u,θ (x,y)=g σ (x,y).exp[2πju(x cos θ+y sin θ)]， 其中，係一高斯函數，其中G σ,u,θ (x,y) 包括實部係R σ,u,θ (x,y)=g σ (x,y)˙cos[2πu(x cos θ+y sin θ)]、以及虛部係I σ,u,θ (x,y)=g σ (x,y)˙sin[2πu(x cos θ+y sin θ)]。 An embodiment of the present invention is a palm vein recognition method for an adaptive Gabor filter, wherein the palm vein recognition method comprises the steps of: (s1) using an infrared device to capture a texture feature of a palm vein image; (s2) The texture feature of the palm vein image is captured for image processing; (s3) the texture feature of the palm vein image after image processing is captured by an adaptive Gabor filter, and the palm vein image is processed after the image processing Forming a plurality of matrix blocks, the plurality of matrix blocks are MxM, wherein the adaptive Galerian filter selects a direction angle θ , a frequency u, in each independent matrix block of the plurality of matrix blocks And a standard deviation σ (standard deviation); and (s4) using a normalized Hamming dstance for the plurality of matrix block matching comparisons of the palm vein image; wherein the adaptive Gabor filter (Gabor filter), with high-frequency recognition ability in the frequency domain and spatial domain, is G σ , u , θ ( x , y ) = g σ ( x , y ).exp[2 πju ( x cos θ + y sin θ )], among them Is a Gaussian function, where G σ , u , θ ( x , y ) includes the real part R σ , u , θ ( x , y ) = g σ ( x , y ) ̇ cos [2 πu ( x cos θ + y sin θ )], and the imaginary part I σ , u , θ ( x , y ) = g σ ( x , y ) ̇ sin [2 πu ( x cos θ + y sin θ )].
s1~s4‧‧‧步驟 S1~s4‧‧‧step
s21~s22‧‧‧步驟 S21~s22‧‧‧Steps
s31~s33‧‧‧步驟 S31~s33‧‧‧Steps
s211~s213‧‧‧步驟 S211~s213‧‧‧Steps
s221~s222‧‧‧步驟 S221~s222‧‧‧Steps
s311~s314‧‧‧步驟 S311~s314‧‧‧Steps
s3211~s323‧‧‧步驟 S3211~s323‧‧‧Steps
s331~s334‧‧‧步驟 S331~s334‧‧‧Steps
s41~s42‧‧‧步驟 S41~s42‧‧‧Steps
圖1 手掌靜脈識別方法的主要步驟。 Figure 1 The main steps of the palm vein identification method.
圖2 步驟(s2)的方法。 Figure 2 The method of step (s2).
圖3 步驟(s3)的方法。 Figure 3 The method of step (s3).
圖4 該步驟(s21)的方法。 Figure 4 shows the method of this step (s21).
圖5 步驟(s22)的方法。 Figure 5 The method of step (s22).
圖6 步驟(s31)的該方向場(orientation field)估計方法。 Figure 6 is an estimation method of the orientation field of step (s31).
圖7 步驟(s32)的該中心頻率(center frequency)估計方法。 Figure 7 shows the center frequency estimation method of step (s32).
圖8 步驟(s33)的該標準差σ(standard deviation)估計方法。 Figure 8 shows the standard deviation σ (standard deviation) estimation method of step (s33).
圖9 步驟(s4)的方法。 Figure 9 The method of step (s4).
圖10 不同加伯濾波器。 Figure 10 Different Gabor filters.
圖11 4種方法的真正接受率(genuine acceptance rate)比較。 Figure 11 Comparison of the true acceptance rate of the four methods.
附件1 使用一紅外線裝置擷取一手掌靜脈影像的紋路特徵。 Attachment 1 uses an infrared device to capture the texture features of a palm vein image.
附件2 不同手掌靜脈影像。 Annex 2 Image of different palm veins.
附件3(a) 原手掌靜脈影像。 Annex 3 (a) Original palm vein image.
附件3(b) 使用奧圖的臨界值法(Otsu’s thresholding)，估計該手掌區域(palm region)。 Annex 3(b) Estimate the palm region using Otto's thresholding.
附件3(c) 使用內部邊界追蹤法(Sonka，Hlavac，Boyle的inner border tracing algorithm)，計算該手掌區域的內部邊界。 Annex 3(c) uses the internal boundary tracking method (Sonka, Hlavac, Boyle's inner border tracing algorithm) to calculate the inner boundary of the palm region.
附件3(d) 由小指和無名指之間的谷點作為第1基準點(P1)，由食指和中指之間的谷點作為第2基準點(P2)；第1基準點(P1)的X軸、Y軸直角座標為(X P1,Y P1)，第2基準點(P2)的X軸、Y軸直角座標為(X P2,Y P2)，該方向角θ係：θ=tan-1(Y P2-Y P1)/(X P2-X P1)。 Annex 3(d) The valley point between the little finger and the ring finger is used as the first reference point (P1), the valley point between the index finger and the middle finger is used as the second reference point (P2); and the X of the first reference point (P1) The right-angle coordinate of the axis and the Y-axis is ( X P 1 , Y P 1 ), and the coordinate of the X-axis and the Y-axis of the second reference point (P2) is ( X P 2 , Y P 2 ), and the direction angle θ is: θ =tan -1 ( Y P 2 - Y P 1 )/( X P 2 - X P 1 ).
附件3(e) 使用第1基準點(P1)、以及第2基準點(P2)的邊長作為一正方形的一邊，該正方形的4個轉角係C1、C2、C3、C4，其中第1基準點(P1)係轉角C1、第2基準點(P2) 係轉角C2，邊長係，該正方形係形成該影像範圍(ROI)。 Annex 3 (e) The length of the side using the first reference point (P1) and the second reference point (P2) As one side of a square, the four corners of the square are C1, C2, C3, and C4, wherein the first reference point (P1) is a corner C1 and the second reference point (P2) is a corner C2, and the side length is system The square forms the image range (ROI).
附件3(f) 其中，由5個手指頂端為5個峰點，相鄰兩個手指的底部鄰接點係為谷點，共有4個谷點。 Attachment 3(f), where the top of the five fingers is 5 peaks, the bottom adjacent points of the two adjacent fingers are valley points, and there are 4 valley points.
附件4(a) 原手掌靜脈影像。 Annex 4 (a) Original palm vein image.
附件4(b) 背景照明。 Annex 4 (b) Background lighting.
附件4(c) 原手掌靜脈影像減去背景照明，以轉換成均勻光(uniform light)。 Attachment 4(c) Original palm vein image minus background illumination for conversion to uniform light.
附件5(a) MXM複數個矩陣區塊，其中M=8。 Annex 5(a) MXM complex matrix blocks, where M=8.
附件5(b) 穩定區(stable area)，頻率u=0，當該變異差(variance)<H1。 Annex 5 (b) Stable area, frequency u = 0, when the variation is <H1.
附件5(c) 緩變區(slow change area)，頻率u=0.08，當H1<該變異差(variance)<H2。 Annex 5 (c) Slow change area, frequency u = 0.08, when H1 < the variance <H2.
附件5(d) 速變區(fast change area)，頻率u=0.2，當該變異差(variance)係屬於其他範圍。 Annex 5 (d) Fast change area, frequency u = 0.2, when the variance is in other ranges.
附件6 8個區域方向型(local direction patterns,LDPs)，每個區域方向型(local direction patterns,LDPs)係具有3x3的矩陣視窗作為遮罩(mask)。 Annex 6 8 local direction patterns (LDPs), each local direction patterns (LDPs) have a 3x3 matrix window as a mask.
附件7(a) 原手掌靜脈影像。 Annex 7 (a) Original palm vein image.
附件7(b) 取樣點(x,y)的離散函數f(x,y)，摺積後解碼為兩種靜脈碼(VeinCode)的實部。 Annex 7(b) The discrete function f(x, y) of the sampling point (x, y) is decomposed and decoded into the real part of the two vein codes (VeinCode).
附件7(c) 取樣點(x,y)的離散函數f(x,y)，摺積後解碼為兩種靜脈碼(VeinCode)的實部。 Annex 7(c) The discrete function f(x,y) of the sampling point (x,y), which is decomposed and decoded into the real part of the two vein codes (VeinCode).
附件7(d) 經由該自適應加伯濾波器後的影像。 本發明一實施例，係一種自適應加伯濾波器的手掌靜脈識別方法，如圖1所示，其中該手掌靜脈識別方法，包括步驟：(s1)使用一紅外線裝置擷取一手掌靜脈影像的紋路特徵，如附件1，附件2；(s2)將擷取該手掌靜脈影像的紋路特徵，做影像處理；(s3)經由自適應加伯濾波器，擷取影像處理後該手掌靜脈影像的紋路特徵，並將該影像處理後該手掌靜脈影像切割成複數個矩陣區塊，該複數個矩陣區塊系MxM個，其中該自適應加伯濾波器於該複數個矩陣區塊的各自獨立矩陣區塊中，選取一方向角θ、一頻率u、以及一標準差σ(standard deviation)；以及(s4)使用正規化漢明距離(normalized Hamming distance)，作該手掌靜脈影像的該複數個矩陣區塊匹配比對；其中該自適應加伯濾波器(Gabor filter)，具有頻域、以及空間域的高速識別能力，係G σ,u,θ (x,y)=g σ (x,y).exp[2πju(x cos θ+y sin θ)]， 其中，係一高斯函數，其中G σ,u,θ (x,y)包 括實部係R σ,u,θ (x,y)=g σ (x,y)˙cos[2πu(x cos θ+y sin θ)]、以及虛部係I σ,u,θ (x,y)=g σ (x,y)˙sin[2πu(x cos θ+y sin θ)]。 Annex 7(d) The image after passing through the adaptive Gabor filter. An embodiment of the present invention is an adaptive Jiabo filter for palm vein recognition, as shown in FIG. 1 , wherein the palm vein recognition method comprises the steps of: (s1) using an infrared device to capture a palm vein image. The texture features, such as Annex 1, Annex 2; (s2) will capture the texture features of the palm vein image for image processing; (s3) through the adaptive Gabor filter, capture the image of the palm vein image after image processing Feature, and the image of the palm vein is cut into a plurality of matrix blocks, the plurality of matrix blocks are MxM, wherein the adaptive Gabor filter is in each independent matrix region of the plurality of matrix blocks In the block, a direction angle θ , a frequency u, and a standard deviation σ (standard deviation) are selected; and (s4) a normalized Hamming distance is used as the plurality of matrix regions of the palm vein image. Block matching alignment; wherein the adaptive Gabor filter has a high frequency recognition capability in the frequency domain and the spatial domain, and is G σ , u , θ ( x , y ) = g σ ( x , y ) .exp[2 πju ( x Cos θ + y sin θ )], where Is a Gaussian function, where G σ , u , θ ( x , y ) includes the real part R σ , u , θ ( x , y ) = g σ ( x , y ) ̇ cos [2 πu ( x cos θ + y sin θ )], and the imaginary part I σ , u , θ ( x , y ) = g σ ( x , y ) ̇ sin [2 πu ( x cos θ + y sin θ )].
其中該步驟(s2)的方法，如圖2所示，包括步驟：(s21)擷取手掌靜脈影像的一影像範圍(ROI)；以及(s22)影像強化(image enhancement)。 The method of the step (s2), as shown in FIG. 2, includes the steps of: (s21) extracting a range of images (ROI) of the palm vein image; and (s22) image enhancement.
其中該步驟(s3)的方法，如圖3所示，包括步驟：(s31)一方向場(orientation field)估計；(s32)一中心頻率(center frequency)估計；以及(s33)該標準差σ(standard deviation)估計。 The method of the step (s3), as shown in FIG. 3, includes the steps of: (s31) an orientation field estimation; (s32) a center frequency estimation; and (s33) the standard deviation σ (standard deviation) estimate.
其中該步驟(s21)的方法，如圖4所示，其係將一手掌區域(palm region)的物件(object)，與一背景(background)分離，如附件 3(a)，係原手掌靜脈影像。該步驟(s21)的方法，包括：(s211)使用奧圖的臨界值法(Otsu’s thresholding)，估計該手掌區域(palm region)，如附件3(b)；(s212)使用內部邊界追蹤法(Sonka，Hlavac，Boyle的inner border tracing algorithm)，計算該手掌區域的內部邊界，如附件3(c)。其中，由5個手指頂端為5個峰點，相鄰兩個手指的底部鄰接點係為谷點，共有4個谷點，如附件3(f)，由小指和無名指之間的谷點作為第1基準點(P1)，由食指和中指之間的谷點作為第2基準點(P2)；第1基準點(P1)的X軸、Y軸直角座標為(X P1,Y P1)，第2基準點(P2)的X軸、Y軸直角座標為(X P2,Y P2)，該方向角θ係：θ=tan-1(Y P2-Y P1)/(X P2-X P1)，如附件3(d)；以及(s213)使用第1基準點(P1)、以及第2基準點(P2)的邊長作為一正方形的一邊，該正方形的4個轉角係C1、C2、C3、C4，其中第1基準點(P1)係轉角C1、第2基準點(P2)係轉角C2，邊長係，該正方形係形成該影像範圍(ROI)，如附件3(e)。 The method of the step (s21), as shown in FIG. 4, separates an object of a palm region from a background, such as the attachment 3(a), which is the original palm vein. image. The method of the step (s21) includes: (s211) estimating the palm region using an Otsu's thresholding method, such as Annex 3(b); (s212) using an internal boundary tracking method ( Sonka, Hlavac, Boyle's inner border tracing algorithm), calculates the inner boundary of the palm area, as in Annex 3 (c). Among them, the top of the five fingers is 5 peaks, and the bottom adjacent points of the two adjacent fingers are valley points. There are 4 valley points, as in Annex 3 (f), the valley between the little finger and the ring finger is used as the valley point. The first reference point (P1) is defined by the valley point between the index finger and the middle finger as the second reference point (P2); the X-axis and Y-axis orthogonal coordinates of the first reference point (P1) are ( X P 1 , Y P 1 The X-axis and Y-axis orthogonal coordinates of the second reference point (P2) are ( X P 2 , Y P 2 ), and the direction angle θ is: θ =tan -1 ( Y P 2 - Y P 1 )/( X P 2 - X P 1 ), as in Annex 3 (d); and (s213) using the first reference point (P1) and the side length of the second reference point (P2) As one side of a square, the four corners of the square are C1, C2, C3, and C4, wherein the first reference point (P1) is a corner C1 and the second reference point (P2) is a corner C2, and the side length is system The square forms the image range (ROI), as in Annex 3(e).
其中該步驟(s22)的方法，如圖5所示，如附件4(a)，係原手掌靜脈影像。該步驟(s22)的方法，其係於該複數個矩陣區塊中，(s221)使用雙立方差值法(bicubic interpolation)，以平均值(mean value)粗估該複數個矩陣區塊的背景照明，如附件4(b)；以及(s222)該複數個矩陣區塊，經由減去原始的背景照明，以轉換成均勻光(uniform light)，如附件4(c)。 The method of the step (s22), as shown in FIG. 5, is as shown in Annex 4(a), which is the original palm vein image. The method of the step (s22) is performed in the plurality of matrix blocks, and (s221) uses a bicubic interpolation method to roughly estimate the background illumination of the plurality of matrix blocks by a mean value. As in Annex 4(b); and (s222) the plurality of matrix blocks are converted to uniform light by subtracting the original background illumination, as in Annex 4(c).
其中該步驟(s31)該方向場(orientation field)估計，如圖6所示，其係包括：(s311)使用區域梯度來計算區域的該方向場(orientation field)，其中該自適應加伯濾波器(Gabor filter)具有8個區域方向型(local direction patterns,LDPs)，每個區域方向型(local direction patterns,LDPs)係具有3x3的矩陣視窗作為遮罩(mask)，如附件6；(s312)經由8個區域方向型(local direction patterns,LDPs)的掃描，總合相應該複數個矩陣區塊的像素(pixel)強度，產生8個特徵值， 係相應的方向強度L P ，其中係該 複數個矩陣區塊的像素(pixel)，由8個區域方向型(local direction patterns,LDPs)的3x3的矩陣視窗遮罩(mask)，形成摺積(convoluted)；(s313)經由3x3的矩陣視窗遮罩(mask)，形成摺積 (convoluted)後，相鄰畫素位置關係的強度加總為； 以及(s314)相應的該方向強度L P 中的最大值，係該自適應加伯濾波器(Gabor filter)的最佳方向值。 Wherein the step (s31) the orientation field estimation, as shown in FIG. 6, includes: (s311) calculating the orientation field of the region using the region gradient, wherein the adaptive filtering filter Gabor filter has 8 local direction patterns (LDPs), and each local direction patterns (LDPs) has a 3x3 matrix window as a mask, as shown in Annex 6; (s312) According to the scanning of 8 local direction patterns (LDPs), the pixel intensity of the plurality of matrix blocks is combined to generate 8 eigenvalues, corresponding to the directional intensity L P , wherein a pixel of the plurality of matrix blocks, which is convoluted by a 3x3 matrix window mask of 8 local direction patterns (LDPs); (s313) via 3x3 The matrix window mask (mask), after forming a convoluted, the intensity of the positional relationship of adjacent pixels is added to And (s314) the corresponding maximum value of the direction strength L P is the optimum direction value of the adaptive Gabor filter.
其中該步驟(s32)的該中心頻率(center frequency)估計，如圖7所示，其係依據一變異差(variance)，將相鄰畫素灰階形成的改變，視作弦波中的頻率，依該變異差(variance)的隨機正規分佈，形成3組的變異差(variance)範圍，如附件5(a)，係MXM複數個矩陣區塊，其中M=8。其中3組的變異差(variance)範圍包括：(s321)穩定區(stable area)，頻率u=0，當該變異差(variance)<H1，如附件5(b)；(s322)緩變區(slow change area)，頻率u=0.08，當H1<該變異差(variance)<H2，如附件5(c)；以及(s323)速變區(fast change area)，頻率u=0.2，當該變異差(variance)係其他範圍，如附件5(d)；其中，H1=I u -I σ ，H2=I u +I σ ，I u 係該變異差(variance)分佈的平均值，I σ 係該標準差σ(standard deviation)分佈的平均值。 The center frequency of the step (s32) is estimated, as shown in FIG. 7, which is based on a variation, and the change of the adjacent pixel gray scale is regarded as the frequency in the sine wave. According to the random normal distribution of the variance, three sets of variation range are formed, as in Annex 5 (a), which is a plurality of matrix blocks of MXM, where M=8. The variation range of the three groups includes: (s321) stable area, frequency u=0, when the variation is <H1, as in Annex 5(b); (s322). (slow change area), frequency u = 0.08, when H1 < the variance <H2, as in Annex 5 (c); and (s323) fast change area, frequency u = 0.2, when Variances are other ranges, as in Annex 5(d); where H1 = I u - I σ , H2 = I u + I σ , I u is the mean of the distribution of the variance, I σ The average of the standard deviation σ (standard deviation) distribution.
其中該步驟(s33)的該標準差σ(standard deviation)估計，如圖 8所示，其係二維正規高斯分佈，其係包括；(s331)使該自適應加伯濾波器的手掌靜脈識別方法，最佳化的該標準差σ(standard deviation)係使用2；(s332)決定最佳化的該方向角θ、該頻率u、以及該標準差σ(standard deviation)；(s333)取樣點(x,y)的離散函數f(x,y)摺積(convolutions)， 其具有實部，，其具有虛 部，；(s334)經由該自適應 加伯濾波器，取樣點(x,y)的離散函數f(x,y)，摺積後解碼為兩種靜脈碼(VeinCode)，其中，V R (x,y)=1，當C R (x,y) σ,u,θ 0，其中，V R (x,y)=0，當C R (x,y) σ,u,θ <0，其中，V I (x,y)=1，當C I (x,y) σ,u,θ 0，其中，V I (x,y)=0，當C I (x,y) σ,u,θ <0，如附件7(a)，係原手掌靜脈影像。如附件7(b)，係取樣點(x,y)的離散函數f(x,y)，摺積後解碼為兩種靜脈碼(VeinCode)的實部。如附件7(c)，係取樣點(x,y)的離散函數f(x,y)，摺積後解碼為兩種靜脈碼(VeinCode)的實部。如附件7(d)，係經由該自適應加伯濾波器後的影像。 Wherein the standard deviation σ (standard deviation) of the step (s33) is estimated as shown in FIG. 8, which is a two-dimensional normal Gaussian distribution, which includes; (s331) the palm vein recognition of the adaptive Gabor filter Method, the standard deviation of the standard deviation σ (standard deviation) is used 2 (s332) determining the direction angle θ of the optimization, the frequency u, and the standard deviation σ (standard deviation); (s333) the discrete function f(x, y) of the sampling point (x, y) is ( Convolutions), which has real parts, , it has an imaginary part, (s334) via the adaptive Galbergian filter, the discrete function f(x, y) of the sampling point (x, y) is decomposed and decoded into two vein codes (VeinCode), where V R ( x , y )=1, when C R ( x , y ) σ , u , θ 0, where V R ( x , y )=0, when C R ( x , y ) σ , u , θ <0, where V I ( x , y )=1, when C I ( x , y ) σ , u , θ 0, where V I ( x , y ) = 0, when C I ( x , y ) σ , u , θ <0, as in Annex 7 (a), is the original palm vein image. For example, in Annex 7(b), the discrete function f(x, y) of the sampling point (x, y) is decomposed and decoded into the real part of the two vein codes (VeinCode). For example, in Annex 7(c), the discrete function f(x, y) of the sampling point (x, y) is decomposed and decoded into the real part of the two vein codes (VeinCode). As shown in Annex 7(d), the image after passing through the adaptive Gabor filter.
其中該步驟(s4)，如圖9所示，係包括；(s41)使用正規化漢明距離(normalized Hamming distance)，其係：
如圖10所示，係單一加伯濾波器、複數個加伯濾波器、以及 本發明的自適應加伯濾波器，使用錯誤拒絕率(false rejection rate,FRR)、以及錯誤接收率(false acceptance rate,FAR)來比對效能。 As shown in FIG. 10, a single Gabor filter, a plurality of Gabor filters, and The adaptive Gabor filter of the present invention uses a false rejection rate (FRR) and a false acceptance rate (FAR) to compare performance.
如圖11所示，係本發明、拉普拉司掌紋(Laplacian palm)(Wang,Yau,et al.,2008)、多個多解濾波器(multi resolution filters,MRFs)(Lin & Fan,2004)，以及習知等4種的真正接受率(genuine acceptance rate)比較。 As shown in Fig. 11, the present invention, Laplacian palm (Wang, Yau, et al., 2008), and multiple multi-resolution filters (MRFs) (Lin & Fan, 2004) ), and the comparison of the true acceptance rate of the four kinds of knowledge.
以上所述，乃僅記載本發明為呈現解決問題所採用的技術手段之較佳實施方式或實施例而已，並非用來限定本發明專利實施之範圍。即凡與本發明專利申請範圍文義相符，或依本發明專利範圍所做的均等變化與修飾，皆為本發明專利範圍所涵蓋。 The above description is only intended to describe the preferred embodiments or embodiments of the present invention, which are not intended to limit the scope of the invention. That is, the equivalent changes and modifications made in accordance with the scope of the patent application of the present invention or the scope of the invention are covered by the scope of the invention.
s1~s4‧‧‧步驟 S1~s4‧‧‧step
Claims (9)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| TW102120313A TW201447625A (en) | 2013-06-07 | 2013-06-07 | Palm vein recognition method using adaptive Gabor filter |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| TW102120313A TW201447625A (en) | 2013-06-07 | 2013-06-07 | Palm vein recognition method using adaptive Gabor filter |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| TW201447625A true TW201447625A (en) | 2014-12-16 |
Family
ID=52707504
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| TW102120313A TW201447625A (en) | 2013-06-07 | 2013-06-07 | Palm vein recognition method using adaptive Gabor filter |
Country Status (1)
| Country | Link |
|---|---|
| TW (1) | TW201447625A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107980140A (en) * | 2017-10-16 | 2018-05-01 | 厦门中控智慧信息技术有限公司 | A kind of recognition methods of vena metacarpea and device |
| CN108509886A (en) * | 2018-03-26 | 2018-09-07 | 电子科技大学 | Vena metacarpea recognition methods based on the judgement of vein pixel |
| CN113034741A (en) * | 2021-03-02 | 2021-06-25 | 桂林电子科技大学 | Palm vein intelligent lock based on DWT-DCT (discrete wavelet transform-discrete cosine transform) transform encryption algorithm |
| CN114241534A (en) * | 2021-12-01 | 2022-03-25 | 佛山市红狐物联网科技有限公司 | Rapid matching method and system for full-palmar venation data |
-
2013
- 2013-06-07 TW TW102120313A patent/TW201447625A/en unknown
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107980140A (en) * | 2017-10-16 | 2018-05-01 | 厦门中控智慧信息技术有限公司 | A kind of recognition methods of vena metacarpea and device |
| CN107980140B (en) * | 2017-10-16 | 2021-09-14 | 厦门熵基科技有限公司 | Palm vein identification method and device |
| CN108509886A (en) * | 2018-03-26 | 2018-09-07 | 电子科技大学 | Vena metacarpea recognition methods based on the judgement of vein pixel |
| CN108509886B (en) * | 2018-03-26 | 2021-08-17 | 电子科技大学 | Palm vein identification method based on vein pixel point judgment |
| CN113034741A (en) * | 2021-03-02 | 2021-06-25 | 桂林电子科技大学 | Palm vein intelligent lock based on DWT-DCT (discrete wavelet transform-discrete cosine transform) transform encryption algorithm |
| CN114241534A (en) * | 2021-12-01 | 2022-03-25 | 佛山市红狐物联网科技有限公司 | Rapid matching method and system for full-palmar venation data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Han et al. | Palm vein recognition using adaptive Gabor filter | |
| Lee | A novel biometric system based on palm vein image | |
| Ma et al. | Iris recognition based on multichannel Gabor filtering | |
| US7899217B2 (en) | Multibiometric multispectral imager | |
| Zhang et al. | Palm vein extraction and matching for personal authentication | |
| WO2017059591A1 (en) | Finger vein identification method and device | |
| WO2010129074A1 (en) | System and method for identifying a person with reference to a sclera image | |
| CA2671561A1 (en) | Method and apparatus for extraction and matching of biometric detail | |
| WO2013087026A1 (en) | Locating method and locating device for iris | |
| Trabelsi et al. | A new multimodal biometric system based on finger vein and hand vein recognition | |
| Ramteke et al. | Extraction of palmprint texture features using combined DWT-DCT and local binary pattern | |
| TW201447625A (en) | Palm vein recognition method using adaptive Gabor filter | |
| Shah et al. | Palm vein pattern-based biometric recognition system | |
| Estudillo-Romero et al. | The Hermite transform: An alternative image representation model for iris recognition | |
| Podgantwar et al. | Extraction of finger vein patterns using gabor filter in finger vein image profiles | |
| Zhang et al. | Reflection-type finger vein recognition for mobile applications | |
| Al-Tamimi | H. A survey on the vein biometric recognition systems: Trends and challenges | |
| Swati et al. | Iris recognition using Gabor | |
| Mohammed et al. | Conceptual analysis of Iris Recognition Systems | |
| Soltany et al. | Daugman’s Algorithm Enhancement for Iris Localization | |
| Sharma et al. | Improved human identification using finger vein images | |
| Bennet et al. | Fingerprint matching using hierarchical level features | |
| Patil | Iris recognition based on Gaussian-Hermite moments | |
| Safronova et al. | Palm vein recognition algorithm using multilobe differential filters | |
| Tajbakhsh et al. | Feature fusion as a practical solution toward noncooperative iris recognition |