1. A fingertip blood sampling point positioning method based on vein segmentation and angular point detection is characterized by comprising the following steps: acquiring a fingertip vein image; extracting the characteristics of the fingertip vein images, and fusing the extracted characteristics to obtain a fused characteristic diagram; segmenting the fusion characteristic graph to obtain a fingertip vein area; carrying out vein network refinement and corner point detection on the fingertip vein area, and selecting the corner point closest to the center point of the fingertip blood sampling area as a blood sampling point.

2. The method for positioning the fingertip blood sampling point based on the vein segmentation and the angular point detection is characterized in that the process of obtaining the fusion feature map comprises the following steps: performing edge 0 complementing operation on the fingertip vein image to enable the pixel of the image to be 256 multiplied by 256; dividing the finger vein image after 0 supplementation into 16 multiplied by 16 sub-images, and calculating the differential excitation of each sub-image; setting the central window width of a Gabor filter according to the differential excitation, and performing filtering processing on each sub-image by adopting the Gabor filter with the set central window width to obtain finger vein region characteristic value data in different directions and window widths; and performing weighted summation processing on the obtained characteristic values to obtain characteristic values in different directions on the same filter window width, and aggregating all the fused characteristic values to obtain a fused characteristic diagram.

3. The fingertip blood sampling method based on vein segmentation and angular point detection as claimed in claim 2A point localization method, wherein the process of computing differential excitations for each sub-pattern comprises: randomly selecting a vein subimage block I₀(u, v) in which I₀(u, v) represents the sub image blocks in the u row and the v column; using a gradient template W_xAnd W_yConvolving with the selected vein sub-image blocks respectively to obtain gradient components g of each pixel point in the horizontal direction and the vertical direction_xAnd g_y(ii) a According to the gradient component g of each pixel point in the horizontal direction_xAnd a gradient component g in the vertical direction_yCalculating differential excitation of the sub-image blocks; the expression for computing the differential excitation is:

where dif (u, v) represents the differential excitation of the sub-image blocks in the u-th row and the v-th column, g_xRepresenting the gradient component, g, of the pixel in the horizontal direction_yRepresenting the gradient component in the vertical direction of the pixel, I₀And (u, v) represents the u-th row and v-th column sub image block.

4. The method for locating the blood sampling point of the fingertip based on the vein segmentation and the angular point detection as claimed in claim 2, wherein the formula for setting the central window width of the Gabor filter according to the differential excitation is as follows:

w represents the central window width of the Gabor filter bank, and dif (u, v) represents the differential excitation of the sub-image blocks in the u-th row and the v-th column.

5. The method for positioning the fingertip blood sampling point based on the vein segmentation and the angular point detection as claimed in claim 2, wherein the process of filtering each sub-image by using the Gabor filter group with the well-set central window width comprises: connecting 8 Gabor filters in different directions and 5 Gabor filters with different window widths to form a Gabor filter bank; and performing convolution filtering on the finger vein sub-image blocks by adopting a Gabor filter bank, extracting characteristic parameters, wherein 40 characteristic parameters can be obtained from each pixel point, and the window widths of 5 Gabor filters with different window widths in the Gabor filter bank are W, W +/-2 and W +/-4 respectively.

6. The method for locating the blood sampling point of the fingertip based on the vein segmentation and the angular point detection as claimed in claim 2, wherein the process of performing the weighted summation processing on the obtained feature values comprises: for 40 characteristic values extracted from each pixel on the image, a local binary mode is adopted to fuse 8 characteristics in different directions on the same window width, and the Gabor characteristic of one pixel point is P_s,tS e (1, …,8) represents the direction, t e (1, …,5) represents the window width, and the average value of the 8 direction feature amplitudes is taken as the feature value weighting quantity avg; carrying out binarization processing on the characteristic amplitude of each azimuth according to the characteristic value weighting quantity to obtain a characteristic value after amplitude characteristic fusion;

the formula for obtaining the weighted quantity avg of the fused eigenvalue is as follows:

avg＝(P_1,t+P_2,t…+P_8,t)/8

the amplitude features in 8 directions are expressed as:

wherein T(s) represents amplitude characteristics after fusion of 8 directions, P_s,tAnd the Gabor characteristic of one pixel point is represented, and the avg represents the characteristic value weighting quantity.

7. The method for positioning the fingertip blood sampling point based on the vein segmentation and the angular point detection as claimed in claim 1, wherein the process of segmenting the fused feature map comprises: constructing a vein image energy function, calculating the energy of the fusion characteristic diagram by adopting the vein image energy function, and introducing a weight factor beta into the energy function for weighting; adopting an undirected graph G which is < V, E > to represent an energy graph of the vein image, wherein V is a vertex set, and E is an edge set; calculating pixels in the graph according to an undirected graph, wherein when the difference between two adjacent pixels is larger, the energy is smaller, minimizing an energy function by adopting minimized graph segmentation, setting a label of a target as 1 and a label of a background as 0, segmenting the boundary of the target and the background according to the set label, and performing binarization processing on the segmented graph to obtain a fingertip vein region; the energy function of the vein image is defined as:

wherein R is_p(gamma) represents the distribution of multiscale venous features, S_a,b(gamma) represents Gaussian probability distribution, p represents vein features F corresponding to one point in the image after multi-scale transformation, and a and b represent adjacent areas with point adjacent areas of 2 and 4 respectively; gamma denotes the segmentation labels of the foreground and background; n denotes the set of all pairs of contiguous pixels in the image and β denotes the weighting factor.

8. The method for positioning the fingertip blood sampling point based on the vein segmentation and the angular point detection as claimed in claim 1, wherein the process of performing the vein network refinement and the angular point detection processing on the fingertip vein area comprises: extracting the central line of the segmented vein image, thinning the vein image, reducing the influence of burrs in the segmented image on angular point detection, and calculating the gray scale generated after the thinned image translates (u, v) pixel points in any direction to be converted into E (u, v); performing second-order Taylor expansion on the gray scale transformation E (u, v), and simplifying to obtain an autocorrelation matrix M of the E (u, v), wherein the eigenvalue of the M is lambda₁And λ₂(ii) a Setting a threshold value K, calculating a corner response value C of each pixel point according to the characteristic value and the threshold value, calculating the corner response value of each pixel point, and if the response value C of each pixel point is reached>0.1 max (c), the pixel is considered as a corner point, otherwise it is not.

9. The method for locating the blood sampling point of the fingertip based on the vein segmentation and the angular point detection as claimed in claim 8, wherein the formula for calculating the gray level variation is as follows:

wherein w (x, y) represents a Gaussian function, I (x, y) represents the gray value of the refined vein image in the window at the position of the corresponding pixel point (x, y), M is a 2 x 2 matrix, and the expression is as follows:

wherein, I_xIs the gradient of the pixel points in the window in the x direction, I_yThe gradient of the pixel points in the window in the y direction is shown.

10. The method for positioning the fingertip blood sampling point based on the vein segmentation and the angular point detection as claimed in claim 1, wherein the process of selecting the blood sampling point comprises: selecting an area within 1.5cm from a fingertip as a blood sampling area, calculating a central rectangular coordinate of the blood sampling area according to a finger vein image, comparing the coordinates of all corner points of the blood sampling area with the central rectangular coordinate, and selecting the corner point closest to the blood sampling area as a blood sampling point.

Background

The blood sampling from the tip of a fingertip is a very common blood collection inspection mode, and the traditional fingertip blood sampling mode is mainly characterized in that the blood is manually collected after the fingertip vein is judged by medical staff. Compare with medical personnel's manual work blood sampling, the blood sampling of intelligence fingertip can reduce medical personnel's work load, promotion work efficiency by a wide margin, can avoid medical personnel to infect the infectious disease simultaneously. In the fingertip blood collection process, because the capillary vessels of the fingertip are less, the fingertip is often required to be squeezed when the bleeding amount is insufficient, and the tissue fluid can permeate into the blood to cause the blood sample to be diluted, so that the detection result is inaccurate. The fingertip vein junction is selected as a blood sampling point, so that the non-extrusion bleeding amount is greatly increased, and the blood detection accuracy is improved. How to acquire the finger tip vein junction is a problem to be solved urgently at present.

Disclosure of Invention

In order to solve the problems in the prior art, the invention provides a fingertip blood sampling point positioning method based on vein segmentation and angular point detection, which comprises the following steps: acquiring a fingertip vein image; extracting the characteristics of the fingertip vein images, and fusing the extracted characteristics to obtain a fused characteristic diagram; segmenting the fusion characteristic graph to obtain a fingertip vein area; carrying out vein network refinement and corner point detection on the fingertip vein area, and selecting the corner point closest to the center point of the fingertip blood sampling area as a blood sampling point.

Preferably, the process of obtaining the fused feature map includes: performing edge 0 complementing operation on the fingertip vein image to enable the pixel of the image to be 256 multiplied by 256; dividing the finger vein image after 0 supplementation into 16 multiplied by 16 sub-images, and calculating the differential excitation of each sub-image; setting the central window width of a Gabor filter according to the differential excitation, and performing filtering processing on each sub-image by adopting the Gabor filter with the set central window width to obtain finger vein region characteristic value data in different directions and window widths; and performing weighted summation processing on the obtained characteristic values to obtain characteristic values in different directions on the same filter window width, and aggregating all the fused characteristic values to obtain a fused characteristic diagram.

Further, the process of calculating the differential excitation of each sub-pattern includes: selecting a vein subimage block I₀(u, v) wherein I₀(u, v) represents the sub image block of the u row and the v column, and adopts a gradient template W_xAnd W_yRespectively convolving with the selected vein sub-image block to obtain the gradient components g of each pixel point in the horizontal direction and the vertical direction_xAnd g_y(ii) a According to the gradient component g of each pixel point in the horizontal direction_xAnd a gradient component g in the vertical direction_yCalculating differential excitation of the sub-image blocks; the expression for computing the differential excitation is:

where dif (u, v) represents the differential excitation of the sub-image blocks in the u-th row and the v-th column, g_xRepresenting the gradient component, g, of the pixel in the horizontal direction_yRepresenting the gradient component in the vertical direction of the pixel, I₀And (u, v) represents the u-th row and v-th column sub image block.

Further, the formula for setting the central window width of the Gabor filter according to the differential excitation is:

w represents the central window width of the Gabor filter bank, and dif (u, v) represents the differential excitation of the sub-image blocks in the u-th row and the v-th column.

Further, the process of performing filtering processing on each sub-image by using the Gabor filter group with the set center window width includes: connecting 8 Gabor filters in different directions and 5 Gabor filters with different window widths to form a Gabor filter bank; and performing convolution filtering on the finger vein sub-image blocks by adopting a Gabor filter bank, extracting characteristic parameters, wherein 40 characteristic parameters can be obtained from each pixel point, and the window widths of 5 Gabor filters with different window widths in the Gabor filter bank are W, W +/-2 and W +/-4 respectively.

Further, the process of performing weighted summation processing on the obtained feature values includes: for 40 features extracted from each pixel on the image, a local binary mode is adopted to fuse 8 features in different directions on the same window width, and the Gabor feature of one pixel point is P_s，tS ∈ (1,.., 8) represents a direction, t ∈ (1,.., 5) represents a window width, an average value of 8 direction feature amplitudes is used as a threshold value to binarize the feature amplitude of each direction, and a fused feature value weighting quantity avg is as follows: avg ═ P_1，t+P_2，t…+P_8，t) The amplitude feature fusion in 8, 8 directions is expressed as:

wherein T(s) represents amplitude characteristics after fusion of 8 directions, P_s，tAnd the Gabor characteristic of one pixel point is represented, and the avg represents the characteristic value weighting quantity.

Preferably, the process of segmenting the fused feature map includes: constructing a vein image energy function, calculating the energy of the fusion characteristic diagram by adopting the vein image energy function, and introducing a weight factor beta into the energy function for weighting; adopting an undirected graph G ═ V, E > to represent an energy graph of the vein image, wherein V and E are the sets of vertexes and edges respectively; when the difference between two adjacent pixels is larger, the energy is smaller, the minimized energy function is cut by adopting a minimized image, the label of the target is set as 1, the label of the background is set as 0, the boundary of the target and the background is divided, and the image after the division processing is subjected to binarization processing to obtain a fingertip vein area; the energy function of the vein image is defined as:

wherein R is_p(gamma) represents the distribution of multiscale venous features, S_a，b(gamma) represents Gaussian probability distribution, p represents vein features F corresponding to one point in the image after multi-scale transformation, and a and b represent adjacent areas with point adjacent areas of 2 and 4 respectively; γ denotes a division index of the foreground and the background, γ ═ 1 denotes the foreground, and γ ═ 0 denotes the background; n represents a set formed by all adjacent pixel pairs in the image, beta is a weight factor, the larger the value of the target with a single shape and concentrated areas is, and the smaller weight factor is more suitable for the target with a complex and relatively discrete local detail shape.

Preferably, the process of performing vein network refinement and corner detection processing on the fingertip vein region includes: extracting the central line of the segmented vein image, thinning the vein image, reducing the influence of burrs in the segmented image on angular point detection, and calculating the gray scale generated after the thinned image translates (u, v) pixel points in any direction to be converted into E (u, v); performing second-order Taylor expansion on the gray scale transformation E (u, v), and simplifying to obtain an autocorrelation matrix M of the E (u, v), wherein the eigenvalue of the M is lambda₁And λ₂(ii) a Setting a threshold value K, calculating an angular point response value C of each pixel point according to the characteristic value and the threshold value, if the response value C of the pixel point is more than 0.1 × Max (C), considering the pixel as an angular point, otherwise, not considering the pixel as an angular point.

Further, the formula for calculating the gray level variation is as follows:

wherein w (x, y) represents a Gaussian function, I (x, y) represents the gray value of the refined vein image in the window at the position of the corresponding pixel point (x, y), M is a 2 x 2 matrix, and the expression is as follows:

wherein, I_xIs the gradient of the pixel points in the window in the x direction, I_yThe gradient of the pixel points in the window in the y direction is shown.

Preferably, the process of selecting the blood sampling point comprises the following steps: selecting an area within 1.5cm from a fingertip as a blood sampling area, calculating a central rectangular coordinate of the blood sampling area according to a finger vein image, comparing the coordinates of all corner points of the blood sampling area with the central rectangular coordinate, and selecting the corner point closest to the blood sampling area as a blood sampling point.

The invention has the beneficial effects that:

1. according to the method, parameters of a Gabor filter bank are set adaptively by adopting local differential excitation of finger vein images, the space-frequency resolution is adjusted dynamically, and more vein network characteristic details can be extracted while noise interference is avoided;

2. according to the method, finger vein image segmentation is realized by adopting a weighted energy function according to the vein image, and the integrity of a segmentation target area of the finger vein image and the continuity and visual consistency of a boundary are better;

3. the blood sampling point is the intersection of the vein at the corner detection and identification part, so that the amount of non-extruded blood is more during the blood sampling of the fingertip, and the accuracy of the fingertip blood detection is effectively improved.

Drawings

FIG. 1 is a flow chart of a fingertip blood sampling point positioning method of the present invention;

FIG. 2 is a finger vein image acquired by near infrared imaging in accordance with the present invention;

FIG. 3 is a flow chart of fingertip vein feature extraction according to the present invention;

FIG. 4 is a flowchart of the finger vein image segmentation of the present invention;

FIG. 5 is an image generated after vein refinement according to the present invention;

FIG. 6 is a flow chart of blood sampling point selection according to the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

A fingertip blood sampling point positioning method based on vein segmentation and corner detection, as shown in fig. 1, the method includes: acquiring a fingertip vein image; extracting the characteristics of the fingertip vein images, and fusing the extracted characteristics to obtain a fused characteristic diagram; segmenting the fusion characteristic graph to obtain a fingertip vein area; carrying out vein network refinement and corner point detection on the fingertip vein area, and selecting the corner point closest to the center point of the fingertip blood sampling area as a blood sampling point.

In the process of collecting the finger vein image, firstly, the finger is irradiated by the infrared lamp, and then the finger vein image is obtained by using the near infrared imaging, and the obtained value vein image is shown in fig. 2.

The process of extracting the features of the fingertip vein image and fusing the extracted features is shown in fig. 3, and the method comprises the following steps: performing edge 0 complementing operation on the fingertip vein image to enable the pixel of the image to be 256 multiplied by 256; dividing the finger vein image after 0 supplementation into 16 multiplied by 16 sub-images, and calculating the differential excitation of each sub-image; setting the central window width of a Gabor filter according to the differential excitation, and performing filtering processing on each sub-image by adopting the Gabor filter with the set central window width to obtain finger vein region characteristic value data in different directions and window widths; and performing weighted summation processing on the obtained characteristic values to obtain characteristic values in different directions on the same filter window width, and aggregating all the fused characteristic values to obtain a fused characteristic diagram.

The process of setting the central window width of the Gabor filter according to the differential excitation includes: selecting a vein subimage block I₀(u, v), (u, v) represent the u row and v column sub image block, and adopt the gradient template W_x，W_yConvolved with an imageGradient component g to each pixel point in horizontal and vertical directions_xAnd g_y

g_x＝I₀(u，v)*W_x

g_y＝I₀(u，v)*W_y

W_x、W_yGradient templates representing the x and y directions, respectively, the differential excitation dif (u, v) of a sub-image block is represented as:

the formula for setting the center window width of the Gabor filter based on differential excitation is:

wherein if (u, v) represents the differential excitation of the sub image blocks in the u-th row and the v-th column, g_xRepresenting the gradient component, g, of the pixel in the horizontal direction_yRepresenting the gradient component in the vertical direction of the pixel, I₀(u, v) denotes the u-th row and v-th column sub image block, W denotes the central window width of the Gabor filter bank, and dif (u, v) denotes the differential excitation of the u-th row and v-th column sub image block.

And taking W as a center, forming a Gabor filter bank by using 8 different directions, a central window W and the positive and negative directions of the central window W and a total of 5 different scales with 2 as step lengths to perform convolution filtering on the finger vein sub-image blocks, extracting characteristic parameters, and obtaining 40 characteristic parameters by each pixel point.

Fusing 8 features in different directions on the same window width by adopting a local binary mode for 40 features extracted from each pixel on the image, wherein the Gabor feature of one pixel point is P_s，tS ∈ (1,..., 8) denotes a direction, and t ∈ (1,.., 5) denotes a window widthAnd using the average value of the 8 directional characteristic amplitudes as a threshold value to binarize the characteristic amplitude of each azimuth, wherein the fused characteristic value weighting quantity avg is as follows: avg ═ P_1，t+P_2，t…+P_8，t) The amplitude feature fusion in 8, 8 directions is expressed as:

the finger vein image segmentation process is as shown in fig. 4, a vein image energy function is constructed, energy of a fusion feature map is calculated by adopting the vein image energy function, a weighting factor beta is introduced into the energy function for weighting, and the energy function of the vein image is defined as:

R_p(gamma) represents the distribution of multiscale venous features, S_a，b(gamma) representing Gaussian probability distribution, wherein p is a vein feature F corresponding to one point in the image after multi-scale transformation, and a and b are adjacent areas with point adjacent areas of 2 and 4; gamma is a segmentation label of the foreground and the background, wherein gamma is 1 to represent the foreground, and gamma is 0 to represent the background; n is a set formed by all adjacent pixel pairs in the image, beta is a weight factor, the object with a single shape and concentrated areas has a larger value, and a smaller weight factor is more suitable for the object with a complex and relatively discrete local detail shape. The vein image is segmented, and the process comprises the following steps: using undirected graph G ═<V，E>And V and E are respectively a set of a vertex and an edge, when the difference between two adjacent pixels is larger, the energy is smaller, the energy function is minimized by minimizing graph segmentation, the label of the target is set as 1, the label of the background is set as 0, the boundary of the target and the background is segmented, and the graph after segmentation is subjected to binarization processing to obtain a fingertip vein region.

The refined and binarized vein network diagram is shown in fig. 5, and the flow is as follows: extracting the central line of the segmented vein image, thinning the vein image to reduce the influence of burrs in the segmented image on the angular point detection,

calculating gray scale transformation E (u, v) generated after (u, v) pixel points of the thinned image are translated in any direction,

the w (x, y) window function adopts a Gaussian function, I (x, y) is the gray value of the refined vein image in the window at the position of the corresponding pixel point (x, y), and the binary Taylor is obtained after expansion and simplification:

I_x，I_ythe gradients of the pixel points in the window in the x and y directions respectively,

e (u, v) of an autocorrelation matrix M with an eigenvalue of λ₁And λ₂Wherein M is

E is an identity matrix, and a corner response value C-lambda is calculated according to the characteristic value₁λ₂+K(λ₁+λ₂) And setting a threshold value K to be 0.05, calculating the angular point response value of each pixel point, and judging the pixel point as the angular point if the response value C of the pixel point is more than 0.1 Max (C).

The process of positioning the blood sampling points according to the processing result is shown in fig. 6, an area within 1.5cm from the fingertip is selected as a blood sampling area, the central rectangular coordinate of the blood sampling area is calculated according to the finger vein image, the distances between the coordinates of all the angular points of the blood sampling area and the central rectangular coordinate are compared, and the angular point closest to the coordinate is selected as the blood sampling point.

The above-mentioned embodiments, which further illustrate the objects, technical solutions and advantages of the present invention, should be understood that the above-mentioned embodiments are only preferred embodiments of the present invention, and should not be construed as limiting the present invention, and any modifications, equivalents, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.