Auricular point identification system based on big data

Abstract

The invention discloses an auricular point recognition system based on big data, and relates to the technical field of auricular point recognition, a judgment and recognition unit is arranged to judge the recognition mode of human ear image data, and model recognition or image recognition is selected to carry out auricular point recognition; the result analysis module analyzes the plurality of auricular point positioning images, quickly screens out neighbor images with relatively high contour similarity from historical data based on a plurality of randomly selected ear contours, analyzes the similarity of regional contours of the auricular points, extracts deviation auricular points, divides a plurality of registration image sets and correspondingly generates registration recognition data; according to the method, image recognition can be carried out on a conventional ear, further distinguishing can be carried out for the situation that the overall contour is conventional but local auricular point structures are inconsistent, matching and auricular point recognition can be rapidly completed, the auricular point recognition and positioning efficiency is improved, dependence on an auricular point recognition model is reduced, and the user experience is improved. And the calculation resource consumption is reduced while the precision is ensured.

Description

TECHNICAL FIELD

[0001] The application relates to the ear acupoint recognition technical field, in particular to an ear acupoint recognition system based on big data. BACKGROUND

[0002] In the development process of modernization of traditional Chinese medicine diagnosis and treatment technology, ear acupoint recognition as an important auxiliary diagnosis and treatment method has been widely concerned, at present, an ear acupoint recognition model is used to recognize and position ear acupoints, which is a common method;

[0003] The method can improve the accuracy of ear acupoint recognition to a certain extent by training a model to learn the characteristics of ear acupoints, however, most people's ears belong to a relatively conventional type, for this kind of conventional ear, the image comparison method can be used to realize the recognition and positioning of ear acupoints, if the ear acupoint recognition model is still used at this time, the waste of computing resources will be caused, and the operation cost and time cost of the system will be increased;

[0004] In order to solve the above problems, the application provides a solution. SUMMARY

[0005] The application aims to provide an ear acupoint recognition system based on big data, in order to solve the problems in the background art.

[0006] The application provides an ear acupoint recognition system based on big data, which comprises:

[0007] A determination and recognition unit is used to determine the recognition mode of the human ear image data according to a preset determination step after receiving the human ear image data of a target patient, and the recognition mode comprises model recognition and image recognition;

[0008] The determination and recognition unit is also used to frame the region contour of a plurality of ear acupoints in the human ear image data of the target patient when the human ear image data of the target patient is determined to be image recognition, so as to obtain an ear acupoint positioning image of the target patient.

[0009] A model recognition unit is used to input the human ear image data determined to be model recognition into a pre-stored ear acupoint recognition model for ear acupoint recognition to obtain an ear acupoint positioning image of the target patient.

[0010] A result analysis module is used to analyze all the ear acupoint positioning images transmitted by the model recognition unit to obtain determination and recognition data of the contours of a plurality of ears.

[0011] Further, the analysis steps of the determination and recognition data of the contours of a plurality of ears are as follows:

[0012] S11: Mark all ear acupoint positioning images stored in the result analysis module as A1, A2,..., Aa, a≥1, and mark all ear acupoints that can be recognized by the ear acupoint recognition model as B1, B2,..., Bb, b≥1;

[0013] S12: Extract the contour of the ear in the ear acupoint positioning image A1 using an image contour extraction algorithm, select the contour as the reference contour, and select the ear acupoint positioning image A1 as the reference image. Similarly, extract the contours C1, C2,..., Ca-1 of the ears in the ear acupoint positioning images A2, A3,..., Aa in turn;

[0014] S13: Calculate the similarity between the contour C1 and the reference contour using a similarity algorithm. If the similarity is greater than or equal to P1, then the ear acupoint positioning image A2 from which the contour C1 is extracted is considered as a similar image of the reference image. P1 is a pre-set image similarity threshold. Similarly, calculate the similarity between the contours C2, C3,..., Ca-1 and the reference contour in turn. Compare the similarity with P1, and based on the comparison result, obtain several similar images of the reference image;

[0015] S14: According to the pre-set screening rule, obtain all neighboring images of the reference image relative to the ear acupoints B1, B2,..., Bb from all similar images of the reference image. The screening rule is as follows:

[0016] S141: Mark all similar images of the reference image as D1, D2,..., Dd in turn according to the order of being similar images of the reference image, 1≤d≤a-1;

[0017] S142: Extract the region contour E1 of the ear acupoint B1 in the reference image using an image contour extraction algorithm. Similarly, extract the region contours F1, F2,..., Fd of the ear acupoint B1 in the similar images D1, D2,..., Dd in turn;

[0018] S143: Determine whether the similar image D1 is a neighboring image of the reference image relative to the ear acupoint B1: calculate the similarity between the region contour F1 and the region contour E1 using a similarity algorithm, and compare the similarity with P2. If the value of the similarity is greater than or equal to P2, then the similar image D1 is determined to be a neighboring image of the reference image relative to the ear acupoint B1, otherwise, the similar image D1 is determined not to be a neighboring image of the reference image relative to the ear acupoint B1. P2 is a pre-set ear acupoint similarity threshold based on the ear acupoint B1;

[0019] S144: Determine whether the similar images D2, D3,..., Dd are neighboring images of the reference image relative to the ear acupoint B1 in turn according to S143, and based on the determination result, obtain all neighboring images of the reference image relative to the ear acupoint B1.

[0020] S145: obtaining all the neighboring images of the reference image relative to the ear points B2, B3,..., Bb in sequence according to S141 to S144;

[0021] S15: traversing all the neighboring images of the reference image relative to the ear points B1, B2,..., Bb and extracting all the neighboring images G1, G2,..., Gg, g≥1, whose occurrence frequencies are greater than or equal to P3 from the neighboring images, where the occurrence frequency refers to the number of repeated occurrences, and P3 is a preset frequency scalar;

[0022] S16: creating an empty absolute image set H1 of the reference image, and adding a number of neighboring images into the absolute image set H1 from the neighboring images G1, G2,..., Gg according to a preset screening and adding rule to obtain the final absolute image set H1 of the reference image, and the screening and adding rule is as follows:

[0023] S161: obtaining the occurrence frequency H1 of the neighboring image G1 from all the neighboring images of the reference image relative to the ear points B1, B2,..., Bb, and if H1=b, then adding the neighboring image G1 into the absolute image set H1, otherwise, not doing any processing;

[0024] S162: obtaining the occurrence frequencies of the neighboring images G2, G3,..., Gg in sequence according to S161, comparing the occurrence frequency with b each time the occurrence frequency is obtained, and selecting whether to add the corresponding neighboring image into the absolute image set H1 based on the comparison result;

[0025] S17: obtaining the total number I1 of all the neighboring images in the neighboring images G1, G2,..., Gg that are not added into the absolute image set H1, comparing I1 with O, and if I1<0, then generating the registration image set L1, L2,..., Ll of the reference image according to a preset first generation rule;

[0026] S18: if the number of the registration image set of the reference image is 0, then generating the determination recognition data of the reference contour according to the reference image, and if the number of the registration image set of the reference image is not 0, then generating the determination recognition data of the reference contour according to the reference image, the positioning reference points of the registration image set L1, L2,..., Ll and the l pieces of registration recognition data;

[0027] S19: eliminating the auricular point positioning image A1 and all the neighboring images of the relative auricular points B1, B2,..., Bb selected as the reference image when the auricular point positioning image A1 is selected as the reference image from the auricular point positioning images A1, A2,..., Aa, randomly selecting one auricular point positioning image from the remaining auricular point positioning images after the elimination as the reference image, and generating the contour determination and recognition data of the ear of the auricular point positioning image according to steps S13 to S18;

[0028] S110: repeating the step S19 until all the auricular point positioning images A1, A2,..., Aa are completely eliminated to obtain the contour determination and recognition data of the ear.

[0029] Further, S17, the first generation rule of the registration image set L1, L2,..., Ll of the reference image is as follows:

[0030] SS11: marking all the neighboring images in the neighboring image set G1, G2,..., Gg that are not selected as K1, K2,..., Kk, 1≤k

[0031] SS12: determining all the deviation auricular points of the neighboring image K1 based on S143 to S145, wherein the deviation auricular point refers to the auricular point based on which the neighboring image K1 is determined not to be the reference image when the neighboring image K1 is determined to be the similar image;

[0032] SS13: sequentially determining all the deviation auricular points of the neighboring images K2, K3,..., Kk according to SS12;

[0033] SS14: generating the registration image set L1, L2,..., Ll, 1≤l

[0034] SS15: generating one piece of registration recognition data of the reference image based on the registration image set L1 according to the preset second generation rule;

[0035] SS16: sequentially generating l-1 pieces of registration recognition data of the reference contour according to the registration image sets L2, L3,..., Ll according to SS15, wherein the registration image sets L2, L3,..., Ll each have one positioning reference point.

[0036] Further, the determination and recognition unit determines the recognition mode of the human ear image data of the target patient as follows:

[0037] S21: extracting the contour of the ear of the target patient from the human ear image data by using an image contour extraction algorithm, calculating the similarity of the contour and the contour of all ears stored in the judgment recognition unit respectively, and extracting the highest similarity T1 from all the calculated similarities;

[0038] S22: comparing T1 and P5 in size, if T1

[0039] S23: if T1≥P5, the judgment recognition data of the contour of the ear participating in the calculation of the similarity T1 is extracted in the judgment recognition unit at this time;

[0040] If the judgment recognition data only contains one ear point positioning image, it is determined that the recognition mode of the human ear image data of the target patient is image recognition, otherwise, the ear point recognition image and all registration recognition data N1, N2,..., Nn, n≥1 are first extracted from the judgment recognition data, and then steps S24 to S25 are executed;

[0041] S24: determining the recognition mode of the human ear image data of the target patient based on the registration recognition data N1 according to the preset judgment steps, and the judgment steps are as follows:

[0042] S241: marking all the position information contained in the registration recognition data N1 as U1, U2,..., Uu, u≥1 respectively;

[0043] S242: extracting the contour of the ear from the ear point recognition image, and removing the area contour of the deviation ear point U1, U2,..., Uu from the contour to obtain the advanced recognition contour of the human ear image data relative to the registration recognition data N1;

[0044] S243: according to the position information of the remaining ear points in the ear point recognition image except the deviation ear points U1, U2,..., Uu, the area contour of the corresponding ear point is framed in the human ear image data;

[0045] S244: removing all the area contours that are not framed in the contour of the ear of the human ear image data, and taking the remaining contour of the ear after removal as the relative recognition contour of the human ear image data relative to the registration recognition data N1;

[0046] S245: calculating the similarity of the relative recognition contour and the advanced recognition contour, if the value of the similarity is greater than or equal to P6, it is determined that the recognition mode of the human ear image data of the target patient is image recognition;

[0047] At this time, the position information of the deviation ear acupoints U1, U2,..., Uu in the ear acupoint recognition image is used to frame the area contour of the corresponding deviation ear acupoints in the human ear image data, and vice versa, no processing is performed, and after the area contours of all the deviation ear acupoints are framed, the ear acupoint positioning image of the target patient is obtained.

[0048] S25: Determine the recognition mode of the human ear image data of the target patient based on the registration recognition data N2, N3,..., Nn in turn according to S24, if the recognition mode of the human ear image data of the target patient based on the registration recognition data N1, N2,..., Nn is not image recognition, it is determined that the recognition mode of the human ear image data of the target patient is model recognition.

[0049] Further, the area contour of the ear acupoint in the human ear image data is framed according to the position information of any one ear acupoint, and the content is as follows:

[0050] The area of the ear contour in the ear acupoint recognition image and the area of the ear contour in the human ear image data are determined, and the proportion of the ear acupoint recognition image and the human ear image data is determined based on the proportion and the position information of the ear acupoint.

[0051] Compared with the prior art, the present application has the following beneficial effects:

[0052] The present application acquires the human ear image data of the target patient by setting the image acquisition module, determines the recognition mode of the human ear image data by setting the determination and recognition unit, and selects model recognition or image recognition for ear acupoint recognition of the human ear image data based on the determination result, thereby avoiding the limitation of single recognition.

[0053] The application analyzes the auricular point positioning image output by the auricular point recognition model through the result analysis module, quickly screens the neighbor images with high contour similarity from the historical data based on the ear contour of the randomly selected reference auricular point positioning image, then extracts the deviation auricular points of the neighbor images and the reference image by analyzing the similarity of the regional contours of each auricular point, and divides the neighbor images into several registration image sets according to the consistency features of the deviation auricular points, so that the application can not only be used for the regular ear with the contour and auricular point distribution conforming to the typical features, but also can be used for the ear with the overall contour being regular but the local auricular point structure being inconsistent, and can further distinguish the deviation auricular point combination with similarity and stability, quickly complete the matching and auricular point recognition through the pre-stored registration recognition data, strengthen the image recognition capability, improve the auricular point recognition positioning efficiency, further screen the input data of the auricular point recognition model, optimize the use of the auricular point recognition model, ensure the accuracy and reduce the calculation resource consumption. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 The system block diagram of the application. DETAILED DESCRIPTION

[0055] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0056] Please refer to Figure 1 The application provides an auricular point recognition system based on big data, which comprises an image acquisition module, an auricular point recognition module, an identification display module and a result analysis module.

[0057] The image acquisition module is used to acquire the human ear image data of a target patient and transmit it to the auricular point recognition module. In the application, the human ear image data is acquired by a device with optical imaging function, and the device comprises an image sensor (such as a CCD or CMOS sensor) and an optical lens. The device is used to take pictures of the outer ear part (including the auricle structure) of the human ear in the visible light wave band (400-760 nm).

[0058] The auricular point recognition module is used to select a corresponding recognition mode for the received human ear image data of a target patient to perform auricular point recognition. The auricular point recognition module comprises a model recognition unit and a judgment recognition unit.

[0059] The ear point recognition module transmits the human ear image data of the target patient to the judgment recognition unit after receiving the transmission;

[0060] The judgment recognition unit judges the recognition mode of the human ear image data of the target patient according to the preset judgment steps after receiving the transmission, and the recognition mode includes model recognition and image recognition in the present application, and the judgment steps are as follows:

[0061] S21: The contour of the ear of the target patient is extracted from the human ear image data by using an image contour extraction algorithm, the similarity of the contour and the contours of all ears stored in the judgment recognition unit is calculated respectively, and the highest similarity T1 is extracted from all the calculated similarities;

[0062] S22: T1 and P5 are compared in size, if T1

[0063] S23: If T1≥P5, the judgment recognition data of the contour of the ear participating in the calculation of the similarity T1 is extracted in the judgment recognition unit at this time, and P5 is a preset method recognition threshold;

[0064] If the judgment recognition data only contains one ear point positioning image, the recognition mode of the human ear image data of the target patient is image recognition, and the human ear image data of the target patient is transmitted to the comparison recognition unit;

[0065] If the judgment recognition data contains more than one ear point positioning image, first, the ear point recognition image and all registration recognition data N1, N2,..., Nn, n≥1 are extracted from the judgment recognition data, and then steps S24 to S25 are executed;

[0066] S24: The recognition mode of the human ear image data of the target patient is judged based on the registration recognition data N1 according to the preset judgment steps, and the judgment steps are as follows:

[0067] S241: All position information contained in the registration recognition data N1 is marked as U1, U2,..., Uu, u≥1, respectively.

[0068] S242: The contour of the ear is extracted from the ear point recognition image, and the area contour of the deviation ear point U1, U2,..., Uu is removed from the contour to obtain the advanced recognition contour of the human ear image data relative to the registration recognition data N1;

[0069] S243: According to the position information of the remaining auricular points in the auricular point recognition image except for the deviation auricular points U1, U2, …, Uu, the area contour of the corresponding auricular point is framed in the human ear image data. The position information of an auricular point includes the area of the auricular point in the auricular point recognition image, the center point coordinates, the position coordinates of the independent positioning point, and the included angle between the long-distance line of the area contour of the auricular point and the x-axis, wherein the independent positioning point is an ear positioning point randomly selected from several preselected ear positioning points, which is located in a position not belonging to the area contour of the auricular point and not belonging to the deviation auricular points U1, U2, …, Uu;

[0070] It should be noted here that the determination of the position information of any auricular point is first to take the lower left corner of the auricular point recognition image as the coordinate origin to establish a plane rectangular coordinate system, select the horizontal right direction as the x-axis direction, and extend from the coordinate origin, and select the vertical upward direction as the y-axis direction, and extend from the coordinate origin. In the established plane rectangular coordinate system, the value of each scale of the x-axis and the y-axis is the length of a unit cell, that is, 1px;

[0071] According to the position information of any auricular point, the area contour of the auricular point is framed in the human ear image data as follows:

[0072] According to the area of the ear contour in the auricular point recognition image and the area of the ear contour in the human ear image data, the proportion of the auricular point recognition image and the human ear image data is determined;

[0073] Based on the proportion and the position information of the auricular point, the area contour of the auricular point in the human ear image data is determined. Specifically, the area, center point coordinates, and position coordinates of the independent positioning point of the area contour of the auricular point in the human ear image data can be determined through the proportion. Based on the included angle, the center point coordinates, the position coordinates of the independent positioning point, and the area contour of the auricular point, the area contour of the auricular point in the human ear image data can be determined. A selected color is used to frame it, and the corresponding auricular point name is labeled in the framed area;

[0074] S244: In the contour of the ear in the human ear image data, all the area contours that are not framed are removed, and the remaining contour of the ear after removal is taken as the relative recognition contour of the relative registration recognition data N1 of the human ear image data;

[0075] S245: The similarity between the relative recognition contour and the advanced recognition contour is calculated. If the value of the similarity is greater than or equal to P6, it is determined that the recognition mode of the human ear image data of the target patient is image recognition;

[0076] At this time, the position information of the deviation ear points U1, U2,..., Uu in the ear point recognition image is used to frame the area contour of the corresponding deviation ear point in the human ear image data, and otherwise no processing is performed. The content of framing the area contour of any one deviation ear point is as follows:

[0077] The area, center point coordinate, and positioning reference point coordinate of the area contour of the deviation ear point in the human ear image data can be determined by the proportion of the ear point recognition image and the human ear image data. Based on the included angle, the center point coordinate, the positioning reference point coordinate, and the area contour of the deviation ear point, the area contour of the deviation ear point in the human ear image data can be determined. One color is selected to frame it, and the corresponding deviation ear point name is labeled in the framed area;

[0078] After the area contours of all deviation ear points are framed, the framed human ear image data is transmitted to the recognition display module as the ear point positioning image of the target patient;

[0079] S25: Determine the recognition mode of the human ear image data of the target patient based on the registration recognition data N2, N3,..., Nn in turn according to S24. It should be noted that S24 and S25 are executed synchronously;

[0080] If the recognition mode of the human ear image data of the target patient based on the registration recognition data N1, N2,..., Nn is not image recognition, it is determined that the recognition mode of the human ear image data of the target patient is model recognition, and the human ear image data of the target patient is transmitted to the model recognition unit;

[0081] The ear point recognition model that has been trained is pre-stored in the model recognition unit. After receiving the transmitted human ear image data of the target patient, the model recognition unit inputs it into the ear point recognition model to obtain the ear point positioning image of the target patient. The ear point positioning image is transmitted to the recognition display module and the result analysis module, respectively. In the ear point positioning image, the area contour of each ear point is framed with different colors and the corresponding ear point name is labeled;

[0082] In this application, the ear point types that can be recognized by the ear point recognition model strictly follow the ear point name system and standard positioning specification specified in the national standard "Ear Point Name and Positioning" (GB / T 13734-2008);

[0083] In this application, the ear point types that can be recognized by the ear point recognition model are selected by professional personnel according to the needs of diagnosis and treatment;

[0084] An identification display module is configured to display the auricular point positioning image of the target patient after receiving the auricular point positioning image of the target patient, and the identification display module displays the transmitted auricular point positioning image of the target patient after receiving the auricular point positioning image of the target patient;

[0085] A result analysis module is configured to analyze all the auricular point positioning images stored therein, and the analysis steps are as follows:

[0086] S11: All the auricular point positioning images stored in the result analysis module are marked as A1, A2,..., Aa respectively, a≥1, and all the auricular points that can be recognized by the auricular point recognition model are marked as B1, B2,..., Bb respectively, b≥1;

[0087] S12: The contour of the ear in the auricular point positioning image A1 is extracted by using an image contour extraction algorithm, the contour is selected as a reference contour, and the auricular point positioning image A1 is selected as a reference image;

[0088] Similarly, the contours C1, C2,..., Ca-1 of the ears in the auricular point positioning images A2, A3,..., Aa are extracted in sequence by using the image contour extraction algorithm;

[0089] S13: The similarity between the contour C1 and the reference contour is calculated by using a similarity algorithm, if the similarity is greater than or equal to P1, the auricular point positioning image A2 from which the contour C1 is extracted is taken as a similar image of the reference image, P1 is a preset image similarity threshold, otherwise no processing is performed;

[0090] Similarly, the similarities between the contours C2, C3,..., Ca-1 and the reference contour are calculated in sequence, and each calculated similarity is compared with P1, and a plurality of similar images of the reference image are obtained based on the comparison result;

[0091] S14: All the nearest neighbor images of the reference image relative to the auricular points B1, B2,..., Bb are selected from all the similar images of the reference image according to a preset selection rule, and the selection rule is as follows:

[0092] S141: All the similar images of the reference image are marked as D1, D2,..., Dd in sequence according to the order of the similar images taken as the reference image, 1≤d≤a-1;

[0093] S142: Extract the region contour E1 of the ear point B1 in the reference image using an image contour extraction algorithm, and similarly extract the region contours F1, F2,..., Fd of the ear point B1 in the similar images D1, D2,..., Dd in turn. It should be noted that the regions of different ear points in the reference image and the similar images D1, D2,..., Dd are framed with different colors, and the corresponding ear points are labeled with text. In this application, the region contour of the corresponding ear point can be extracted by using the image contour extraction algorithm based on the text and the corresponding color frame to obtain the region contour of the corresponding ear point.

[0094] S143: Determine whether the similar image D1 is a near-neighbor image of the reference image relative to the ear point B1, and the determination content is as follows:

[0095] Calculate the similarity between the region contour F1 and the region contour E1 using a similarity algorithm, and compare the similarity with P2. If the value of the similarity is greater than or equal to P2, it is determined that the similar image D1 is a near-neighbor image of the reference image relative to the ear point B1, otherwise it is determined that the similar image D1 is not a near-neighbor image of the reference image relative to the ear point B1. P2 is a preset ear point similarity threshold value based on the ear point B1.

[0096] S144: Determine whether the similar images D2, D3,..., Dd are near-neighbor images of the reference image relative to the ear point B1 in turn according to S143, and obtain all near-neighbor images of the reference image relative to the ear point B1 based on the determination results.

[0097] S145: Obtain all near-neighbor images of the reference image relative to the ear points B2, B3,..., Bb in turn according to S141 to S144.

[0098] S15: Traverse all near-neighbor images of the reference image relative to the ear points B1, B2,..., Bb and extract all near-neighbor images G1, G2,..., Gg, g≥1 that satisfy the preset frequency condition from them.

[0099] The frequency condition is as follows: the appearance frequency is greater than or equal to P3, and the appearance frequency refers to the number of repeated appearances. P3 is a preset frequency scalar.

[0100] Explanation: If a near-neighbor image appears repeatedly in all near-neighbor images of the reference image relative to the ear points B1, B2,..., Bb more than or equal to P3 times, it is extracted.

[0101] S16: Create an empty absolute image set H1 of the reference image, and select a number of near-neighbor images from the near-neighbor images G1, G2,..., Gg according to the preset selection and addition rule to add them into the absolute image set H1 to obtain the final absolute image set H1 of the reference image. The selection and addition rule is as follows:

[0102] S161: Obtain the occurrence frequency H1 of the neighboring image G1 in all neighboring images of the reference image relative to the auricular points B1, B2, ..., Bb. If H1 = b, then add the neighboring image G1 to the absolute image set H1; otherwise, do nothing.

[0103] S162: Obtain the occurrence frequencies of neighboring images G2, G3, ..., Gg in sequence according to S161. For each occurrence frequency obtained, compare the occurrence frequency with b, and determine whether to add the corresponding neighboring image to the absolute image set H1 based on the comparison result.

[0104] S17: Obtain the total number I1 of all neighbor images G1, G2, ..., Gg that have not been filtered and added to the absolute image set H1, compare I1 with O, and if I1 < 0, generate the registration image set L1, L2, ..., L1 of the reference image according to the preset first generation rule. The first generation rule is as follows:

[0105] SS11: All neighbor images in the neighbor images G1, G2, ..., Gg that are not filtered and added to the absolute image set H1 are marked as K1, K2, ..., Kk, 1≤k <g;

[0106] SS12: Determine all deviation ear points of the neighboring image K1 based on S143 to S145, where the deviation ear points refer to the ear points based on which the neighboring image K1 is determined not to be a neighboring image of the reference image when serving as a similar image. For example:

[0107] If the neighboring image K1 is determined not to be a neighboring image of the reference image relative to the ear point B1 when serving as a similar image, then the ear point B1 is a deviation ear point of the neighboring image K1;

[0108] SS13: determine all deviation auricular points of neighboring images K2, K3, ..., Kk in sequence according to SS12;

[0109] SS14: Generate the registration image set L1, L2, ..., Ll of the reference image based on all deviation ear points of the neighboring images K1, K2, ..., Kk, 1≤l <k,此时生成的所述配准图像集L1、L2、...、Ll中任意一个配准图像集满足预设的第一、二和三划定规则:

[0110] The first delineation rule is as follows: any one of the generated registered image sets L1, L2, ..., L1 contains several neighboring images among the neighboring images K1, K2, ..., Kk;

[0111] The second demarcation rule is as follows: all deviation ear points of each adjacent image contained in any one of the generated registration image sets L1, L2,..., Ll are consistent, wherein the consistency refers to that the names and quantities of all deviation ear points of any two adjacent images are consistent;

[0112] The third demarcation rule is as follows: the number of adjacent images contained in any one of the generated registration image sets L1, L2,..., Ll is greater than or equal to P4, and P4 is a preset registration demarcation index;

[0113] It should be noted that if the number of the generated registration image sets of the reference image is 0, the SS15-SS16 are not executed, and the following steps are directly executed;

[0114] SS15: generating one piece of registration identification data of the reference image based on the registration image set L1 according to a preset second generation rule, and the second generation rule is as follows:

[0115] SS21: marking all deviation ear points of the adjacent images contained in the registration image set L1 as M1, M2,..., Mm, respectively, and 1≤m<b;

[0116] It should be noted that all deviation ear points of all adjacent images in the registration image set L1 are consistent, and therefore, only the deviation ear points of one adjacent image are marked;

[0117] SS22: randomly selecting, based on the deviation ear points M1, M2,..., Mm, a positioning point of an ear region contour of an ear point whose position does not belong to any one of the deviation ear points M1, M2,..., Mm as a positioning reference point of the registration image set L1 from several preselected ear positioning points;

[0118] It should be noted that the ear positioning point is a positioning point selected by a professional on an ear based on an ear structure to serve as a reference mark and provide a positioning function. In the present application, the preselected ear positioning points include an antitragus tip, an upper end of a helix, a contralateral antitragus tip, a lobe top point, a posterior edge midpoint of a concha, a central point of a concha cavity, a triangular fossa top triangular fossa, a contralateral upper helix bifurcation point, a suprathiac midpoint, and a helix nodule highest point.

[0119] Tip of the free edge of the tragus (Tragus tip): the highest point of the free edge of the tragus (i.e. the top end of the tragus); Origin of the crus of the helix: the starting point where the helix goes into the cymba concha; Antitragus tip: the tip of the antitragus; Lobe apex: the lowest end of the lobe; Midpoint of the posterior edge of the cymba concha: the midpoint of the posterior edge of the cymba concha (upper part of the concha); Geometric center of the concha cavity: the geometric center point of the concha cavity (lower part of the concha); Apex of the triangular fossa: the topmost point of the triangular fossa (between the upper and lower branches of the antihelix) near the upper branch of the antihelix; Branching point of the upper branch of the antihelix: the branching point at the starting point of the upper branch of the antihelix; Midpoint of the supratragus incisure: the midpoint of the depression between the superior edge of the tragus and the helix (supratragus incisure); Highest point of the helical tubercle: the top end of the small nodular elevation in the upper posterior part of the helix (helical tubercle);

[0120] SS23: Randomly select one of all the neighboring images contained in the registration image set L1, establish a plane rectangular coordinate system with the lower left corner of the neighboring image as the coordinate origin, select the direction to the right as the x-axis direction, and extend from the coordinate origin, select the direction vertically upward as the y-axis direction, and extend from the coordinate origin, wherein the value of each scale of the x-axis and y-axis in the established plane rectangular coordinate system is the length of one unit cell, i.e. 1px;

[0121] SS24: Determine the position coordinates O1(x1, y1) of the positioning reference point in the plane rectangular coordinate system;

[0122] Obtain the center point coordinates O2(x2, y2) of the region profile of the deviation ear point M1 in the randomly selected neighboring image in SS23, and obtain the included angle O3 of the long-distance line of the region profile of the deviation ear point M1 and the x-axis in the plane rectangular coordinate system, wherein the long-distance line refers to the connecting line of the two pixel points with the farthest distance in the region profile of the deviation ear point M1;

[0123] Calculate the shortest straight line distance R1 between O1 and O2, wherein the calculation formula of R1 is

[0124] SS25: Generate the position information of the registration image set L1 relative to the deviation ear point M1 according to the included angle O3, the shortest straight line distance R1, and the region profile of the deviation ear point M1 in the randomly selected neighboring image in SS23;

[0125] SS26: Generate the position information of the registration image set L1 relative to the deviation ear points M2, M3,..., Mm in turn according to SS21 to SS25;

[0126] SS27: Calculate the proportion R2 of the area of the reference contour O4 and the area of the ear contour O5 in the randomly selected neighboring image in SS23, and the calculation formula of R21 is R2=O4 / O5.

[0127] SS28: removing the area profile of the deviated ear point M1, M2,..., Mm from the contour of the ear of the randomly selected neighbor image in SS23, and generating one piece of registration identification data of the reference profile according to the remaining contour of the ear after the removal, the proportion R1, and the position information of the registration image set L1 relative to the deviated ear point M1, M2,..., Mm;

[0128] SS16: generating l-1 pieces of registration identification data of the reference profile according to the registration image sets L2, L3,..., Ll in turn according to SS15, wherein each of the registration image sets L2, L3,..., Ll has one positioning reference point;

[0129] S18: if the number of the registration image sets of the reference image is 0, generating the determination identification data of the reference profile according to the reference image, and if the number of the registration image sets of the reference image is not 0, generating the determination identification data of the reference profile according to the reference image, the positioning reference points of the registration image sets L1, L2,..., Ll, and the l pieces of registration identification data of the reference image generated based on the registration image sets L1, L2,..., Ll;

[0130] S19: removing the ear point positioning image A1 and all the neighbor images of the relative ear point B1, B2,..., Bb selected when the ear point positioning image A1 is selected as the reference image from the ear point positioning images A1, A2,..., Aa;

[0131] randomly selecting one ear point positioning image from all the ear point positioning images remaining after the removal as the reference image, and generating the determination identification data of the contour of the ear of the ear point positioning image according to steps S13 to S18;

[0132] S110: repeating the steps of S19 until all the ear point positioning images A1, A2,..., Aa are removed completely, and obtaining several pieces of determination identification data of the contour of the ear;

[0133] The result analysis module transmits all the determination identification data of the contour of the ear obtained to the ear point identification module, which receives and transmits the data to the determination identification unit for storage;

[0134] Some data in the above formula are dimensionless for numerical calculation, and the contents not described in detail in the specification are all prior art known to those skilled in the art.

[0135] The above examples are only used to illustrate the technical method of the present application but not limit the present application. Although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present application can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present application.

Claims

1. An ear acupoint recognition system based on big data, characterized in that: include: a determination and recognition unit, configured to determine, after receiving the human ear image data of the target patient, an identification method of the human ear image data according to a preset determination step, the identification method including model recognition and image recognition; The determination and recognition unit is further configured to, when the recognition mode of the human ear image data of the target patient is determined to be image recognition, frame the regional outlines of several auricular points in the human ear image data to obtain an auricular point positioning image of the target patient; The model recognition unit is used to input the human ear image data determined to be the model recognition recognition mode into the pre-stored ear acupoint recognition model to perform ear acupoint recognition to obtain the ear acupoint positioning image of the target patient; The result analysis module is used to analyze all the auricular point positioning images transmitted by the model recognition unit to obtain the judgment and recognition data of the contours of several ears.

2. The ear acupoint recognition system based on big data according to claim 1, characterized in that: The analysis steps for analyzing the judgment and recognition data of the contours of several ears are as follows: S11: Mark all auricular acupoint positioning images stored in the result analysis module as A1, A2, ..., Aa, where a≥1, and mark all auricular acupoints that can be identified by the auricular acupoint recognition model as B1, B2, ..., Bb, where b≥1; S12: Using an image contour extraction algorithm to extract the contour of the ear in the auricular point positioning image A1, selecting the contour as a reference contour, selecting the auricular point positioning image A1 as a reference image, and similarly extracting the contours C1, C2, ..., Ca-1 of the ear in the auricular point positioning images A2, A3, ..., Aa in sequence; S13: Calculate the similarity between the contour C1 and the reference contour using a similarity algorithm. If the similarity is greater than or equal to P1, extract the auricular point positioning image A2 of the contour C1 as a similar image to the reference image. P1 is a preset image similarity standard threshold. Similarly, calculate the similarity between the contours C2, C3, ..., Ca-1 and the reference contour in sequence. For each similarity calculated, compare the similarity with P1, and obtain several similar images of the reference image based on the comparison results. S14: All neighboring images of the reference image relative to the auricular points B1, B2, ..., Bb are obtained from all similar images of the reference image according to a preset screening rule. The screening rule is as follows: S141: Marking all similar images of the reference image as D1, D2, ..., Dd in sequence according to the order of the similar images used as the reference image, where 1≤d≤a-1; S142: extracting the regional contour E1 of the auricular point B1 in the reference image using an image contour extraction algorithm, and similarly extracting the regional contours F1, F2, ..., Fd of the auricular point B1 in the similar images D1, D2, ..., Dd in sequence; S143: Determine whether the similar image D1 is a neighboring image of the reference image relative to the auricular point B1: Calculate the similarity between the region outline F1 and the region outline E1 using a similarity algorithm, and compare the similarity with P2. If the similarity is greater than or equal to P2, determine that the similar image D1 is a neighboring image of the reference image relative to the auricular point B1; otherwise, determine that the similar image D1 is not a neighboring image of the reference image relative to the auricular point B1. P2 is a threshold value of auricular point similarity preset based on the auricular point B1. S144: determining in sequence whether the similar images D2, D3, ..., Dd are neighboring images of the reference image relative to the auricular point B1 according to S143, and obtaining all neighboring images of the reference image relative to the auricular point B1 based on the determination results; S145: Obtain all neighboring images of the reference image relative to auricular points B2, B3, ..., Bb in sequence according to S141 to S144; S15: Traverse all neighboring images of the reference image relative to the auricular points B1, B2, ..., Bb and extract all neighboring images G1, G2, ..., Gg with an appearance frequency greater than or equal to P3, where g≥1, where the appearance frequency refers to the number of repeated appearances, and P3 is a preset frequency scalar; S16: Create an empty absolute image set H1 of the reference image, select a number of neighboring images from the neighboring images G1, G2, ..., Gg according to the preset screening and adding rules, and add them to the absolute image set H1 to obtain the final absolute image set H1 of the reference image. The screening and adding rules are as follows: S161: Obtain the occurrence frequency H1 of the neighboring image G1 in all neighboring images of the reference image relative to the auricular points B1, B2, ..., Bb. If H1 = b, then add the neighboring image G1 to the absolute image set H1; otherwise, do nothing. S162: Obtain the occurrence frequencies of neighboring images G2, G3, ..., Gg in sequence according to S161. For each occurrence frequency obtained, compare the occurrence frequency with b, and determine whether to add the corresponding neighboring image to the absolute image set H1 based on the comparison result. S17: Obtain the total number I1 of all neighbor images G1, G2, ..., Gg that have not been filtered and added to the absolute image set H1, compare I1 with O, and if I1 < 0, generate the registration image set L1, L2, ..., L1 of the reference image according to the preset first generation rule; S18: If the number of registered image sets used to generate the reference image is zero, generating reference contour determination and recognition data based on the reference image; if the number of registered image sets used to generate the reference image is not zero, generating reference contour determination and recognition data based on the reference image, the positioning reference points of the registered image sets L1, L2, ..., L1, and one piece of registration and recognition data; S19: Exclude the auricular point location images A1 that have been selected as reference images and all the neighboring images of the relative auricular points B1, B2, ..., Bb obtained when it was selected as the reference image from the auricular point location images A1, A2, ..., Aa. Randomly select an auricular point location image from the remaining auricular point location images after exclusion as the reference image, and generate the determination and recognition data of the ear contour of the auricular point location image according to steps S13 to S18; S110: Repeat the steps of S19 until all the auricular point location images in the auricular point location images A1, A2, ..., Aa are completely excluded, and obtain the determination and recognition data of the contours of several ears.

3. The ear acupoint recognition system based on big data according to claim 2, characterized in that: S17, the first generation rule for generating the registration image sets L1, L2, ..., Ll of the reference image is as follows: SS11: Mark the neighboring images G1, G2, ..., Gg that have not been screened and added to the absolute image set H1 as K1, K2, ..., Kk respectively, where 1 ≤ k < g; SS12: Determine all the deviated auricular points of the neighboring image K1 based on S143 to S145, where the deviated auricular points refer to the auricular points based on which the neighboring image K1 is determined not to be a neighboring image of the reference image when it is used as a similar image; SS13: Sequentially determine all the deviated auricular points of the neighboring images K2, K3, ..., Kk according to SS12; SS14: Generate the registration image sets L1, L2, ..., Ll of the reference image according to all the deviated auricular points of the neighboring images K1, K2, ..., Kk, where 1 ≤ l < k. At this time, any one of the generated registration image sets L1, L2, ..., Ll satisfies the preset first, second, and third delimitation rules: SS15: Generate a registration recognition data of the reference image based on the registration image set L1 according to the preset second generation rule; SS16: Sequentially generate l - 1 registration recognition data of the reference contour according to the registration image sets L2, L3, ..., Ll according to SS15, where each of the registration image sets L2, L3, ..., Ll has a positioning reference point.

4. The ear acupoint recognition system based on big data according to claim 3, characterized in that: In SS14, the first delimitation rule is as follows: Any one of the generated registration image sets L1, L2, ..., Ll contains several neighboring images among the neighboring images K1, K2, ..., Kk. The second delimitation rule is as follows: All the deviated auricular points of each neighboring image contained in any one of the generated registration image sets L1, L2, ..., Ll are the same. The third delimitation rule is as follows: The number of neighboring images contained in any one of the generated registration image sets L1, L2, ..., Ll is greater than or equal to P4, and P4 is a preset registration delimitation index.

5. The ear acupoint recognition system based on big data according to claim 3, characterized in that: SS15, the second generation rule for generating a registration recognition data of the reference image based on the registration image set L1 is as follows: SS21: Mark all the deviated auricular points of the neighboring images contained in the registration image set L1 as M1, M2, ..., Mm respectively, where 1 ≤ m < b; SS22: Randomly select a positioning point whose position does not belong to the regional contour of any of the deviation auricular points M1, M2, ..., Mm from several pre-selected ear positioning points as the positioning reference point of the registration image set L1; SS23: Randomly select a neighboring image from all the neighboring images included in the registration image set L1, and establish a plane rectangular coordinate system with the lower left corner of the neighboring image as the coordinate origin; SS24: Determine the position coordinates O1(x1, y1) of the positioning reference point in the plane rectangular coordinate system; obtain the center point coordinates O2(x2, y2) of the regional contour of the deviation auricular point M1 in the neighboring image, and obtain the angle O3 between the long-distance line of the regional contour of the deviation auricular point M1 and the x-axis in the plane rectangular coordinate system. The long-distance line refers to the line connecting the two pixel points with the farthest distance in the regional contour of the deviation auricular point M1, and calculate the shortest straight-line distance R1 between O1 and O2; SS25: Generate the position information of the registration image set L1 relative to the deviation auricular point M1 according to the angle O3, the shortest straight-line distance R1, and the regional contour of the deviation auricular point M1 in the neighboring image; SS26: Generate the position information of the registration image set L1 relative to the deviation auricular points M2, M3, ..., Mm in sequence according to SS21 to SS25; SS27: Calculate and obtain the ratio R2 of the neighboring image to the reference image. The calculation formula of R2 is R2 = O4 / O5, where O4 is the area of the reference contour and O5 is the area of the ear contour in the neighboring image; SS28:剔除偏差耳穴M1、M2、...、Mm的区域轮廓,依据剔除了剩余的耳朵的轮廓、比例占比R1、配准图像集L1相对偏差耳穴M1、M2、...、Mm的位置信息生成参照轮廓的一条配准识别数据。 6. The ear acupoint recognition system based on big data according to claim 2, characterized in that: The determination steps for the determination recognition unit to determine the recognition method of the ear image data of the target patient are as follows: S21: Use the image contour extraction algorithm to extract the contour of the ear of the target patient from the ear image data, calculate the similarity between the contour and all the ear contours stored in the determination recognition unit respectively, and extract the highest similarity value T1 from all the calculated similarities; S22: Compare the magnitudes of T1 and P5. If T1 < P5, it is determined that the recognition method of the ear image data of the target patient is model recognition; S23; If T1 ≥ P5, at this time, extract the determination recognition data of the ear contour participating in the calculation to obtain the similarity T1 in the determination recognition unit; If there is only one auricular point positioning image in the determination recognition data, it is determined that the recognition method of the ear image data of the target patient is image recognition. Otherwise, first extract the auricular point recognition image and all the registration recognition data N1, N2, ..., Nn, n ≥ 1 from the determination recognition data, and then execute steps S24 to S25; It should be noted that there is an unclear part in the original text for item . The translation is based on the best understanding of the overall context, but this part may need further clarification in the original text for a more accurate translation. S24: Determine the recognition method of the target patient's ear image data based on the registration recognition data N1 according to the preset determination steps. The determination steps are as follows: S241: Mark the deviation ear points corresponding to all the position information contained in the registration recognition data N1 as U1, U2, ..., Uu, where u≥1; S242: Extracting the ear contour from the ear point recognition image, and removing the regional contours of the deviation ear points U1, U2, ..., Uu from the contour to obtain an advanced recognition contour of the human ear image data relative to the registration recognition data N1; S243: Selecting the area outlines of the corresponding ear points in the human ear image data according to the position information of the remaining ear points except the deviation ear points U1, U2, ..., Uu in the ear point recognition image; S244: Eliminating all unselected area contours in the ear contour of the ear image data, and using the remaining ear contour as the relative recognition contour of the ear image data relative to the registration recognition data N1; S245: Calculating the similarity between the relative recognition profile and the advanced recognition profile. If the similarity value is greater than or equal to P6, determining that the recognition method of the human ear image data of the target patient is image recognition; At this time, based on the position information of the deviated ear points U1, U2, ..., Uu in the ear point recognition image, the area contours of the corresponding deviated ear points are framed in the human ear image data. Otherwise, no processing is performed. After the area contours of all the deviated ear points are framed, the ear point positioning image of the target patient is obtained; S25: According to S24, the recognition method of the human ear image data of the target patient is determined based on the registration recognition data N2, N3, ..., Nn in sequence. If the recognition method of the human ear image data of the target patient based on the registration recognition data N1, N2, ..., Nn is not image recognition, then the recognition method of the human ear image data of the target patient is determined to be model recognition.

7. The ear acupoint recognition system based on big data according to claim 6, characterized in that: In S243, the position information of an ear point includes the area of ​​the ear point, the center point coordinates, the position coordinates of the independent positioning point, and the angle between the long-distance line of the regional contour of the ear point and the x-axis in the ear point recognition image, wherein the independent positioning point is an ear positioning point randomly selected from a number of pre-selected ear positioning points, the position of which does not belong to the ear point and does not belong to the regional contour of the deviation ear points U1, U2, ..., Uu.

8. The ear acupoint recognition system based on big data according to claim 6, characterized in that: The content of selecting the area outline of any ear point in the human ear image data according to the position information of the ear point is as follows: determining a ratio of the ear point recognition image to the human ear image data based on an area of ​​a contour of the ear in the ear point recognition image and an area of ​​a contour of the ear in the human ear image data; The regional contour of the ear point in the human ear image data is determined based on the proportion and the position information of the ear point.

Images