WO2018107872A1 - 一种体型预测方法及设备 - Google Patents

一种体型预测方法及设备 Download PDF

Info

Publication number
WO2018107872A1
WO2018107872A1 PCT/CN2017/104673 CN2017104673W WO2018107872A1 WO 2018107872 A1 WO2018107872 A1 WO 2018107872A1 CN 2017104673 W CN2017104673 W CN 2017104673W WO 2018107872 A1 WO2018107872 A1 WO 2018107872A1
Authority
WO
WIPO (PCT)
Prior art keywords
current time
curvature
contour
current
predicted
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2017/104673
Other languages
English (en)
French (fr)
Inventor
辛奇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangzhou Shiyuan Electronics Thecnology Co Ltd
Original Assignee
Guangzhou Shiyuan Electronics Thecnology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guangzhou Shiyuan Electronics Thecnology Co Ltd filed Critical Guangzhou Shiyuan Electronics Thecnology Co Ltd
Publication of WO2018107872A1 publication Critical patent/WO2018107872A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7271—Specific aspects of physiological measurement analysis
    • A61B5/7275—Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing

Definitions

  • the present invention relates to the field of data analysis technologies, and in particular, to a body type prediction method and device.
  • the existing physical health index collection devices have only a simple data collection and recording function or can simply evaluate the current health status of the individual based on the collected data, and cannot predict the future body state. Therefore, the existing physical health index collection device and measurement method cannot allow the user to predict the future body shape in advance, thereby making it difficult for the user to adjust the fitness maintenance plan according to the actual situation, thereby affecting the user experience.
  • an object of the present invention is to provide a body type prediction method and apparatus capable of predicting a body type and having a good user experience.
  • an aspect of the present invention provides a body type prediction method, which includes the following steps: acquiring a standard posture map of a predetermined part of a user collected at a current time, and acquiring a corresponding one according to the standard posture map of the current time.
  • a contour map of the current time performing boundary sampling on the contour map at the current moment to acquire corresponding sampling points, and fitting the contour curve of the current time according to all the sampled points obtained; calculating and saving each of the sampling points
  • the curvature on the contour curve at the current time corresponding to the same sampling point at the current time and the predetermined time period before the current time Curvature, obtain the curvature function of each sampling point; predict the curvature of each sampling point at a predetermined future time according to the curvature function corresponding to each sampling point; obtain the predicted contour map according to the curvature of each sampling point obtained by the prediction .
  • the step of “obtaining a predicted contour map according to the curvature of each sample point obtained by prediction” is specifically: drawing a tangent having a curvature corresponding to the sampling point at each sampling point to obtain two adjacent ones An intersection of tangent lines of the sampling points; the intersection points are sequentially connected to form the predicted contour map.
  • the body shape prediction method further includes: when determining that the information amount of the current time is greater than a preset threshold, obtaining each curvature according to a curvature corresponding to the same sampling point at a current time and a predetermined time period before the current time a curvature function of the sampling point; wherein the information amount includes the number of contour maps saved in the preset contour library and the curvature of the same sampling point at the current time and a predetermined time period before the current time The difference between the two of the corresponding curvatures.
  • the body type prediction further includes: acquiring and saving body index data of the user collected at the current time; wherein the body index data is specifically index data of a body index of the user, and the body index includes at least the following A: body mass index, body mass index, body fat rate index and muscle density index; according to the same body index at the current time and the current time before the current time in the predetermined time period of the index data, obtain the corresponding body index function; The body index function predicts the body index data of the user at a predetermined future time.
  • the method further comprises: training the body index data of the current time according to the preset data training model with the current time
  • the contour map establishes a mapping relationship and saves the mapping relationship
  • the preset data training model includes a plurality of groups of corresponding preset contour maps and preset body index data; acquiring the predicted contour map and obtaining a mapping with the body index data at the current moment.
  • the contour map of the current moment of the relationship ; obtaining the final predicted contour map according to the acquired current moment and the predicted contour map.
  • the method further includes: obtaining the final predicted contour according to the acquired current moment and the predicted contour map.
  • the step of “acquiring the final predicted contour map according to the acquired current moment and the predicted contour map” is specifically: acquiring the predicted contour map and the current contour of the current moment
  • the current contour map includes a current contour curve and a current reference point, and the current contour curve is obtained by analyzing, by a preset edge detection algorithm, the standard posture map of the current time, the current reference image
  • the point is obtained by analyzing the standard attitude map of the current moment by a preset image algorithm
  • the predicted contour map includes a predicted contour curve and a predicted reference point, and the predicted contour curve is obtained according to the prediction Obtaining a curvature of the sampling point, the predicted reference point is obtained according to the current reference point; acquiring the current reference point and the predicted reference point, and overlapping the current reference point with the predicted reference point So that the contour map of the current moment overlaps with the predicted contour map; obtaining the weight of the current contour curve and the predicted contour curve a region and a non-coincident region, rendering the coincident region as a preset color, and
  • a body type preset device including: a contour map acquiring module, configured to acquire a standard posture map of a predetermined part of a user collected at a current time, and according to the standard posture map of the current time Obtaining a contour map of the current current time; a contour curve acquiring module, configured to perform boundary sampling on the contour map at the current time to obtain a corresponding sampling point, and obtain a contour curve of the current time according to all the sampling points obtained by fitting a curvature calculation module, configured to calculate and save a curvature of each of the sampling points on the contour curve at a current time; a curvature function acquisition module, configured to The curvature of each sampling point is obtained by the curvature of the same sampling point at the current time and the predetermined time before the current time; the curvature prediction module is configured to predict each sampling according to the curvature function corresponding to each sampling point.
  • a contour map prediction module for obtaining a predicted contour map
  • the contour curve obtaining module includes: a coordinate establishing unit configured to establish a two-dimensional coordinate system in the contour map at a current time; and a sampling unit configured to use the predetermined pixel pitch to the contour map of the current time Perform boundary sampling to obtain corresponding sampling points, and save the acquired sampling points in the form of coordinates; the fitting unit is configured to curve the coordinates corresponding to each sampling point according to a preset fitting algorithm. Combine to obtain the contour curve of the current moment.
  • the body type prediction method and the body type prediction device provided by the present invention acquire a standard posture map of a predetermined part of the user collected at the current time, and acquire a contour map of the current current time according to the standard posture map of the current time. And performing boundary sampling on the contour map at the current moment to obtain corresponding sampling points, and fitting the contour curves of the current time according to all the sampled points obtained; then calculating and saving each of the sampling points at the current moment The curvature on the contour curve; then obtaining the curvature function of each sampling point according to the curvature corresponding to the same sampling point at the current time and the predetermined time period before the current time; and then corresponding to each sampling point The curvature function predicts the curvature of each sample point at a predetermined future time; finally, the predicted contour map is obtained according to the curvature of each sample point obtained by the prediction, thereby realizing the process of the user's body shape prediction. Therefore, the present invention can predict the size of the user, so that the user can adjust his or her fitness plan according
  • FIG. 1 is a flowchart of a body shape prediction method according to an embodiment of the present invention.
  • FIG. 2 is a schematic structural diagram of a body type prediction apparatus according to an embodiment of the present invention.
  • FIG. 3 is a schematic structural diagram of a contour curve acquiring module provided in FIG. 2.
  • FIG. 3 is a schematic structural diagram of a contour curve acquiring module provided in FIG. 2.
  • an aspect of the present invention provides a body shape prediction method, which includes steps S10 to S15:
  • S10 Obtain a standard posture map of a predetermined part of the user collected at the current time, and obtain a contour map of the current current time according to the standard posture map of the current time.
  • the standard attitude map refers to an image taken by the camera to the user in a standard posture state.
  • the standard posture state may be: the user is barefoot, the legs are standing straight, the legs are the same width as the shoulders, and the two arms are slightly raised at a 45 degree angle to the body side; wherein, the men wear Shorts, women wear shorts and sports tops, and both men and women need to expose all faces, arms, legs, waist and abdomen.
  • the front side of the user faces a body-high volume prediction device (such as a mirror display device) and an HD camera of the body type prediction device.
  • the predetermined part of the user is preferably the whole body of the user in the embodiment of the present invention, and may be the hand, the face or the leg of the user, and is not specifically limited herein.
  • the standard posture should be able to fully display the morphological features of the predetermined part and the user can conveniently make the action (convenient for the user to perform repeated actions), which is not specifically limited by the present invention.
  • the contour map at the current moment is obtained by performing image contour edge detection on the standard posture map at the current moment according to a preset edge detection algorithm.
  • the preset The edge detection algorithm is the Canny algorithm.
  • the preset edge detection algorithm may also be a Roberts algorithm, a Prewitt algorithm, a Sobel algorithm, or a Log algorithm, and is not specifically limited herein.
  • the standard posture map is an image formed by a closed region formed by a body contour curve of a predetermined portion of the user (which may further include a reference point of a predetermined portion of the user), the body contour curve ( Or the pupil center point is different from the color of the background area in the current body contour map, for example, the body contour curve may be black, the background area is white or the body contour curve may be red The background area is yellow or the like.
  • S11 Perform boundary sampling on the contour map at the current time to obtain corresponding sampling points, and obtain a contour curve of the current time according to all the sampled points obtained.
  • step S11 specifically includes step S110 to step S112:
  • the contour map of the current moment of the user's frontal body is exemplified: Referring to the figure, the same tangent of the bottom of the two legs in the contour map is taken as the X-axis (the same tangent at the bottom of both feet is taken as X) The axis is because the body shape of the human body changes, the plane of the bottom of the two feet of the human body is hardly changed.
  • the center line of the line segment formed by the bottom of the two feet is used as the Y axis (at the bottom of the two feet)
  • the center line of the formed line segment is taken as the Y-axis because the human body is almost symmetrical with respect to the center line, thereby establishing a two-dimensional coordinate system of the contour map at the current time.
  • the contour map is a contour map of the user's head
  • the line connecting the center points of the two pupils of the user may be used as a reference line, and then a line parallel to the reference line and tangent to the bottom of the head may be made.
  • a first tangent, the first tangent as an X-axis, and then a second tangent perpendicular to the X-axis and tangent to the side of the head, and the second tangent as the Y-axis thereby establishing a two-dimensional coordinate system of the head.
  • the pupil center point of the head contour map is obtained by analyzing a head image captured by a camera by a preset image algorithm, for example, the pupil center point of the head contour map can pass
  • the FaceDetector control in the Open Source Computer Vision Library (OpenCV) acquires, that is, when the FaceDetector control is called on the head image, the FaceDetector control automatically locates two pupil center pixels in the head image.
  • the two pupil center pixels are in the head contour map The two center points of the pupil.
  • the process of establishing a two-dimensional coordinate system may refer to the above two examples, and details are not described herein again.
  • S111 Perform boundary sampling on the contour map of the current time according to a predetermined pixel pitch to acquire a corresponding sampling point, and save the acquired sampling point in the form of coordinates.
  • the contour map for example, the leftmost or lowermost pixel in the contour map
  • the row Or the pixels of the column are sampled one by one until all the pixels in the row (or the column) are sampled, and the pixels of the contour curve in the contour map acquired in the row (or the column) are in coordinates
  • the form is saved in the preset sampling point matrix in the order of sampling time.
  • the pixels of the next row (or the next column) in the contour map are acquired according to the predetermined pixel pitch (for example, one pixel or two pixels, etc.) based on the row (or the column), and the row is in the same sampling order.
  • the pixels in (or the column) are sampled one by one, and the pixels of the contour curve in the contour map acquired in the row (or the column) are saved in the form of coordinates in the order of sampling time in the preset sampling. Point in the dot matrix.
  • the corresponding sampling points of the contour curve in the contour map can be obtained.
  • Each of the acquired contour maps has a preset sampling point lattice corresponding to it.
  • an example of the frontal full body contour of the user who has established the two-dimensional coordinate system is as follows: the y coordinate of the front body contour map is set to 0, and the x coordinate is left by the front body contour map.
  • the edge is incremented from pixel to pixel (ie, coordinate point) toward the right edge of the frontal body contour map (ie, increasing from the x-axis negative direction toward the x-axis positive direction) until the left and right edges of the current body contour map are All pixel points with a y coordinate of 0 are retrieved one by one, and the sampling points of the contour pixels retrieved are represented in the form of coordinates, for example, P01(x1,0), P02(x2,0), P03(x3 , 0), P04 (x4, 0), etc.; then all the contour pixels of the Y-axis with a coordinate of 1 between the left and right edges of the current body contour map are retrieved, and the contours are retrieved in turn
  • the sampling points of the pixels are expressed in the
  • the obtained sampling points in each of the preset sampling point lattices are fitted by a fitting function, that is, all the coordinate data (x, y) of all the sampling points to be fitted are imported into
  • Each of the contour curves corresponds to each of the preset sampling point lattices, that is, each of the contour curves corresponds to each contour map of the predetermined portion.
  • each of the sampling point lattices corresponding to the contour map of the current time is calculated on the contour curve of the current time Curvature.
  • the process of obtaining the curvature of each sampling point in the preset sampling point lattice is:
  • the order of the save times in the sampling point lattice (ie, the order of the sampling times of the sampling points) is stored in the preset curvature database.
  • the preset sampling point lattice of each of the contour maps has a one-to-one correspondence with the preset curvature database.
  • the curvature corresponding to the corresponding sampling points of the contour map before the contour map at the current moment is also saved in the preset curvature database in the order of sampling time.
  • the curvature corresponding to the sampling point in the preset curvature database at the current time and the predetermined time period before the current time is obtained.
  • the corresponding curvature of the storage time of the sampling point P01 in the preset curvature database in the time period from t1 to tn is obtained by statistics, and sampling is obtained.
  • Point P01's corresponding curvature database Ktn [Kt1, Kt2..., Ktn], put the curvature database of the sampling point P01 into the corresponding curve fitting tool, select the smooth curve to fit, and obtain the sampling point.
  • the predicted curvature of the sampling point at a predetermined future time can be obtained.
  • the predicted curvature of each sample point in each sample dot matrix can be obtained.
  • step S15 specifically includes steps S150 to S151:
  • Each of the sampling points in the current body contour map draws a tangent having a corresponding curvature such that the intersection of the tangent lines of two adjacent sampling points can be acquired.
  • the adjacent two intersection points are sequentially connected in a smooth curve manner to form the predicted contour map.
  • the obtained contour map can be displayed by a display screen of a related device (for example, a full-surface mirror display device).
  • the standard posture map of the predetermined part of the user collected at the current time is acquired, and the contour map of the current current time is acquired according to the standard posture map of the current time; and the contour of the current time is obtained.
  • the image is subjected to boundary sampling to obtain corresponding sampling points, and the contour curve of the current time is obtained according to all the sample points obtained; then the curvature of each of the sampling points on the contour curve at the current time is calculated and saved; Then, according to the curvature corresponding to the same sampling point at the current time and the predetermined time before the current time, the curvature function of each sampling point is obtained; and then each sampling point is predicted according to the curvature function corresponding to each sampling point.
  • the present invention can predict the size of the user, so that the user can adjust his or her fitness plan according to his or her predicted body shape, so the present invention has a good user experience.
  • the body shape prediction method further includes:
  • the amount of information includes the number of contour maps saved in the preset contour library and the curvature of the same sampling point at the current moment and the curvature corresponding to one of the predetermined time periods before the current time. Difference value.
  • the curvature corresponding to the same sample point at the current time and the predetermined time period before the current time is obtained.
  • the curvature function of the sampling points; or when the difference between the curvature corresponding to the same sampling point at the current time and the curvature corresponding to a certain time before the current time is greater than At ten o'clock, the curvature function of each sampling point is obtained according to the curvature corresponding to the same sampling point at the current time and the predetermined time before the current time.
  • the process of acquiring the curvature of the pixel sampling point please refer to step S120 and step S121 described above, and details are not described herein again.
  • the existing corresponding information amount does not reach the preset threshold, it indicates one of the following two situations: 1. It indicates that the number of existing acquired contour maps is less, which will The curvature function is relatively inaccurate, so when the number of acquired contour maps is greater than a predetermined threshold, the curvature function of the sampling point is calculated, and a more accurate curvature function can be obtained, thereby obtaining a more accurate curvature function. It is possible to obtain a more accurate predicted contour map; 2.
  • the curvature corresponding to the same sampling point at the current time and the predetermined time before the current time is obtained, and each is obtained.
  • the curvature function of the sampling point which can obtain a more accurate curvature function or improve the efficiency of the CPU.
  • the body shape prediction method further includes:
  • the body index data is specifically index data of a user's body index
  • the body index includes at least one of the following: a body mass index, a body mass index , body fat rate index and muscle density index.
  • the body index data is obtained by a corresponding body detecting device, for example, body mass index data, body fat index data, and muscle density index data are obtained by body fat weighing. It should be noted that the body index data may further include other types of body index data, such as a protein index (obtained by a protein measuring instrument) or a skin moisture index (measured by a moisture sensor), etc., and are not specifically limited herein. .
  • S18 Acquire a corresponding body index function according to the index data corresponding to the same body index at a current time and a predetermined time period before the current time.
  • the corresponding body mass index function is obtained.
  • the process of obtaining the corresponding body index function of the other indexes of the body index is the same as the body mass index obtaining body mass index function, and will not be described herein.
  • the corresponding body index data at a certain moment in the future can be obtained; for example, inputting a certain moment in the future into the body mass index function can Get the corresponding body mass index data for the future time. It should be noted that when the body index data of the predetermined future time is acquired, it is possible to display the body index data of the predetermined future time on the display screen of the corresponding body type preset device.
  • the body index data of the user collected at the current time is acquired and saved, and the corresponding body is obtained according to the index data corresponding to the same body index at the current time and the predetermined time period before the current time.
  • the exponential function finally predicts the body index data of the user at a predetermined future time according to the body index function, thereby implementing a prediction process of the user's corresponding body index. Therefore, the preferred embodiment can predict the body index of the user's future time, thereby allowing the user to adjust his or her fitness plan according to the body index of his future time, thereby further improving the user experience.
  • the method further includes:
  • the body index data of the current time is mapped to the contour map of the current time, and the mapping relationship is saved; wherein the preset data training model includes several groups. One-to-one corresponding preset contour map and preset body index data.
  • the specific process of establishing the mapping relationship between the body index data of the current time and the contour map of the current time according to the preset data training model is: the obtained current time point and the acquired current time
  • the body index data is placed in the preset data training model for traversal comparison, and a set of matching Map ⁇ Key, Image> is found to represent the mapping relationship between the current body contour map and the current body index data.
  • the Key in Map ⁇ Key, Image> represents a set of preset body index data (including body mass index data, body fat index data, body mass index data, etc.), and Image represents a preset. Outline map.
  • the current body contour map corresponding to the current body index data can be acquired by inputting the current body index data, or the current body contour map can be acquired by inputting the current body contour map.
  • the data training model is obtained by pre-sampling a large number of samples, and the data training model includes body state posture states of different age groups, different genders, different heights, different weights, and different ethnic users.
  • the current time of the current time at which the body index data has a mapping relationship may be acquired by inputting the body index data of the current time in the preset data training model while acquiring the predicted contour map.
  • the outline map may be acquired by inputting the body index data of the current time in the preset data training model while acquiring the predicted contour map.
  • step “acquiring the final predicted contour map according to the acquired current moment and the predicted contour map” is specifically step S220 to step S224:
  • S220 Acquire the predicted contour map and the current contour map of the current time; wherein the current contour map includes a current contour curve and a current reference point, where the current contour curve is current by a preset edge detection algorithm Obtaining the standard attitude map of the moment, the current reference point is obtained by analyzing the standard attitude map of the current moment by a preset image algorithm; the predicted contour map includes a predicted contour curve And the predicted reference point, the predicted contour curve is obtained according to the curvature of each sample point obtained according to the prediction, and the predicted reference point is obtained according to the current reference point.
  • the preset edge detection algorithm is a Canny algorithm.
  • the preset edge detection algorithm may also be a Roberts algorithm, a Prewitt algorithm, a Sobel algorithm, or a Log algorithm, and is not specifically limited herein.
  • the current reference point when the contour map is the front body contour map, the reference point is preferably the above-mentioned pupil center point
  • the current reference point can be obtained by the above-mentioned FaceDetector control.
  • S221 Acquire the current reference point and the predicted reference point, and overlap the current reference point with the predicted reference point, so that the contour map of the current time overlaps with the predicted contour map.
  • S222 Acquire an overlapping area and a non-coincident area of the current contour curve and the predicted contour curve, render the overlapping area as a preset color, and determine whether the overlapping area is located in the predicted contour curve. In the closed area.
  • the non-coincident region between the current contour curve and the predicted contour curve is located in a closed region surrounded by the predicted contour curve, indicating that the future body shape of the user is larger than the existing body shape ( For example, getting fat); and when the current contour curve and the predicted contour curve are between The non-coincident region is located in a closed region surrounded by the current contour curve, indicating that the user's body size is becoming smaller (eg, thinner).
  • the specific process of rendering is: when the non-coincident region is located in a closed region surrounded by the predicted contour curve, the current contour curve is closest to the contour curve of the non-coincident region and the coincident region Taking the pixel point P(x, y) as an example, the gray value of the pixel point P is obtained, the gray value is multiplied by 70% to obtain a new gray value, and the new gray value is assigned to the pixel point P. 1 to 3 overlapping area pixels or non-coincident area pixels, and so on, until all the pixel points of the overlapping area and all the pixels of the non-coincident area obtain new gray value, and finally realize the overlapping area Rendering with the non-coincident area.
  • the pixels of the non-coincident region are adjusted to an image (eg, the contour map of the current moment) background color (eg, the image background color is pure White, the pixels of the non-coincident area are adjusted to white, and in effect, the portion of the contour map at the current time that exceeds the predicted contour map is erased.
  • an image eg, the contour map of the current moment
  • background color eg, the image background color is pure White
  • the body index data of the current time is mapped to the contour map of the current time according to a preset data training model, and then the predicted contour map is acquired and acquired with the current time.
  • the body index data has the contour map of the current moment of the mapping relationship, and finally obtains the final predicted contour map (ie, the rendered contour map) according to the acquired current moment and the predicted contour map. . Therefore, the preferred embodiment can realize the prediction of the user's body shape, so that the user can adjust his or her fitness plan according to his or her predicted body shape, so the present invention has a good user experience.
  • the method further includes:
  • the step of “acquiring the final predicted contour map according to the acquired current moment and the predicted contour map” is specifically the steps S220 to S222 of the foregoing steps.
  • steps S220 to S222 of the foregoing steps For details, refer to the above content. , will not repeat them here.
  • the final predicted body contour map (ie, the rendered contour map) is derived from the current body contour map and the predicted body contour map acquired at the current time. Therefore, the preferred embodiment can realize the prediction of the user's body shape, so that the user can adjust his or her fitness plan according to his or her predicted body shape, so the present invention has a good user experience.
  • another aspect of the present invention further provides a body type prediction device, which includes: a contour map obtaining module 10, configured to acquire a standard posture map of a predetermined part of a user collected at a current time, and according to the current time
  • the standard attitude map acquires a contour map of the current current time;
  • the contour curve obtaining module 11 is configured to perform boundary sampling on the contour map at the current time to obtain corresponding sampling points, and fit according to all the sample points obtained.
  • the curvature calculation module 12 is configured to calculate and save the curvature of each of the sampling points on the contour curve at the current time;
  • the curvature function obtaining module 13 is configured to use the same sampling point at the current Curvature function of each sampling point is obtained by the time and the curvature corresponding to the predetermined time period before the current time;
  • the curvature prediction module 14 is configured to predict each sampling point in a predetermined future according to the curvature function corresponding to each sampling point.
  • a contour map prediction module 15 configured to obtain a predicted wheel based on the predicted curvature of each sample point Profile.
  • the contour curve obtaining module 11 includes: a coordinate establishing unit 110, configured to establish a two-dimensional coordinate system in the contour map at the current time; and a sampling unit 111, configured to follow a predetermined pixel pitch Performing boundary sampling on the contour map at the current moment to acquire corresponding sampling points, and storing the acquired sampling points in the form of coordinates; the fitting unit 112 is configured to each according to a preset fitting algorithm The coordinates corresponding to the sampling points are curve-fitted to obtain the contour curve at the current time.
  • the contour map acquiring module 10 acquires a standard posture map of a predetermined part of the user collected at the current time, and obtains a corresponding posture according to the standard posture map of the current time. a contour map of the previous moment; and the boundary curve acquisition module 11 performs boundary sampling on the contour map of the current time to obtain a corresponding sampling point, and obtains a contour curve of the current time according to all the sampled points obtained; Then, the curvature calculation module 12 calculates and saves the curvature of each of the sampling points on the contour curve at the current time; and then passes the curvature function acquisition module 13 according to the same sampling point before the current time and the current time.
  • the curvature corresponding to the predetermined time period, the curvature function of each sampling point is obtained; and then the curvature prediction module 14 predicts the curvature of each sampling point at a predetermined future time according to the curvature function corresponding to each sampling point. Finally, the contour map prediction module 15 obtains the predicted contour map according to the predicted curvature of each sampling point, thereby implementing the process of the user's body shape prediction. Therefore, the present invention can predict the size of the user, so that the user can adjust his or her fitness plan according to his or her predicted body shape, so the present invention has a good user experience.
  • the storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Surgery (AREA)
  • Medical Informatics (AREA)
  • Veterinary Medicine (AREA)
  • Public Health (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Animal Behavior & Ethology (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Psychiatry (AREA)
  • Physiology (AREA)
  • Signal Processing (AREA)

Abstract

一种体型预测方法及设备，该方法包括以下步骤：获取当前时刻采集到的用户的预定部位的标准姿态图，并根据当前时刻的标准姿态图获取对应的当前时刻的轮廓图（S10）；对当前时刻的轮廓图进行边界采样，以获取相应的采样点，并根据获取的所有采样点拟合得到当前时刻的轮廓曲线（S11）；计算并保存每个采样点在当前时刻的轮廓曲线上的曲率（S12）；根据同一个采样点在当前时刻以及当前时刻之前的预定时刻阶段内所对应的曲率，得到每个采样点的曲率函数（S13）；根据每个采样点所对应的曲率函数预测每个采样点在预定的未来时刻的曲率（S14）；根据预测得到的每个采样点的曲率得到预测的轮廓图（S15）。该方法及设备能够预测用户的体型，从而提高了用户的体验。

Description

一种体型预测方法及设备 技术领域
本发明涉及数据分析技术领域，尤其涉及一种体型预测方法及设备。
背景技术
现如今，人们越来越关注自己的身体健康状态，尤其是自己的体型。但是现在人们只可以通过一些健康指标采集设备(例如体脂称)来获取自己当前的一些身体健康指标(例如体重或者身高等)而让人们可以方便获知自己的当前健康状态(例如体重是否增加或者身高是否增加等)，并不能获取用户的体型图像数据。而且现有的这些身体健康指标采集设备只有简单的数据采集记录的功能或者是只能根据采集到的数据对个人目前的健康状态进行简单评估，而不能对未来的体型状态进行预测。因此现有的身体健康指标采集设备与测量方法不能让用户提前预知自己未来的体型状态，从而让用户很难根据实际情况调整健身保养计划，进而影响了用户的体验。
发明内容
针对上述问题，本发明的目的在于提供一种能够预测体型且具有良好用户体验的体型预测方法及设备。
为了实现上述目的，本发明一方面提供了一种体型预测方法，其包括以下步骤：获取当前时刻采集到的用户的预定部位的标准姿态图，并根据当前时刻的所述标准姿态图获取对应的当前时刻的轮廓图；对当前时刻的所述轮廓图进行边界采样，以获取相应的采样点，并根据获取的所有采样点拟合得到当前时刻的轮廓曲线；计算并保存每个所述采样点在当前时刻的所述轮廓曲线上的曲率；根据同一个采样点在当前时刻以及当前时刻之前的预定时刻阶段内所对应 的曲率，得到每个采样点的曲率函数；根据每个采样点所对应的曲率函数预测每个采样点在预定的未来时刻的曲率；根据预测得到的每个采样点的曲率得到预测的轮廓图。
进一步地，所述步骤“对当前时刻的所述轮廓图进行边界采样，以获取相应的采样点，并根据获取的所有采样点拟合得到当前时刻的轮廓曲线”具体为：在当前时刻的所述轮廓图中建立二维坐标系；按照预定的像素间距对当前时刻的所述轮廓图进行边界采样，以获取相应的采样点，并将获取到的采样点以坐标的形式进行保存；根据预设的拟合算法对每一个采样点所对应的坐标进行曲线拟合，以获取当前时刻的所述轮廓曲线。
进一步地，所述步骤“根据预测得到的每个采样点的曲率得到预测的轮廓图”具体为：在每一个采样点绘制具有与所述采样点对应的曲率的切线，以获取相邻的两个所述采样点的切线的交点；依次连接所述交点，以形成所述预测的轮廓图。
进一步地，所述体型预测方法还包括：当判断当前时刻的信息量大于预设的阈值时，根据同一个采样点在当前时刻以及当前时刻之前的预定时刻阶段内所对应的曲率，得到每个采样点的曲率函数；其中，所述信息量包括保存在预设的轮廓图库中的轮廓图的数量以及同一个采样点在当前时刻所对应的曲率与当前时刻之前的预定时刻阶段内的某一个对应的曲率的两者之间的差异值。
进一步地，所述体型预测还包括：获取并保存当前时刻采集到的用户的身体指数数据；其中，所述身体指数数据具体为用户的身体指数的指数数据，所述身体指数至少包括以下其中之一：体重指数、身体质量指数、体脂率指数以及肌肉密度指数；根据同一个身体指数在当前时刻以及当前时刻之前的预定时刻阶段内所对应的指数数据，获取相应的身体指数函数；根据所述身体指数函数预测用户在预定的未来时间的身体指数数据。
进一步地，在所述步骤“根据所述身体指数函数预测用户在预定的未来时间的身体指数数据”之后还包括：根据预设的数据训练模型将当前时刻的所述身体指数数据与当前时刻的所述轮廓图建立映射关系，并保存所述映射关系； 其中，所述预设的数据训练模型包括若干组一一对应的预设的轮廓图和预设的身体指数数据；获取所述预测的轮廓图以及获取与当前时刻的所述身体指数数据具有映射关系的当前时刻的所述轮廓图；根据获取到的当前时刻所述轮廓图与所述预测的轮廓图获取最终的预测的轮廓图。
进一步地，所述步骤“根据预测得到的每个采样点的曲率得到预测的轮廓图”之后还包括：根据获取到的当前时刻所述轮廓图与所述预测的轮廓图获取最终的预测的轮廓图。
进一步地，所述步骤“根据获取到的当前时刻所述轮廓图与所述预测的轮廓图获取最终的预测的轮廓图”具体为：获取所述预测的轮廓图和当前时刻的所述当前轮廓图；其中，所述当前轮廓图包括当前轮廓曲线与当前参照点，所述当前轮廓曲线是由预设的边缘检测算法对当前时刻的所述标准姿态图进行分析得出的，所述当前参照点是由预设的图像算法对当前时刻的所述标准姿态图进行分析得出的；所述预测的轮廓图包括预测轮廓曲线与预测参照点，所述预测轮廓曲线是根据预测得到的每个采样点的曲率而得到的，所述预测参照点是根据所述当前参照点得到的；获取所述当前参照点与所述预测参照点，并将所述当前参照点与所述预测参照点重合，以使得当前时刻的所述轮廓图与所述预测的轮廓图重叠；获取所述当前轮廓曲线与所述预测轮廓曲线的重合区域与非重合区域，将所述重合区域渲染为预设的颜色，并判断所述重合区域是否位于所述预测轮廓曲线围成的闭合区域中；若是则将所述非重合区域渲染为所述预设的颜色；若否则将所述非重合区域渲染为图像背景色。
本发明另外一方面还提供了一种体型预设设备，其包括：轮廓图获取模块，用于获取当前时刻采集到的用户的预定部位的标准姿态图，并根据当前时刻的所述标准姿态图获取对应的当前时刻的轮廓图；轮廓曲线获取模块，用于对当前时刻的所述轮廓图进行边界采样，以获取相应的采样点，并根据获取的所有采样点拟合得到当前时刻的轮廓曲线；曲率计算模块，用于计算并保存每个所述采样点在当前时刻的所述轮廓曲线上的曲率；曲率函数获取模块，用于根据 同一个采样点在当前时刻以及当前时刻之前的预定时刻阶段内所对应的曲率，得到每个采样点的曲率函数；曲率预测模块，用于根据每个采样点所对应的曲率函数预测每个采样点在预定的未来时刻的曲率；轮廓图预测模块，用于根据预测得到的每个采样点的曲率得到预测的轮廓图。
进一步地，所述轮廓曲线获取模块包括：坐标建立单元，用于在当前时刻的所述轮廓图中建立二维坐标系；采样单元，用于按照预定的像素间距对当前时刻的所述轮廓图进行边界采样，以获取相应的采样点，并将获取到的采样点以坐标的形式进行保存；拟合单元，用于根据预设的拟合算法对每一个采样点所对应的坐标进行曲线拟合，以获取当前时刻的所述轮廓曲线。
本发明提供的所述体型预测方法及所述体型预测设备，通过获取当前时刻采集到的用户的预定部位的标准姿态图，并根据当前时刻的所述标准姿态图获取对应的当前时刻的轮廓图；并且对当前时刻的所述轮廓图进行边界采样，以获取相应的采样点，并根据获取的所有采样点拟合得到当前时刻的轮廓曲线；接着计算并保存每个所述采样点在当前时刻的所述轮廓曲线上的曲率；然后根据同一个采样点在当前时刻以及当前时刻之前的预定时刻阶段内所对应的曲率，得到每个采样点的曲率函数；再然后根据每个采样点所对应的曲率函数预测每个采样点在预定的未来时刻的曲率；最后根据预测得到的每个采样点的曲率得到预测的轮廓图，从而实现用户的体型预测的过程。因此本发明能够预测出用户的体型，从而可以让用户根据自己的预测体型调整自己的健身计划，所以本发明具有良好的用户体验。
附图说明
为了更清楚地说明本发明的技术方案，下面将对实施方式中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本发明的一些实施方式，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1是本发明实施例提供的一种体型预测方法流程图；
图2是本发明实施例提供的一种体型预测设备的结构示意图；
图3是图2提供的轮廓曲线获取模块的结构示意图。
具体实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
请参见图1，本发明一方面提供了一种体型预测方法，其包括步骤S10至步骤S15：
S10，获取当前时刻采集到的用户的预定部位的标准姿态图，并根据当前时刻的所述标准姿态图获取对应的当前时刻的轮廓图。
其中，所述标准姿态图指的是摄像头对用户在标准姿态状态下所拍摄出来的图像。例如，如果是全身拍摄，则所述标准姿态状态可以为：用户赤足、双腿伸直站立，两腿跨度与肩部同宽，两手臂微抬与体侧呈45度角；其中，男士穿短裤，女士穿运动短裤和运动短上衣，且男士与女士均需露出全部面部、手臂、腿部、腰腹部区域。以便所述摄像头对所述预定的部位进行图像采集。此外，在所述摄像头进行图像采集的过程中，用户的正面面对全身高的体型预测设备(例如镜面显示设备)以及所述体型预测设备的高清摄像头。
需要说明的是，所述用户的预定部位在本发明实施例中优选为用户的全身，当然还可以为用户的手部，面部或者腿部等，在此不做具体限定。
此外，所述标准姿态应只要能全面展示预定部位的形态特征且用户能方便做出该动作即可(方便用户进行重复动作)，本发明不做具体限定。
需要说明的是，当前时刻的所述轮廓图是根据预设的边缘检测算法对当前时刻的所述标准姿态图进行图像轮廓边缘检测而得出的。优选地，所述预设的 边缘检测算法为Canny算法。此外，所述预设的边缘检测算法还可以为Roberts算法、Prewitt算法、Sobel算法或者Log算法等，在此不做具体限定。此外，优选地，所述标准姿态图为用户的预定部位的身体轮廓曲线(还可以包括用户的预定部位的参照点)所构成的一个闭合区域而形成的一幅图像，所述身体轮廓曲线(或者所述瞳孔中心点)与所述当前身体轮廓图中的背景区域的颜色是不相同的，例如所述身体轮廓曲线可以为黑色，所述背景区域为白色或者所述身体轮廓曲线可以为红色，所述背景区域为黄色等。
S11，对当前时刻的所述轮廓图进行边界采样，以获取相应的采样点，并根据获取的所有采样点拟合得到当前时刻的轮廓曲线。
其中，优选地，步骤S11具体包括步骤S110至步骤S112：
S110，在当前时刻的所述轮廓图中建立二维坐标系。
在这里，以用户的正面全身的当前时刻的所述轮廓图举例说明：请参见图，将所述轮廓图中的两脚的底部的同一切线作为X轴(以两脚底部的同一切线作为X轴是因为人体的体型在发生变化的过程中，人体的两脚的底部所在的平面是几乎不发生变化的)，以两脚的底部所构成的线段的中线作为Y轴(以两脚的底部所构成的线段的中线作为Y轴是因为人的身体相对于该中线几乎对称)，以此建立当前时刻的所述轮廓图的二维坐标系。又例如当所述轮廓图为用户的头部轮廓图时，可以以用户的两个瞳孔中心点的连线作为基准线，然后作一条平行于所述基准线的且与头部的底部相切的第一切线，将所述第一切线作为X轴，再然后作一条垂直于该X轴且与头部的侧边相切的第二切线，并将所述第二切线作为Y轴，从而建立头部的二维坐标系。需要说明的是，所述头部轮廓图的瞳孔中心点是由预设的图像算法对摄像头拍摄出来的头部图像进行分析而获取出来的，例如所述头部轮廓图的瞳孔中心点可以通过开源计算机视觉库(Open Source Computer Vision Library，OpenCV)中的FaceDetector控件来获取，即对所述头部图像调用FaceDetector控件时，FaceDetector控件会自动定位所述头部图像中的两个瞳孔中心像素，该两个瞳孔中心像素即为所述头部轮廓图中 的两个所述瞳孔中心点。
当所述轮廓图为用户其他身体部位的轮廓图时，建立二维坐标系的过程可以参考上述两个例子，在此不再赘述。
S111，按照预定的像素间距对当前时刻的所述轮廓图进行边界采样，以获取相应的采样点，并将获取到的采样点以坐标的形式进行保存。
即，以所述轮廓图中的某一行(或某一列)的像素中的某一个像素(例如所述轮廓图中的最左边或者最下边的像素)为起始采样点，然后对该行(或该列)的像素逐一进行采样，直到将该行(或该列)中的所有像素点采样完毕，将该行(或该列)中获取到的轮廓图中的轮廓曲线的像素以坐标的形式按照采样时间的先后顺序保存在预设的采样点点阵中。接着以该行(或该列)为基准按照预定的像素间距(例如一个像素或者两个像素等)获取所述轮廓图中下一行(或下一列)的像素，按照相同的采样顺序对该行(或该列)中的像素进行逐一采样，将该行(或该列)中获取到的轮廓图中的轮廓曲线的像素以坐标的形式按照采样时间的先后顺序保存在所述预设的采样点点阵中。以此类推，可以获取出所述轮廓图中的轮廓曲线的相应的采样点。其中，每一个获取到的轮廓图都有一个与其唯一对应的预设的采样点点阵。
在此，以上述的已经建立二维坐标系的用户的正面全身的轮廓图举一个例子：将所述正面全身轮廓图的y坐标设为0，而x坐标由所述正面全身轮廓图的左边缘向所述正面全身轮廓图的右边缘逐个像素点(即坐标点)递增(即由x轴负方向向x轴正方向递增)，直到将所述当前身体轮廓图的左边缘与右边缘的之间的y坐标为0的所有像素点逐一检索完，并依次将检索到轮廓像素的采样点以坐标的形式进行表示，例如P01(x1，0)，P02(x2，0)，P03(x3，0)，P04(x4，0)等；然后对所述当前身体轮廓图的左边缘与右边缘的之间的Y轴的坐标为1的所有轮廓像素点进行检索，并依次将检索到轮廓像素的采样点以坐标的形式进行表示，例如P11(x1，1)，P12(x2，1)，P13(x3，1)，P14(x4，1)等；依次将y的坐标以1的增量递增，直到当在所述当前身体轮廓图的左边缘与右边缘的之间没有检索到轮廓像素点时，停止检索与采样，最终将 所述当前时刻的正面全身轮廓图得到的所有的采样点以坐标的形式依次存放在预设的采样点点阵中。可以理解的是，用户的其他身体部位的轮廓图的边界采样与所述正面全身轮廓图的边界采样的过程相同，在此不再赘述。
S112，根据预设的拟合算法对每一个采样点所对应的坐标进行曲线拟合，以获取当前时刻的所述轮廓曲线。
将得到的每一个所述预设的采样点点阵中的所述采样点利用拟合函数进行拟合，即，通过将要拟合的所有所述采样点的坐标数据(x，y)全部导入到拟合函数polyfit中，可以得到当前时刻的所述轮廓图的轮廓曲线函数：y＝f(x)，该轮廓曲线函数即代表该轮廓图的轮廓曲线。其中，每一个所述轮廓曲线与每一个所述预设的采样点点阵对应，即，每一个所述轮廓曲线与该预定部位的每一张轮廓图对应。
S12，计算并保存每个所述采样点在当前时刻的所述轮廓曲线上的曲率。
当当前时刻的所述轮廓图获取到相应的轮廓曲线时，计算与当前时刻的所述轮廓图对应的所述预设的采样点点阵中的每一个采样点在当前时刻的所述轮廓曲线上的曲率。其中，所述预设的采样点点阵中的每一个采样点的曲率的获取过程为：
首先将当前时刻的所述轮廓曲线的函数转换成相应的参数方程形式：即将y＝f(x)转换成
Figure PCTCN2017104673-appb-000001
其中，x，y分别代表每一个所述轮廓曲线的函数的x、y坐标。然后根据以下公式求出所述预设的采样点点阵中的每一个采样点P(x，y)的在所述轮廓曲线上的曲率：
Figure PCTCN2017104673-appb-000002
Figure PCTCN2017104673-appb-000003
将获取到的与每一个采样点的对应的所述曲率按照每一个采样点在所述预 设的采样点点阵中的保存时间的先后顺序(即采样点的采样时间的先后顺序)存储在预设的曲率数据库中。其中，每一个所述轮廓图的所述预设的采样点点阵与所述预设的曲率数据库一一对应。其中，在当前时刻的所述轮廓图之前的轮廓图的相应采样点所对应的曲率也是按照采样时间的先后顺序保存在所述预设的曲率数据库中的。
S13，根据同一个采样点在当前时刻以及当前时刻之前的预定时刻阶段内所对应的曲率，得到每个采样点的曲率函数。
即获取所述预设的曲率数据库中的同一个所述采样点在当前时刻以及当前时刻之前的预定时刻阶段内所对应的曲率。在此，以上述的采样点P01(x1，y1)为例，通过统计获得采样点P01在所述预设的曲率数据库中的保存时间在t1～tn时间段内的所对应的曲率，获得采样点P01的相应的曲率数据库Ktn＝[Kt1，Kt2......，Ktn]，将采样点P01的曲率数据库放入相应的曲线拟合工具中，选择平滑曲线进行拟合，获得采样点P01在保存时间范围内用于表示其所对应的曲率变化情况的曲率函数表达式K＝f(t)。依次类推，可以获取出每一个所述采样点点阵中的每一个采样点的所对应的曲率函数。
S14，根据每个采样点所对应的曲率函数预测每个采样点在预定的未来时刻的曲率。
即，通过将每一个采样点点阵中的同一个采样点的预定的未来时刻输入到对应的曲率函数中，即可获取出该采样点在预定的未来时刻的预测的曲率。以此类推，可获取出每一个采样点点阵中的每一个采样点的预测的曲率。
S15，根据预测得到的每个采样点的曲率得到预测的轮廓图。
优选地，步骤S15具体包括步骤S150至步骤S151：
S150，在每一个采样点绘制具有与所述采样点对应的曲率的切线，以获取相邻的两个所述采样点的切线的交点。
根据所述当前身体轮廓图中的每一个采样点所对应的预测的曲率，在所述 当前身体轮廓图中的每一个所述采样点绘制具有相应曲率的切线，这样可以获取相邻的两个所述采样点的切线的交点。
S151，依次连接所述交点，以形成所述预测的轮廓图。
以平滑曲线的方式依次连接相邻的两个交点以形成所述预测的轮廓图。需要说明的是，获取到的所述预测的轮廓图可以通过相关设备(例如全身高的镜面显示设备)的显示屏进行显示。
在本发明实施例中，通过获取当前时刻采集到的用户的预定部位的标准姿态图，并根据当前时刻的所述标准姿态图获取对应的当前时刻的轮廓图；并且对当前时刻的所述轮廓图进行边界采样，以获取相应的采样点，并根据获取的所有采样点拟合得到当前时刻的轮廓曲线；接着计算并保存每个所述采样点在当前时刻的所述轮廓曲线上的曲率；然后根据同一个采样点在当前时刻以及当前时刻之前的预定时刻阶段内所对应的曲率，得到每个采样点的曲率函数；再然后根据每个采样点所对应的曲率函数预测每个采样点在预定的未来时刻的曲率；最后根据预测得到的每个采样点的曲率得到预测的轮廓图，从而实现用户的体型预测的过程。因此本发明能够预测出用户的体型，从而可以让用户根据自己的预测体型调整自己的健身计划，所以本发明具有良好的用户体验。
为了便于对本发明实施例的理解，在此提供本发明实施例的一些优选实施例：
第一种优选实施例：
所述体型预测方法还包括：
S16，当判断当前时刻的信息量大于预设的阈值时，根据同一个采样点在当前时刻以及当前时刻之前的预定时刻阶段内所对应的曲率，得到每个采样点的曲率函数；其中，所述信息量包括保存在预设的轮廓图库中的轮廓图的数量以及同一个采样点在当前时刻所对应的曲率与当前时刻之前的预定时刻阶段内的某一个对应的曲率的两者之间的差异值。
例如，当所述预设的身体轮廓图库中的身体轮廓图的数量大于10张时，这时才根据同一个采样点在当前时刻以及当前时刻之前的预定时刻阶段内所对应的曲率，得到每个采样点的曲率函数；或者是当检测出同一个采样点在当前时刻所对应的曲率与当前时刻之前的预定时刻阶段内的某一个对应的曲率的两者之间的差异值大于百分之十时，这时才就根据同一个采样点在当前时刻以及当前时刻之前的预定时刻阶段内所对应的曲率，得到每个采样点的曲率函数。其中，所述像素采样点的曲率的获取过程具体请参考上述的步骤S120与步骤S121，在此不再赘述。需要说明的是，因为如果现有的相应的信息量达不到预设的阈值时，表明如下两种情况的其中一种：1、表明现有获取过的轮廓图的数量偏少，这样会得不到相对准确的所述曲率函数，因此当获取到的轮廓图的数量大于预定的阈值时才开始计算采样点的曲率函数，可以获取到更加准确的曲率函数，从而根据更加准确的曲率函数可以获取到更加准确的预测的轮廓图；2、或者是表明现有的身体轮廓图相对于以前获取的轮廓图的同一个采样点的曲率的变化幅度不大，即说明用户的体型的变化幅度不大，因此就无需再进行曲率函数的计算，而是直接显示出当前时刻的所述轮廓图，这样可以减少不必要的程序运行和数据运算，从而提高了相应的体型预测设备的CPU的工作效率。
在本优选实施例中，通过当判断出当前时刻的信息量大于预设的阈值时，才根据同一个采样点在当前时刻以及当前时刻之前的预定时刻阶段内所对应的曲率，并得到每个采样点的曲率函数，这样可以获取出更加准确的曲率函数或者是提高CPU的工作效率。
第二种优选实施例：
所述体型预测方法还包括：
S17，获取并保存当前时刻采集到的用户的身体指数数据；其中，所述身体指数数据具体为用户的身体指数的指数数据，所述身体指数至少包括以下其中之一：体重指数、身体质量指数、体脂率指数以及肌肉密度指数。
其中，所述身体指数数据由相应的身体检测设备来获取的，例如体重指数数据、体脂率指数数据以及肌肉密度指数数据由体脂称来获取。需要说明的是，所述身体指数数据还可以包括其他类型的身体指数数据，例如蛋白质指数(由蛋白质测量仪来获取)或者皮肤水分指数(由水分传感器测量得到)等，在此不做具体限定。
S18，根据同一个身体指数在当前时刻以及当前时刻之前的预定时刻阶段内所对应的指数数据，获取相应的身体指数函数。
例如根据体重指数在当前时刻以及当前时刻之前的预定时刻阶段内所对应的指数数据，获取相应的体重指数函数。同理，所述身体指数的其他指数获取相应的身体指数函数的过程与体重指数获取体重指数函数相同，在此不再赘述。
S19，根据所述身体指数函数预测用户在预定的未来时间的身体指数数据。
即将未来的某一时刻输入到相应的所述身体指数函数中，就可以获取出未来某一时刻的相应的身体指数数据；例如，将未来的某一时刻输入到所述体重指数函数中，可以获取出相应的未来时间的体重指数数据。需要说明的是，当获取出预定的未来时间的身体指数数据时，这是可以将预定的未来时间的身体指数数据在相应的体型预设设备的显示屏上进行显示。
在本优选实施例中，通过获取并保存当前时刻采集到的用户的身体指数数据，并根据同一个身体指数在当前时刻以及当前时刻之前的预定时刻阶段内所对应的指数数据，获取相应的身体指数函数，最后根据所述身体指数函数预测用户在预定的未来时间的身体指数数据，从而实现用户的相应的身体指数的预测过程。因此本优选实施例能够预测出用户的未来时间的身体指数，从而可以让用户根据自己的未来时间的身体指数调整自己的健身计划，进而进一步提高用户的使用体验。
第三种优选实施例：
在所述步骤“根据所述身体指数函数预测用户在预定的未来时间的身体指数数据”之后还包括：
S20，根据预设的数据训练模型将当前时刻的所述身体指数数据与当前时刻的所述轮廓图建立映射关系，并保存所述映射关系；其中，所述预设的数据训练模型包括若干组一一对应的预设的轮廓图和预设的身体指数数据。
根据预设的数据训练模型将当前时刻的所述身体指数数据与当前时刻的所述轮廓图建立映射关系的具体过程为：将获取到的当前时刻所述轮廓图与获取到的当前时刻所述身体指数数据放到所述预设的数据训练模型中进行遍历对比，找到一组匹配的Map<Key，Image>，以此代表所述当前身体轮廓图和所述当前身体指数数据的映射关系。其中，Map<Key，Image>中的Key代表的是一组预设的身体指数数据(即包括体重指数数据、体脂率指数数据以及身体质量指数数据等)，Image代表的是一张预设的轮廓图。这样，通过输入所述当前身体指数数据就可以获取出与所述当前身体指数数据对应的所述当前身体轮廓图，或者是输入所述当前身体轮廓图就可以获取出与所述当前身体轮廓图对应的所述当前身体指数数据。需要说明的是，所述数据训练模型是通过预先的大量取样而获得的，所述数据训练模型包含各个年龄段、不同性别、不同身高、不同体重、不同人种用户的身体标准姿态状态下的预设的轮廓图和与所述预设的轮廓图对应的预设的身体指数数据，其中，训练模型中每一组所述预设的身体指数数据对应一张所述所述预设的轮廓图。
S21，获取所述预测的轮廓图以及获取与当前时刻的所述身体指数数据具有映射关系的当前时刻的所述轮廓图。
即，在获取所述预测的轮廓图的同时通过在所述预设的数据训练模型中输入当前时刻的所述身体指数数据可以获取与当前时刻的所述身体指数数据具有映射关系的当前时刻的所述轮廓图。
S22，根据获取到的当前时刻所述轮廓图与所述预测的轮廓图获取最终的预 测的轮廓图。
其中，优选地，所述步骤“根据获取到的当前时刻所述轮廓图与所述预测的轮廓图获取最终的预测的轮廓图”具体为步骤S220至步骤S224：
S220，获取所述预测的轮廓图和当前时刻的所述当前轮廓图；其中，所述当前轮廓图包括当前轮廓曲线与当前参照点，所述当前轮廓曲线是由预设的边缘检测算法对当前时刻的所述标准姿态图进行分析得出的，所述当前参照点是由预设的图像算法对当前时刻的所述标准姿态图进行分析得出的；所述预测的轮廓图包括预测轮廓曲线与预测参照点，所述预测轮廓曲线是根据根据预测得到的每个采样点的曲率而得到的，所述预测参照点是根据所述当前参照点得到的。
优选地，所述预设的边缘检测算法为Canny算法。此外，所述预设的边缘检测算法还可以为Roberts算法、Prewitt算法、Sobel算法或者Log算法等，在此不做具体限定。根据预测得到的每个采样点的曲率而得到的所述预测轮廓曲线的过程请参考上述的步骤S150至步骤S151，在此不再赘述。需要说明的是，所述当前参照点(当轮廓图为所述正面全身轮廓图时，所述参照点优选为上述的瞳孔中心点)可以通过上述的FaceDetector控件来获取。
S221，获取所述当前参照点与所述预测参照点，并将所述当前参照点与所述预测参照点重合，以使得当前时刻的所述轮廓图与所述预测的轮廓图重叠。
S222，获取所述当前轮廓曲线与所述预测轮廓曲线的重合区域与非重合区域，将所述重合区域渲染为预设的颜色，并判断所述重合区域是否位于所述预测轮廓曲线围成的闭合区域中。
S223，若是则将所述非重合区域渲染为所述预设的颜色。
S224，若否则将所述非重合区域渲染为图像背景色。
其中，当所述当前轮廓曲线与所述预测轮廓曲线之间的所述非重合区域位于所述预测轮廓曲线围成的闭合区域中，表明用户未来的体型相对于现有的体型在变大(例如变胖)；而当所述当前轮廓曲线与所述预测轮廓曲线之间的所 述非重合区域位于所述当前轮廓曲线围成的闭合区域中，表明用户的体型在变小(例如变瘦)。
优选地，渲染的具体过程为：当所述非重合区域位于所述预测轮廓曲线围成的闭合区域中时，以所述当前轮廓曲线最靠近由所述非重合区域与重合区域的轮廓曲线的像素点P(x，y)为例，获取像素点P的灰度值，将所述灰度值乘以70％得到新的灰度值，将新的灰度值赋值给像素点P临近的1～3个重合区域像素或者非重合区域像素，以此类推，直至所述重合区域的所有像素点和所述非重合区域的所有像素点都获得新的灰度值，最终实现所述重合区域与所述非重合区域的渲染。当所述非重合区域位于所述当前轮廓曲线围成的闭合区域中时，将所述非重合区域的像素调整为图像(例如当前时刻的所述轮廓图)背景色(例如图像背景色是纯白，就将所述非重合区域的像素调整为白色，从效果上看是把当前时刻的所述轮廓图中的超出所述预测的轮廓图的部分抹除)。
在本优选实施例中，通过根据预设的数据训练模型将当前时刻的所述身体指数数据与当前时刻的所述轮廓图建立映射关系，然后获取所述预测的轮廓图以及获取与当前时刻的所述身体指数数据具有映射关系的当前时刻的所述轮廓图，最后根据获取到的当前时刻所述轮廓图与所述预测的轮廓图获取最终的预测的轮廓图(即经过渲染的轮廓图)。因此本优选实例可以实现对用户体型的预测，从而可以让用户根据自己的预测体型调整自己的健身计划，所以本发明具有良好的用户体验。
第四种优选实施例：
在所述步骤“根据预测得到的每个采样点的曲率得到预测的轮廓图”之后还包括：
S22，根据获取到的当前时刻所述轮廓图与所述预测的轮廓图获取最终的预测的轮廓图。
其中，优选地，所述步骤“根据获取到的当前时刻所述轮廓图与所述预测的轮廓图获取最终的预测的轮廓图”具体为上述步骤的步骤S220至步骤S222，具体请参考上述内容，在此不再赘述。
在本优选实施例中，通过所述当前身体轮廓图与当前时刻获取到的所述预测的身体轮廓图得出最终的预测的身体轮廓图(即经过渲染的轮廓图)。因此本优选实例可以实现对用户体型的预测，从而可以让用户根据自己的预测体型调整自己的健身计划，所以本发明具有良好的用户体验。
请参见图2，本发明另外一方面还提供了一种体型预测设备，其包括：轮廓图获取模块10，用于获取当前时刻采集到的用户的预定部位的标准姿态图，并根据当前时刻的所述标准姿态图获取对应的当前时刻的轮廓图；轮廓曲线获取模块11，用于对当前时刻的所述轮廓图进行边界采样，以获取相应的采样点，并根据获取的所有采样点拟合得到当前时刻的轮廓曲线；曲率计算模块12，用于计算并保存每个所述采样点在当前时刻的所述轮廓曲线上的曲率；曲率函数获取模块13，用于根据同一个采样点在当前时刻以及当前时刻之前的预定时刻阶段内所对应的曲率，得到每个采样点的曲率函数；曲率预测模块14，用于根据每个采样点所对应的曲率函数预测每个采样点在预定的未来时刻的曲率；轮廓图预测模块15，用于根据预测得到的每个采样点的曲率得到预测的轮廓图。
优选地，请参见图3，所述轮廓曲线获取模块11包括：坐标建立单元110，用于在当前时刻的所述轮廓图中建立二维坐标系；采样单元111，用于按照预定的像素间距对当前时刻的所述轮廓图进行边界采样，以获取相应的采样点，并将获取到的采样点以坐标的形式进行保存；拟合单元112，用于根据预设的拟合算法对每一个采样点所对应的坐标进行曲线拟合，以获取当前时刻的所述轮廓曲线。
在本发明实施例中，通过所述轮廓图获取模块10获取当前时刻采集到的用户的预定部位的标准姿态图，并根据当前时刻的所述标准姿态图获取对应的当 前时刻的轮廓图；并且通过所述轮廓曲线获取模块11对当前时刻的所述轮廓图进行边界采样，以获取相应的采样点，并根据获取的所有采样点拟合得到当前时刻的轮廓曲线；接着通过所述曲率计算模块12计算并保存每个所述采样点在当前时刻的所述轮廓曲线上的曲率；然后通过所述曲率函数获取模块13根据同一个采样点在当前时刻以及当前时刻之前的预定时刻阶段内所对应的曲率，得到每个采样点的曲率函数；再然后通过所述曲率预测模块14根据每个采样点所对应的曲率函数预测每个采样点在预定的未来时刻的曲率；最后通过所述轮廓图预测模块15根据预测得到的每个采样点的曲率得到预测的轮廓图，从而实现用户的体型预测的过程。因此本发明能够预测出用户的体型，从而可以让用户根据自己的预测体型调整自己的健身计划，所以本发明具有良好的用户体验。
以上所揭露的仅为本发明一些较佳实施例而已，当然不能以此来限定本发明之权利范围，本领域普通技术人员可以理解实现上述实施例的全部或部分流程，并依本发明权利要求所作的等同变化，仍属于发明所涵盖的范围。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机程序来指令相关的硬件来完成，所述的程序可存储于一计算机可读取存储介质中，该程序在执行时，可包括如上述各方法的实施例的流程。其中，所述的存储介质可为磁碟、光盘、只读存储记忆体(Read-Only Memory，ROM)或随机存储记忆体(Random Access Memory，RAM)等。

Claims (10)

  1. 一种体型预测方法，其特征在于，包括如下步骤：
    获取当前时刻采集到的用户的预定部位的标准姿态图，并根据当前时刻的所述标准姿态图获取对应的当前时刻的轮廓图；
    对当前时刻的所述轮廓图进行边界采样，以获取相应的采样点，并根据获取的所有采样点拟合得到当前时刻的轮廓曲线；
    计算并保存每个所述采样点在当前时刻的所述轮廓曲线上的曲率；
    根据同一个采样点在当前时刻以及当前时刻之前的预定时刻阶段内所对应的曲率，得到每个采样点的曲率函数；
    根据每个采样点所对应的曲率函数预测每个采样点在预定的未来时刻的曲率；
    根据预测得到的每个采样点的曲率得到预测的轮廓图。
  2. 根据权利要求1所述的体型预测方法，其特征在于，所述步骤“对当前时刻的所述轮廓图进行边界采样，以获取相应的采样点，并根据获取的所有采样点拟合得到当前时刻的轮廓曲线”具体为：
    在当前时刻的所述轮廓图中建立二维坐标系；
    按照预定的像素间距对当前时刻的所述轮廓图进行边界采样，以获取相应的采样点，并将获取到的采样点以坐标的形式进行保存；
    根据预设的拟合算法对每一个采样点所对应的坐标进行曲线拟合，以获取当前时刻的所述轮廓曲线。
  3. 根据权利要求1所述的体型预测方法，其特征在于，所述步骤“根据预测得到的每个采样点的曲率得到预测的轮廓图”具体为：
    在每一个采样点绘制具有与所述采样点对应的曲率的切线，以获取相邻的两个所述采样点的切线的交点；
    依次连接所述交点，以形成所述预测的轮廓图。
  4. 根据权利要求1所述的体型预测方法，其特征在于，还包括：
    当判断当前时刻的信息量大于预设的阈值时，根据同一个采样点在当前时刻以及当前时刻之前的预定时刻阶段内所对应的曲率，得到每个采样点的曲率函数；其中，所述信息量包括保存在预设的轮廓图库中的轮廓图的数量以及同一个采样点在当前时刻所对应的曲率与当前时刻之前的预定时刻阶段内的某一个对应的曲率的两者之间的差异值。
  5. 根据权利要求1所述的体型预测方法，其特征在于，还包括：
    获取并保存当前时刻采集到的用户的身体指数数据；其中，所述身体指数数据具体为用户的身体指数的指数数据，所述身体指数至少包括以下其中之一：体重指数、身体质量指数、体脂率指数以及肌肉密度指数；
    根据同一个身体指数在当前时刻以及当前时刻之前的预定时刻阶段内所对应的指数数据，获取相应的身体指数函数；
    根据所述身体指数函数预测用户在预定的未来时间的身体指数数据。
  6. 根据权利要求5所述的体型预测方法，其特征在于，在所述步骤“根据所述身体指数函数预测用户在预定的未来时间的身体指数数据”之后还包括：
    根据预设的数据训练模型将当前时刻的所述身体指数数据与当前时刻的所述轮廓图建立映射关系，并保存所述映射关系；其中，所述预设的数据训练模型包括若干组一一对应的预设的轮廓图和预设的身体指数数据；
    获取所述预测的轮廓图以及获取与当前时刻的所述身体指数数据具有映射关系的当前时刻的所述轮廓图；
    根据获取到的当前时刻所述轮廓图与所述预测的轮廓图获取最终的预测的轮廓图。
  7. 根据权利要求1所述的体型预测方法，其特征在于，在所述步骤“根据预测得到的每个采样点的曲率得到预测的轮廓图”之后还包括：
    根据获取到的当前时刻所述轮廓图与所述预测的轮廓图获取最终的预测的轮廓图。
  8. 根据权利要求6或7所述的体型预测方法，其特征在于，所述步骤“根据获取到的当前时刻所述轮廓图与所述预测的轮廓图获取最终的预测的轮廓图”具体为：
    获取所述预测的轮廓图和当前时刻的所述当前轮廓图；其中，所述当前轮廓图包括当前轮廓曲线与当前参照点，所述当前轮廓曲线是由预设的边缘检测算法对当前时刻的所述标准姿态图进行分析得出的，所述当前参照点是由预设的图像算法对当前时刻的所述标准姿态图进行分析得出的；所述预测的轮廓图包括预测轮廓曲线与预测参照点，所述预测轮廓曲线是根据预测得到的每个采样点的曲率而得到的，所述预测参照点是根据所述当前参照点得到的；
    获取所述当前参照点与所述预测参照点，并将所述当前参照点与所述预测参照点重合，以使得当前时刻的所述轮廓图与所述预测的轮廓图重叠；
    获取所述当前轮廓曲线与所述预测轮廓曲线的重合区域与非重合区域，将所述重合区域渲染为预设的颜色，并判断所述重合区域是否位于所述预测轮廓曲线围成的闭合区域中；
    若是，则将所述非重合区域渲染为所述预设的颜色；
    若否，则将所述非重合区域渲染为图像背景色。
  9. 一种体型预测设备，其特征在于，包括：
    轮廓图获取模块，用于获取当前时刻采集到的用户的预定部位的标准姿态图，并根据当前时刻的所述标准姿态图获取对应的当前时刻的轮廓图；
    轮廓曲线获取模块，用于对当前时刻的所述轮廓图进行边界采样，以获取相应的采样点，并根据获取的所有采样点拟合得到当前时刻的轮廓曲线；
    曲率计算模块，用于计算并保存每个所述采样点在当前时刻的所述轮廓曲线上的曲率；
    曲率函数获取模块，用于根据同一个采样点在当前时刻以及当前时刻之前的预定时刻阶段内所对应的曲率，得到每个采样点的曲率函数；
    曲率预测模块，用于根据每个采样点所对应的曲率函数预测每个采样点在预定的未来时刻的曲率；
    轮廓图预测模块，用于根据预测得到的每个采样点的曲率得到预测的轮廓图。
  10. 根据权利要求9所述的体型预测设备，其特征在于，所述轮廓曲线获取模块包括：
    坐标建立单元，用于在当前时刻的所述轮廓图中建立二维坐标系；
    采样单元，用于按照预定的像素间距对当前时刻的所述轮廓图进行边界采样，以获取相应的采样点，并将获取到的采样点以坐标的形式进行保存；
    拟合单元，用于根据预设的拟合算法对每一个采样点所对应的坐标进行曲线拟合，以获取当前时刻的所述轮廓曲线。
PCT/CN2017/104673 2016-12-15 2017-09-29 一种体型预测方法及设备 Ceased WO2018107872A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201621377971 2016-12-15
CN201621377971.9 2016-12-15

Publications (1)

Publication Number Publication Date
WO2018107872A1 true WO2018107872A1 (zh) 2018-06-21

Family

ID=62557901

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/104673 Ceased WO2018107872A1 (zh) 2016-12-15 2017-09-29 一种体型预测方法及设备

Country Status (2)

Country Link
CN (1) CN106780530B (zh)
WO (1) WO2018107872A1 (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110533682A (zh) * 2019-08-30 2019-12-03 福建省德腾智能科技有限公司 一种基于曲率滤波的图像边缘实时提取方法
CN111202494A (zh) * 2018-11-22 2020-05-29 松下知识产权经营株式会社 皮肤分析装置、皮肤分析方法以及记录介质
CN112419479A (zh) * 2020-11-10 2021-02-26 广州二元科技有限公司 一种基于体重身高及身体影像的体型数据计算方法
CN116520822A (zh) * 2023-03-13 2023-08-01 中国人民解放军国防科技大学 一种光滑的曲率参数化表示的路径生成方法

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106780530B (zh) * 2016-12-15 2019-06-14 广州视源电子科技股份有限公司 一种体型预测方法及设备
CN107958311A (zh) * 2017-12-11 2018-04-24 新疆普昂信息技术有限公司 一种镜面显示方法及镜面显示设备
CN109393614B (zh) * 2018-09-03 2021-02-19 海盐昆植生物技术有限公司 一种量体裁衣尺寸测量的系统
CN109459119B (zh) * 2018-10-17 2020-06-05 京东数字科技控股有限公司 一种体重测量方法、设备和计算机可读存储介质
CN112783949B (zh) * 2021-01-29 2024-04-09 北京百度网讯科技有限公司 人体数据预测方法、装置、电子设备和存储介质
CN113628233A (zh) * 2021-08-02 2021-11-09 嘉兴九全科技服务有限公司 基于互联网的体型管理方法、装置、设备及存储介质

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101989326A (zh) * 2009-07-31 2011-03-23 三星电子株式会社 人体姿态识别方法和装置
CN102178530A (zh) * 2011-01-24 2011-09-14 天津大学 基于三维点云数据的人体尺寸自动测量方法
JP2012048649A (ja) * 2010-08-30 2012-03-08 Citizen Holdings Co Ltd 体形変化予測装置
CN106780530A (zh) * 2016-12-15 2017-05-31 广州视源电子科技股份有限公司 一种体型预测方法及设备

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101499177A (zh) * 2008-01-28 2009-08-05 上海西门子医疗器械有限公司 一种三维模型的建立方法和系统
CN101750031B (zh) * 2009-12-09 2011-08-10 华中科技大学 一种二维轮廓形状的测量方法及测量装置
CN101763649B (zh) * 2009-12-30 2012-07-25 北京航空航天大学 一种增强模型轮廓的表面点绘制方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101989326A (zh) * 2009-07-31 2011-03-23 三星电子株式会社 人体姿态识别方法和装置
JP2012048649A (ja) * 2010-08-30 2012-03-08 Citizen Holdings Co Ltd 体形変化予測装置
CN102178530A (zh) * 2011-01-24 2011-09-14 天津大学 基于三维点云数据的人体尺寸自动测量方法
CN106780530A (zh) * 2016-12-15 2017-05-31 广州视源电子科技股份有限公司 一种体型预测方法及设备

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111202494A (zh) * 2018-11-22 2020-05-29 松下知识产权经营株式会社 皮肤分析装置、皮肤分析方法以及记录介质
CN110533682A (zh) * 2019-08-30 2019-12-03 福建省德腾智能科技有限公司 一种基于曲率滤波的图像边缘实时提取方法
CN110533682B (zh) * 2019-08-30 2023-02-14 福建省德腾智能科技有限公司 一种基于曲率滤波的图像边缘实时提取方法
CN112419479A (zh) * 2020-11-10 2021-02-26 广州二元科技有限公司 一种基于体重身高及身体影像的体型数据计算方法
CN112419479B (zh) * 2020-11-10 2022-09-06 广州二元科技有限公司 一种基于体重身高及身体影像的体型数据计算方法
CN116520822A (zh) * 2023-03-13 2023-08-01 中国人民解放军国防科技大学 一种光滑的曲率参数化表示的路径生成方法
CN116520822B (zh) * 2023-03-13 2023-09-19 中国人民解放军国防科技大学 一种光滑的曲率参数化表示的路径生成方法

Also Published As

Publication number Publication date
CN106780530A (zh) 2017-05-31
CN106780530B (zh) 2019-06-14

Similar Documents

Publication Publication Date Title
WO2018107872A1 (zh) 一种体型预测方法及设备
Berssenbrügge et al. 2D and 3D analysis methods of facial asymmetry in comparison
WO2020156148A1 (zh) Smpl参数预测模型的训练方法、计算机设备及存储介质
WO2020103417A1 (zh) 一种bmi评测方法、装置及计算机可读存储介质
CN114470719B (zh) 一种全自动姿态矫正训练方法及系统
CN106408566B (zh) 一种胎儿超声图像质量控制方法及系统
CN113033481A (zh) 结合长宽比优先的一阶全卷积目标检测（fcos）算法的手持棍棒的检测方法
CN113782184A (zh) 一种基于面部关键点与特征预学习的脑卒中辅助评估系统
CN111402338B (zh) 一种穴位定位方法
CN116469174A (zh) 一种基于单目摄像头的深度学习坐姿测量检测方法
CN111383212B (zh) 一种盆底超声视频图像的分析方法
CN116434952A (zh) 阅读能力评估方法、装置、设备及存储介质
CN109549619B (zh) 眼底盘沿宽度确定方法、青光眼疾病诊断装置和系统
CN111652300A (zh) 脊柱曲度分类方法、计算机设备和存储介质
Jia et al. Image-based volume estimation for food in a bowl
CN115393351B (zh) 一种基于朗格汉斯细胞判断角膜免疫状态的方法及装置
CN120089336A (zh) 基于数据和知识融合驱动的肢端肥大患者早期筛查系统
CN112528874A (zh) 一种面向人的sar图象多目标视觉识别能力量化方法、系统及计算机可读介质
CN113327236B (zh) 新型冠状病毒抗体快速检测试剂的识别方法及系统
CN117745902A (zh) 一种用于康复演示的数字人生成方法及装置
Wang et al. Automatic image segmentation and cobb measurement of spine base on u-net
CN113781453B (zh) 一种基于x线片的脊柱侧凸进展预测及方法和装置
CN114648512A (zh) 舌下图像分析方法、装置、计算机设备和存储介质
CN120392525B (zh) 一种稳定识别人体背部穴位的方法及智能终端
CN113850130A (zh) 一种验光仪瞳孔中心定位的方法及系统

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17881216

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 25.10.2019)

122 Ep: pct application non-entry in european phase

Ref document number: 17881216

Country of ref document: EP

Kind code of ref document: A1