CN117292060A - Three-dimensional human body touchless reconstruction method based on sound wave backscattering - Google Patents

Three-dimensional human body touchless reconstruction method based on sound wave backscattering Download PDF

Info

Publication number
CN117292060A
CN117292060A CN202311229155.8A CN202311229155A CN117292060A CN 117292060 A CN117292060 A CN 117292060A CN 202311229155 A CN202311229155 A CN 202311229155A CN 117292060 A CN117292060 A CN 117292060A
Authority
CN
China
Prior art keywords
human body
characteristic
far
field data
dimensional
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.)
Pending
Application number
CN202311229155.8A
Other languages
Chinese (zh)
Inventor
尹伟石
孟品超
刘宏宇
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.)
Changchun University of Science and Technology
Original Assignee
Changchun University of Science and Technology
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 Changchun University of Science and Technology filed Critical Changchun University of Science and Technology
Priority to CN202311229155.8A priority Critical patent/CN117292060A/en
Publication of CN117292060A publication Critical patent/CN117292060A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S15/00—Systems using the reflection or reradiation of acoustic waves, e.g. sonar systems
    • G01S15/02—Systems using the reflection or reradiation of acoustic waves, e.g. sonar systems using reflection of acoustic waves
    • G01S15/06—Systems determining the position data of a target
    • G01S15/42—Simultaneous measurement of distance and other co-ordinates
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras
    • G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10028—Range image; Depth image; 3D point clouds
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Remote Sensing (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Geometry (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Acoustics & Sound (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Software Systems (AREA)
  • Computer Graphics (AREA)
  • Image Analysis (AREA)

Abstract

A three-dimensional human body touchless reconstruction method based on sound wave backscattering belongs to the technical field of three-dimensional modeling, and human body characteristic parameters are extracted according to scanned human body point cloud data acquired by equipment; automatically extracting human body characteristic points and characteristic section point clouds by utilizing the principle of anthropometry, extracting characteristic boundary lines by the characteristic boundary to obtain characteristic boundary lines, and calculating the characteristic size of the human body; according to the human body manifold principle, a group of Fourier coefficients are adopted to represent the geometric shape, and due to the fact that the human body shape is complex, limited characteristic values are considered, far-field data of all characteristic parameters of the human body shape are obtained; obtaining far-field data samples corresponding to a human body through a neural network construction model, inverting the shape of the human body through a Fourier method, and establishing a three-dimensional model. The invention obtains the characteristic parameters of the human body through sound wave back scattering and can quickly reconstruct the three-dimensional human body shape.

Description

Three-dimensional human body touchless reconstruction method based on sound wave backscattering
Technical Field
The invention belongs to the technical field of three-dimensional modeling, and particularly relates to a three-dimensional human body touchless reconstruction method based on sound wave backscattering.
Background
Along with the integration of electronic information technology and production life, the three-dimensional human modeling technology is widely applied to the fields of online fitting, movie animation production, game modeling and the like, the existing three-dimensional modeling is realized by means of technologies such as computer graphics, computer vision, virtual reality and the like, and then a three-dimensional model is artificially generated by using geometric modeling software. Therefore, a new solution is needed in the prior art to solve the above-mentioned problems.
Disclosure of Invention
The technical problems to be solved by the invention are as follows: the three-dimensional human body touchless reconstruction method based on the sound wave back scattering is provided, the defects of complex three-dimensional human body reconstruction process, high cost for scanning human body data and the like in the prior art are overcome, human body characteristic parameters are obtained through the sound wave back scattering, and the three-dimensional human body shape can be quickly reconstructed.
The three-dimensional human body touchless reconstruction method based on acoustic wave backscattering comprises the following steps which are sequentially carried out,
step one, acquiring scanned human body point cloud data according to equipment, and extracting human body characteristic parameters;
step two, automatically extracting human body characteristic points and characteristic section point clouds by utilizing the principle of anthropometry, extracting characteristic boundary lines by the characteristic boundary to obtain characteristic boundary lines, and calculating the characteristic size of the human body;
step three, according to the human body manifold principle, a group of Fourier coefficients are adopted to represent the geometric shape, and because the human body shape is complex, limited characteristic values are considered, far-field data of each characteristic parameter of the human body shape are obtained;
and fourthly, constructing a one-to-one correspondence relation between human body characteristic parameters and far field data through a neural network, taking human body size characteristics as model input, taking far field data of each characteristic parameter of human body shape as model output, performing neural network model training, obtaining far field data samples corresponding to the human body, and inverting the human body shape through far field data through a Fourier method to establish a three-dimensional model.
The method for extracting the human body characteristic parameters in the first step comprises the following steps:
firstly, preprocessing point cloud data, converting a human body center point into a coordinate origin, enabling a human body to be positioned on a reference surface, calculating a center point coordinate of a human body point cloud model according to geometric center coordinates, and according to the center point coordinate and a cameraAffine transformation matrix M for angular rotation around Z axis m The method comprises the steps of carrying out a first treatment on the surface of the According to the coordinate conversion relation, the human body three-dimensional point cloud model is adjusted to a coordinate axis taking a human body center point as a coordinate origin;
after the point cloud model of coordinate conversion is obtained, the Z-axis direction is the height direction of the human body, and the maximum value Z of the Z-axis direction is calculated max And a minimum value Z min Calculate height h= |z max -Z min Normalizing the height of the human body according to the proportional relation between the characteristic points and the height of the human body, wherein the estimated position of the characteristic parameter points of the human body is h position The ratio of the position of the characteristic point to the height is lambda chara The approximate height of the human body characteristic point is h position ＝λ chara H+Z min 。
The third human body shape characteristic parameters comprise height, arm length, shoulder width, neck circumference, chest circumference, waistline and hip circumference.
The method for obtaining far-field data of each characteristic parameter of the human body shape comprises the following steps:
active source corresponding to human body shapeIs a tightly packed function, +.>Wherein->Is a border area, ++> Is the current source intensity of the node in the region D, and the scattered wave generated by the active source f is formed byThe following equation controls
Wherein,the limiting form in formula (1) is called Sommerfeld radiation condition, where i imaginary units, k wave numbers, Δ is Laplace operator, +.>For partial derivative r, r is the spatial distance and d is the spatial dimension;
the far field pattern of the scattered wave u is
Wherein the method comprises the steps ofAnd->R is real space, S is curved space, and F represents Fourier transform of F.
The training data are segmented in the training process of the neural network model, the human body size characteristic parameters and the segmented far-field data are trained by using a plurality of identical neural networks respectively, and the segmented data are trained in parallel through forward propagation of the networks and reverse errors of the errors for a plurality of times to obtain proper predicted values; exchanging parameters between the neural networks, and storing related parameters of the network model and predicted far-field data information in order; and splicing the predicted far-field data together according to the sequence of the corresponding observation points to form a far-field data sample corresponding to the human body.
And firstly, acquiring human body point cloud data by using a deep perception camera.
Through the design scheme, the invention has the following beneficial effects: a three-dimensional human body touchless reconstruction method based on sound wave backscattering can reconstruct a three-dimensional shape through human body characteristic parameters without using equipment to scan a human body, thereby realizing rapid reconstruction of the three-dimensional human body shape and reducing reconstruction cost.
The invention is further described in connection with the following detailed description:
Detailed Description
A three-dimensional human body touchless reconstruction method based on sound wave backscattering uses a deep-sensing camera to acquire human body point cloud data, and calculates human body characteristic parameters after preprocessing the point cloud data of a real human body. For different body types, normal and obese body types are identified based on a body shape analysis method, human body characteristic points are located through layered exploration, characteristic section information is obtained through the human body characteristic points, characteristic contour boundary points are extracted through an edge identification method, and real characteristic parameter information of people with different body types is effectively calculated.
Since the geometric shape can be represented by Fourier coefficients, the more complex the shape, the more feature values corresponding to the geometric shape, and since the shape of the human body is complex, we consider limited feature parameters. We regard height, arm length, shoulder width, neck circumference, chest circumference, waist circumference, hip circumference, etc. as characteristic parameters.
And obtaining far-field data of the sound wave through calculation according to the Helmholtz system satisfied by the sound wave. Because the dimension difference between the geometric parameters and the far-field data is overlarge, the training data are segmented in the training process of the neural network, the human body dimension characteristic parameters and the segmented far-field data are respectively trained by using a plurality of homologous neural networks, the network parameters are adjusted through reverse error propagation, and corresponding far-field data predicted values are obtained for the input human body parameters. And splicing the predicted far-field data together according to the sequence of the corresponding observation points to form a far-field data sample corresponding to the human body.
And inverting the shape of the human body from the far-field data corresponding to the human body by using a Fourier method. And carrying out grid division on the region where the human body is located by giving limited wave numbers and corresponding far-field data to obtain sampling grid points in the region. And calculating a Fourier coefficient corresponding to the far-field data of the three-dimensional human body at each sampling grid node, and reconstructing the shape of the human body by utilizing Fourier expansion.
Specifically, the human body is scanned by using equipment to acquire human body point cloud information, preprocessing is carried out, relevant characteristic parameters of the human body are calculated according to a human body measurement principle, the characteristic parameters of the human body are determined according to a human body manifold correlation theory, then the human body shape is used as an L2 source to obtain a corresponding far-field data result, a corresponding relation between the human body characteristic parameters and the far-field data is constructed in a parallel neural network mode to obtain a network training result, and further the corresponding far-field data is predicted according to the human body characteristic parameters.
And acquiring scanned human body point cloud data according to the equipment, and extracting human body characteristic parameters. Firstly, preprocessing point cloud data, and converting a human body center point into a coordinate origin so that a human body is positioned on a reference surface. Calculating the center point coordinates of the human body point cloud model according to the geometric center coordinates, and obtaining an affine transformation matrix M rotating around the Z axis according to the center point coordinates and the camera angle m . According to the coordinate transformation relation, the human body three-dimensional point cloud model can be adjusted to a coordinate axis taking the human body center point as the origin of coordinates.
After the point cloud model of coordinate conversion is obtained, the Z-axis direction is the height direction of the human body, and the maximum value Z of the Z-axis direction is calculated max And a minimum value Z min Calculate height h= |z max -Z min And normalizing the height of the human body according to the proportional relation between the characteristic points and the height of the human body. The estimated position of the human body characteristic parameter point is h position The ratio of the position of the characteristic point to the height is lambda chara The approximate height of the human body characteristic point is h position ＝λ chara H+Z min 。
Human body characteristic points and characteristic section point clouds which are automatically extracted by utilizing the principle of anthropometry are extracted through characteristic boundaries to obtain characteristic boundary lines, so that the human body characteristic size is calculated.
Taking waistline parameter calculation process as an example, determining a waistline searching area according to the proportional relation between human body characteristic points and height, then obtaining a series of point cloud sets of characteristic sections according to a searching plane with a designated step length, obtaining a set of characteristic boundaries by using an edge recognition method, obtaining a maximum value and a minimum value of boundary girth, and finally determining waistline parameters of a human body according to body types.
According to the related knowledge of the manifold of the human body, the geometry can be represented by a set of Fourier coefficients, the more complex the shape is, the more feature values corresponding to the geometry are, and the feature values for describing the human body can be infinite due to the complex shape of the human body, so that for practical reasons, we consider the limited feature values. We consider height, arm length, shoulder width, neck circumference, chest circumference, waist circumference, and hip circumference as characteristic parameters.
Active source corresponding to human body shapeIs a tightly packed function, +.>Wherein->Is a border area, ++> Is the current source intensity of the node in the region D, and the scattered wave generated by the active source fControlled by the following equation
Wherein,the limiting form in equation (1) is called the Sommerfeld radiation condition, the far field mode of the scattered wave u is
Wherein the method comprises the steps ofAnd->
F represents the Fourier transform of F.
After obtaining the human body characteristic parameters and far-field data, constructing a one-to-one correspondence relationship between the human body characteristic parameters and the far-field data through a neural network, wherein the human body dimension characteristic parameters are x= (x) 1 ，x 2 ，…x T ) As an input of the model, far-field data y= (y) corresponding to the human body 1 ，y 2 ，…y M ) As an output of the model. For the input vector x= (x) 1 ，x 2 ，…x T ) Each element x of (2) T The characteristics are extracted through a neural network, the extracted characteristics are used as intermediate characteristics to carry out characteristic memory, and the memorized characteristics y are used M Training and learning are carried out through the neural network, so that training of the whole neural network model is completed.
Because the dimension difference between the input data and the output data is huge, the training data are segmented in the process of training the neural network, the human body dimension characteristic parameters and the segmented far-field data are trained by using a plurality of identical neural networks respectively, and the appropriate predicted value is obtained through the forward propagation of the network and the reverse error of the error and the parallel training of the data of each segment through multiple times of adjustment. And exchanging parameters between the neural networks, storing related parameters of the network model and storing predicted far-field data information in order. And splicing the predicted far-field data together according to the sequence of the corresponding observation points to form a far-field data sample corresponding to the human body.
A new far-field data set is predicted for a given new set of characteristic parameters from a training model derived from a neural network based on the relationship between the previous geometry and the characteristic parameters. By the relationship between the far field data and the Fourier coefficients,
the human body shape can be represented by the following Fourier-expanded truncated form:
wherein a represents the size of the mesh region where the human body is located
Is a Fourier basis function. Thus inverting the shape of the human body by the Fourier method from the far field data.

Claims (6)

1. A three-dimensional human body touchless reconstruction method based on sound wave backscattering is characterized in that: comprising the following steps, which are sequentially carried out,
step one, acquiring scanned human body point cloud data according to equipment, and extracting human body characteristic parameters;
step two, automatically extracting human body characteristic points and characteristic section point clouds by utilizing the principle of anthropometry, extracting characteristic boundary lines by the characteristic boundary to obtain characteristic boundary lines, and calculating the characteristic size of the human body;
step three, according to the human body manifold principle, a group of Fourier coefficients are adopted to represent the geometric shape, and because the human body shape is complex, limited characteristic values are considered, far-field data of each characteristic parameter of the human body shape are obtained;
and fourthly, constructing a one-to-one correspondence relation between human body characteristic parameters and far field data through a neural network, taking human body size characteristics as model input, taking far field data of each characteristic parameter of human body shape as model output, performing neural network model training, obtaining far field data samples corresponding to the human body, and inverting the human body shape through far field data through a Fourier method to establish a three-dimensional model.
2. The three-dimensional human body touchless reconstruction method based on sound wave backscattering according to claim 1, wherein the method is characterized in that: the method for extracting the human body characteristic parameters in the first step comprises the following steps:
firstly, preprocessing point cloud data, converting a human body center point into a coordinate origin, enabling a human body to be positioned on a reference plane, calculating a center point coordinate of a human body point cloud model according to a geometric center coordinate, and obtaining an affine transformation matrix M rotating around a Z axis according to the center point coordinate and a camera angle m The method comprises the steps of carrying out a first treatment on the surface of the According to the coordinate conversion relation, the human body three-dimensional point cloud model is adjusted to a coordinate axis taking a human body center point as a coordinate origin;
after the point cloud model of coordinate conversion is obtained, the Z-axis direction is the height direction of the human body, and the maximum value Z of the Z-axis direction is calculated max And a minimum value Z min Calculate height h= |z max -Z min Normalizing the height of the human body according to the proportional relation between the characteristic points and the height of the human body, wherein the estimated position of the characteristic parameter points of the human body is h position The ratio of the position of the characteristic point to the height is lambda chara The approximate height of the human body characteristic point is h position ＝λ chara H+Z min 。
3. The three-dimensional human body touchless reconstruction method based on sound wave backscattering according to claim 1, wherein the method is characterized in that: the third human body shape characteristic parameters comprise height, arm length, shoulder width, neck circumference, chest circumference, waistline and hip circumference.
4. The three-dimensional human body touchless reconstruction method based on sound wave backscattering according to claim 1, wherein the method is characterized in that: the method for obtaining far-field data of each characteristic parameter of the human body shape comprises the following steps:
assume that the active source f corresponding to the shape of the human body:is oneTightly-packed functions, ++>Wherein->Is a border area, ++> Is the current source intensity of the node in the region D, and the scattered wave generated by the active source f is formed byThe following equation controls
Wherein, r= |x|,the limiting form in formula (1) is called Sommerfeld radiation condition, where i imaginary units, k wave numbers, Δ is Laplace operator, +.>For partial derivative r, r is the spatial distance and d is the spatial dimension;
the far field pattern of the scattered wave u is
Wherein the method comprises the steps ofAnd->R is real space, S is curved space, and F represents Fourier transform of F.
5. The three-dimensional human body touchless reconstruction method based on sound wave backscattering according to claim 1, wherein the method is characterized in that: the training data are segmented in the training process of the neural network model, the human body size characteristic parameters and the segmented far-field data are trained by using a plurality of identical neural networks respectively, and the segmented data are trained in parallel through forward propagation of the networks and reverse errors of the errors for a plurality of times to obtain proper predicted values; exchanging parameters between the neural networks, and storing related parameters of the network model and predicted far-field data information in order; and splicing the predicted far-field data together according to the sequence of the corresponding observation points to form a far-field data sample corresponding to the human body.
6. The three-dimensional human body touchless reconstruction method based on sound wave backscattering according to claim 1, wherein the method is characterized in that: and firstly, acquiring human body point cloud data by using a deep perception camera.
CN202311229155.8A 2023-09-22 2023-09-22 Three-dimensional human body touchless reconstruction method based on sound wave backscattering Pending CN117292060A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311229155.8A CN117292060A (en) 2023-09-22 2023-09-22 Three-dimensional human body touchless reconstruction method based on sound wave backscattering

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311229155.8A CN117292060A (en) 2023-09-22 2023-09-22 Three-dimensional human body touchless reconstruction method based on sound wave backscattering

Publications (1)

Publication Number Publication Date
CN117292060A true CN117292060A (en) 2023-12-26

Family

ID=89252954

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311229155.8A Pending CN117292060A (en) 2023-09-22 2023-09-22 Three-dimensional human body touchless reconstruction method based on sound wave backscattering

Country Status (1)

Country Link
CN (1) CN117292060A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117935837A (en) * 2024-03-25 2024-04-26 中国空气动力研究与发展中心计算空气动力研究所 Time domain multi-sound source positioning and noise processing method

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117935837A (en) * 2024-03-25 2024-04-26 中国空气动力研究与发展中心计算空气动力研究所 Time domain multi-sound source positioning and noise processing method
CN117935837B (en) * 2024-03-25 2024-05-24 中国空气动力研究与发展中心计算空气动力研究所 Time domain multi-sound source positioning and noise processing method

Similar Documents

Publication Publication Date Title
CN110443885B (en) Three-dimensional human head and face model reconstruction method based on random human face image
CN105654483B (en) The full-automatic method for registering of three-dimensional point cloud
CN110807829A (en) Method for constructing three-dimensional heart model based on ultrasonic imaging
JP5950486B1 (en) Aging prediction system, aging prediction method, and aging prediction program
CN117292060A (en) Three-dimensional human body touchless reconstruction method based on sound wave backscattering
CN110866969A (en) Engine blade reconstruction method based on neural network and point cloud registration
CN109118455B (en) Ancient human skull craniofacial interactive restoration method based on modern soft tissue distribution
CN112330813A (en) Wearing three-dimensional human body model reconstruction method based on monocular depth camera
CN111260702B (en) Laser three-dimensional point cloud and CT three-dimensional point cloud registration method
CN111583408A (en) Human body three-dimensional modeling system based on hand-drawn sketch
CN112465966A (en) Cliff three-dimensional modeling method integrating oblique photogrammetry and three-dimensional laser scanning
CN112991300B (en) Single wood skeleton extraction and visualization method based on neighborhood characteristics
CN114596409A (en) Method for obtaining human face three-dimensional grid model with texture
CN116740703B (en) Wheat phenotype parameter change rate estimation method and device based on point cloud information
CN116725563B (en) Eyeball salience measuring device
CN117094925A (en) Pig body point cloud completion method based on point agent enhancement and layer-by-layer up-sampling
CN110009726B (en) Method for extracting plane from point cloud according to structural relationship between plane elements
CN112184912A (en) Multi-metric three-dimensional face reconstruction method based on parameterized model and position map
CN116580148A (en) Parameterized human body model fitting method based on human body point cloud and measurement parameters
CN116310103A (en) Human body posture estimation and grid recovery method based on skin multi-person linear model
CN112598669B (en) Lung lobe segmentation method based on digital human technology
Liou et al. Conformal self-organizing map for a genus-zero manifold
CN111968113B (en) Brain image two-dimensional convolution deep learning method based on optimal transmission mapping
CN112381952B (en) Face contour point cloud model reconstruction method and device based on multiple cameras
Barazzetti et al. Vault modeling with neural networks

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination