CN102920459A - Human body circumference parameter measuring method based on three-dimensional point cloud - Google Patents

Abstract

The invention discloses a human body circumference parameter measuring method based on a three-dimensional point cloud. In the conventional method, certain point cloud data positioned on both sides of a human body are lost due to causes such as shield of the human body, incomplete acquired information, scanning dead corners and the like. The method comprises the following steps of: firstly, combining all human body three-dimensional point cloud data under the same three-dimensional coordinate system, converting three-dimensional human body section point cloud data of a specified height into a two-dimensional plane point set, and automatically completing classification of section point cloud data by performing horizontal projection analysis on section point cloud data on the two-dimensional plane point set; secondly, analyzing the positions of various types of data loss part points to obtain point cloud data to be fitted; thirdly, selecting a secondary curve parameter equation, fitting point cloud data by adopting a least squares method, and completing gap point filling work by using an obtained fitting curve; and lastly, calculating the circumference of a body position which corresponds to each classified datum. According to the method, a measured result of the body circumference of a human body can be closer to a real value.

Description

A kind of human body degree of enclosing measurement method of parameters based on three-dimensional point cloud

Technical field

The present invention relates to the human engineering technical field, is a kind of human body degree of enclosing measurement method of parameters based on three-dimensional point cloud specifically.

Background technology

Three-dimensional body dimension is measured the difference according to metering system, can be divided into two kinds of contact type measurement and non-contact measurements.And the non-contact 3-D metering system that utilizes optics and vision technique enjoys favor because of height response and high-resolution.Wherein based on the contactless humanbody three-dimensional measurement of structured light with its wide range, large visual field, high accuracy, the real-time and advantage such as controlled initiatively, day by day come into one's own and in human body dimension measurement, be used widely.Yet at finite structure light collection device scan and when rebuilding human body three-dimensional point cloud, often owing to human body self blocks, Information Monitoring is incomplete and the scanning reason such as dead angle so that some are positioned at the cloud data disappearance of human body both sides, directly cause the failure of human body section size measurement links section Points cloud Fitting.

Summary of the invention

The present invention is directed to the deficiencies in the prior art, a kind of human body degree of enclosing measurement method of parameters based on three-dimensional point cloud is provided.The inventive method is specifically:

Step (1) with all human body three dimensional point cloud amalgamations to the same three-dimensional system of coordinate, vertical height in conjunction with appointment, by the projection on two dimensional surface, the 3 D human body section cloud data of specified altitude assignment is converted into the two dimensional surface point set, namely obtain the section cloud data.

The geometrical constraint condition that step (2) is obtained in conjunction with the 3 D human body cloud data by the section cloud data on the two dimensional surface point set is carried out the floor projection analysis, is finished the classification of section cloud data automatically.

Step (3) is analyzed Various types of data disappearance part and is put the position sorted data, and the point around the utilization disappearance part obtains and treats the match cloud data.

Step (4) after step (3), is selected the conic section parametric equation to the data of each class, and adopts the least square fitting cloud data, utilizes the matched curve that obtains to finish the work that breach is mended point.

Step (5) on the basis of step (4), adopt algorithm of convex hull calculate each categorical data the degree of enclosing of corresponding body part.

The invention provides a kind of human body degree of enclosing measurement method of parameters based on three-dimensional point cloud, can make the measurement result of human body degree of enclosing more near actual value, have a wide range of applications.

Description of drawings

Fig. 1 is method flow diagram of the present invention;

Fig. 23 D human body section point cloud automatic cluster and comparison, wherein:

(a) sorting technique signal, (b) k means clustering method result, (c) projecting method of the present invention;

Preprocessing process sketch map before Fig. 3 least square fitting; Wherein:

(a) clockwise sampling, (b) point cloud local enlarged drawing, (c) breach sketch map, (d) overall situation sampling;

Fig. 4 be the present invention to the fitting result figure of human body section point cloud, wherein:

(a) original section point cloud, (b) fitting result of the present invention.

The specific embodiment

The invention will be further described below in conjunction with accompanying drawing.

As shown in Figure 1, a kind of human body degree of enclosing measurement method of parameters based on three-dimensional point cloud comprises:

(1) with all human body three dimensional point cloud amalgamations to the same three-dimensional system of coordinate, in conjunction with vertical height of appointment, by the projection on two dimensional surface, the 3 D human body section cloud data of specified altitude assignment is converted into the two dimensional surface point set.

(2) the geometrical constraint condition of obtaining in conjunction with the 3 D human body cloud data by the cloud data on the two dimensional surface point set is carried out the floor projection analysis, is finished the classification of section cloud data automatically.

(3) to sorted data, to analyze Various types of data disappearance part and put the position, the point around the utilization disappearance part obtains and treats the match cloud data.

(4) to the data of each class, after step (3), select the conic section parametric equation, and adopt the least square fitting cloud data.Utilize the matched curve that obtains to finish the work that breach is mended point.

(5) on the basis of step (4), adopt algorithm of convex hull calculate each categorical data the degree of enclosing of corresponding body part.

With all human body three dimensional point cloud amalgamations to the same three-dimensional system of coordinate, vertical height in conjunction with appointment, by the projection on two dimensional surface, the detailed process that the 3 D human body section cloud data of specified altitude assignment is converted into the two dimensional surface point set may further comprise the steps:

1. the point data splitting that different acquisition equipment is obtained is to same three-dimensional system of coordinate x-y-z, and wherein the y axle is the vertical of human body, and x-z is the human body horizontal plane;

2. according to the z value of appointment, 3 D human body section cloud data is converted into the two dimensional surface point set.

In conjunction with the geometrical constraint condition that the 3 D human body cloud data obtains, by the cloud data on the two dimensional surface point set is carried out the floor projection analysis, the detailed process of automatically finishing the classification of section cloud data may further comprise the steps:

1. at first the section cloud data is sorted along X direction is ascending;

Be the sorting technique signal such as Fig. 2 (a), Fig. 2 (b) is k means clustering method result.Shown in Fig. 2 (c), before 3 D human body is carried out tangent plane, must rotate human body point cloud direction, make human body towards parallel with the z axle.Therefore can put cloud to section carries out floor projection along the x axle, namely becomes the one dimension figure by X-Y scheme and classifies.At first each point is sorted along the x axle is ascending.

With a threshold value T as standard value, the quantity that the difference of adjacent transverse axis coordinate components is surpassed threshold value T at 2 is added up, if the quantity number is designated as C, namely thinks sample to be divided into the C+1 class.

Calculate the difference of adjacent x at 2.Begin to calculate adjacent distance at 2 from first point, every when running into distance greater than two points of threshold value, first the last point to these two points from sample is divided into a class.If the number of such point, then is considered as noise less than threshold value T=10 and removes, remaining o'clock repeats current operation as a new sample, to the last till point.

3. determine sample section source by the class number of statistics at last, as be three classes that must come from two arms and health, two classes then come from both legs etc.

To sorted data, analyze Various types of data disappearance part and put the position, the point around the utilization disappearance part, obtain the detailed process for the treatment of the match cloud data and may further comprise the steps:

1. at first, along clockwise direction the section cloud data is sorted;

After classification finishes, obtain the central point of each class, utilize central point that such is sorted in a clockwise direction by angle.Central point is such center of gravity, and the coordinate of establishing central point O is (x _o, y _o), solution formula is suc as formula shown in (1), wherein f (x _i, y _i)=1.Ordering be according to OA along carrying out to the angle theta of OB is ascending clockwise, shown in Fig. 3 (a), the finding the solution suc as formula shown in (2) of θ angle.

$x_o=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}(x_i\·f(x_i,y_i))}{n},$ $y_o=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}(y_i\·f(x_i,y_i))}{n}---\left(1\right)$

$\mathrm{\&theta;}=\left\{\begin{array}{cc}(\arcsin \left(\frac{z_a-z_o}{\sqrt{{(x_a-x_o)}^2+{(z_a-z_o)}^2}}\right)+2\mathrm{\&pi;})\mathrm{mod}2& ,x_a-x_o<0\\ (-\arcsin \left(\frac{z_a-z_o}{\sqrt{{(x_a-x_o)}^2+{\left(z_a-z_o\right)}^2}}\right)+\mathrm{\&pi;})& ,x_a-x_o\&GreaterEqual;0\end{array}\right.---\left(2\right)$

2. secondly, determine disappearance part position;

Ask for the distance between adjacent 2.Shown in Fig. 3 (b), L is adjacent distance at 2, and then threshold value is set, and is disappearance part position if 2 adjacent distance, then illustrates this position of 2 greater than threshold value.Shown in Fig. 3 (c), greater than threshold value, illustrate then that section is disappearance part position between C point and the D point such as the distance between the CD.

3. last, reach clockwise 30 sample points of counterclockwise respectively sampling on the gap position edge.

Such as Fig. 3 (d), get along clockwise direction 30 points from D point beginning, suppose to get in the counterclockwise direction 30 points from the C point again till the E point, suppose till the F point, utilize at last these points of these both sides, position to carry out match.

To the data of each class, after step (3), select the conic section parametric equation, and adopt the least square fitting cloud data.The detailed process of utilizing the matched curve that obtains to finish the work of breach benefit point may further comprise the steps:

1. select fitting function f (x, A), A=(a ₀, a ₁..., a _n) be a series of undetermined parameters;

F (x, A) is called model of fit, A=(a ₀, a ₁..., a _n) be a series of undetermined parameters.Way be select parameter A so that model of fit and actual observed value in the difference e of each point _i=z _i-f (x _i, weighted sum of squares A) is minimum.Namely ask f ^*(x), so that:

$\underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}\left\{w\left(x_i\right){(f^{*}\left(x_i\right)-z_i)}^2\right\}\&LeftArrow;\min \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}\{w\left(x_i\right){(f\left(x_i\right)-z_i)}^2---\left(3\right)$

W (x wherein _i) 〉=0 is called power, and it reflects data (x _i, z _i) proportion of shared data in experiment.The curve of using this method match is called the least square fitting curve.Ask matched curve at first will determine the model of fit f (x, z) of conic section=a with method of least square ₀x ²+ a ₁z ²+ a ₂Xz+a ₃X+a ₄Z+a ₅, a wherein ₀=0, a ₁=1, w (x _i, z _i)=1.

2. the sample that obtains of integrating step (3) utilizes method of least square to calculate undetermined parameter;

According to extreme value theorem, order $G(a_0,a_2,a_3,a_4,a_5)=\underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}{(f(x_i,z_i)-Y)}^2,$ Then by $\frac{\&PartialD;G}{{\&PartialD;a}_0}=0,$ $\frac{\&PartialD;G}{{\&PartialD;a}_2}=0,$ $\frac{\&PartialD;G}{{\&PartialD;a}_3}=0,$ $\frac{\&PartialD;G}{{\&PartialD;a}_4}=0,$ $\frac{\&PartialD;G}{\&PartialD;a_5}=0.$ Obtain:

$\left[\begin{array}{ccccc}\underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}{x_i}^4& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}\left({x_i}^3{z}_i\right)& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}{x_i}^3& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}\left({x_i}^2{z}_i\right)& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}{x_i}^2\\ \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}\left({x_i}^3{z}_i\right)& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}\left({x_i}^2{z_i}^2\right)& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}\left({x_i}^2{z}_i\right)& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}\left(x_i{z_i}^2\right)& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}\left(x_i{y}_i\right)\\ \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}{x_i}^3& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}\left(x_i{z_i}^2\right)& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}{x_i}^2& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}\left(x_i{z}_i\right)& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}{x}_i\\ \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}\left({x_i}^2{z}_i\right)& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}\left(x_i{z_i}^2\right)& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}\left(x_i{z}_i\right)& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}{z_i}^2& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}{z}_i\\ \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}{x_i}^2& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}\left(x_i{z}_i\right)& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}{x}_i& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}{z}_i& \underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}1\end{array}\right]\&CenterDot;\left[\begin{array}{c}{a}_0\\ a_2\\ a_3\\ a_4\\ a_5\end{array}\right]=\left[\begin{array}{c}-\underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}\left({x_i}^2{z_i}^2\right)\\ -\underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}\left(x_i{z_i}^3\right)\\ -\underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}\left(x_i{z_i}^2\right)\\ -\underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}{z_i}^3\\ -\underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}{z_i}^2\end{array}\right]---\left(4\right)$

Because mention hereinbefore clockwise and counterclockwise respectively get 30 points, so the value of m gets 60, and through type (4) solves a again ₀, a ₂, a ₃, a ₄, a ₅Value.

3. utilize the matched curve that obtains and the work of finishing breach benefit point perpendicular to the intersection point of x axle straight line.

Utilize the intersection point of a series of straight lines perpendicular to the x axle and this matched curve to finish the some work of mending.

On the basis of step (4), adopt algorithm of convex hull calculate each categorical data the detailed process of degree of enclosing work of corresponding body part may further comprise the steps:

1. on the basis of step (4), obtain the section cloud data through each position of human body behind the benefit point;

2. adopt algorithm of convex hull calculate all kinds of cloud datas the degree of enclosing of corresponding body part.

In order to verify accuracy of the present invention and practicality, same 3 D human body point cloud is tested, such as Fig. 4 (a) with (b), from left to right be followed successively by the again profile of match of first pretreatment that profile before the match and the present invention propose, be followed successively by from top to bottom the sample section from chest, ancon, waist and thigh.Measurement parameter result to human body degree of enclosing after the match of section point cloud is as shown in table 1.Can find out that from table 1 the data obtained the result that the present invention obtains is near actual value.

Table 1

。

Claims (6)

1. human body degree of the enclosing measurement method of parameters based on three-dimensional point cloud is characterized in that the method comprises the steps:

Step (1) with all human body three dimensional point cloud amalgamations to the same three-dimensional system of coordinate, vertical height in conjunction with appointment, by the projection on two dimensional surface, the 3 D human body section cloud data of specified altitude assignment is converted into the two dimensional surface point set, namely obtain the section cloud data;

The geometrical constraint condition that step (2) is obtained in conjunction with the 3 D human body cloud data by the section cloud data on the two dimensional surface point set is carried out the floor projection analysis, is finished the classification of section cloud data automatically;

Step (3) is analyzed Various types of data disappearance part and is put the position sorted data, and the point around the utilization disappearance part obtains and treats the match cloud data;

Step (4) after step (3), is selected the conic section parametric equation to the data of each class, and adopts the least square fitting cloud data, utilizes the matched curve that obtains to finish the work that breach is mended point;

Step (5) on the basis of step (4), adopt algorithm of convex hull calculate each categorical data the degree of enclosing of corresponding body part.

2. a kind of human body degree of enclosing measurement method of parameters based on three-dimensional point cloud according to claim 1 is characterized in that the detailed process of step (1) is as follows:

1) point data splitting that different acquisition equipment is obtained is to same three-dimensional system of coordinate x-y-z, and wherein the y axle is the vertical of human body, and x-z is the human body horizontal plane;

2) according to the z value of appointment, 3 D human body section cloud data is converted into the two dimensional surface point set.

3. a kind of human body degree of enclosing measurement method of parameters based on three-dimensional point cloud according to claim 1 is characterized in that the detailed process of step (2) is as follows:

1) at first the section cloud data is sorted along X direction is ascending;

2) with a threshold value T as standard value, the quantity that the difference of adjacent transverse axis coordinate components is surpassed threshold value T at 2 is added up, and establishes the quantity number and is designated as C, namely sample can be divided into the C+1 class.

4. a kind of human body degree of enclosing measurement method of parameters based on three-dimensional point cloud according to claim 1 is characterized in that the detailed process of step (3) is as follows:

1) along clockwise direction the section cloud data is sorted;

2) determine disappearance part position;

3) reach clockwise 30 sample points of counterclockwise respectively sampling on the gap position edge.

5. a kind of human body degree of enclosing measurement method of parameters based on three-dimensional point cloud according to claim 1 is characterized in that the detailed process of step (4) is as follows:

1) selects fitting function

, A=

A series of undetermined parameters;

2) cloud data that obtains of integrating step (3) utilizes method of least square to calculate undetermined parameter;

3) utilize the matched curve that obtains and the work of finishing perpendicular to the intersection point of x axle straight line section cloud data breach benefit point.

6. a kind of human body degree of enclosing measurement method of parameters based on three-dimensional point cloud according to claim 1 is characterized in that the detailed process of step (5) is as follows:

1) on the basis of step (4), obtains the cloud data through each position section of human body behind the benefit point;

2) adopt the permanent algorithm of convex hull of Ge Li calculate all kinds of cloud datas the degree of enclosing of corresponding body part.

Images