CN1971274A - Method for assessing smoothness level of dress material based on point model - Google Patents

Abstract

The invention discloses a rating method of garments material flatness based on the point pattern that comprises steps as following: 1. point data set: the point pattern of garments material is obtained by 3-D non-contact high-accuracy scanner; 2. data analysis: a. the coordinate data in direction of Z are analyzed in aspect of concentration tendency or discrete tendency of data distribution and shape of data distribution; b. the characteristic points which reflects the fluctuation feature of material are sifted from point-set, neighboring regions of characteristic points are constructed; c. curved surface in maximum peak is rebuilt; d. the curvature values in characteristic points are analyzed based on the curved surface theory of discrete differential geometry; 3. evaluation model: the evaluation model of garments material flatness is built based on neural net model. The invention adopts the analytical method of point model to build the evaluation system of garments material flatness with any color and pattern, and the classes can be accurately separated, the system accuracy reaches about 96 %.

Description

Garment material smoothness grade assessment method based on point model

Technical field

The present invention relates to a kind of properties of textile field tests, specifically, is a kind of based on objective ranking method point model, evaluation garment material wrinkle grade.

Background technology

That bafta has is ventilative, moisture absorption and superior function such as comfortable and easy to wear, is widely used in the processing of underwear, shirt, Western-style trousers and sheet, quilt cover etc.But bafta elasticity is relatively poor, and is wrinkling easily in dress or the washing process.In recent years, along with popularizing of rinsing machine, increasing consumer requires the textile clothes after the noniron finish to have the smooth performance of good washing.Therefore, in the back arrangement and clothing process of bafta, to the lining planarization can check and Study on Forecast more and more cause the attention of clothes processing enterprise.

The wrinkle grade of garment material is the important indicator of its smooth performance of evaluation.At present, mainly adopting standard specimen counter point (U.S. AATCC-124 standard) for the assessment mode of garment material wrinkle grade, is the overall visual impression to qualitativeization of garment material.Because scale costs an arm and a leg, evaluation environment harshness is difficult in the actual production popularize.In addition, in the evaluation procedure, the blind spot that human eye exists, psychological factor all are inevitable influence factor during the artificial vision detects to the influence of visual effect, environment to the polysemy of the influence of vision, human eye vision effect and fault-tolerance or the like.Therefore the result of subjective assessment garment material wrinkle grade often has uncertain and uniqueness not.

In recent years, maturation along with computer image processing technology, people such as G.Stylios, X.Bu, Youngjoo Na, tight Hao scape, Huang Xiubao, Wang Liming and Yang Xiaobo serve as the research basis with the plane picture of 3D solid, and the application image treatment technology has been obtained certain progress to the research that the wrinkle grade of garment material carries out objective evaluation.But view data is subject to influences such as lining color, flower type and even institutional framework.In the processes such as reconstruction, recovery, enhancing, compression and feature extraction of image also losing of image fault and useful data may appear.In addition, when utilizing image processing techniques to distinguish the approaching lining of wrinkle grade value, the grading accuracy decreases.

Recently, the laser scanner technique based on optical triangulation theory has obtained widespread use in 3 D non-contacting type is measured.J.Amirbayat, M.J.Alagha, B.Xu, Jie Su, Christopher N.Turner, Amirbayat J., Tae Jin Kang and Fan Jin soil, people such as Lv Dongfeng, Liu Fu utilize three-dimensional laser scanning technique all to obtain the three-D profile of lining comparatively exactly, objective evaluation the wrinkle grade of garment material.Laser scanning method is not subjected to the influence of fabric color and decorative pattern, the accuracy rate of measuring height.But, need scan, gather and handle multiple image during laser measurement, both consuming time, be difficult to realize real-time measurement again.In addition, be subject to the influence at lining surface diffuse reflectance and inclination angle during laser measurement, for LASER Light Source irradiation less than the position blind spot also appears scanning, and can not clearly to distinguish wrinkle grade be lining between 3 to 3.5.

The three-dimensional model that obtains along with scanning becomes increasingly complex with meticulous, and the point model technology becomes a research focus.The point model that 3-D scanning generates is exactly implicitly to represent solid with the discrete point of solid surface intensive sampling.The most basic information that each discrete point comprises is the coordinate position that scanning is obtained.Because the coordinate information of putting is original, the most real data during three-dimensional data is obtained, not needing explicitly to carry out topology between the discrete point connects, in data handling procedure, only need to store, handle and transmit with the form of pure point, so both made required space-time cost lower, avoided many limitation and restriction conditions of forcing for the correctness that guarantees topological relation in the Digital Geometry Processing again, the geometric manipulations of data is easier.

Summary of the invention

The technical problem to be solved in the present invention provides a kind of garment material smoothness grade assessment method based on point model, and this method is applicable to the wrinkle grade of the garment material of any color and decorative pattern, and accuracy reaches about 96%.

For solving the problems of the technologies described above, technical scheme of the present invention is:

Garment material smoothness grade assessment method based on point model comprises following concrete steps:

1. data point set

Obtain the point model of garment material by 3 D non-contacting type high precision scanner;

2. data analysis

A. analyze putting the coordinate data of concentrating on the each point Z direction from the concentrated or discrete trend of DATA DISTRIBUTION, the aspects such as form of DATA DISTRIBUTION;

B. concentrate the unique point that filters out reflection lining fluctuation characteristic from putting, and the structural attitude neighborhood of a point;

C. rebuild the curved surface at very big crest place;

D. adopt the curved surface theory in the discrete differential geometry, the curvature value at analytical characteristic point place;

3. rating model

Utilize neural network module, set up the rating model of estimating the lining wrinkle grade.

Described point model is the set that each sample point coordinate value is accurate to the three-dimensional data points of eight of radix point back.

The analysis of described coordinate data is meant interquartile range, mean difference, standard deviation and the kurtosis in the mathematical statistics, wherein:

A. interquartile range R _d

R _d＝Q _U-Q _L (1)

Interquartile range R _dBe upper quartile Q _UWith lower quartile Q _LDifference, reflection is in the deviation value of data in 50% on the ordered data centre position, is the measured value of data discrete degree;

B. mean difference R _a

$R_a=\frac{1}{N}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}|Z_i-\stackrel{-}{Z}|---\left(2\right)$

In the formula: Z _i---the coordinate figure of any 1 Z direction in the some cloud, mm;

---all assembly averages in the Z direction at random in the some cloud, mm;

N---the quantity of point at random in the some cloud;

Mean difference R _aSize reflected in the dispersion point cloud departure degree of face in the each point and some cloud, measured cloud data in the short transverse dispersion of a distribution.Mean difference is big more, and the comprehensive dispersion degree of data is bigger;

C. standard deviation R _q

$R_q=\sqrt{\frac{1}{N}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(Z_i-\stackrel{-}{Z})}^2}---\left(3\right)$

Standard deviation R _qWhat reflect is the absolute dispersion degree on the Z direction at random, than mean difference R _aMore given prominence to the influence of the variation of each point to statistic;

D. kurtosis R _k

$R_k=\frac{1}{N}\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(Z_i-\stackrel{-}{Z})}^4}{{R_q}^4}---\left(4\right)$

Kurtosis R _kBe that ebb or the peak form that metric data distributes departs from estimating of normal distribution degree.

Adopt the bead model in the screening process of described unique point.

The curved surface at the very big crest of described reconstruction place is the elevation information that extracts ten maximum wave peak dots in the point model, gets its mean value, and note is done

The curved surface at described very big crest place is the least square curved surface, and its toroidal function form is:

z＝a ₀+a ₁x+a ₂y+a ₃x ²+a ₄xy+a ₅y ² (5)

Described curved surface theory is based on the curved surface theory of discrete differential geometry.Setting up an office, (x y) is the function of its corresponding x, y coordinate, that is: for the z=f of each point in the model

$\left\{\begin{array}{c}x=x(u,v)\\ y=y(u,v)\\ z=z(u,v)\end{array}\right.$

If r _u=(x _u, y _u, z _u) and r _v=(x _v, y _v, z _v) represent curved surface respectively u to v to cutting arrow, make E=r _u ², F=r _ur _v, G=r _v ², and remember that the arc length of curved surface upper curve is s, have:

(ds) ²＝E(du) ²+2Fdudv+G(dv) ² (6)

E in the formula, F and G be curved surface u to v to the continuous function of cutting arrow.

If vowing, the master of the unit method of curved surface P point place curve is m, the per unit system of curved surface vow to be n, and both angles are φ, curvature k, and make L=-r _un _u=nr _Uu, $M=-\frac{1}{2}(r_u{n}_u+r_v{n}_u)=n{r}_{\mathrm{uv}},$ N=-r _vn _v=nr _Vv, have:

kcosφ(ds) ²＝L(du) ²+2Mdudv+N(dv) ² (7)

L in the formula, M and N also be curved surface u to v to the continuous function of cutting arrow.Then curved surface at the curvature value H at P point place is:

$H=\frac{\mathrm{EN}+\mathrm{GL}-2\mathrm{FM}}{2(\mathrm{EG}-F^2)}---\left(8\right)$

Described curvature information is meant the curvature information of ten maximum wave peak dots in the point model, gets its mean value, and note is done

Described neural network module is based on multilayer feedforward network BP learning algorithm.

Beneficial effect of the present invention:

The present invention adopts the analytical approach of point model, has set up the rating system of the garment material wrinkle grade that is applicable to any color and decorative pattern, and can correctly distinguish each grade, and system's accuracy reaches about 96%.

Description of drawings

Fig. 1 is based on the general flow chart of the garment material flatness rating system of point model;

Fig. 2 is the point model distribution plan of the lining of different subjective grading values;

Fig. 3 is the kurtosis curve map that the Z value distributes under the point model;

Fig. 4 is a feature neighborhood of a point structure;

Fig. 5 be fitting surface cut arrow figure;

Fig. 6 is the master of the unit method arrow figure of curve and curved surface;

Fig. 7 is the distribution plan of geometric feature of the sample of different subjective grading values;

Fig. 8 is the unique point and the neighborhood figure thereof of the sample of different subjective grading values;

Fig. 9 is the maximum wave peak dot average height and the mean curvature value distribution plan thereof of the sample of different subjective grading values;

Figure 10 is a BP network error curve map.

Embodiment

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

Point model of the present invention is that the discrete point with the model surface intensive sampling comes the surface of representation model implicitly, does not deposit or safeguard any topology information.Fig. 1 is based on the garment material flatness rating system general flow chart of point model.

Fig. 2 is the point model distribution plan of 6 garment materials that obtained by 3 D non-contacting type high precision scanner, and the subjective grading value of its wrinkle grade is respectively 1,2,3,3.5,4 and 5 grade.

Adopt the analytical approach evaluation garment material wrinkle grade value of point model to comprise following 5 steps:

A. analyze putting the coordinate data of concentrating on the each point Z direction from the concentrated or discrete trend of DATA DISTRIBUTION, the aspects such as form of DATA DISTRIBUTION;

B. concentrate the unique point that filters out reflection lining fluctuation characteristic from putting, and the structural attitude neighborhood of a point;

C. rebuild the curved surface at very big crest place;

D. adopt the curved surface theory in the discrete differential geometry, the curvature value at analytical characteristic point place;

E. utilize neural network module, set up the rating model of estimating the lining wrinkle grade.

Wherein:

(1) point model is the set that each sample point coordinate value is accurate to the three-dimensional data points of eight of radix point back.

(2) analysis of coordinate data is meant interquartile range, mean difference, standard deviation and the kurtosis in the mathematical statistics, wherein:

A. interquartile range R _d

R _d＝Q _U-Q _L (1)

Interquartile range R _dBe upper quartile Q _UWith lower quartile Q _LDifference, reflection is in the deviation value of data in 50% on the ordered data centre position, is the measured value of data discrete degree;

B. mean difference R _a

$R_{a=}\frac{1}{N}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}|Z_i-\stackrel{-}{Z}|---\left(2\right)$

In the formula: Z _i---the coordinate figure of any 1 Z direction in the some cloud, mm;

---all assembly averages in the Z direction at random in the some cloud, mm;

N---the quantity of point at random in the some cloud;

Mean difference R _aSize reflected in the dispersion point cloud departure degree of face in the each point and some cloud, measured cloud data in the short transverse dispersion of a distribution.Mean difference is big more, and the comprehensive dispersion degree of data is bigger;

C. standard deviation R _q

$R_q=\sqrt{\frac{1}{N}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(Z_i-\stackrel{-}{Z})}^2}---\left(3\right)$

Standard deviation R _qWhat reflect is the absolute dispersion degree on the Z direction at random.Standard deviation is big more, shows that the dispersion degree of a cloud is big more, and it is low uneven more promptly to put the cloud level, and distribution range is wide more; Standard deviation is more little, shows that the dispersion degree of a cloud is more little, promptly puts low neat more the concentrating of the cloud level, and distribution range is more little.Standard deviation is sensitiveer than the mean difference reaction, has more given prominence to the influence of the variation of discrete point Z value to statistic;

D. kurtosis R _k

$R_k=\frac{1}{N}\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(Z_i-\stackrel{-}{Z})}^4}{{R_q}^4}---\left(4\right)$

Kurtosis R _kBe that ebb or the peak form that metric data distributes departs from estimating of normal distribution degree.When kurtosis was 3, the kurtosis of data was consistent with normal distribution; Kurtosis value was less than 3 o'clock, and the expression data are that platykurtosis distributes; Kurtosis value was greater than 3 o'clock, and the expression data are that high kurtosis distributes, as shown in Figure 3.

(3) adopt the bead model in the screening process of unique point.

Owing to do not have any topology information between the point in the point model, can only rely on the analysis of unique point and to analyze its regional area.So, be the center with each number of scans strong point P, construct the ball that radius is r, fall into the point set (as shown in Figure 4) that the interior point of bead constitutes this vertex neighborhood, judge the Z coordinate figure of each point in the ball.If the Z coordinate figure of P is the very big or minimal value of point set, then the P point is a unique point, and calculated characteristics point is concentrated the Euclidean distance of each point and this point, and sorts according to the size of distance value.The point sequence called after PP of the unique point P that ordering forms ₁P ₂LLP _m, wherein the P distance P is nearest, P _mDistance P farthest.Big more with the near more point of unique point distance to the influence of curved surface that will match.

(4) curved surface of rebuilding very big crest place be since influence garment material wrinkle grade value often should the zone in several maximum crest fold degree, so the present invention only extracts the elevation information of ten maximum wave peak dots in the point model, get its mean value, remember work

(5) greatly the curved surface at crest place is the least square curved surface, and its toroidal function form is:

z＝a ₀+a ₁x+a ₂y+a ₃x ²+a ₄xy+a ₅y ² (5)

(6) the curved surface theory is based on the curved surface theory of discrete differential geometry.

(x y) is the function of its corresponding x, y coordinate, that is: to the z=f of each point in the point model

$\left\{\begin{array}{c}x=x(u,v)\\ y=y(u,v)\\ z=z(u,v)\end{array}\right.$

If r _u=(x _u, y _u, z _u) and r _v=(x _v, y _v, z _v) represent curved surface respectively u to v to cutting arrow, make E=r _u ², F=r _ur _v, G=r _v ², and remember that the arc length of curved surface upper curve is s, have:

(ds) ²＝E(du) ²+2Fdudv+G(dv) ² (6)

E in the formula, F and G be curved surface u to v to the continuous function of cutting arrow.

If vowing, the master of the unit method of curved surface P point place curve is m, the per unit system of curved surface vow to be n, and both angles are φ, curvature k, and make L=-r _un _u=nr _Uu, $M=-\frac{1}{2}(r_u{n}_u+r_v{n}_u)=n{r}_{\mathrm{uv}}$ , N=-r _vn _v=nr _Vv, have:

kcosφ(ds) ²＝L(du) ²+2Mdudv+N(dv) ² (7)

L in the formula, M and N also be curved surface u to v to the continuous function of cutting arrow.Then curved surface at the curvature value H at P point place is:

$H=\frac{\mathrm{EN}+\mathrm{GL}-2\mathrm{FM}}{2(\mathrm{EG}-F^2)}---$ $\left(8\right)$

As seen, the u that curvature H depends on curved surface to v to cutting arrow, by the inwardness decision of curved surface.The present invention only extracts the curvature value of 10 maximum wave peak dots in the point model, gets its mean value, and note is done

(7) neural network module is based on multilayer feedforward network BP learning algorithm.

Each parameter of BP training network of garment material wrinkle grade is:

Network input parameter (7): R _d, R _a, R _q, R _k,

With subjectivity grading value VG

Hide layer: 5

Learning rate: 0.01

Error: 0.01

Maximum frequency of training: 500

Output layer (1): objective grading value OG

Training sample: 120

The present invention utilizes the some cloud of the high precision garment material that the 3 D non-contacting type scanner obtains, and introduces the notion of point model, with continuous garment material surface discretize, extracts the geometric feature of reflection data centralization or Discrete Distribution and DATA DISTRIBUTION form.Concentrate unique point and the neighborhood point thereof that filters out reflection garment material fluctuation characteristic at point then, the least square curved surface of structural attitude point in conjunction with the curved surface theory in the discrete differential geometry, is set up the curvature characteristic quantity.By the neural network module of MatLAB, create a cover at last based on rating model point model, objective evaluation garment material wrinkle grade.The present invention is applicable to the wrinkle grade of the garment material of any color and decorative pattern, and degree of accuracy reaches about 96%.

Below according to Fig. 1～10; provide a better embodiment of the inventive method; and in conjunction with description to embodiment; further provide the ins and outs of the inventive method; enabling that technical characterictic of the present invention and functional characteristics are described better, but not to be used for limiting protection domain before the right of the present invention.

At first, regulation according to AATCC-124, to the pure cotton fabric cyclic washing of 32 300 * 300mm and dry 5 times, on the 3 D non-contacting type scanner, obtain high-precision fabric point set, the more uniform 50 * 50mm of intercepting fold zone is totally 120 conduct sample datas to be graded.The scan-data of part sample through pretreated three-dimensional point cloud atlas such as denoisings as shown in Figure 2.The quantity of putting on each sample is about 16000.

1. utilize formula (1)～(4), respectively 120 sample datas are carried out the geometric properties quantitative statistics, statistics as shown in Figure 7.Can see from Fig. 7 a～7d, when wrinkle grade is low, the interquartile range R of coordinate figure on the Z direction under the garment material point model _d, mean difference T _aWith standard deviation R _qAll bigger, and kurtosis value R _kLess.Illustrate that the lower garment material surface undulation of wrinkle grade is bigger, the point under its point model relatively disperses.Along with the raising of wrinkle grade, R _d, R _aAnd R _qSignificantly descend R _kIncrease gradually.Explanation is along with the raising of wrinkle grade value, and the fluctuating quantity of garment material descends gradually, and the point under its point model is tending towards normal distribution gradually, even is high kurtosis distribution.When not considering that wrinkle grade is 3.5 grades the geometric feature of lining, all meet the relation of match preferably between above-mentioned 4 parameters and the subjective flatness grading value.

2. on the point model of sample, extract all extreme points as unique point, and be the center with each extreme point, structure extreme point neighborhood, the extreme point of the point set that the subjective grading value of 6 wrinkle grades is different and neighborhood thereof are shown in Fig. 8 a～8f, wherein red area is wave crest point and neighborhood thereof, and blue region is trough point and neighborhood thereof.

3. on each point set, choose maximum 10 disjunct crests, get its height flat average

The result is shown in 9a.

4. extract the coordinate information of maximum 10 maximum wave peak dots and neighborhood point thereof, utilize formula (5), the least square curved surface at structure crest place; Utilize formula (6)～(8), calculate the curvature value at each very big crest place, get its mean value

The result is shown in Fig. 9 b.

Fig. 9 a and 9b show, subjective mean value is that 1 grade the crest of fabric is higher, and greatly the curvature absolute value at crest place is bigger, illustrates that subjective grading value is that fold on 1 grade the fabric is high and sharp-pointed fold.Along with the raising of wrinkle grade, greatly the height at crest place begins to descend.Subjective grading value is that the fold on 2 grades the fabric is higher and smooth big wrinkle, and subjective grading value is the lower fold that distributing on 3 grades the fabric, and the sharp-pointed degree of fold is sharp-pointed but smooth than the fold on 1 grade of fabric than the fold on 2 grades of fabrics.Subjective grading value is that 3.5 grades fabric sample is compared with 3 grades, the projection that distributing obviously but more smooth big wrinkle.Subjective grading value is that the fold on 4 grades the fabric sample is tiny rugula, and the sharp-pointed degree of its fold is sharp-pointed than the fold of 3.5 grades of fabrics, but smooth than the fold on 2 grades of fabrics.Though subjective grading value is to have filtered out more extreme point on 5 grades the fabric sample,, miss that the protruding point of the most yarn that is caused by the weaving mode of fabric is considered as extreme point because 3 D non-contacting type scanner scanning precision is high.

5.BP the training error curve of network as shown in figure 10.After about 120 step iteration, network model is restrained substantially, and related coefficient reaches 0.9684.

Claims (8)

1. the garment material smoothness grade assessment method based on point model is characterized in that, comprises following concrete steps:

(1) data point set

Obtain the point model of garment material by 3 D non-contacting type high precision scanner;

(2) data analysis

A. analyze putting the coordinate data of concentrating on the each point Z direction from the concentrated or discrete trend of DATA DISTRIBUTION, the aspects such as form of DATA DISTRIBUTION;

B. concentrate the unique point that filters out reflection lining fluctuation characteristic from putting, and the structural attitude neighborhood of a point;

C. rebuild the curved surface at very big crest place;

D. adopt the curved surface theory in the discrete differential geometry, the curvature value at analytical characteristic point place;

(3) rating model

Utilize neural network module, set up the rating model of estimating the lining wrinkle grade.

2. the garment material smoothness grade assessment method based on point model according to claim 1 is characterized in that: described point model is the set that each sample point coordinate value is accurate to the three-dimensional data points of eight of radix point back.

3. the garment material smoothness grade assessment method based on point model according to claim 1, it is characterized in that: the analysis of described coordinate data is meant interquartile range, mean difference, standard deviation and the kurtosis in the mathematical statistics, wherein:

A. interquartile range R _d

R _d＝Q _U-Q _L (1)

Interquartile range R _dBe upper quartile Q _UWith lower quartile Q _LDifference, reflection is in the deviation value of data in 50% on the ordered data centre position, is the measured value of data discrete degree;

B. mean difference R _a

$R_a=\frac{1}{N}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}|Z_i-\stackrel{\‾}{Z}|---\left(2\right)$

In the formula: Z _i---the coordinate figure of any 1 Z direction in the some cloud, mm;

---all assembly averages in the Z direction at random in the some cloud, mm;

N---the quantity of point at random in the some cloud;

Mean difference R _aSize reflected in the dispersion point cloud departure degree of face in the each point and some cloud, measured cloud data in the short transverse dispersion of a distribution.Mean difference is big more, and the comprehensive dispersion degree of data is bigger;

C. standard deviation R _q

$R_q=\sqrt{\frac{1}{N}\underset{i=1}{\overset{1}{\mathrm{\Σ}}}{(Z_i-\stackrel{\‾}{Z})}^2}---\left(3\right)$

Standard deviation R _qWhat reflect is the absolute dispersion degree on the Z direction at random, than mean difference R _aMore given prominence to the influence of the variation of each point to statistic;

D. kurtosis R _k

$R_k=\frac{1}{N}\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(Z_i-\stackrel{\‾}{Z})}^4}{{R_q}^4}---\left(4\right)$

Kurtosis R _kBe that ebb or the peak form that metric data distributes departs from estimating of normal distribution degree.

4. the garment material smoothness grade assessment method based on point model according to claim 1 is characterized in that: adopt the bead model in the screening process of described unique point.

5. according to the described garment material smoothness grade assessment method based on point model of claim l, it is characterized in that: the curved surface at the very big crest of described reconstruction place is the elevation information that extracts ten maximum wave peak dots in the point model, gets its mean value, and note is done

6. the garment material smoothness grade assessment method based on point model according to claim 1 is characterized in that: the curved surface at described very big crest place is the least square curved surface, and its toroidal function form is:

z=a ₀+a ₁x+a ₂y+a ₃x ²+a ₄xy+a ₅y ² (5)

7. the garment material smoothness grade assessment method based on point model according to claim 1 is characterized in that: described curved surface theory is based on the curved surface theory of discrete differential geometry,

(x y) is the function of its corresponding x, y coordinate, that is: to the z=f of each point in the point model

$\left\{\begin{array}{c}x=x(u,v)\\ y=y(u,v)\\ z=z(u,v)\end{array}\right.$

If r _u=(x _u, y _u, z _u) gentle=(x _v, y _v, z _v) represent curved surface respectively u to v to cutting arrow, make E=r _u ², F=r _ur _v, G=r _v ², and remember that the arc length of curved surface upper curve is s, have:

(ds) ²＝E(du) ²+2Fdudv+G(dv) ² (6)

E in the formula, F and G be curved surface u to v to the continuous function of cutting arrow.

If vowing, the master of the unit method of curved surface P point place curve is m, the per unit system of curved surface vow to be n, and both angles are φ, curvature k, and make L=-r _un _u=nr _Uu, $M=-\frac{1}{2}(r_u{n}_u+r_u{n}_u)=n{r}_{\mathrm{uv}}$ , N=-r _vn _v=nr _Vv, have:

kcosφ(ds) ²＝L(du) ²+2Mdudv+N(dv) ² (7)

L in the formula, M and N also be curved surface u to v to the continuous function of cutting arrow.Then curved surface at the curvature value H at P point place is:

$H=\frac{\mathrm{EN}+\mathrm{GL}-2\mathrm{FM}}{2(\mathrm{EG}-F^2)}---\left(8\right)$

8. the garment material smoothness grade assessment method based on point model according to claim 1 is characterized in that: described neural network module is based on multilayer feedforward network BP learning algorithm.

Images