CN105894586A - Method for extracting single-side page of three-dimensional book model - Google Patents

Abstract

The invention relates to a method for extracting a single-side page of a three-dimensional book model. The method includes the steps of forming a minimal cuboid enclosing box based on coordinate extreme values of a point cloud on X, Y and Z directions, establishing a point cloud topological structure based on spatial unit grid, solving the blank plane equation expression dividing left and right pages of a book based on geometric characteristics of the book model, plugging the coordinate values of the point cloud into the equation expression, determining the position of the point cloud, extracting the left or right point cloud of the book model, solving a background plane equation expression based on the minimal value and maximum value points of spatial coordinate x and y of the point cloud and a plurality points in the surrounding areas of the four points, plugging the coordinate values of the point cloud into the equation expression for calculation, determining the position of the point cloud, reserving the point cloud above the plane, conducting book opening cutting, and finally updating and storing the new three-dimensional page model file. The invention further provides a method for automation extracting single-side page three-dimensional model from a three-dimensional book model.

Description

A kind of unilateral page extraction of three-dimensional books model

Technical field

The invention belongs to threedimensional model processing technology field, be specifically related to a kind of to that obtain from spatial digitizer or from three The whole three-dimensional books model that dimension module file obtains processes, it is thus achieved that the processing method of the threedimensional model of the unilateral page.

Background technology

Optical character recognition (OCR) currently for distortion file and picture remains a more scabrous problem, due to Being distorted by document and affected, the Text region rate of two dimensional image often has the biggest decline.Owing to OCR character recognition is to pass through The character image obtained is contrasted with the character image in standard character library and is identified, including to document row Multiple links such as cutting, the cutting of character, the calculating of multiple geometric properties of single character picture and contrast.Due to file and picture Distortion, two dimension file and picture in word guild be distorted, word can deform upon, so causing OCR character recognition software Literal line cutting to file and picture makes a mistake often, and the cutting of single character can occur inaccurate situation the most often, and And, the deformation of single character image also makes the value of calculation of its multiple geometric properties, with the standard character image in standard character library The value of calculation generation relatively large deviation of multiple geometric properties, None-identified, or the situation of wrong identification finally occur.

Due to the appearance of 3-D scanning technology, the character recognition problem for distortion file and picture provides a new solution Approach.Can first obtain the threedimensional model of file and picture, i.e. obtain tens0000 and arrive hundreds of thousands, or up to a million some clouds, and The little image of each grid, then by geological information, calculates the coefficient of torsion of the page, carries out three-dimensional to distortion document Virtual planish, thus obtain more straight file and picture, be greatly improved the OCR character identification rate of distortion file and picture.For The virtual image planished is substantially identical to the character image on smooth paper, and therefore, literal line is neat, and Character deformation does not almost have Having, the calculating of multiple geometrical characteristics of character is the most consistent with the geometrical characteristic of the standard character image in standard character library, Therefore character identification rate is greatly improved.

And for two dimension warp image if it is desired to carry out correction and the recovery in space, although each pixel has X, Y-axis 2 Coordinate on individual direction, but owing to lacking the information of degree of depth Z axis, so cannot reason about the picture calculated when distorting the page and being flattened The position of vegetarian refreshments, and that coordinate on getable X, 2 directions of Y-axis be only also after perspective imaging on imaging plane Coordinate, these 2 values be both not equal to the coordinate on X in real world, 2 directions of Y-axis, can not be according to fixed proportion equal to true In the real world, the coordinate on X, 2 directions of Y-axis, therefore relies solely on this 2 coordinate figures, on the straight paper that reasoning and calculation goes out The position of each pixel is the biggest with its actual position often deviation.So needing the data by threedimensional model to make inferences meter Calculate.

To the 3-d recovery of distortion document firstly the need of being syncopated as the threedimensional model of the clean single page, but general feelings Under condition, from spatial digitizer scanning, obtain is the threedimensional model to whole the books opened, and namely includes the three of 2 pages Dimension module, and this model the most inevitably includes that spatial context, fore-edge etc. need not and can be for the most three-dimensional twisted Recovering to bring interference and troublesome some cloud and texture, if by artificial cognition, and craft is syncopated as not comprising background dot cloud and book The threedimensional model of the single page of mouth point cloud, is not only cumbersome, and is difficult to accurately control.

Summary of the invention

It is an object of the invention to for from spatial digitizer or the three-dimensional books model that obtains from three dimensional file, it is proposed that A kind of automatization's extracting method carrying out single-page threedimensional model.The method include a generation for cloud topological structure, the page draw Point, the rejecting of background, the link such as generation of the excision of fore-edge and new threedimensional model file.

Specifically, the present invention adopts the following technical scheme that

A kind of unilateral page extraction of three-dimensional books model, its step includes:

1) the some cloud based on books model coordinate extreme value formation one on three directions of X, Y, Z axis is parallel with coordinate axes Minimum cuboid bounding box, set up some cloud topological structure based on space cell lattice；

2) geometric characteristic based on books model, the empty facial plane equation obtaining the segmentation books left and right page is expressed Formula, brings the coordinate figure of a cloud into equation expression formula, it is judged that the position of some cloud, extracts on the left of books model or the point on right side Cloud；

3) in point cloud model, based on cloud space coordinates x, the minima of y, maximum of points and this four some peripheral regions If territory is done, obtain background plane equation expression formula, the coordinate figure of a cloud brought in equation expression formula and calculates, it is judged that The position of some cloud, retains the some cloud above background plane；

4) to step 3) the left-hand point cloud that obtains, along some cloud right side boundary direction, it is divided into from right side boundary border to the left A plurality of, carry out fore-edge cutting along most one of page left-half point cloud quantity, retain the some cloud on the right side of demarcation line, obtain Books verso cake cloud；To step 3) the right-hand point cloud that obtains also can in like manner alignment processing；

5) the space vertex coordinate index value that amendment planar grains coordinated indexing is corresponding, updates and preserves new three-dimensional page Model file.

The method of the present invention may be used for the image of three-dimensional twisted document and recovers, but is not limited only to the figure of three-dimensional twisted document As recovering, also can fit for other range of application needing the threedimensional model extracting single-page from complete three-dimensional books model With.

Accompanying drawing explanation

Fig. 1 is the Complete three-dimensional books model of the embodiment of the present invention.

Fig. 2 is the books point cloud model of the embodiment of the present invention.

Fig. 3 is the books minimum cuboid bounding box of the embodiment of the present invention.

Fig. 4 is the middle segmentation plane of the left and right page of the embodiment of the present invention.

Fig. 5 is the left-hand point cloud after the middle segmentation of the embodiment of the present invention.

Fig. 6 is the background segment plane of the embodiment of the present invention.

Fig. 7 is the left page point cloud after the background segment of the embodiment of the present invention.

Fig. 8 is the fore-edge cutting planes of the embodiment of the present invention.

Fig. 9 is the left-hand point cloud page after the fore-edge cutting of the embodiment of the present invention.

Figure 10 is the part v section list of the new model file of the embodiment of the present invention.

Figure 11 is the part vt section list of the new model file of the embodiment of the present invention.

Figure 12 is the part f section list of the new model file of the embodiment of the present invention.

Figure 13 is the left page threedimensional model after the final cutting of the embodiment of the present invention.

Detailed description of the invention

Below by embodiment and accompanying drawing, the present invention is described in detail.

For certain three-dimensional books model, as being intended to extract three-dimensional page on the left of it, extraction step based on this method includes:

Step 1: form a minimum parallel with coordinate axes based on cloud coordinate extreme value on three directions of X, Y, Z axis Cuboid bounding box, sets up some cloud topological structure based on space cell lattice.Specifically comprise the following steps that

1-1) read in the related data of books threedimensional model from spatial digitizer or three-dimensional data file, note dough sheet summit is total Number is P, and X, Y, Z axis (the Z axis positive direction substantially books frontal) coordinate of dough sheet summit p is x_p、y_p、z_p, dough sheet top The index value of point is index_vp, U, V axial coordinate on its corresponding texture summit is u_p、v_p(U axle is that texture summit is in texture maps Transverse axis in Xiang, V axle is the texture summit longitudinal axis in texture image), the index value on texture summit is index_tp, dough sheet q institute Corresponding dough sheet vertex index values is { index_vq1,index_vq2......index_vqrAnd texture vertex index unit {index_tq1,index_tq2......index_tqr}.Otherwise for each summit p, increase relevant deletion and identify unit flag_p, with rope Draw adjustment unit decrease_p。flag_pWith decrease_pInitial value is 0.

Such as, we read in an exemplary three-dimensional document model file example.obj, as shown in Figure 1.Its dough sheet summit Total number P=105895.Its point cloud model is as shown in Figure 2.

1-2) scan each dough sheet summit p (x successively_p,y_p,z_p), obtain dough sheet summit minima in the X, Y, Z direction, Big value.If on three directions, min coordinates value is: min_x、min_yAnd min_z, maximum coordinate value among is: max_x、max_yAnd max_z.According to The extreme value in these three direction, calculates the width of sub-cuboid and length subsize according to formula (1):

$subsize=\sqrt[3]{\frac{({\max }_x-{\min }_x)*({\max }_y-{\min }_y)*({\max }_z-{\min }_z)}{P}}---\left(1\right)$

Such as, for this example model, its dough sheet summit minima in the X, Y, Z direction, maximum be:The result of calculation of formula (1) is: subsize=5.5081213577693333.

1-3) according to min_x、min_y、min_z、max_x、max_yAnd max_z, a cloud is formed a minimum parallel with coordinate axes Cuboid bounding box.And according to sub-cuboid length and width value subsiz_e, cuboid bounding box is pressed tri-coordinates of X, Y, Z Direction marks off L*M*1 sub-cuboid.L, M are given according to formula (2):

$\left\{\begin{array}{c}L=\mathrm{int}\left(\frac{{\max }_x-{\min }_x}{subsize}\right)+1\\ M=\mathrm{int}\left(\frac{{\max }_y-{\min }_y}{subsize}\right)+1\end{array}\right.---\left(2\right)$

Wherein, int is bracket function.

Such as, for this example model, minimum cuboid bounding box, as it is shown on figure 3, the result of calculation of its formula (2) is:

1-4) for every sub-cuboid, setting up one group of data cell, storage is positioned at each top of this sub-cuboid respectively The index value index of point_vp.It is (x for D coordinates value_p,y_p,z_p) some p, then the sub-cuboid in its place is at two coordinates of X, Y Index value on direction (l, m) calculates according to formula (3):

$\left\{\begin{array}{c}l=\mathrm{int}\left(\frac{x_p-{\min }_x}{subsize}\right)\\ m=\mathrm{int}\left(\frac{y_p-{\min }_y}{subsize}\right)\end{array}\right.---\left(3\right)$

Step 2: geometric characteristic based on books model, obtains the empty facial plane equation table of the segmentation books left and right page Reach formula, bring the coordinate figure of a cloud into equation expression formula, it is judged that the position of some cloud, extract on the left of books model or the point on right side Cloud.Specifically comprise the following steps that

The center cell lattice call number of Y-axis 2-1) is calculated according to formula (4)

$m_{center}=\mathrm{int}\left(\frac{M}{2}\right)---\left(4\right)$

Such as, for this example model, the result of calculation of the center cell lattice call number of its Y-axis is: m_center=34.

2-2) according to the centre coordinate apparent position of formula (5) calculating Y-axis:

y_center=min_y+m_center*subsize (5)

Such as, for this example model, the result of calculation of the centre coordinate apparent position of its Y-axis is: y_center=- 3.9651152909207781。

2-3) antithetical phrase cuboid { (0, m_center),(1,m_center),......,(c,m_center),......,(L-1, m_center), the Z axis coordinate average { zave on each dough sheet summit in calculating sub-cuboid respectively₀,zave₁,......, zave_c,......,zave_L-1, and remember that the X-axis index value of the sub-cuboid in Largest Mean place is maxl.If Go to step 2-4)；If otherwiseThen go to step 2-5).

Such as, for this example model, the Search Results of the X-axis index value of its sub-cuboid in Largest Mean place is: Maxl=36,Go to step 2-4).

2-4) scan { zave successively_maxl,zave_maxl+1,......,zave_cl,......,zave_L-2, and depend on according to formula (6) The difference of secondary calculating Z axis coordinate average:

delta_cl=zave_cl-zave_Cl+1 (6)

Obtain first difference value delta less than 0_c0(delta_c0＜ 0), and remember zave_cc=zave_c0。

Such as, for this example model, when it obtains first difference value less than 0: c0=44, and assignment: zave_cc =-1.8424575415750344

2-5) scan zave successively_maxl,zave_maxl-1,......,zave_cr,......,zave₁, and according to formula (7) successively The difference of calculating Z axis coordinate average:

delta_cr=zave_cr-zave_cr-1 (7)

Obtain first difference value delta less than 0_c0(delta_c0＜ 0), and remember zave_cc=zave_c0。

2-6) according to formula (8) structure point p_c:

(min_x+ c0*subsize, min_y+ m_center*subsize,zave_cc) (8)

In like manner, order respectively:

$\left\{\begin{array}{c}{m}_{center}=\mathrm{int}\left(\frac{M}{2}\right)-e*4\\ \mathrm{......}\\ m_{center}=\mathrm{int}\left(\frac{M}{2}\right)-2*4\\ m_{center}=\mathrm{int}\left(\frac{M}{2}\right)-1*4\\ m_{center}=\mathrm{int}\left(\frac{M}{2}\right)+1*4\\ m_{center}=\mathrm{int}\left(\frac{M}{2}\right)+2*4\\ \mathrm{......}\\ m_{center}=\mathrm{int}\left(\frac{M}{2}\right)+e*4\end{array}\right.$

Bring step 2-3 into), 2-4) or 2-3), 2-5), obtain 2e new value to (c0, zave_cc), the value of e typically may be used For [1,5], construct 2e new point (p according to formula (8)_c-e,......,p_c-2,p_c-1,p_C+1,p_C+2,......,p_C+e)。

Such as, for this example model, its value: e=2, (p_c-2,p_c-1,p_c,p_C+1,p_C+2) structure result be:

$\left\{\begin{array}{c}{p}_{c-2}=(-4.9482285550901395,-48.030086153075445,-15.508145419034092)\\ p_{c-1}=(6.0680141604485129,-25.997600721998140,-4.5445390560410237)\\ p_c=(11.576135518217853,-3.9651152909207781,-1.8424575415750344)\\ p_c=(17.084256875987194,18.067370140156527,-0.92686784480299267)\\ p_{c+2}=(28.100499591525875,40.099855571233888,0.59222989777723944)\end{array}\right..$

2-7) to (p_c-e,......,p_c-2,p_c-1,p_c,p_C+1,p_C+2,......,p_c+e) calculate central point according to formula (9) pcenter(x_pcenter,y_pcenter,z_pcenter):

$\left\{\begin{array}{c}{\mathrm{pcenter}}_x=\frac{\underset{ci=c-e}{\overset{c+e}{\Σ}}{x}_{ci}}{2e+1}\\ {\mathrm{pcenter}}_y=\frac{\underset{ci=c-e}{\overset{c+e}{\Σ}}{y}_{ci}}{2e+1}\\ {\mathrm{pcenter}}_z=\frac{\underset{ci=c-e}{\overset{c+e}{\Σ}}{z}_{ci}}{2e+1}\end{array}\right.---\left(9\right)$

And calculate vector according to formula (10)

$\left\{\begin{array}{c}{w}_x=x_{c+e}-x_{c-e}\\ w_y=y_{c+e}-y_{c-e}\\ w_z=z_{c+e}-z_{c-e}\end{array}\right.---\left(10\right)$

Such as, for this example model, its pcenter (x_pcenter,y_pcenter,z_pcenter) result of calculation be:Its vectorResult of calculation be:

Value 2-8) taking space coordinates X is min_xWith max_x, and the value of space coordinates Y is min_yWith max_yPoint, and Other each point in this sub-cuboid region, four some places, constitutes the point set { (x of I point altogether₀,y₀,z₀),(x₁,y₁, z₁),......,(x_i,y_i,z_i),......(x_I-1,y_I-1,z_I-1), the value including space coordinates X is min_xWith max_x, with And the value of space coordinates Y is min_yWith max_y4 points, (x_i,y_i,z_i) be summit i in institute's reconnaissance point set X, Y, Z axis position sit Mark.The normal vector of background plane is calculated according to formula (11)

$\{\begin{array}{c}{h}_x=\underset{hi=0}{\overset{I-1}{\Σ}}(y_{hi}-y_{next\left(hi\right)})(z_{hi}+z_{next\left(hi\right)})\\ h_y=\underset{hi=0}{\overset{I-1}{\Σ}}(z_{hi}-z_{next\left(hi\right)})(x_{hi}+x_{next\left(hi\right)})\\ h_z=\underset{hi=0}{\overset{I-1}{\Σ}}(x_{hi}-x_{next\left(hi\right)})(y_{hi}+y_{next\left(hi\right)})\end{array}---\left(11\right)$

Wherein next (hi) is the sequence number on next summit of summit hi, calculates according to formula (12):

Next (hi)=(hi+1) mod I (12)

Wherein, mod is complementation function.

Such as, for this example model, its vectorResult of calculation be:

The law vector of segmentation plane 2-9) is calculated according to formula (13) cross product

$\stackrel{\→}{b}=\stackrel{\→}{w}\×\stackrel{\→}{h}---\left(13\right)$

Such as, for this example model, its vectorFor:

2-10) calculate segmentation plane according to formula (14) and point to the normal vector of X-axis forward

$\left\{\begin{array}{cc}{b^{\&prime;}}_x=h_x& b_x\&GreaterEqual;0\\ {b^{\&prime;}}_y=h_y& b_x\&GreaterEqual;0\\ {b^{\&prime;}}_z=h_z& b_x\&GreaterEqual;0\\ {b^{\&prime;}}_x=-h_x& b_x<0\\ {b^{\&prime;}}_y=-h_y& b_x<0\\ {b^{\&prime;}}_z=-h_z& b_x<0\end{array}\right.---\left(14\right)$

Such as, for this example model, its vectorResult of calculation be:

2-11) to each dough sheet summit p (x_p,y_p,z_p), calculate each dough sheet summit p (x according to formula (15)_p,y_p,z_p) for segmentation The eigenvalue g of plane relative position_cp:

g_cp=b'_x(x_p-pcenter_x)+b'_y(y_p-pcenter_y)+b'_z(z_p-pcenter_z) (15)

If left page to be extracted, update flag according to formula (16)_p:

$\{\begin{array}{cc}{\mathrm{flag}}_p=1& g_{cp}>0\\ {\mathrm{flag}}_p={\mathrm{flag}}_p& g_{cp}\&le;0\end{array}---\left(16\right)$

If being intended to extract the right side page, update flag according to formula (17)_p:

$\{\begin{array}{cc}{\mathrm{flag}}_p=1& g_{cp}<0\\ {\mathrm{flag}}_p={\mathrm{flag}}_p& g_{cp}\&GreaterEqual;0\end{array}---\left(17\right)$

Such as, for this example model, it splits the plane of the left and right page as shown in Figure 4, left part page point cloud after segmentation As shown in Figure 5.

Step 3: in point cloud model, based on cloud space coordinates x, the minima of y, maximum of points and this four some weeks If enclosing region to do, obtaining background plane equation expression formula, the coordinate figure of a cloud being brought in equation expression formula and calculates, sentencing The position of breakpoint cloud, retains the some cloud above plane.Specifically comprise the following steps that

Value 3-1) taking space coordinates X is min_xWith max_x, and the value of space coordinates Y is min_yWith max_yPoint, and Other each point in this sub-cuboid region, four some places, constitutes the point set ((x of I point altogether₀,y₀,z₀),(x₁,y₁, z₁),......,(x_i,y_i,z_i),......(x_I-1,y_I-1,z_I-1)), the value including space coordinates X is min_xWith max_x, with And the value of space coordinates Y is min_yWith max_y4 points, (x_i,y_i,z_i) be summit i in institute's reconnaissance point set X, Y, Z axis position sit Mark.To selected point set, calculate the average coordinates a (a of institute's reconnaissance according to formula (18)_x,a_y,a_z):

$\left\{\begin{array}{c}{a}_x=\frac{\underset{ai=0}{\overset{I-1}{\&Sigma;}}{x}_{ai}}{I}\\ a_y=\frac{\underset{ai=0}{\overset{I-1}{\&Sigma;}}{y}_{ai}}{I}\\ a_z=\frac{\underset{ai=0}{\overset{I-1}{\&Sigma;}}{z}_{ai}}{I}\end{array}\right.---\left(18\right)$

3-2) calculate background plane according to formula (19) and point to the normal vector of Z axis forward

$\left\{\begin{array}{cc}{h^{\&prime;}}_x=h_x& h_z\&GreaterEqual;0\\ {h^{\&prime;}}_y=h_y& h_z\&GreaterEqual;0\\ {h^{\&prime;}}_z=h_z& h_z\&GreaterEqual;0\\ {h^{\&prime;}}_x=-h_x& h_z<0\\ {h^{\&prime;}}_y=-h_y& h_z<0\\ {h^{\&prime;}}_z=-h_z& h_z<0\end{array}\right.---\left(19\right)$

Such as, for this example model, the result of calculation of its vector is:

3-3) to each dough sheet summit p (x_p,y_p,z_p), calculate dough sheet summit p (x according to formula (20)_p,y_p,z_p) for background plane The eigenvalue g of position relatively_bp:

g_bp=h'_x(x_p-a_x)+h'_y(y_p-a_y)+h'_z(z_p-a_z) (20)

And update flag according to formula (21)_p:

$\{\begin{array}{cc}{\mathrm{flag}}_p=1& g_{bp}<0\\ {\mathrm{flag}}_p={\mathrm{flag}}_p& g_{bp}\&GreaterEqual;0\end{array}---\left(21\right)$

Such as, for this example model, the background plane of its left part page as shown in Figure 6, above the background plane of extraction Some cloud as shown in Figure 7.

Step 4: to step 3) the left-hand point cloud that obtains, along some cloud right side boundary direction, from right side boundary border to the left It is divided into a plurality of, carries out fore-edge cutting along most one of page left-half point cloud quantity, retain the some cloud on the right side of demarcation line, Obtain books verso cake cloud.Can be to step 3) the right-hand point cloud that obtains also can in like manner alignment processing.Specifically comprise the following steps that

To step 3) left page that obtains, along some cloud right side boundary direction, it is divided into D from right side boundary border to the left (0,1 ... d ... 2D-1) article (the 0th article at rightmost, the 2D-1 article at Far Left), along page left-half midpoint The most maxd of cloud quantity carries out fore-edge cutting, retains the some cloud on the right side of segmentation lines, obtains books verso cake cloud.To step Rapid 3) the right side page obtained, along some cloud left border right direction, be divided into from left border border to the right D (0, 1 ... d ... 2D-1) article (the 0th article at Far Left, the 2D-1 article at rightmost), along page right half part point cloud The most maxd of quantity carries out cutting, retains the some cloud on the left of segmentation lines, obtains books recto cake cloud.The value model of D Enclose and may generally be [15,25].Concretely comprising the following steps of described step 4:

4-1) for step 3) page that obtains, remember flag_pThe X-axis minima on the dough sheet summit of=0 is min'_x, flag_p The X-axis maximum on the dough sheet summit of=0 is max'_x, calculate every step-length step in the X-axis direction according to formula (22):

$step=\frac{{{\max }^{\&prime;}}_x-{{\min }^{\&prime;}}_x}{2D}---\left(22\right)$

Such as, for this example model, taking D=20, the result of calculation of every step-length step in the X-axis direction is: Step=5.5455702781677250.

4-2) to (D, D+1 ... d ... 2D-1) each bar, compose number num of available point in bar according to formula (23)_dJust Value is 0:

num_d=0 (23)

4-3) to (D, D+1 ... d ... 2D-1) each bar, calculate fall in every d number num of available point_d。 Method particularly includes: scan each dough sheet summit p (x_p,y_p,z_p), for step 3) left page that obtains, calculate dough sheet according to formula (24) Summit p (x_p,y_p,z_p) for when preceding article d is relative to the eigenvalue g of position_npdrWith g_npdl:

$\{\begin{array}{c}{g}_{npdr}={b^{\&prime;}}_x(x_p-{\mathrm{pcenter}}_x+step*d)+{b^{\&prime;}}_y(y_p-{\mathrm{pcenter}}_y)+{b^{\&prime;}}_z(z_p-{\mathrm{pcenter}}_z)\\ g_{npdl}={b^{\&prime;}}_x(x_p-{\mathrm{pcenter}}_x+step*(d+1\left)\right)+{b^{\&prime;}}_y(y_p-{\mathrm{pcenter}}_y)+{b^{\&prime;}}_z(z_p-{\mathrm{pcenter}}_z)\end{array}---\left(24\right)$

And update num according to formula (25)_d:

Here " ∧ " be " and " relation, " ∨ " is "or" relation.

For step 3) the right side page that obtains, calculate dough sheet summit p (x according to formula (26)_p,y_p,z_p) for when preceding article d phase Eigenvalue g to position_npdrWith g_npdl:

$\{\begin{array}{c}{g}_{npdr}={b^{\&prime;}}_x(x_p-{\mathrm{pcenter}}_x-step*(d+1\left)\right)+{b^{\&prime;}}_y(y_p-{\mathrm{pcenter}}_y)+{b^{\&prime;}}_z(z_p-{\mathrm{pcenter}}_z)\\ g_{npdl}={b^{\&prime;}}_x(x_p-{\mathrm{pcenter}}_x-step*d)+{b^{\&prime;}}_y(y_p-{\mathrm{pcenter}}_y)+{b^{\&prime;}}_z(z_p-{\mathrm{pcenter}}_z)\end{array}---\left(26\right)$

And update num according to formula (27)_d:

4-4) to (0,1 ... d ... D-1) each bar, scan each num_d, take num_dThe bar number of maximum be maxd.For step 3) left page that obtains, to each dough sheet summit p (x_p,y_p,z_p), calculate dough sheet summit p according to formula (28) (x_p,y_p,z_p) for fore-edge segmentation plane relative to the eigenvalue g of position_ep:

g_ep=b'_x(x_p-pcenter_x+step*maxd)+b'_y(y_p-pcenter_y)+b'_z(z_p-pcenter_z) (28)

And update flag according to formula (29)_p:

$\{\begin{array}{cc}{\mathrm{flag}}_p={\mathrm{flag}}_p& g_{ep}>0\\ {\mathrm{flag}}_p=1& g_{ep}\&le;0\end{array}---\left(29\right)$

For step 3) the right side page that obtains, to each dough sheet summit p (x_p,y_p,z_p), calculate dough sheet summit p according to formula (30) (x_p,y_p,z_p) for fore-edge segmentation plane relative to the eigenvalue g of position_ep:

g_ep=b'_x(x_p-pcenter_x-step*maxd)+b'_y(y_p-pcenter_y)+b'_z(z_p-pcenter_z) (30)

And update flag according to formula (31)_p:

$\{\begin{array}{cc}{\mathrm{flag}}_p=1& g_{ep}\&GreaterEqual;0\\ {\mathrm{flag}}_p={\mathrm{flag}}_p& g_{ep}<0\end{array}---\left(31\right)$

Such as, for this example model, its scanning result is: maxd=35.The fore-edge cutting planes such as Fig. 8 institute obtained Showing, the some cloud plane after fore-edge cutting is as shown in Figure 9.

Step 5: the space vertex coordinate index value that amendment planar grains coordinated indexing is corresponding, updates and preserves new three-dimensional Page model file.Specifically comprise the following steps that

5-1) press dough sheet vertex index values index_vpOrder from small to large scans each dough sheet summit p (x successively_p,y_p,z_p), If flag_p=1, the most no longer by its front D coordinates value (x_p,y_p,z_p) it is stored in the vertex list of model file, i.e. v section list.If flag_p=0, then by its front D coordinates value (x_p,y_p,z_p) be stored in the vertex list of model file, statistical index value be 1, 2,......j......,index_vpFlag in the dough sheet summit of-1}_pNumber s of=1, and calculate according to formula (32)

${\mathrm{decrease}}_{{\mathrm{index}}_{vp}}=s---\left(32\right)$

It is index for index value_vpThe index adjustment unit value on dough sheet summit.

Such as, for this example model, the part v section list finally given is as shown in Figure 10.

It is 5-2) index by texture vertex index values_tpOrder from small to large scans each texture summit p (u successively_p,v_p), And by two-dimensional coordinate value (u_p,v_p) it is stored in the texture image list of coordinates of model file, i.e. vt section list.

Such as, for this example model, the part vt section list finally given is as shown in figure 11.

5-3) scan successively by dough sheet vertex index index forming dough sheet_vWith texture vertex index index_tConstituted Each dough sheet.

For dough sheet q, its corresponding dough sheet vertex index values is { index_vq1,index_vq2......index_vqr, stricture of vagina Reason vertex index is { index_tq1,index_tq2......index_tqr, if index value { index_vq1, index_vq2......index_vqrIn dough sheet summit corresponding to }, there is certain summit p (x_p,y_p,z_p), its flag_p=1, then face The corresponding dough sheet vertex index values { index of sheet q_vq1,index_vq2......index_vqrAnd texture vertex index values {index_tq1,index_tq2......index_tqrThe face list of model file of not restoring, i.e. f section list.

Otherwise, for dough sheet q, its corresponding dough sheet vertex index values is { index_vq1, index_vq2......index_vqr, texture vertex index is { index_tq1,index_tq2......index_tqr, if index value {index_vq1,index_vq2......index_vqrIn dough sheet summit corresponding to }, all summit p (x_p,y_p,z_p), all there is flag_p =0, then according to the corresponding new index value in dough sheet summit of formula (33) calculating dough sheet q:

$\left\{\begin{array}{c}{{\mathrm{index}}^{\&prime;}}_{vq1}={\mathrm{index}}_{vq1}-{\mathrm{decrease}}_{{\mathrm{index}}_{vq1}}\\ {{\mathrm{index}}^{\&prime;}}_{vq2}={\mathrm{index}}_{vq2}-{\mathrm{decrease}}_{{\mathrm{index}}_{vq2}}\\ \mathrm{......}\\ \mathrm{......}\\ {{\mathrm{index}}^{\&prime;}}_{vqr}={\mathrm{index}}_{vqr}-{\mathrm{decrease}}_{{\mathrm{index}}_{vqr}}\end{array}\right.---\left(33\right)$

And by dough sheet q its corresponding dough sheet new vertex index { index'_vq1,index'_vq2......index'_vqr, stricture of vagina Reason vertex index { index_tq1,index_tq2......index_tqr, all it is stored in the face list of model file, i.e. f section list.

Such as, for this example model, the part f section list finally given is as shown in figure 12.The attachment stricture of vagina finally given The extraction page of reason is as shown in figure 13.

Though various threedimensional model file formats have difference, but ultimate principle is as above, during the storage of concrete model file, can according to tool Body file actual format is adjusted.

Above-described embodiment is only in order to illustrate the know-why of the present invention, not in order to limit the present invention.The technology of this area Technical scheme can be made equal change and amendment by personnel, and protection scope of the present invention should be with the limit of claims It is set to standard.

Claims (6)

1. the unilateral page extraction of a three-dimensional books model, it is characterised in that comprise the following steps:

1) based on books model some cloud coordinate extreme value on three directions of X, Y, Z axis, formed one parallel with coordinate axes Minimum cuboid bounding box, sets up some cloud topological structure based on space cell lattice；

2) geometric characteristic based on books model, obtains the empty facial plane equation expression formula of the segmentation books left and right page, will The coordinate figure of some cloud brings equation expression formula into, it is judged that the position of some cloud, extracts on the left of books model or the some cloud on right side；

3) in point cloud model, if based on cloud space coordinates x, the minima of y, maximum of points and this four some peripheral regions Do, obtain background plane equation expression formula, the coordinate figure of a cloud is brought in equation expression formula and calculates, it is judged that some cloud Position, retain the some cloud above background plane；

4) to step 3) the left-hand point cloud that obtains, along some cloud right side boundary direction, it is divided into a plurality of from right side boundary border to the left, Carry out fore-edge cutting along most one of page left-half point cloud quantity, retain the some cloud on the right side of demarcation line, obtain books Verso cake cloud；To step 3) the right-hand point cloud the most in like manner alignment processing that obtains；

5) the space vertex coordinate index value that the planar grains coordinated indexing of amendment books model is corresponding, updates and preserves new three Dimension page model file.

2. the method for claim 1, it is characterised in that described step 1) concretely comprise the following steps:

The related data of books threedimensional model, the total number in note dough sheet summit 1-1) is read in from spatial digitizer or three-dimensional data file X, Y, Z axis (Z axis positive direction substantially books frontal) coordinate for P, dough sheet summit p is x_p、y_p、z_p, dough sheet summit Index value is index_vp, U, V axial coordinate on its corresponding texture summit is u_p、v_p, the index value on texture summit is index_tp, Dough sheet vertex index values corresponding to dough sheet q is { index_vq1,index_vq2......index_vqrAnd texture vertex index unit {index_tq1,index_tq2......index_tqr}；Otherwise for each summit p, increase relevant deletion and identify unit flag_p, with rope Draw adjustment unit decrease_p；flag_pWith decrease_pInitial value is 0；

1-2) scan each dough sheet summit p (x successively_p,y_p,z_p), obtain dough sheet summit minima in the X, Y, Z direction, maximum Value；If on three directions, min coordinates value is: min_x、min_yAnd min_z, maximum coordinate value among is: max_x、max_yAnd max_z；According to this The extreme value in three directions, calculates the width of sub-cuboid and length subsize according to formula (1):

1-3) according to min_x、min_y、min_z、max_x、max_yAnd max_z, a cloud is formed a minimum parallel with coordinate axes rectangular Body bounding box；And according to sub-cube length subsize, cuboid bounding box is marked off L* by tri-coordinate directions of X, Y, Z M*1 sub-cuboid；L, M are given according to formula (2):

Wherein, int is bracket function；

1-4) for every sub-cuboid, setting up one group of data cell, storage is positioned at each summit of this sub-cuboid respectively Index value index_vp；It is (x for D coordinates value_p,y_p,z_p) some p, then the sub-cuboid in its place is at two coordinate directions of X, Y On index value (l, m) according to formula (3) calculate:

3. method as claimed in claim 2, it is characterised in that described step 2) concretely comprise the following steps:

The center cell lattice call number of Y-axis 2-1) is calculated according to formula (4)

2-2) according to the centre coordinate apparent position of formula (5) calculating Y-axis:

y_center=min_y+m_center*subsize (5)

2-3) antithetical phrase cuboid { (0, m_center),(1,m_center),......,(c,m_center),......,(L-1,m_center), The Z axis coordinate average { zave on each dough sheet summit in calculating sub-cube respectively₀,zave₁,......,zave_c,......, zave_L-1, and remember that the X-axis index value of the sub-cuboid in Largest Mean place is maxl；IfGo to step 2-4)；No If thenThen go to step 2-5)；

2-4) scan { zave successively_maxl,zave_maxl+1,......,zave_cl,......,zave_L-2, and count successively according to formula (6) The difference of calculation Z axis coordinate average:

delta_cl=zave_cl-zave_cl+1 (6)

Obtain first difference value delta less than 0_c0(delta_c0＜ 0), and remember zave_cc=zave_c0；

2-5) scan zave successively_maxl,zave_maxl-1,......,zave_cr,......,zave₁, and calculate Z successively according to formula (7) The difference of axial coordinate average:

delta_cr=zave_cr-zave_cr-1 (7)

Obtain first difference value delta less than 0_c0(delta_c0＜ 0), and remember zave_cc=zave_c0；

2-6) according to formula (8) structure point p_c:

(min_x+c0*subsize,min_y+m_center*subsize,zave_cc) (8)

In like manner, order respectively:

Bring step 2-3 into), 2-4) or 2-3), 2-5), obtain 2e new value to (c0, zave_cc), construct 2e according to formula (8) New point (p_c-e,......,p_c-2,p_c-1,p_c+1,p_c+2,......,p_c+e)；

2-7) to (p_c-e,......,p_c-2,p_c-1,p_c,p_c+1,p_c+2,......,p_c+e) calculate central point pcenter according to formula (9) (x_pcenter,y_pcenter,z_pcenter):

And calculate vector according to formula (10)

Value 2-8) taking space coordinates X is min_xWith max_x, and the value of space coordinates Y is min_yWith max_yPoint, and this four Other each point in sub-cuboid region, individual some place, constitutes the point set { (x of I point altogether₀,y₀,z₀),(x₁,y₁,z₁),......, (x_i,y_i,z_i),......(x_I-1,y_I-1,z_I-1), the value including space coordinates X is min_xWith max_x, and space coordinates Y Value be min_yWith max_y4 points, (x_i,y_i,z_i) it is the X, Y, Z axis position coordinates of summit i in institute's reconnaissance point set；According to formula (11) Calculate the normal vector of background plane

Wherein next (hi) is the sequence number on next summit of summit hi, calculates according to formula (12):

Next (hi)=(hi+1) modI (12)

Wherein, mod is complementation function；

The law vector of segmentation plane 2-9) is calculated according to formula (13) cross product

2-10) calculate segmentation plane according to formula (14) and point to the normal vector of X-axis forward

2-11) to each dough sheet summit p (x_p,y_p,z_p), calculate each table sheet cake p (x according to formula (15)_p,y_p,z_p) for segmentation plane The eigenvalue g of position relatively_cp:

g_cp=b'_x(x_p-pcenter_x)+b'_y(y_p-pcenter_y)+b'_z(z_p-pcenter_z) (15)

If left page to be extracted, update flag according to formula (16)_p:

If being intended to extract the right side page, update flag according to formula (17)_p:

4. method as claimed in claim 3, it is characterised in that described step 3) concretely comprise the following steps:

Value 3-1) taking space coordinates X is min_xWith max_x, and the value of space coordinates Y is min_yWith max_yPoint, and this four Other each point in sub-cube region, individual some place, constitutes the point set ((x of I point altogether₀,y₀,z₀),(x₁,y₁,z₁),......, (x_i,y_i,z_i),......(x_I-1,y_I-1,z_I-1)), the value including space coordinates X is min_xWith max_x, and space coordinates Y Value be min_yWith max_y4 points, (x_i,y_i,z_i) it is the X, Y, Z axis position coordinates of summit i in institute's reconnaissance point set；To institute's reconnaissance Collection, calculates the average coordinates a (a of institute's reconnaissance according to formula (18)_x,a_y,a_z):

3-2) calculate background plane according to formula (19) and point to the normal vector of Z axis forward

3-3) to each dough sheet summit p (x_p,y_p,z_p), according to formula (20) computational chart sheet cake p (x_p,y_p,z_p) relative for background plane The eigenvalue g of position_bp:

g_bp=h'_x(x_p-a_x)+h'_y(y_p-a_y)+h'_z(z_p-a_z) (20)

And update flag according to formula (21)_p:

5. method as claimed in claim 4, it is characterised in that described step 4) method particularly includes: to step 3) obtain Left page, along some cloud right side boundary direction, be divided into from right side boundary border to the left D (0,1 ... d ... 2D-1) Bar, wherein the 0th article at rightmost, and the 2D-1 article at Far Left；Along the maxd that page left-half point cloud quantity is most Carry out fore-edge cutting, retain the some cloud on the right side of segmentation lines, obtain books verso cake cloud；To step 3) the right side page that obtains, Along some cloud left border direction, be divided into from left border border to the right D (0,1 ... d ... 2D-1) bar, Qi Zhong Article 0, at Far Left, the 2D-1 article at rightmost；A maxd most along page right half part point cloud quantity carries out cutting, Retain the some cloud on the left of segmentation lines, obtain books recto cake cloud；Described step 4) concretely comprise the following steps:

4-1) for step 3) page that obtains, remember flag_pThe X-axis minima on the dough sheet summit of=0 is min'_x, flag_p=0 The X-axis maximum on dough sheet summit is max'_x, calculate every step-length step in the X-axis direction according to formula (22):

4-2) to (D, D+1 ... d ... 2D-1) each bar, compose number num of available point in bar according to formula (23)_dInitial value is 0:

num_d=0 (23)

4-3) to (D, D+1 ... d ... 2D-1) each bar, calculate fall in every d number num of available point_d；Specifically Method is: scan each dough sheet summit p (x_p,y_p,z_p), for step 3) left page that obtains, according to formula (24) computational chart sheet cake p(x_p,y_p,z_p) for when preceding article d is relative to the eigenvalue g of position_npdrWith g_npdl:

And update num according to formula (25)_d:

Here " ∧ " be " and " relation, " ∨ " is "or" relation；

For step 3) the right side page that obtains, according to formula (26) computational chart sheet cake p (x_p,y_p,z_p) for when the phase para-position of preceding article d The eigenvalue g put_npdrWith g_npdl:

And update num according to formula (27)_d:

4-4) to (0,1 ... d ... D-1) each bar, scan each num_d, take num_dThe bar number of maximum be maxd；Right In step 3) left page that obtains, to each dough sheet summit p (x_p,y_p,z_p), according to formula (28) computational chart sheet cake p (x_p,y_p,z_p) For fore-edge segmentation plane relative to the eigenvalue g of position_ep:

g_ep=b'_x(x_p-pcenter_x+step*maxd)+b'_y(y_p-pcenter_y)+b'_z(z_p-pcenter_z) (28)

And update flag according to formula (29)_p:

For step 3) the right side page that obtains, to each dough sheet summit p (x_p,y_p,z_p), according to formula (30) computational chart sheet cake p (x_p, y_p,z_p) for fore-edge segmentation plane relative to the eigenvalue g of position_ep:

g_ep=b'_x(x_p-pcenter_x-step*maxd)+b'_y(y_p-pcenter_y)+b'_z(z_p-pcenter_z) (30)

And update flag according to formula (31)_p:

6. method as claimed in claim 5, it is characterised in that described step 5) concretely comprise the following steps:

5-1) press dough sheet vertex index values index_vpOrder from small to large scans each dough sheet summit p (x successively_p,y_p,z_p), if flag_p=1, the most no longer by its front D coordinates value (x_p,y_p,z_p) it is stored in the vertex list of model file, i.e. v section list；If flag_p=0, then by its front D coordinates value (x_p,y_p,z_p) be stored in the vertex list of model file, statistical index value be 1, 2,......j......,index_vpFlag in the dough sheet summit of-1}_pNumber s of=1, and calculate according to formula (32)

It is index for index value_vpThe index adjustment unit value on dough sheet summit；

It is 5-2) index by texture vertex index values_tpOrder from small to large scans each texture summit p (u successively_p,v_p), and will Two-dimensional coordinate value (u_p,v_p) it is stored in the texture image list of coordinates of model file, i.e. vt section list；

5-3) scan successively by dough sheet vertex index index forming dough sheet_vWith texture vertex index index_tEach constituted Sheet；

For dough sheet q, its corresponding dough sheet vertex index values is { index_vq1,index_vq2......index_vqr, texture top Point index is { index_tq1,index_tq2......index_tqr, if index value { index_vq1,index_vq2......index_vqr} In corresponding dough sheet summit, there is certain summit p (x_p,y_p,z_p), its flag_p=1, then the corresponding dough sheet summit of dough sheet q Index value { index_vq1,index_vq2......index_vqrAnd texture vertex index values { index_tq1, index_tq2......index_tqrThe face list of model file of not restoring, i.e. f section list；

Otherwise, for dough sheet q, its corresponding dough sheet vertex index values is { index_vq1,index_vq2......index_vqr, Texture vertex index is { index_tq1,index_tq2......index_tqr, if index value { index_vq1, index_vq2......index_vqrIn dough sheet summit corresponding to }, all summit p (x_p,y_p,z_p), all there is flag_p=0, then depend on The corresponding new index value in dough sheet summit of formula (33) calculating dough sheet q:

And by dough sheet q its corresponding dough sheet new vertex index { index'_vq1,index'_vq2......index'_vqr, texture top Point index { index_tq1,index_tq2......index_tqr, all it is stored in the face list of model file, i.e. f section list.