• Nenhum resultado encontrado

Symmetry constraints derived from loops over spherical faces . 121

5.3 Symmetry constraints originated from face boundary loops

5.3.5 Symmetry constraints derived from loops over spherical faces . 121

sphere Os. If BSp is the boundary of Sp and Γi P t1, . . . , nLu its loops, all the areas bounded by Γi are finite and so are their complement, i.e. AreapSpq AreapSp|Γiq. This property shows that the configurations 1 and 2 set for cylinders and cones are no longer applicable here since the loop status over Sp does not refer to the concept of external/internal. Differing from cylinders and cones, a sphere has no particular direction for symmetry planes. Hence, symmetry planes overSp rely more strongly on the symmetry properties ofBSp, which can be characterized from symmetry and loca- tion properties of facesFa adjacent toSp. These properties are detailed in section5.4 where the influence of references surfaces will be detailed.

5.3.6 Synthesis about loops and loop symmetry CSP

From the analysis of symmetry constraints related to face boundaries (see sec- tion 5.3), the importance of loops and the diversity of CSPs ΠOLBLS becomes clear in a surface symmetry analysis process. Whether a face F has an external loop or not, the number of loops and their relative position over F, all these informations are key elements influencing the symmetry properties of F and give bounds on the maximum number of symmetry planes depending on loop configurations.

Each loop Γi of F has its own symmetry plane set, namely self symmetry planes:

slfΓ

i s. These symmetry planes cross Γi, which indicates that they can be either of type orthogonal rΠOs or loop bisector rΠLBs only. Now, when considering different loops Γij PF, their common self symmetry planes may come from the intersection betweenrΠslfΓ

i sandrΠslfΓ

js. Because loops overF must be either disconnected or touch each other at isolated points only to stay consistent with B-Rep model fundamentals, rΠslfΓi sXrΠslfΓjsmay not produce all the symmetry planes set by the constraints Γij. The relative position of Γi and Γj, as two disconnected infinite point sets, may also satisfy the symmetry condition5.1whereF1 andF2 are substituted by Γi and Γj. Consequently, the corresponding symmetry plane Π is calledloop symmetrycandidate symmetry plane and noted ΠLS. Indeed, ΠLS being derived from loops Γi and Γj, it can be noted more precisely ΠLS,Γi,j. Regarding their relative position in MM I,

Γi and Γj cannot touch each other more than once, otherwise they define a face Fk, homologous toF, lying in between the facesFi andFj that could be associated to Γi

and Γj respectively. Each contact point being surrounded by more than three faces Fi,Fj,FkandF, and no more than two loops in contact can be attached to two faces (see section 5.4.2), e.g. Fi Fj Fij, then Fij crosses Fk (Fk F). Hence, other loops (faces) touchingFi or Fj or Fk are not homologous with these ones, therefore these faces are not crossingF andFij and the manifold vertex splitting operator will duplicate these contact points but the edge merging one will remove them, leaving no more than one contact point representing the crossing configuration inMM AX. The CSP attached to that point is necessarily of type ΠLB not ΠLS. All the contact points associated to non crossing configurations will be removed and represented by ΠLS not ΠLB.

This analysis shows that ΠLS are strictly attached to loop entities and, hence, disconnected from the hypergraphs Gij, which contain vertices and maximal edges only. This justifies the existence of loop datastructures derived from the hypergraphs to complement the faces, edges and vertices.

Let rΓexs, rΓins and rΓuks be the sets of loops representing the external loops, internal loopsandloops of unknown statusofF, respectively. The loops Γi, Γj and Γk

belong to either set rΓexs orrΓins orrΓuks without influencing the existence of ΠLS. However, ΠLS cannot be defined by loops Γi and Γjbelonging to either set: Γi P rΓexs, Γj P rΓins, for example. Consequently to the above analysis of connected/disconnected loops, one pair of loops only has a unique ΠLS plane. Because ΠLSΓ

pi,jq is not crossing any other loop, ΠLSΓ

pi,jq R rΠslfΓ

islfΓ

js. Therefore, the set of symmetry planes of Γi

and Γj satisfy:

Γpi,jqs prΠslfΓ

i s X rΠslfΓ

jsq YΠLSΓpi,jq. (5.2) In case a faceF is bounded by external loops rΓexsas well as internal ones rΓins, because rΓexs can be said as the mandatory limitation of F, i.e. the loops that avoid havingF unbounded, and they are distinct from rΓins, whatever the symmetry properties ofrΓins, ifrΓexsis not symmetric,F is not symmetric. As a result,rΠBFs € rΠΓexs. Now, consideringrΓins, the symmetry planes ofF must be contained inrΠΓins: rΠBFs € rΠΓins. Finally, we have:

BFs rΠΓexs X rΠΓins. (5.3) To unify eq. (5.3), when Γex does not exist, rΠΓexs means all possible symmetry planes ofF inIR3. Similarly, if Γindoes not exist,rΠΓinsmeans all possible symmetry planes of F in IR3, too. In these cases, symmetry plane sets are infinite. When F contains loops of unknown status, it means thatrΠΓexsdoes not exist. Indeed, rΓuks

Symmetry constraints originated from face boundary loops 123 behaves likerΠΓinsand can be included in it.

As an example, regarding cylinders, their maximal number of external loops is two. Then, with the eq. (5.2), rΠΓexsis expressed as:

Γexs rΠΓex

p1,2qs prΠslfΓex

1 s X rΠslfΓex

2 sq YΠLSΓex

p1,2q. (5.4) If the number of internal loops is smaller than two, eq. (5.4) has to be filtered with the set of symmetry planes for the internal loop but there is no ΠLS plane involved in rΠΓins. However, if the number of internal loops is greater than or equal to two, there is no criterion for filtering the symmetry planes. rΠΓins can be bounded by the union of the symmetry properties of each pair of internal loops. As a result, eq. (5.4) changes to:

Γins € ¤

i1...n j1...n i j

Γin

pi,jqs. (5.5)

Finally, for a surface having two external loops and several internal loops, the symmetry planes combine as follows:

BFs € prΠΓexs X ¤

i1...n j1...n i j

Γin

pi,jqs

q, (5.6)

BFs € r

prΠslfΓex

1 s X rΠslfΓex

2 sq YΠLSΓex

p1,2q

X ¤

i1...n j1...n i j

prΠslfΓin

i s X rΠslfΓin

j sq YΠLSΓin pi,jq

s. (5.7)

The boundary loops ofF are strong symmetry constraints. There are two different boundary loop configurations: external ones Γex and internal ones Γin. Reference surfaces having at least one external loop, such as planes, cylinders and cones, the CSPs ofF are constrained by the external loops first. If the number of internal loops is less than two, symmetry planes are defined by the intersection between external and internal symmetry plane groups only, i.e. there is no ΠLS plane. Otherwise, the symmetry planes are defined by the intersection between external loops symmetry planes and the union of internal loops symmetry planes including the ΠLS planes belonging to both sets. If F has no external loop, the internal loops are the only constraints.

Right now, the symmetry planes associated with loops stand for symmetry infor- mation attached to a face boundary as well as neighboring faces adjacent to edges and vertices of the loop. As a result, it incorporates symmetry properties of these faces:

ΠO and ΠLB symmetry planes. For example in Figure 5.9, a boundary loop of the cylinder is a planar curve and it is an ellipse. The ellipse has two symmetry planes orthogonal to the plane containing this curve. Back to the symmetry properties of a cylinder, valid symmetry planes either pass through the axis or are orthogonal to it. In this case, the symmetry plane of the ellipse coinciding with its minor axis is invalid. Consequently, the combination of symmetry constraints of faces with loop structures reduces the number of CSPs compared to independent processing of loops and surfaces.

Figure 5.9: Combination of face and its boundary loop symmetry properties.