Difference between revisions of "User:Plussai/GSoC 2012 log"

From BRL-CAD
(6 July 2012)
(1 Augest 2012)
 
(21 intermediate revisions by the same user not shown)
Line 15: Line 15:
 
     ON_BrepTrimArray m_T;//trims
 
     ON_BrepTrimArray m_T;//trims
 
     ON_BrepVertexArray m_V;//vertexs
 
     ON_BrepVertexArray m_V;//vertexs
    figure:[http://brlcad.org/w/images/5/5e/Brep.jpg]
+
[[Image:brep.jpg]]   
 +
figure:[http://brlcad.org/w/images/5/5e/Brep.jpg]
  
 
== 9 May 2012 ==
 
== 9 May 2012 ==
Line 220: Line 221:
  
 
=== 6 July 2012 ===
 
=== 6 July 2012 ===
*add a trimcurve for ell.brep and hyp.brep manually to test
+
*add a trimcurve for ell.brep and hyp.brep manually to test trimming
left:top view     right:left view
+
*ell:left:top view,right:left view
[[Image:Ell1.png]]  [[Image:Ell2.png]]
+
[[Image:Ell1.png]]  [[Image:Ell2.png]]
left:local view
+
*ell:left:local view,right:uv region
  [[Image:Ell3.png]]
+
[[Image:Ell3.png]]        [[Image:Ell4.png]]
 +
 
 +
*hyp:
 +
[[Image:HYP1.png]]        [[Image:HYP2.png]]
 +
[[Image:HYP3.png]]        [[Image:HYP4.png]]
 +
 
 +
=== 8 July 2012 ===
 +
*continue to debug 'doTrimming'.There exists a situation that some trim leafs' bboxings overlap and it is difficult  to find the  nearest  trim  leaf above a surface leaf's bboxing in uv. Function 'prepTrim' mark the surface leaf in such situation as 'm_checkTrim = true'.But in some cases, the surface leaf is trimmed out.I changed the subdivision depth to 10 manually and tested the rhino0.s in mged.Some surface leaves which should be trimmed out are still remain.So, need a more accuracy method to check it.
 +
*image: some BBoxing in uv which should be trimmed are still remain(right bottom conner).
 +
[[Image:rhino.png]]        [[Image:rhino1.png]]
 +
 
 +
=== 10 July 2012 ===
 +
*find a algorithm for 'TrimLoop' triangulation,algorithm has the steps below:
 +
a) make the node in 'TrimLoop' to form a linked list.
 +
 
 +
b)for each triangle form by p0,p1,p2 successive in linked list,calculate the minimum inside angle.
 +
 
 +
c)find the triangle which has the max minimum inside angle, then record  the  triangle and remove the point p1 in the triangle
 +
 
 +
d)come to (a) until there only exists three points in the linked list
 +
 
 +
=== 12 July 2012 ===
 +
*Update my patches.
 +
 
 +
==''Week 10''==
 +
 
 +
=== 1 Augest 2012 ===
 +
*because of pressure of my paper schedule, I need to leave for a long time.I must
 +
say sorry that I will not complete the final-evaluation for gsoc.
 +
 
 +
==''Week 11''==
 +
 
 +
=== 6 Augest 2012 ===
 +
*I'v package my patch into three part after cleaning up and adding some comments.
 +
I think the subdividion is OK some more test must be done to ensure my trimming algorithm is accurate enough.
 +
*because of large amount of code need to be cleaned up and test but time is up, I decide to handle my paper firstly.
 +
*But I will come back and complete my work finally.

Latest revision as of 08:31, 28 October 2012

Community bonding period[edit]

7 May 2012[edit]

  • build the brl-cad svn on VM successfully.
  • build the brlcad-7.18.4 on windows(vs2005),but some link error occured from libtclcad.lib(solved go_init(..)undefined)

8 May 2012[edit]

  • get familiar with the Brep structure.
    ON_CurveArray m_C2;//Pointers to parameter space trimming curves (used by trims)
    ON_CurveArray m_C3;//Pointers to 3d curves (used by edges)
    ON_BrepEdgeArray m_E;//edges
    ON_BrepFaceArray m_F;//faces
    ON_BrepLoopArray m_L;//loops
    ON_SurfaceArray m_S;//Pointers to parametric surfaces (used by faces)
    ON_BrepTrimArray m_T;//trims
    ON_BrepVertexArray m_V;//vertexs

Brep.jpg figure:[1]

9 May 2012[edit]

  • solve the link problem when build brlcad_18.4 in vs2005.(function go_init(..) undefinded)
  • understand some important point about NURBS. 3D NURBS is not tensor product surface,but it can be represented into tensor product surface by mapping 3D NURBS into 4D using homogeneous coordinate.(The NURBS BOOK page 30)
  • circle Arc can be represented by vector{(1,1,0,1),(1,1,1,1),(2,0,2,2)}in yz plane

10 May 2012[edit]

  • build brlcad-SVN successfully on my ubantu.(PS:/sh/make_deb.sh is a good tool)
  • study some about Nonuniform Scaling of Surfaces.

12 May 2012[edit]

  • find that the ehy.brep is lack of face and loop in brep structure.add them to make a correct brep.
   int surfindex = (*b)->m_S.Count();
   (*b)->NewFace(surfindex - 1);
   int faceindex = (*b)->m_F.Count();
   (*b)->NewOuterLoop(faceindex-1);

13 May 2012[edit]

  • find that computing the longest edge size is in the bspile/nurb_tess.c and the nurbs is represented by the form of struct face_g_snurb that is different from the ON_BrepSurface.It needs to implement a new interface for the longest edge size.

15 May 2012[edit]

  • configure some plug-in unit for my vim (e.g ctags,cscope,winmanger)

19 May 2012[edit]

  • submit a patch of brep_debug.cpp.check if the index is in the range and string of letter.

Coding period[edit]

Week 1[edit]

21 May 2012[edit]

  • read the code of brep_debug.cpp which has many functions to debug the geometry in struct of Brep in mged.

23 May 2012[edit]

  • update the patch of brep_debug.cpp.add the sub command 'info T','info E' for display the information of trims and edges in brep.Also, make the 'info' dispaly the information in a range of indexed elements for keeping sync with the 'plot'.

25 May 2012[edit]

  • add a subcommand 'plot SCV' to display nurbs control point net

27 May 2012[edit]

  • waste a lot of time to test the opennurbs' API of:
 int CreateMesh(const ON_MeshParameters& mp, ON_SimpleArray<ON_Mesh*>& mesh_list) const; 
  • but,it seems difficult to make a good ON_MeshParameters to create mesh successfully.read the sources of opennurbs_mesh.cpp
ON_Mesh* ON_BrepFace::CreateMesh( 
           const ON_MeshParameters& mp,
           ON_Mesh* mesh
           ) const
{
 return 0;
}

Week 2[edit]

29 May 2012[edit]

  • do some works on converting nurbs curve and surface to bezier

31 May 2012[edit]

  • do some works on comput the intersection between a line and a bezier curve/surface
  • some materials: the nurbs book and Curves and Surfaces for Computer Aided Geometric Design: A Practical Guide.

1 June 2012[edit]

  • implement RT_ADD_VLIST(vhead, pt1, BN_VLIST_POINT_DRAW) in src/libdm/dm-plot.c but it seems that dm-plot.c has nothing relation with the brep_debug.cpp

2 June 2012[edit]

  • implement RT_ADD_VLIST(vhead, pt1, BN_VLIST_POINT_DRAW) in src/libdm/dm-ogl.c and upload the patch.
  • make progress for subcommand 'plot SCV' in mged.The patch can display not only the control net,but the control point.

Week 3[edit]

4 June 2012[edit]

  • one step that make 2d trimming curve into 3d can be skip because the BREP struct has the edge in 3d corresponding to the 3d trimming curve.
  • study the qd tree and kd tree for subdivision of nurbs surface in paper and the approximation error for the subdivision.

6 June 2012[edit]

  • figure out the coving triangles to avoid cracks when making boundary curve tessellation.
  • figure out the trimming curve tracing when splitting the trimming curve into bezier patch.(the active edge,active leaftree)

8 June 2012[edit]

  • read the opennurbs API and find the ON_NurbsCurve::MakePiecewiseBezier() can clamps ends and adds knots so the NURBS curve has bezier spans (all distinct knots have multiplitity = degree) ,the ON_NurbsSurface also has the same member method.It is a better way to convert trimming curve and surface into bezier
  • read the opennurbs API and find the ON_NurbsCurve::InsertKnot() can make the conversion between Nurbs and Bezier if setting the parameter multiplicity equal to degree.,the ON_NurbsSurface also has the same member method.

10 June 2012[edit]

  • add 'info TB' 'SB' to display information piecewise bezier of trimcurve and surface.patch has been upload
  • add 'plot TESS index plotres' to draw the trimesh of a nurbs surface. the tess has not taken apporimmation into consideration and just simply subdivide u,v into plotres steps;

Week 4[edit]

12 June 2012[edit]

  • upload the patch to implement the adaptive bezier subdivision and triangle mesh display which includes the function bezierApporixmation(...),bezierSubdivision_kdtree..
  • read opennurbs_ext.cpp,especially figuring out the function of
 BBNode* SurfaceTree::surfaceBBox(...)
 BBNode* initialBBox(...)
 BBNode*SurfaceTree::subdivideSurfaceByKnots(...)
 BBNode* SurfaceTree::subdivideSurface(...)
  • SurfaceTree::subdivideSurfaceByKnots and BBNode* SurfaceTree::subdivideSurface can help me to create the kd/qdtree as a represent of bezier subdivision.But the recursion termination condition is different.

14 June 2012[edit]

  • In function of SurfaceTree::subdivideSurface() at src/opennurbs_exp.cpp,line 1528 to 1531:
           const ON_Surface* surf = m_face->SurfaceOf();
            localsurf->FrameAt(u.Mid() - uq, v.Mid() - vq, frames[5]);
            localsurf->FrameAt(u.Mid() - uq, v.Mid() + vq, frames[6]); 
            localsurf->FrameAt(u.Mid() + uq, v.Mid() - vq, frames[7]);
            localsurf->FrameAt(u.Mid() + uq, v.Mid() + vq, frames[8]);

Interval u,v is in gloable area, So I think the four lines above need to be modified.

  • Read the BBNode structure in opennurbs_exp.cpp, BBNode will be used to represent the node in qd/kd tree and handling the trimcurve.

16 June 2012[edit]

  • upload a patch in opennurbs_ext.cpp
  • add three functions below to subdivide bezier,in opennurbs_ext.cpp
           BBNode* subdivideSurfaceToBezier(...)
           BBNode* subdivideBezier(...)
           double getBezierApporixmation(...)
  • add two functions below to create BezierTree root as the struct of BBNode
          bool setFace(ON_BrepFace* face, bool removeTrimmed);
          BBNode* createBezierTreeNode();

Week 5[edit]

18 June 2014[edit]

  • compare surface split between
           BBNode* subdivideSurfaceToBezier(...)
           BBNode* subdivideBezier(...)
and 
          BBNode*SurfaceTree::subdivideSurfaceByKnots(...)
          BBNode* SurfaceTree::subdivideSurface(...)

the original method seem to be unsymmetric when split ell.brep sphere.brep. But the new method has not taken the edge length into consideration,so the ratio of lengths in u,v direction may be too long or too small.

  • understand SurfaceTree::subdivideSurface() at src/opennurbs_exp.cpp,line 1528 to 1531, it comes to the same result when replacing localsurf into surf because the localsurf's domain is change along with the split.

20 June 2012[edit]

  • figure out the functions below in brep_debug.cpp which is used to draw trimmed BBNode of surface.But these functions seem not be able to remark the intersections of BRNode and trimcurve
           void drawBBNode(....)
           void drawisoU(....)
           void drawisoV(....)
           void drawisoUCheckForTrim(.....)
           void drawisoVCheckForTrim(.....)
  • figure out the trimflag of BBNode: m_trimmed,m_checkTrim and the function of preptrim() in opennurbs_ext.pp.

when a BBNode is remark as m_trimmed is true, the BBNode has been remove out from the surface. When a BBNode is remark as m_trimmed is true, the BBNode in uv region has intersections with the BRNode of trimcurve,so it need to be check again and compute the interections if needed.

22 June 2012[edit]

  • figure out the functions below in opennurbs_ext.h which can help me to comput the intersection between BBNode and trimcurve.
      fastf_t BANode<BA>::getCurveEstimateOfU(fastf_t v, fastf_t tol) const
      fastf_t BANode<BA>::getCurveEstimateOfV(fastf_t v, fastf_t tol) const
  • add a function below to return the parameter t of the intersection based on the two functions above
      template<class BA>
      fastf_t BANode<BA>::getCurveEstimateOfT(fastf_t value, bool direction, fastf_t& t, fastf_t tol) const
  • add a function to sort the BRNode in CurveTree by the t in parameter region.
     sortT(BRNode* first, BRNode* second);

24 June 2012[edit]

  • implement TrimNode and TrimLoop to represent the struct of untrimmed regions in parameter area.some functions included below:
      TrimLoop(ON_Interval m_u,ON_Interval m_v):tol(0.000001);//construct
      TrimNode* findPos(TrimNode* n);//find the position to insert for a new trimnode
      TrimNode* findSamePoint(const ON_2dPoint& pt);//find the node whose point in parameter is equal to pt
      bool linkOnTrimNode(TrimNode* pre,TrimNode* next);//link two trimnode which both within the bboxing
      bool insertInEdge(TrimNode* pos, TrimNode* n);//insert a new node
      TrimNode* travelLoop(TrimNode* n,int step);//travel from a trimnode to form a close loop

Week 6[edit]

26 June 2012[edit]

  • start to implement BVNode<BV>::doTrimming function in opennurbs_ext.cpp.
  • implement getLeavesIntersection(....) to find the bboxings of curveleaves which have intersections with the BBNode in UV region
  • calculate the intersection between edge of bboxing and trimcurve's leaves(bottom,top,left,right)
  • solve some special sitiuations: intersection on the conner of bboxing.

28 June 2012[edit]

  • continue to implement BVNode<BV>::doTrimming function in opennurbs_ext.cpp.
  • deal with a special situation: trim curve is collinear with the bboxing's edge. if the start or end point of trim curve is within bboxing's edge, it is regarded as the intersection between edge and trimcurve.
  • in function of BANode<BA>::isTrimmed(const ON_2dPoint& uv, fastf_t &trimdist) const , ' <= ' is used to compare float number. It may cause error because float's precision error,So I make some improve for it. This function will be useful in doTrimming.
  • link the innerNode in BANode's BBoxing of uv by call the function inline bool TrimLoop::linkOnTrimNode(TrimNode* pre,TrimNode* next)

30 June 2012[edit]

  • continue to implement BVNode<BV>::doTrimming function in opennurbs_ext.cpp.
  • deal with a special situation: the BVNode covers some trimcurve's bboxing which is remarked as m_checkTrim, however, the BVNode has no intersections with any trimcurve. the method which I use may is not precise enough,So I'll try to make a better one.
  • add two functions for BANode for judging if a point is contained in a BANode
 bool containsUV(const ON_2dPoint& uv);
 bool containsUVClose(const ON_2dPoint& uv);
  • debug for doTrimming() by subdivide plane surface manually in BBNode* SurfaceTree::subdivideSurface()

Week 7[edit]

2 July 2012[edit]

  • continue to implement BVNode<BV>::doTrimming function in opennurbs_ext.cpp.
  • deal with a special situation: if a trimcurve is collinear with a edge of BBNode's uv region and a endpoint of the trimcurve is just on the uv region's conner, the conner point will not to be the intersection.
  • add a flag 'unTrimmed ' in 'TrimLoop' to mark if a BBNode's uv region has been trimmed, situation of trimcurve lies on edge, endpoint just on edge are included;

4 July 2012[edit]

  • insert adjacent subdivide hit into every BBNode's uv region.
  • add member var into class BVNode<BV> and assign these member when subdividing surface
    BVNode<BV>* north;
    BVNode<BV>* south;
    BVNode<BV>* west;
    BVNode<BV>* east; 
    BVNode<BV>* parent;
  • add member function into class BVNode<BV> to get adjacent node
    BVNode<BV>* getAdjacentNorth();
    BVNode<BV>* getAdjacentSouth();
    BVNode<BV>* getAdjacentWest();
    BVNode<BV>* getAdjacentEast();
  • add member function into class BVNode<BV> to insert adjacent subdivide hit
    void insertAdjacentHit();

6 July 2012[edit]

  • add a trimcurve for ell.brep and hyp.brep manually to test trimming
  • ell:left:top view,right:left view

Ell1.png Ell2.png

  • ell:left:local view,right:uv region

Ell3.png Ell4.png

  • hyp:

HYP1.png HYP2.png HYP3.png HYP4.png

8 July 2012[edit]

  • continue to debug 'doTrimming'.There exists a situation that some trim leafs' bboxings overlap and it is difficult to find the nearest trim leaf above a surface leaf's bboxing in uv. Function 'prepTrim' mark the surface leaf in such situation as 'm_checkTrim = true'.But in some cases, the surface leaf is trimmed out.I changed the subdivision depth to 10 manually and tested the rhino0.s in mged.Some surface leaves which should be trimmed out are still remain.So, need a more accuracy method to check it.
  • image: some BBoxing in uv which should be trimmed are still remain(right bottom conner).

Rhino.png Rhino1.png

10 July 2012[edit]

  • find a algorithm for 'TrimLoop' triangulation,algorithm has the steps below:

a) make the node in 'TrimLoop' to form a linked list.

b)for each triangle form by p0,p1,p2 successive in linked list,calculate the minimum inside angle.

c)find the triangle which has the max minimum inside angle, then record the triangle and remove the point p1 in the triangle

d)come to (a) until there only exists three points in the linked list

12 July 2012[edit]

  • Update my patches.

Week 10[edit]

1 Augest 2012[edit]

  • because of pressure of my paper schedule, I need to leave for a long time.I must

say sorry that I will not complete the final-evaluation for gsoc.

Week 11[edit]

6 Augest 2012[edit]

  • I'v package my patch into three part after cleaning up and adding some comments.

I think the subdividion is OK some more test must be done to ensure my trimming algorithm is accurate enough.

  • because of large amount of code need to be cleaned up and test but time is up, I decide to handle my paper firstly.
  • But I will come back and complete my work finally.