Added new version of algorithm for computing normals for point clouds. Approx 8 times faster and works for clouds of a few millions of points...

This commit is contained in:
Paolo Cignoni 2012-11-08 15:33:32 +00:00
parent 26ee5e5246
commit 51a65af2c0
1 changed files with 154 additions and 0 deletions

View File

@ -0,0 +1,154 @@
#ifndef NORMAL_EXTRAPOLATION_H
#define NORMAL_EXTRAPOLATION_H
#include <vcg/space/index/kdtree/kdtree.h>
#include <vcg/space/fitting3.h>
#include <vcg/complex/algorithms/smooth.h>
namespace vcg {
namespace tri {
///
/** \addtogroup trimesh */
/*@{*/
/// Class of static functions to smooth and fair meshes and their attributes.
template <typename MeshType>
class PointCloudNormal {
public:
typedef typename MeshType::VertexType VertexType;
typedef typename MeshType::VertexType::CoordType CoordType;
typedef typename MeshType::VertexPointer VertexPointer;
typedef typename MeshType::VertexIterator VertexIterator;
typedef typename MeshType::ScalarType ScalarType;
class WArc
{
public:
WArc(VertexPointer _s,VertexPointer _t):src(_s),trg(_t),w(fabs(_s->cN()*_t->cN())){}
VertexPointer src;
VertexPointer trg;
float w;
bool operator< (const WArc &a) const {return w<a.w;}
};
static void ComputeUndirectedNormal(MeshType &m, int nn, float maxDist, KdTree<float> &tree,vcg::CallBackPos * cb=0)
{
tree.setMaxNofNeighbors(nn);
int cnt=0;
int step=m.vn/100;
for (VertexIterator vi=m.vert.begin();vi!=m.vert.end();++vi)
{
tree.doQueryK(vi->cP());
if(cb && (++cnt%step)==0) cb(cnt/step,"Fitting planes");
int neighbours = tree.getNofFoundNeighbors();
std::vector<CoordType> ptVec;
for (int i = 0; i < neighbours; i++)
{
int neightId = tree.getNeighborId(i);
if(Distance(vi->cP(),m.vert[neightId].cP())<maxDist)
ptVec.push_back(m.vert[neightId].cP());
}
Plane3f plane;
FitPlaneToPointSet(ptVec,plane);
vi->N()=plane.Direction();
}
}
static void AddNeighboursToHeap( MeshType &m, VertexPointer vp, KdTree<float> &tree, std::vector<WArc> &heap)
{
tree.doQueryK(vp->cP());
int neighbours = tree.getNofFoundNeighbors();
for (int i = 0; i < neighbours; i++)
{
int neightId = tree.getNeighborId(i);
if (neightId < m.vn && (&m.vert[neightId] != vp))
{
if(!m.vert[neightId].IsV())
{
heap.push_back(WArc(vp,&(m.vert[neightId])));
if(heap.back().w < 0.3f) heap.pop_back();
}
}
}
std::push_heap(heap.begin(),heap.end());
}
/*! \brief parameters for the normal generation
*/
struct Param
{
Param():
fittingAdjNum(10),
smoothingIterNum(0),
coherentAdjNum(8),
viewPoint(0,0,0),
useViewPoint(false)
{}
int fittingAdjNum; /// number of adjacent nodes used for computing the fitting plane
int smoothingIterNum; /// number of itaration of a simple normal smoothing (use the same number of ajdacent of fittingAdjNjm)
int coherentAdjNum; /// number of nodes used in the coherency pass
Point3f viewPoint; /// position of a viewpoint used to disambiguate direction
bool useViewPoint; /// if the position of the viewpoint has to be used.
};
static void Compute(MeshType &m, Param p, vcg::CallBackPos * cb)
{
tri::Allocator<MeshType>::CompactVertexVector(m);
if(cb) cb(1,"Building KdTree...");
VertexConstDataWrapper<MeshType> DW(m);
KdTree<float> tree(DW);
ComputeUndirectedNormal(m, p.fittingAdjNum, std::numeric_limits<ScalarType>::max(), tree,cb);
tri::Smooth<MeshType>::VertexNormalPointCloud(m,p.fittingAdjNum,p.smoothingIterNum,&tree);
if(p.coherentAdjNum==0) return;
tree.setMaxNofNeighbors(p.coherentAdjNum+1);
tri::UpdateFlags<MeshType>::VertexClearV(m);
std::vector<WArc> heap;
VertexIterator vi=m.vert.begin();
while(true)
{
// search an unvisited vertex
while(vi!=m.vert.end() && vi->IsV())
++vi;
if(vi==m.vert.end()) return;
if ( p.useViewPoint &&
( vi->N().dot(p.viewPoint- vi->P())<0.0f) )
vi->N()=-(*vi).N();
vi->SetV();
AddNeighboursToHeap(m,&*vi,tree,heap);
while(!heap.empty())
{
std::pop_heap(heap.begin(),heap.end());
WArc a = heap.back();
heap.pop_back();
if(!a.trg->IsV())
{
a.trg->SetV();
if(a.src->cN()*a.trg->cN()<0) a.trg->N()=-a.trg->N();
AddNeighboursToHeap(m,a.trg,tree,heap);
}
}
}
return;
}
};
}//end namespace vcg
}//end namespace vcg
#endif // NORMAL_EXTRAPOLATION_H