BambuStudio/libslic3r/Orient.cpp

545 lines
22 KiB
C++
Raw Permalink Normal View History

2024-12-20 06:44:50 +00:00
#include "Orient.hpp"
#include "Geometry.hpp"
#include <numeric>
#include <ClipperUtils.hpp>
#include <boost/geometry/index/rtree.hpp>
#include <tbb/parallel_for.h>
#if defined(_MSC_VER) && defined(__clang__)
#define BOOST_NO_CXX17_HDR_STRING_VIEW
#endif
#include <boost/log/trivial.hpp>
#include <boost/multiprecision/integer.hpp>
#include <boost/rational.hpp>
#undef MAX3
#define MAX3(a,b,c) std::max(std::max(a,b),c)
#undef MEDIAN
#define MEDIAN3(a,b,c) std::max(std::min(a,b), std::min(std::max(a,b),c))
#ifndef SQ
#define SQ(x) ((x)*(x))
#endif
namespace Slic3r {
namespace orientation {
struct CostItems {
float overhang;
float bottom;
float bottom_hull;
float contour;
float area_laf; // area_of_low_angle_faces
float area_projected; // area of projected 2D profile
float volume;
float area_total; // total area of all faces
float radius; // radius of bounding box
float height_to_bottom_hull_ratio; // affects stability, the lower the better
float unprintability;
CostItems(CostItems const & other) = default;
CostItems() { memset(this, 0, sizeof(*this)); }
static std::string field_names() {
return " overhang, bottom, bothull, contour, A_laf, A_prj, unprintability";
}
std::string field_values() {
std::stringstream ss;
ss << std::fixed << std::setprecision(1);
ss << overhang << ",\t" << bottom << ",\t" << bottom_hull << ",\t" << contour << ",\t" << area_laf << ",\t" << area_projected << ",\t" << unprintability;
return ss.str();
}
};
// A class encapsulating the libnest2d Nester class and extending it with other
// management and spatial index structures for acceleration.
class AutoOrienter {
public:
int face_count_hull;
OrientMesh *orient_mesh = NULL;
TriangleMesh* mesh;
TriangleMesh mesh_convex_hull;
Eigen::MatrixXf normals, normals_quantize, normals_hull, normals_hull_quantize;
Eigen::VectorXf areas, areas_hull;
Eigen::VectorXf is_apperance; // whether a facet is outer apperance
Eigen::MatrixXf z_projected;
Eigen::VectorXf z_max, z_max_hull; // max of projected z
Eigen::VectorXf z_median; // median of projected z
Eigen::VectorXf z_mean; // mean of projected z
std::vector<Vec3f> face_normals;
std::vector<Vec3f> face_normals_hull;
OrientParams params;
std::vector< Vec3f> orientations; // Vec3f == stl_normal
std::function<void(unsigned)> progressind = { }; // default empty indicator function
public:
AutoOrienter(OrientMesh* orient_mesh_,
const OrientParams &params_,
std::function<void(unsigned)> progressind_,
std::function<bool(void)> stopcond_)
{
orient_mesh = orient_mesh_;
mesh = &orient_mesh->mesh;
params = params_;
progressind = progressind_;
params.ASCENT = cos(PI - orient_mesh->overhang_angle * PI / 180); // use per-object overhang angle
// BOOST_LOG_TRIVIAL(info) << orient_mesh->name << ", angle=" << orient_mesh->overhang_angle << ", params.ASCENT=" << params.ASCENT;
// std::cout << orient_mesh->name << ", angle=" << orient_mesh->overhang_angle << ", params.ASCENT=" << params.ASCENT;
preprocess();
}
AutoOrienter(TriangleMesh* mesh_)
{
mesh = mesh_;
preprocess();
}
struct VecHash {
size_t operator()(const Vec3f& n1) const {
return std::hash<coord_t>()(int(n1(0)*100+100)) + std::hash<coord_t>()(int(n1(1)*100+100)) * 101 + std::hash<coord_t>()(int(n1(2)*100+100)) * 10221;
}
};
Vec3f quantize_vec3f(const Vec3f n1) {
return Vec3f(floor(n1(0) * 1000) / 1000, floor(n1(1) * 1000) / 1000, floor(n1(2) * 1000) / 1000);
}
Vec3d process()
{
orientations = { { 0,0,-1 } }; // original orientation
area_cumulation_accurate(face_normals, normals_quantize, areas, 10);
area_cumulation_accurate(face_normals_hull, normals_hull_quantize, areas_hull, 14);
add_supplements();
if(progressind)
progressind(20);
remove_duplicates();
if (progressind)
progressind(30);
std::unordered_map<Vec3f, CostItems, VecHash> results;
BOOST_LOG_TRIVIAL(info) << CostItems::field_names();
std::cout << CostItems::field_names() << std::endl;
for (int i = 0; i < orientations.size();i++) {
auto orientation = -orientations[i];
project_vertices(orientation);
auto cost_items = get_features(orientation, params.min_volume);
float unprintability = target_function(cost_items, params.min_volume);
results[orientation] = cost_items;
BOOST_LOG_TRIVIAL(info) << std::fixed << std::setprecision(4) << "orientation:" << orientation.transpose() << ", cost:" << std::fixed << std::setprecision(4) << cost_items.field_values();
std::cout << std::fixed << std::setprecision(4) << "orientation:" << orientation.transpose() << ", cost:" << std::fixed << std::setprecision(4) << cost_items.field_values() << std::endl;
}
if (progressind)
progressind(60);
typedef std::pair<Vec3f, CostItems> PAIR;
std::vector<PAIR> results_vector(results.begin(), results.end());
sort(results_vector.begin(), results_vector.end(), [](const PAIR& p1, const PAIR& p2) {return p1.second.unprintability < p2.second.unprintability; });
if (progressind)
progressind(80);
//To avoid flipping, we need to verify if there are orientations with same unprintability.
Vec3f n1 = {0, 0, 1};
auto best_orientation = results_vector[0].first;
for (int i = 1; i< results_vector.size()-1; i++) {
if (abs(results_vector[i].second.unprintability - results_vector[0].second.unprintability) < EPSILON && abs(results_vector[0].first.dot(n1)-1) > EPSILON) {
if (abs(results_vector[i].first.dot(n1)-1) < EPSILON*EPSILON) {
best_orientation = n1;
break;
}
}
else {
break;
}
}
BOOST_LOG_TRIVIAL(info) << std::fixed << std::setprecision(6) << "best:" << best_orientation.transpose() << ", costs:" << results_vector[0].second.field_values();
std::cout << std::fixed << std::setprecision(6) << "best:" << best_orientation.transpose() << ", costs:" << results_vector[0].second.field_values() << std::endl;
return best_orientation.cast<double>();
}
void preprocess()
{
int count_apperance = 0;
{
int face_count = mesh->facets_count();
auto its = mesh->its;
face_normals = its_face_normals(its);
areas = Eigen::VectorXf::Zero(face_count);
is_apperance = Eigen::VectorXf::Zero(face_count);
normals = Eigen::MatrixXf::Zero(face_count, 3);
normals_quantize = Eigen::MatrixXf::Zero(face_count, 3);
for (size_t i = 0; i < face_count; i++)
{
float area = its.facet_area(i);
normals.row(i) = face_normals[i];
normals_quantize.row(i) = quantize_vec3f(face_normals[i]);
areas(i) = area;
is_apperance(i) = (its.get_property(i).type == EnumFaceTypes::eExteriorAppearance);
count_apperance += (is_apperance(i)==1);
}
}
if (orient_mesh)
BOOST_LOG_TRIVIAL(debug) <<orient_mesh->name<< ", count_apperance=" << count_apperance;
// get convex hull statistics
{
mesh_convex_hull = mesh->convex_hull_3d();
//mesh_convex_hull.write_binary("convex_hull_debug.stl");
int face_count = mesh_convex_hull.facets_count();
auto its = mesh_convex_hull.its;
face_count_hull = mesh_convex_hull.facets_count();
face_normals_hull = its_face_normals(its);
areas_hull = Eigen::VectorXf::Zero(face_count);
normals_hull = Eigen::MatrixXf::Zero(face_count_hull, 3);
normals_hull_quantize = Eigen::MatrixXf::Zero(face_count_hull, 3);
for (size_t i = 0; i < face_count; i++)
{
float area = its.facet_area(i);
//We cannot use quantized vector here, the accumulated error will result in bad orientations.
normals_hull.row(i) = face_normals_hull[i];
normals_hull_quantize.row(i) = quantize_vec3f(face_normals_hull[i]);
areas_hull(i) = area;
}
}
}
void area_cumulation(const Eigen::MatrixXf& normals_, const Eigen::VectorXf& areas_, int num_directions = 10)
{
std::unordered_map<stl_normal, float, VecHash> alignments;
// init to 0
for (size_t i = 0; i < areas_.size(); i++)
alignments.insert(std::pair(normals_.row(i), 0));
// cumulate areas
for (size_t i = 0; i < areas_.size(); i++)
{
alignments[normals_.row(i)] += areas_(i);
}
typedef std::pair<stl_normal, float> PAIR;
std::vector<PAIR> align_counts(alignments.begin(), alignments.end());
sort(align_counts.begin(), align_counts.end(), [](const PAIR& p1, const PAIR& p2) {return p1.second > p2.second; });
num_directions = std::min((size_t)num_directions, align_counts.size());
for (size_t i = 0; i < num_directions; i++)
{
orientations.push_back(align_counts[i].first);
//orientations.push_back(its_face_normals(mesh->its)[i]);
BOOST_LOG_TRIVIAL(debug) << align_counts[i].first.transpose() << ", area: " << align_counts[i].second;
}
}
//This function is to make sure to return the accurate normal rather than quantized normal
void area_cumulation_accurate( std::vector<Vec3f>& normals_, const Eigen::MatrixXf& quantize_normals_, const Eigen::VectorXf& areas_, int num_directions = 10)
{
std::unordered_map<stl_normal, std::pair<std::vector<float>, Vec3f>, VecHash> alignments_;
Vec3f n1 = { 0, 0, 0 };
std::vector<float> current_areas = {0, 0};
// init to 0
for (size_t i = 0; i < areas_.size(); i++) {
alignments_.insert(std::pair(quantize_normals_.row(i), std::pair(current_areas, n1)));
}
// cumulate areas
for (size_t i = 0; i < areas_.size(); i++)
{
alignments_[quantize_normals_.row(i)].first[1] += areas_(i);
if (areas_(i) > alignments_[quantize_normals_.row(i)].first[0]){
alignments_[quantize_normals_.row(i)].second = normals_[i];
alignments_[quantize_normals_.row(i)].first[0] = areas_(i);
}
}
typedef std::pair<stl_normal, std::pair<std::vector<float>, Vec3f>> PAIR;
std::vector<PAIR> align_counts(alignments_.begin(), alignments_.end());
sort(align_counts.begin(), align_counts.end(), [](const PAIR& p1, const PAIR& p2) {return p1.second.first[1] > p2.second.first[1]; });
num_directions = std::min((size_t)num_directions, align_counts.size());
for (size_t i = 0; i < num_directions; i++)
{
orientations.push_back(align_counts[i].second.second);
BOOST_LOG_TRIVIAL(debug) << align_counts[i].second.second.transpose() << ", area: " << align_counts[i].second.first[1];
}
}
void add_supplements()
{
std::vector<Vec3f> vecs = { {0, 0, -1} ,{0.70710678, 0, -0.70710678},{0, 0.70710678, -0.70710678},
{-0.70710678, 0, -0.70710678},{0, -0.70710678, -0.70710678},
{1, 0, 0},{0.70710678, 0.70710678, 0},{0, 1, 0},{-0.70710678, 0.70710678, 0},
{-1, 0, 0},{-0.70710678, -0.70710678, 0},{0, -1, 0},{0.70710678, -0.70710678, 0},
{0.70710678, 0, 0.70710678},{0, 0.70710678, 0.70710678},
{-0.70710678, 0, 0.70710678},{0, -0.70710678, 0.70710678},{0, 0, 1} };
orientations.insert(orientations.end(), vecs.begin(), vecs.end());
}
/// <summary>
/// remove duplicate orientations
/// </summary>
/// <param name="tol">tolerance. default 0.01 =sin(0.57\degree)</param>
void remove_duplicates(double tol=0.0000001)
{
for (auto it = orientations.begin()+1; it < orientations.end(); )
{
bool duplicate = false;
for (auto it_ok = orientations.begin(); it_ok < it; it_ok++)
{
if (it_ok->isApprox(*it, tol)) {
duplicate = true;
break;
}
}
const Vec3f all_zero = { 0,0,0 };
if (duplicate || it->isApprox(all_zero,tol))
it = orientations.erase(it);
else
it++;
}
}
void project_vertices(Vec3f orientation)
{
int face_count = mesh->facets_count();
auto its = mesh->its;
z_projected.resize(face_count, 3);
z_max.resize(face_count, 1);
z_median.resize(face_count, 1);
z_mean.resize(face_count, 1);
for (size_t i = 0; i < face_count; i++)
{
float z0 = its.get_vertex(i,0).dot(orientation);
float z1 = its.get_vertex(i,1).dot(orientation);
float z2 = its.get_vertex(i,2).dot(orientation);
z_projected(i, 0) = z0;
z_projected(i, 1) = z1;
z_projected(i, 2) = z2;
z_max(i) = MAX3(z0,z1,z2);
z_median(i) = MEDIAN3(z0,z1,z2);
z_mean(i) = (z0 + z1 + z2) / 3;
}
z_max_hull.resize(mesh_convex_hull.facets_count(), 1);
its = mesh_convex_hull.its;
for (size_t i = 0; i < z_max_hull.rows(); i++)
{
float z0 = its.get_vertex(i,0).dot(orientation);
float z1 = its.get_vertex(i,1).dot(orientation);
float z2 = its.get_vertex(i,2).dot(orientation);
z_max_hull(i) = MAX3(z0, z1, z2);
}
}
static Eigen::VectorXi argsort(const Eigen::VectorXf& vec, std::string order="ascend")
{
Eigen::VectorXi ind = Eigen::VectorXi::LinSpaced(vec.size(), 0, vec.size() - 1);//[0 1 2 3 ... N-1]
std::function<bool(int, int)> rule;
if (order == "ascend") {
rule = [vec](int i, int j)->bool {
return vec(i) < vec(j);
};
}
else {
rule = [vec](int i, int j)->bool {
return vec(i) > vec(j);
};
}
std::sort(ind.data(), ind.data() + ind.size(), rule);
return ind;
//sorted_vec.resize(vec.size());
//for (int i = 0; i < vec.size(); i++) {
// sorted_vec(i) = vec(ind(i));
//}
}
// previously calc_overhang
CostItems get_features(Vec3f orientation, bool min_volume = true)
{
CostItems costs;
costs.area_total = mesh->bounding_box().area();
costs.radius = mesh->bounding_box().radius();
// volume
costs.volume = mesh->stats().volume > 0 ? mesh->stats().volume : its_volume(mesh->its);
float total_min_z = z_projected.minCoeff();
// filter bottom area
auto bottom_condition = z_max.array() < total_min_z + this->params.FIRST_LAY_H - EPSILON;
auto bottom_condition_hull = z_max_hull.array() < total_min_z + this->params.FIRST_LAY_H - EPSILON;
auto bottom_condition_2nd = z_max.array() < total_min_z + this->params.FIRST_LAY_H/2.f - EPSILON;
//The first layer is sliced on half of the first layer height.
//The bottom area is measured by accumulating first layer area with the facets area below first layer height.
//By combining these two factors, we can avoid the wrong orientation of large planar faces while not influence the
//orientations of complex objects with small bottom areas.
costs.bottom = bottom_condition.select(areas, 0).sum()*0.5 + bottom_condition_2nd.select(areas, 0).sum();
// filter overhang
Eigen::VectorXf normal_projection(normals.rows(), 1);// = this->normals.dot(orientation);
for (size_t i = 0; i < normals.rows(); i++)
{
normal_projection(i) = normals.row(i).dot(orientation);
}
auto areas_appearance = areas.cwiseProduct((is_apperance * params.APPERANCE_FACE_SUPP + Eigen::VectorXf::Ones(is_apperance.rows(), is_apperance.cols())));
auto overhang_areas = ((normal_projection.array() < params.ASCENT) * (!bottom_condition_2nd)).select(areas_appearance, 0);
Eigen::MatrixXf inner = normal_projection.array() - params.ASCENT;
inner = inner.cwiseMin(0).cwiseAbs();
if (min_volume)
{
Eigen::MatrixXf heights = z_mean.array() - total_min_z;
costs.overhang = (heights.array()* overhang_areas.array()*inner.array()).sum();
}
else {
costs.overhang = overhang_areas.array().cwiseAbs().sum();
}
{
// contour perimeter
#if 1
// the simple way for contour is even better for faces of small bridges
costs.contour = 4 * sqrt(costs.bottom);
#else
float contour = 0;
int face_count = mesh->facets_count();
auto its = mesh->its;
int contour_amout = 0;
for (size_t i = 0; i < face_count; i++)
{
if (bottom_condition(i)) {
Eigen::VectorXi index = argsort(z_projected.row(i));
stl_vertex line = its.get_vertex(i, index(0)) - its.get_vertex(i, index(1));
contour += line.norm();
contour_amout++;
}
}
costs.contour += contour + params.CONTOUR_AMOUNT * contour_amout;
#endif
}
// bottom of convex hull
costs.bottom_hull = (bottom_condition_hull).select(areas_hull, 0).sum();
// low angle faces
auto normal_projection_abs = normal_projection.cwiseAbs();
Eigen::MatrixXf laf_areas = ((normal_projection_abs.array() < params.LAF_MAX) * (normal_projection_abs.array() > params.LAF_MIN) * (z_max.array() > total_min_z + params.FIRST_LAY_H)).select(areas, 0);
costs.area_laf = laf_areas.sum();
// height to bottom_hull_area ratio
//float total_max_z = z_projected.maxCoeff();
//costs.height_to_bottom_hull_ratio = SQ(total_max_z) / (costs.bottom_hull + 1e-7);
return costs;
}
float target_function(CostItems& costs, bool min_volume)
{
float cost=0;
float bottom = costs.bottom;//std::min(costs.bottom, params.BOTTOM_MAX);
float bottom_hull = costs.bottom_hull;// std::min(costs.bottom_hull, params.BOTTOM_HULL_MAX);
if (min_volume)
{
float overhang = costs.overhang / 25;
cost = params.TAR_A * (overhang + params.TAR_B) + params.RELATIVE_F * (/*costs.volume/100*/overhang*params.TAR_C + params.TAR_D + params.TAR_LAF * costs.area_laf * params.use_low_angle_face) / (params.TAR_D + params.CONTOUR_F * costs.contour + params.BOTTOM_F * bottom + params.BOTTOM_HULL_F * bottom_hull + params.TAR_E * overhang + params.TAR_PROJ_AREA * costs.area_projected);
}
else {
float overhang = costs.overhang;
cost = params.RELATIVE_F * (costs.overhang * params.TAR_C + params.TAR_D + params.TAR_LAF * costs.area_laf * params.use_low_angle_face) / (params.TAR_D + params.CONTOUR_F * costs.contour + params.BOTTOM_F * bottom + params.BOTTOM_HULL_F * bottom_hull + params.TAR_PROJ_AREA * costs.area_projected);
}
cost += (costs.bottom < params.BOTTOM_MIN) * 100;// +(costs.height_to_bottom_hull_ratio > params.height_to_bottom_hull_ratio_MIN) * 110;
costs.unprintability = costs.unprintability = cost;
return cost;
}
};
void _orient(OrientMeshs& meshs_,
const OrientParams &params,
std::function<void(unsigned, std::string)> progressfn,
std::function<bool()> stopfn)
{
if (!params.parallel)
{
for (size_t i = 0; i != meshs_.size(); ++i) {
auto& mesh_ = meshs_[i];
progressfn(i, mesh_.name);
//auto progressfn_i = [&](unsigned cnt) {progressfn(cnt, "Orienting " + mesh_.name); };
AutoOrienter orienter(&mesh_, params, /*progressfn_i*/{}, stopfn);
mesh_.orientation = orienter.process();
Geometry::rotation_from_two_vectors(mesh_.orientation, { 0,0,1 }, mesh_.axis, mesh_.angle, &mesh_.rotation_matrix);
BOOST_LOG_TRIVIAL(info) << std::fixed << std::setprecision(3) << "v,phi: " << mesh_.axis.transpose() << ", " << mesh_.angle;
//flush_logs();
}
}
else {
tbb::parallel_for(tbb::blocked_range<size_t>(0, meshs_.size()), [&meshs_, &params, progressfn, stopfn](const tbb::blocked_range<size_t>& range) {
for (size_t i = range.begin(); i != range.end(); ++i) {
auto& mesh_ = meshs_[i];
progressfn(i, mesh_.name);
AutoOrienter orienter(&mesh_, params, {}, stopfn);
mesh_.orientation = orienter.process();
Geometry::rotation_from_two_vectors(mesh_.orientation, { 0,0,1 }, mesh_.axis, mesh_.angle, &mesh_.rotation_matrix);
mesh_.euler_angles = Geometry::extract_euler_angles(mesh_.rotation_matrix);
BOOST_LOG_TRIVIAL(debug) << "rotation_from_two_vectors: " << mesh_.orientation << "; " << mesh_.axis << "; " << mesh_.angle << "; euler: " << mesh_.euler_angles.transpose();
}});
}
}
void orient(OrientMeshs & arrangables,
const OrientMeshs &excludes,
const OrientParams & params)
{
auto &cfn = params.stopcondition;
auto &pri = params.progressind;
_orient(arrangables, params, pri, cfn);
}
void orient(ModelObject* obj)
{
auto m = obj->mesh();
AutoOrienter orienter(&m);
Vec3d orientation = orienter.process();
Vec3d axis;
double angle;
Geometry::rotation_from_two_vectors(orientation, { 0,0,1 }, axis, angle);
obj->rotate(angle, axis);
obj->ensure_on_bed();
}
void orient(ModelInstance* instance)
{
auto m = instance->get_object()->mesh();
AutoOrienter orienter(&m);
Vec3d orientation = orienter.process();
Vec3d axis;
double angle;
Matrix3d rotation_matrix;
Geometry::rotation_from_two_vectors(orientation, { 0,0,1 }, axis, angle, &rotation_matrix);
instance->rotate(rotation_matrix);
}
} // namespace arr
} // namespace Slic3r