cgal/Visibility_2/doc/Visibility_2/visibility_2.txt

162 lines
8.8 KiB
Plaintext

namespace CGAL {
/*!
\mainpage User Manual
\anchor Chapter_2D_Visibility_Computation
\cgalAutoToc
\authors Michael Hemmer, Kan Huang, Francisc Bungiu, Ning Xu
<!-- \cgalFigureBegin{example_figure,visibility-teaser.png}\cgalFigureEnd -->
\section visibility_2_introduction Introduction
This package provides functionality to compute the visibility region within polygons in two dimensions.
The package is based on the package \ref PkgArrangementOnSurface2 and uses CGAL::Arrangement_2
as the fundamental class to specify the input as well as the output.
Hence, a polygon \f$ P \f$ is represented by a bounded arrangement face \f$ f \f$
that does not have any isolated vertices and any edge that is adjacent to \f$ f \f$ separates \f$ f \f$ from another face.
Note that \f$ f \f$ may contain holes.
Similarly, a simple polygon is represented by a face without holes.
Given two points \f$ p \f$ and \f$ q \f$ in \f$ P \f$, they are said to be
visible to each other iff the segment \f$ pq \subset P \f$, where \f$ P \f$ is
considered to be closed, that is, \f$\partial P \subset P\f$.
For a query point \f$ q \in P \f$, the set of points that are visible from
\f$ q \f$ is defined as the visibility region of \f$ q \f$, denoted by \f$ V_q \f$
\subsection visibility_2_degeneracies Degeneracies and Regularization
\cgalFigureBegin{definition-fig, example1.png}
Non-regularized visibility and regularized visibility.
\cgalFigureEnd
As illustrated in \cgalFigureRef{definition-fig} (1) the visibility
region \f$ V_q \f$ of a query point \f$ q \f$ may not be a polygon.
In the figure, all labeled points are collinear, which implies that the
point \f$ c \f$ is visible to \f$ q \f$, that is,
the segment \f$ bc \f$ is part of \f$ V_q \f$.
We call such low dimensional features that are caused by degeneracies `needles`.
However, for many applications these needles are actually irrelevant.
Moreover, for some algorithms it is even more efficient to ignore needles
in the first place.
Therefore, this package offers also
functionality to compute the regularized visibility area
\f$ \overline{V_q} = closure(interior(V_q)) = (V_q\setminus\partial V_q) \cup \partial (V_q\setminus\partial V_q)\f$,
as shown in \cgalFigureRef{definition-fig} (2).
For more information about regularization, refer to Chapter \ref PkgBooleanSetOperations2.
\section visibility_2_classes Classes and Algorithms
Answering visibility queries is, in many ways, similar to answering point-location queries.
Thus, we use the same design used to implement \ref PkgArrangementOnSurface2 point location.
Each of the various visibility class templates employs a different
algorithm or \em strategy for answering
queries\cgalFootnote{The term \em strategy is borrowed from the
design-pattern taxonomy \cgalCite{cgal:ghjv-dpero-95}.
A \em strategy provides the means to define a family of algorithms,
each implemented by a separate class. All classes that implement the various
algorithms are made interchangeable, letting the algorithm in use vary according to the user choice.}.
Similar to the point-location case, some of the strategies require preprocessing. Thus, before a visibility object is used
to answer visibility queries, it must be attached to an arrangement object.
Afterwards, the visibility object observes changes to the attached arrangement.
Hence, it is possible to modify the arrangement after attaching the visibility object.
However, this feature should be used with caution as each change to the arrangement also
requires an update of the auxiliary data structures in the attached object.
An actual query is performed by giving the view point \f$ p \f$ and its containing face \f$ f \f$
(which must represent a valid polygon) to a visibility object.
For more details see the documentation of the overloaded member function `Visibility_2::compute_visibility()`.
The following models of the `Visibility_2` concept are provided:
<CENTER>
Class | Function | Preprocessing | Query |Algorithm
-------------------------------|-----------------------------------------------------|-------------------------------|-----------------------------------|-------------------------------
`Simple_polygon_visibility_2` | simple polygons | No |\f$ O(n) \f$ time and \f$ O(n) \f$ space | B. Joe and R. B. Simpson \cite bjrb-clvpa-87
`Rotational_sweep_visibility_2` | polygons with holes | No | \f$ O(n\log n) \f$ time and \f$ O(n) \f$ space | T. Asano \cite ta-aeafvpprh-85
`Triangular_expansion_visibility_2` | polygons with holes | \f$ O(n) \f$ time and \f$ O(n) \f$ space | \f$ O(nh) \f$ time and \f$ O(n) \f$ space. | Bungiu et al. \cite ecvp-bhhhk-14
</CENTER>
Where \f$ n \f$ denotes the number of vertices of \f$ f \f$ and \f$ h \f$ the number of holes+1.
\section benchmarks Running Time in Practice
\cgalFigureBegin{cathedral-fig, cathedral_2.png}
Example representing a cathedral.
\cgalFigureEnd
The left hand side of Figure \cgalFigureRef{cathedral-fig} depicts the outer boundary of a cathedral,
which is a simple polygon with 565 vertices.
The right hand side shows the cathedral also with its inner pillars, which is a polygon (with holes)
with 1153 vertices.
The following table shows the total running time consumption of the computation of
all visibility polygons for all vertices of the cathedral.
<CENTER>
Boundary Cathedral | total preprocessing time | total query time |
-------------------------------|-----------------------------------------------------|-------------------------------|
`Simple_polygon_visibility_2` | - | 0.38 |
`Rotational_sweep_visibility_2` | - | 1.01 |
`Triangular_expansion_visibility_2` | 0.01 | 0.06 |
</CENTER>
The second table shows the same for the complete cathedral.
The table does not report the time for `Simple_polygon_visibility_2`
as this algorithm can only handle simple polygons.
<CENTER>
Complete Cathedral | total preprocessing time | total query time |
-------------------------------|-----------------------------------------------------|-------------------------------|
`Rotational_sweep_visibility_2` | - | 1.91 |
`Triangular_expansion_visibility_2` | 0.01 | 0.04 |
</CENTER>
Thus, in general we recommend to use `Triangular_expansion_visibility_2` even if the polygon is simple.
The main advantage of the algorithm is its locality.
After the triangle that contains the query point is located in the triangulation,
the algorithm explores neighboring triangles, but only those that are actually seen.
In this sense the algorithm can be considered as output sensitive.
Note that the `Triangular_expansion_visibility_2` algorithm performs better on the full
cathedral since the additional pillars block the view early in many cases.
However, if the simple polygon is rather convex (i.e., nearly all boundary is seen) or
if only one (or very little) queries are required, using one of the algorithms that
does not require preprocessing is advantageous.
\section simple_polygon_visibility_example Example of Visibility in a Simple Polygon
The following example shows how to obtain the regularized and non-regularized visibility regions.
\cgalFigureBegin{simple_example, simple_example.png}
The visibility region of \f$ q \f$ in a simple polygon: (1) non-regularized visibility; and (2) regularized visibility.
\cgalFigureEnd
\cgalExample{Visibility_2/simple_polygon_visibility_2.cpp}
\section general_polygon_example Example of Visibility in a Polygon with Holes
The following example shows how to obtain the regularized visibility region using the model
`Triangular_expansion_visibility_2`, see \cgalFigureRef{general_polygon}.
The arrangement has six bounded faces and an unbounded face.
The query point \f$ q \f$ is on a vertex.
The red arrow denotes the halfedge \f$ \overrightarrow{pq} \f$,
which also identifies the face in which the visibility region is computed.
\cgalFigureBegin{general_polygon, general_polygon_example.png}
The visibility region of \f$ q \f$ in a polygon with two holes.
\cgalFigureEnd
\cgalExample{Visibility_2/general_polygon_example.cpp}
\section implementation_history Implementation History
This package was first developed during Google Summer of Code 2013:
Francisc Bungiu developed the `CGAL::Simple_polygon_visibility_2`,
Kan Huang developed the `CGAL::Rotational_sweep_visibility_2`,
and Michael Hemmer developed the `CGAL::Triangular_expansion_visibility_2`.
During Google Summer of Code 2014 Ning Xu fixed a bug in `CGAL::Simple_polygon_visibility_2` and improved the testsuite.
*/
}
// `Preprocessed_rotational_sweep_visibility_2` | polygons with holes | \f$ O(n^2) \f$ time and \f$ O(n^2) \f$ space | \f$ O(n) \f$ time and \f$ O(n) \f$ space | Takao Asano, Tetsuo Asano etc \cite aaghi-vpsesp-85