Galran/adding projected paraxial lenses #258
Conversation
Fixes #242
added first set of functions for computing fraction of beam energy which passes from the paraxial display lens through an aperture.
Fixes #242
adding code to compute beamenergy.
Fixes #242
added function to compute virtual point for paraxial lens. Modified the shape types that can be used for ParaxialLens so they share a common distancefrom plane function.
Fixes #242
Added functions for computing area of spherical triangles and convex polygons. Spherical polygon area computation does not appear to be correct.
…ygons. Fix tomorrow.
Fixes #242
spherical area for spherical triangles and spherical polygons appears to work. Takes approx. 300ns to compute the area of a four sided spherical polygon. Longer than I would have expected. Optimize later if necessary.
Fixes #242
simplified calculation of area of spherical polygon. now runs about 3x faster 76ns for 4 sided polygon whereas before it was 250ns.
Fixes #242
added functions to project pupil vertices onto lens plane, to intersect the projected pupil polygon with the lens aperture polygon, and to compute the beam energy.
…play. added activebox function which computes which pixels can possibly illuminate a planar polygon located in world space. Used to minimize number of pixels for which have to compute beamenergy.
…tionality needed and has many other features that don't make sense in the paraxial case.
Fixes #242
adding display to lensletassembly.
Fixes #242
moved interface, distancefromplane, normal from all of the surfaces that inherit from PlanarShapes into a generic function defined for PlanarShapes.
added transforms to objects in LesletAssembly so points can be put in the correct coordinate frame.
Fixes #242
wrote test programes for virtual point and projection. Moved beam energy code into lenslet assembly.
Fixes #242
fixed error with missing beamenergy.jl file
Fixes #242
modified planar shapes to return vertices in an SMatrix instead of as SVector{SVector} because LazySet.jl only works with SMatrix and Vector.
Still need to change ConvexPolygon to this form.
started work on testintersection function.
Fixes #242
modified ConvexPolygon to represent polygon points as a matrix. This data structure is more compatible with LazySets which is used to many geometric calculations.
…o represent polygon points. Passed unit tests.
Fixes #242
added * operator for points represented as matrices for better compatibility with LazySets.jl. Unfortunately Julia compiler wasn't smart enough to compile efficient code using high level forms so had to write out the matrix multply loop.
Fixes #242
fixed Ellipse, Hexagon, Rectangle so they all have zero allocations for vertices3D
Fixes #242
fixed Ellipse, Hexagon, Rectangle so they all have zero allocations for vertices3D
…he T and N in the declaration after encountering some problems in the way ParaxialLens is using the ConvexPolygon
… the creation of Paraxial MLAs based on projection onto a Plane, Cylinder or Sphere
…enses # Conflicts: # src/Geometry/Primitives/NonCSG/ConvexPolygon.jl
Codecov Report
@@ Coverage Diff @@
## main #258 +/- ##
==========================================
- Coverage 54.97% 52.95% -2.03%
==========================================
Files 72 77 +5
Lines 6708 6965 +257
==========================================
Hits 3688 3688
- Misses 3020 3277 +257
Continue to review full report at Codecov.
|
| sum = Base.sum(points) | ||
| centroid = Vec3(sum * (1.0 / (length(points)))) | ||
|
|
||
| # Calc full 3x3 covariance matrix, excluding symmetries: |
There was a problem hiding this comment.
there is a function Statistics.cov which I think computes the same thing. Would be simpler to use it, unless you are doing something non-standard. We already use the Statistics package so this wouldn't add a dependency.
Might work more efficiently if points were represented as an SMatrix{3,N}. Otherwise it looks like you'd have to convert from a Vector{SVector} to SMatrix.
| dir = Vec3(xz*yz - xy*zz, det_y, xy*xz - yz*xx) | ||
| else | ||
| dir = Vec3(xy*yz - xz*yy,xy*xz - yz*xx,det_z) | ||
| end |
There was a problem hiding this comment.
Is this computation time sensitive? If not then it might be simpler to use the SVD on a matrix of the points and use the singular vectors as the coordinate frame of the best fitting plane. This would only be a few lines of code:
y= x .- mean(x,dims=2)
U,_,_ = svd(y)
then choose which of the U vectors you want to align with the x,y,z axes. Or just return the smallest singular vector as the normal to the plane and compute the other two axes elsewhere.
Built in svd on 3x100 Matrix takes 10usec. If you use MMatrix instead it takes 6us. Or you can use my JacobiSVD (available on request) on MMatrix and it takes 3usec.
There was a problem hiding this comment.
i'm not sure. i had some stability issues in the past with SVG in extreme cases - not sure if because of numerical errors or other reason. will check it out.
There was a problem hiding this comment.
@galran if you have stability issues then try my JacobiSVD code. It is supposed to compute small singular vectors to higher accuracy than the more common algorithms. And it is twice as fast as the Julia built in SVD.
| c = centroid(poly) | ||
|
|
||
| l2w = local2world(poly.local_frame) | ||
| len = length(poly.local_points) | ||
| len = size(poly.local_points)[2] |
There was a problem hiding this comment.
Is this right? local_points is a Vector{SVector{2,T}}.
size(poly.local_points)[2]
gives an indexing error on my machine.
There was a problem hiding this comment.
poly.local_points is an SMatrix - that is why i changed it from length to size
| center(h::Shape) = h._center | ||
| points(h::Shape) = h._points | ||
| coordinates(h::Shape) = h._coordinates | ||
|
|
There was a problem hiding this comment.
This is more idiomatic Julia and is simpler
getshapes(::Type{Any};blah blah) = @error "Unknown Type ...."
function get_shapes(::Type{Hexagon}, resolution::Tuple{Int,Int}=(2,2), radius=1.5)::Vector{Shape{2, Float64}}
cells = Repeat.hexcellsinbox(resolution[1],resolution[2])
hexbasis = Repeat.HexBasis1()
basic_tile = Repeat.tilevertices(hexbasis) * radius
res = Vector{Shape{2, Float64}}(undef, 0)
for c in cells
center = SVector(hexbasis[c[1], c[2]]) * radius
points = [(SVector(p...) + center) for p in eachrow(basic_tile)]
center = center
push!(res, Shape(center, points, c))
end
return res
end
function get_shapes(::Type{Rectangle}, resolution::Tuple{Int,Int}=(2,2), size=1.5)::Vector{Shape{2, Float64}}
if (typeof(size) == Float64)
w = h = size
elseif (typeof(size) == Tuple{Float64, Float64})
w, h = size
else
@error "Unsupported size format [$size] - can be either a Float64 for a square or (Float64, Float64) for a rectangle"
end
cells = [[c for c in Iterators.product(-resolution[1]:resolution[1], -resolution[2]:resolution[2])]...]
# hexbasis = Repeat.HexBasis1()
# basic_tile = Repeat.tilevertices(hexbasis) * radius
basic_tile = [w h; w -h; -w -h; -w h]
res = Vector{Shape{2, Float64}}(undef, 0)
for c in cells
center = SVector(Float64(c[1])*w*2.0, Float64(c[2])*h*2.0)
points = [(SVector(p...) + center) for p in eachrow(basic_tile)]
center = center
push!(res, Shape(center, points, c))
end
return res
end
|
|
||
|
|
||
| function project(shapes::Vector{Shape{2, T}}, csg::OpticSim.CSGTree{T})::Vector{Shape{3, T}} where {T<:Real} | ||
| ray_origin_distance = -10.0 |
There was a problem hiding this comment.
should ray_origin_distance be an argument to this function? Or does it make sense for it to always be the same? I haven't read the code carefully enough to figure this out. Just asking, not a criticism.
There was a problem hiding this comment.
well, the assumption is that the original shapes are on the XY plane and the CSG surface to project is in the positive Z half space (above the XY plane), so it is safe to shoot from a negative Z position upwards.
IF the CSG surface is complex and moves into the negative Z it might present a problem.
| @@ -21,7 +21,7 @@ The local frame defines the plane (spans by the right and up vectors) with the p | |||
| the local_polygon_points are given with respect to the local frame and are 2D points. | |||
| NOTE: This class uses static vectors to hold the points which will lead to more efficient performance, but should not be used with polygons with more than 20-30 points. | |||
| """ | |||
| struct ConvexPolygon{N,T<:Real} <: PlanarShape{T} | |||
| struct ConvexPolygon{T<:Real, N} <: PlanarShape{T} | |||
There was a problem hiding this comment.
was there some reason you changed the order of T and N? I did it this way so it would be similar to the StaticArrays type declarations. Does having N first cause problems with the code?
There was a problem hiding this comment.
reversing this change after our conversation.
…f the system also removed a couple of commented functions
changed the get_shapes function to rely on the Type instead of a symbol Added an example function: OpticSim.ParaxialAnalysis.HeadEye.Examples.example1()
# Conflicts: # src/Geometry/Primitives/NonCSG/ConvexPolygon.jl # src/Geometry/Primitives/NonCSG/PlanarShape.jl # src/Optical/ParaxialAnalysis/ParaxialAnalysis.jl
added license header
added the HeadEye module inside the ParaxialAnalysis module to support the creation of Paraxial MLAs based on projection onto a Plane, Cylinder or Sphere.
This is a draft version - comments and exports will be added after initial integration to better support the tools and interfaces we will need.