PROCESS FOR CREATING AN OPTICAL COMPONENT FOR GENERATING, FROM A GIVEN LIGHT SOURCE, A GIVEN NEAR-FIELD ILLUMINATION

Abstract

Disclosed is a method for fabricating an optical component that is configured so as to generate on an illumination target in the near-field an illumination that has a determined pattern according to which each point (i) of the illumination target receives a quantity of light (alpha_i) via an illumination generated by an illumination light source that is incident on the optical component, which is placed between the illuminating light source and the illumination target.

Description

BACKGROUND OF THE INVENTION

Field of the Invention

The present invention relates to the creation of optical components, in particular the creation of an optical component in order to generate, from a given light source, a given illumination on a near field target (finite distance).

Description of the Related Art

The field of non-imaging optics (or anidolic optics, non-imaging optics being the English designation) deals with the design of optical components with the goal being to transfer the optical radiation emitted by a light source onto a prescribed target. This issue lies at the core of many applications for which it is desired to optimize the use of light energy by reducing light loss or light pollution. Such problems arise in particular but not exclusively in the design of vehicle headlight beams, public lighting, solar furnaces, or indeed even, for example, for hydroponic cultivation.

For a far-field target (at infinity), the creation of optical components (for example mirrors and lenses) that provide the ability to transport the light energy emitted by a light source to this far field target is a known technical approach: given a source of collimated light (like the sun) or point source (emitted by a point), and a desired far-field illumination target, the optical components created have the property of reflecting (in the case of mirrors) or of refracting (in the case of lenses) the light source on the illumination target (such as a black and white image). The published scientific paper “Light in power: A general and parameter free algorithm for caustic design”, Quentin Merigot, Jocelyn Meyron, Boris Thibert, version 1, 31/07/2017, arXiv:1708.04820, thus describes a method for creating an optical component in order to generate a given illumination from a given light source, on a far-field target.

There are also existing methods for creating optical components in order to generate a given illumination from a given light source, on a near field target. The published scientific paper “High contrast computational caustic design”, Yuliy Schwartzburg, Romain Testuz, Andrea Tagliasacchi, Mark Pauly, ACM, Transactions on Graphics (TOG) 33, 4 (2014), 74, presents the construction of good quality lenses in the case of a collimated light source (that is to say parallel rays, like those of the sun) and a near field (at finite distance) target. The method used is based on heuristics and requires the adjustment of a plurality of parameters, which turns out to be complex in practice.

However, there currently exists no method that is simple, easy to configure or even parameter free, which provides the ability to create an optical component in order to generate on a near field (at a finite distance) target a given illumination from a given light source.

SUMMARY OF THE INVENTION

The present invention thus proposes to meet this need by providing a method based on the iterative solving of a series of far-field non-imaging (anidolic) optics problems.

The Applicant indeed found in a surprising manner that solving a series of far-field non-imaging optics problems converged to a solution of near-field non-imaging optics.

Therefore, an object of the present invention is a method for fabricating an optical component that is configured so as to generate on a near-field illumination target an illumination that has a determined pattern according to which each point (i) of the illumination target receives a quantity of light (alpha_i) via an illumination originating from an illumination light source which is incident on the optical component placed between the illumination light source and the illumination target, characterized in that the method includes the following steps:

• • (a) positioning an origin point between the illumination light source and the illumination target;
  • (b) for each point (i) of the illumination target, computing a direction (dir_i) which corresponds to the direction of the vector connecting the origin point and the point (i) on the illumination target;
  • (c) positioning a reference point between the origin point and the illumination target;
  • (d) creating the optical component whose surface passes by the reference point and which, when the illumination originating from the illumination light source is incident on the optical component, generates an illumination comprised of all the illuminations of direction (dir_i) and quantity of light (alpha_i);
  • (e) for each point (z_i) of the optical component that generates the illumination of direction (dir_i) and quantity of light (alpha_i), computing a corrected direction (dirc_i) which corresponds to the direction of the vector connecting the point (z_i) and the point (i) on the illumination target;
  • (f) determining whether or not, for each point (i), the difference between the direction (dir_i) and the corrected direction (dirc_i) satisfies a predetermined criterion;
  • (g) if, for each point (i), the difference between the direction (dir_i) and the corrected direction (dirc_i) satisfies the predetermined criterion, fabricating the optical component that corresponds to the optical component,
  • (h) if, for each point (i), the difference between the direction (dir_i) and the corrected direction (dirc_i) does not satisfy the predetermined criterion, reiterating the steps (d) to (f) by substituting (dir_i) with (dirc_i).

The creation step (d) for creating an optical component pertains to the creation of an optical component for a far-field target. Any appropriate method may be used to create this optical component for a far field target. Preferably, but not exclusively, the method described in the published scientific paper “Light in power: A general and parameter free algorithm for caustic design” will be used, this method being described in greater detail in the section presenting the detailed description of the invention.

According to one embodiment, the predetermined criterion of the step (f) is that on each point (i) the mean of the norms of the differences between the direction (dir_i) and the corrected direction (dirc_i) is less than a predetermined value, for example but not exclusively less than 10⁻⁶, the difference being calculated by subtracting (dir_i)-(dirc_i).

More precisely, the calculation of this global criterion for N points is given by:

(Norm((dir_1)−(dirc_1))+Norm((dir_2)−(dirc_2))+ . . . +Norm((dir_N)−(dirc_N)))/N,

the norm of a vector v( ) with three coordinates being given by Norm (v)=square root of (v1*v1+v2*v2+v3*v3)

According to one embodiment, the illumination light source is one of a point light source and a collimated light source.

According to one embodiment, the optical component is one of a concave optical component and a convex optical component.

According to one embodiment, the optical component is one of a mirror and a lens.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to better illustrate the object of the present invention, a particular embodiment will be described here below, with reference to the accompanying drawings.

In these drawings:

FIG. 1 is a diagram illustrating a step of the method according to the invention;

FIG. 2 is a diagram illustrating another step of the method according to the invention;

FIG. 3 is a diagram illustrating yet another step of the method according to the invention;

FIG. 4 is a simulation of different iterations of the method of the invention, for a concave lens and a convex mirror.

DESCRIPTION OF THE PREFERRED EMBODIMENTS'

As a preamble to the description of the method of the present invention, the method that makes it possible to create an optical component, in order to generate a given illumination from a given light source on a far-field target (at infinity) will be described.

As previously indicated here above, this method, which corresponds to the method described in the published scientific paper “Light in power: A general and parameter free algorithm for caustic design”, is not the only method that can be used and indeed it is possible for any analogous or equivalent method to be used in the context of the present invention, such that the invention therefore is not in any way limited to this particular method.

In order to illustrate the method, a plurality of various mirror or lens design problems that occur in non-imaging (anidolic) optics are presented. In all the problems, there is a given light source (either collimated or point source) and a desired illumination on a far-field target, subsequent to reflection or refraction. The goal is to design the geometry of a mirror or a lens that transports the energy emitted by the light source on to the target, with the multiple refractions and reflections not being taken into account. Even though the respective problems considered are different from each other, they share a common structure that corresponds to the equation referred to as generalized Monge-Ampere equation, of which the discrete version is given by Equation (1):

∀i∈{1, . . . ,n}∫_Vi(ψ)ρ(x)dx=σ_i  (1)

The method is illustrated for lenses and mirrors, whether concave or convex, with point or collimated light sources.

1—Mirror Design

1.1—Convex Mirror for a Collimated Light Source

For this first problem, the light source is collimated: the light source may be encoded by a light intensity function ρ over a 2D domain. For the sake of simplicity, it is assumed that the domain is included in ²×{0}⊂³ and that all the rays are parallel to the direction z (vertical) and directed in the upward direction. The desired target illumination is in the far field (at infinity) and is described by a set of intensity values σ=(σ_i)_1≤i≤N supported on a finite set of directions Y={y_i, . . . , y_n} included in the unit sphere ². The problem is to find the surface of a mirror that sends the source intensity ρ to the target intensity σ. This problem corresponds to a Monge-Ampére equation in the 2D plane, which corresponds to a quadratic optimal transport problem.

Given that the number of reflected directions is finite, the mirror surface is composed of a finite number of planar facets. R_ψ is defined as the graph of a convex function of the form xmax_ix|p_i−ψ_i, where x|y denotes the scalar product of x and y; for all i∈{1, . . . , N}, p_i is the orthogonal projection of a unit normal of the plane (referred to as slope in the sections that follow) that reflects according to Snell-Descartes law the vertical ray (0,0,1) towards the direction y_i, and ψ_i is a real number that encodes the elevation of the support plane with the slope p_i.

ψ:=(ψ_i)_1≤i≤N denotes the set of elevations. The Visibility cell V_i(ψ) of y_i is defined as a set of localized points x∈²×{0} whose rays are reflected towards the direction y_i, which means that the vertical rays hit the ith facet of R_ψ.

Given the definition of R_ψ, it leads to the following:

V _i(ψ)={x∈²×{0}|∀j,−x|p_i+ψ_i≤−x|p_j+ψ_j}

By construction, the vertical ray emanating from the point x∈V_i(ψ) touches the mirror surface at an altitude x|p_i−ψ_i for a given i and is reflected towards the direction y_i, and as a consequence thereof the quantity of light reflected towards the direction y_i is equal to the integral of ρ over V_i(ψ). We also have ∇R_ψ(x)=p_i if x∈V_i(ψ). The problem of the mirror with collimated light source then amounts to finding (ψ_i) such that:

∀i∈{1, . . . ,n}∫_Vi(ψ)ρ(x)dx=σ_i  (1)

By construction, a solution to Equation (1) provides a parameterization R_ψ of a convex mirror, which sends the collimated light source to the discrete target σ:

R _ψ :x∈ ² (x,max_i(x|p_i−ψ_i) where ²×{0} and ² are identified.

It should be noted that since the mirror is a graph over ²×{0}, the vectors y_i cannot be directed in the upward direction (upward vertical). In practice, it is assumed that:

y _i∈²:={x∈²,x|e_z≤0}.

In addition, the position of the mirror is localized by considering it only above the support:

X _ρ :={x∈ ²×{0},ρ(x)≠0} of ρ.

The same approach also allows for the construction of concave mirrors using a concave function of the form xmin_ix|p_i+ψ_i This amounts to replacing the Visibility cells by:

V _i(ψ)={x∈²×{0}|x|p_i+ψ_i≤x|p_j+ψ_j∀_j}

In this case, a solution to Equation (1) provides a parameterization of a concave mirror R_ψ(x)=(x, min_ix|p_i+ψ_i) that sends the collimated light source ρ to the discrete source σ.

1.2—Concave Mirror for a Point Light Source

In this second design problem, all of the rays are emitted from a single point in space, situated at the origin, and the light source is described by a function of intensity ρ on the unit sphere ².

As in the previous case, the target is in the far field and is described by a set of values σ=(σ_i)_1≤i≤N supported on the finite set of directions Y={y_i . . . , y_N}⊂². The problem which is being considered is that of finding the surface of a mirror that sends the light intensity ρ to the light intensity σ.

Thereafter a concave surface is constructed which is made up of pieces of confocal paraboloids. More precisely, P (y_i, ψ_i) is used to denote the solid paraboloid of which the focal point is at the origin with the focal length ψ_i and with the direction y_i. The surface R_ψ is defined as the boundary of the intersection of the solid paraboloids, that is to say, R_ψ=θ (∩_i P (y_i, ψ_i)). The Visibility cell V_i(ψ) is the set of directions of rays x∈² emanating from the light source that are reflected in the direction y_i. Given that each paraboloid ∂P(y_i, ψ_i) is parameterized over the sphere by xψ_i x/(1−x|y_i), it leads to:

$V_i\left(\psi \right)=\left\{x\in {𝕊}^2|\forall j,\frac{{\psi }_i}{1-\langle x|y_i\rangle }\le \frac{{\psi }_j}{1-\langle x|y_j\rangle }\right\}.$

The point light source mirror problem then amounts to finding (ψ_p) that satisfies the light energy conservation Equation (1). The mirror surface is then parameterized by:

$R_{\psi }\text{:}x\in {𝕊}^2\mapsto {\min }_i\frac{{\psi }_i}{1-\langle x|y_i\rangle }x.$

In practice, it is assumed that the target Y is included in ², that the support X_ρ of ρ is included in ₊²:={x∈², x|e_z≥0} and that the mirror is parameterized over X_ρ.

The mirror surface may also be defined as the boundary of the union (instead of the intersection) of a family of solid paraboloids. The Visibility cell thus then becomes:

$V_i\left(\psi \right)=\left\{x\in {𝕊}^2|\forall j,\frac{{\psi }_i}{1-\langle x|y_i\rangle }\ge \frac{{\psi }_j}{1-\langle x|y_j\rangle }\right\}$

and a solution of Equation (1) provides a parameterization R_ψ(x)=(xmax_iψ_i/(1−x|y_i)) of the surface of the mirror.

2. Lens Design

In this section, the goal is to design lenses that refract a given light source intensity to a desired target. In a manner similar to designing of a mirror, collimated or point light sources are considered.

In this instance n₁ is used to denote the refractive index of the lens, n₂ the refractive index of the ambient space, and K=n₁/n₂ denotes the ratio of the two indices.

Considered here is a collimated light source that is encoded by a function ρ on a 2D domain and a target illumination supported on a finite set:

Y={y ₁ , . . . ,y _N}⊂²,encoded by σ=(σ_i)_1≤i≤N.

The objective is to find the surface of a lens that sends ρ to σ.

It is assumed that the rays emitted by the light source are vertical and that the base or bottom part of the lens is flat and orthogonal to the vertical axis. There is no angle of refraction when the rays enter the lens, and consequently it is thus only necessary to build the top part of the lens.

By means of a simple change of variables, it is shown that this problem is equivalent to that of designing a mirror for a collimated light source. More precisely, for every y_i∈Y, now p_i is defined to be the slope of a plane that refracts the vertical ray (0,0,1) to the direction y_i. As well, is defined as the graph of a convex function of the form xmax_ix|p_i−ψ_i, where ψ=(ψ_i)_1≤i≤N is the set of elevations. The Visibility cell V_i(ψ) is defined as being the set of points x∈²×{0} that are refracted to the direction y_i:

V _i(ψ)={x∈²×{0}|∀j,−x|p_i+ψ_i≤−x|p_j+ψ_j}

The collimated light source lens design problem thus then amounts to finding the weights ψ=(ψ_i)_1≤i≤N that satisfy the Equation (1). In this case, the lens surface is then parameterized by:

R _ψ :x∈ ² (x,max_ix|p_i−ψ_i)

In practice, it is necessary to choose the directions y_i in ₊² and the mirror to be parameterized over the support X_ρ of ρ.

It should be noted that it is also possible to construct concave lenses by taking into consideration the parameterizations with convex functions of the form xmin_ix|p_i+ψ_i.

2.2 Convex Lens for a Point Light Source

The same problem is considered, except that the collimated light source is replaced by a point light source. As in the configuration of the collimated light source, the base or bottom part of the lens is fixed. A piece of sphere centered at the source is chosen, such that the rays are not deviated. The lens is composed of pieces of ellipsoids of constant eccentricities K>1, where K is the ratio of indices of refraction. Each ellipsoid ∂E(y_i, ψ_i) can be parameterized over the sphere by xψ_ix/(1−Kx|y_i)

The Visibility cell is then:

$V_i\left(\psi \right)=\left\{x\in {𝕊}^2|\forall j,\frac{{\psi }_i}{1-K\langle x|y_i\rangle }\le \frac{{\psi }_j}{1-K\langle x|y_j\rangle }\right\}$

The point light source lens problem thus then amounts to finding the weights (ψ_i)_1≤i≤N that satisfy the Equation (1).

The top surface of the lens is then parameterized by:

$R_{\psi }\text{:}x\in {𝕊}^2\mapsto {\min }_i\frac{{\psi }_i}{1-K\langle x|y_i\rangle }x.$

In practice, it is necessary to choose the set of directions y_i so as to belong to ₊² and the lens to be parameterized over the support X_ρ⊂₊² of ρ.

It is also possible to choose to define the lens surface as the boundary of the union (instead of the intersection) of a family of solid ellipsoids. In this case, the Visibility cells are given by:

$V_i\left(\psi \right)=\left\{x\in {𝕊}^2|\forall j,\frac{{\psi }_i}{1-K\langle x|y_i\rangle }\ge \frac{{\psi }_j}{1-K\langle x|y_j\rangle }\right\}$

and a solution to the Equation (1) provides a parameterization R_ψ(x)=x max_iψ_i/(1−Kx|y_i) of the lens surface.

3. General Formulation

Let X be a domain either of the plane ²×{0}, or of the unit sphere ²; ρ: X→ a probability density; and Y={y₁ . . . , y_N}⊂² a set of N points. The function G: ^N→^N g is defined by:

G _i(ψ)=∫_Vi(104)ρ(x)dx

Where G(ψ)=(G_i(ω))_1≤i≤N and V_i(ψ)⊂X is the Visibility cell of y_i, whose definition depends on the non-imaging problem. The use of this notation enables the reformulating of Equation (1) so as to find the weights ψ=(ψ_i)_1≤i≤N such that:

∀i∈{1, . . . ,N},G_i(ψ)=σ_i(2)

4. Visibility and Power Cells

It is therefore necessary to compute the Visibility cells V_i(ψ) associated with each optical modelling. The Visibility cells always have the same structure, making it possible to build a generic algorithm, as detailed here below. In all of the non-imaging optics problems, the Visibility cells are of the following form:

V _i(ψ)=Pow_i(ψ)#X  (3)

For a collimated light source, X denotes the plane ²×{0} and for a point light source, X is the unit sphere ². The sets Pow_i(P) are the usual Power cells of a weighted point cloud P={(p_u, ω_i)}⊂₃×:

Pow_i(P):={x∈³|∀j,∥x−p_i∥²+ω_i≤∥x−p_j∥²+ω_j}.

The expression of the weighted point cloud P={(p_i, ω_i))} depends on the problem. The deduction of the expression for p_i and w in the collimated light source mirror case is explained, with the other formulas being set out in Table 1 for the other cases. In the collimated light source mirror case, the light source is collimated and p_i∈²×{0} is the slope of the plane that reflects (according to Snell-Descartes law) the vertical ray upwards e_z:=(0, 0, 1) towards the direction y_i. A calculation shows that p_i=P² (y_i−e_z) y_i|e_z|e_z, where denotes the orthogonal projection onto ²×{0}.

The Visibility cell of y_i is then given by:

V _i(ψ)={x∈²×{0}|∀j,−|p_i+ψ_i≤−x|p_j+ψ_j}=Pow_i(P)∩(²×{0}),

where ω_i=2ψ_i−∥p_i∥².

It may thus be concluded therefrom that the Visibility cells for a convex mirror of the point light source mirror problem are given by the Equation (3), in which the weighted point cloud is given by the first row of Table 1, which gives the formulas for the weighted points used to define the Power cells in the Equation (3) for various different non-imaging optics problems. In the lens design problem, K>0 is the ratio of the indices of refraction, K>1 in the point source and lens configuration. Ccv signifies concave, Cvx signifies convex, signifies that the optical component converges to a concave component when the discretization tends to infinity, CS signifies collimated light source, PS point light source:

TABLE 1
Type	Points	Weights
Cvx (CS/miroir)	$p_i=\text{?}$	ωi = 2ψi − ||pi||2
Ccv (Cs/miroir)	$p_i=-\text{?}$	ωi = 2ψi − ||pi||2
Cvx (PS/miroir)	$p_i=-\text{?}$	${\omega }_i=-\text{?}-\text{?}$
(PS/miroir)	pi = yi/(2 ln(ψi))	${\omega }_i=\text{?}-\text{?}$
Cvx (CS/lentille)	$p_i=-\text{?}$	ωi = 2ψi − ||pi||2
Ccv (Cs/lentille)	$p_i=\text{?}$	ωi = 2ψi − ||pi||2
Cvx (PS/lentille)	$p_i=-\kappa \text{?}$	${\omega }_i=-\text{?}-\text{?}$
(PS/lentille)	pi = κyi/(2 ln(ψi))	${\omega }_i=\text{?}-\text{?}$
indicates data missing or illegible when filed

5. Generic Algorithm

For each optical design problem, given a light source intensity function, a target light intensity function and an error parameter, Algorithm 1 (here below) provides a triangulation of a mirror or a lens that satisfies the light energy conservation Equation (1).

The main problem is to find weights ψ such that G(ψ)=σ. This is achieved by using a damped Newton algorithm which has a quadratic local convergence rate for optimal transport problems or for Monge-Ampere equations in the plane.

The algorithm comprises three steps:

• • Initialization: the source density is discretized into a piecewise affine density and the target one into a finitely supported measure. This is then followed by constructing the initial weights ψ⁰ satisfying the condition ∀i, G_L(ψ⁽⁰⁾)>0.
  • Damped Newton: a sequence ψ^k is constructed following Algorithm 2 until ∥G(ψ^k−σ)∥_∞≤η.
  • Surface Construction: finally the solution ψ^k∈^N is converted into a triangulation. Depending on the non-imaging optics problem, this amounts to approximating a union (or an intersection) of half-spaces (or solid paraboloids, or ellipsoids) by a triangulation.

Initialization:

Discretization of Light Intensity Functions: The framework of the method makes it possible to support any type of collimated or point light source, or target light intensity functions. It may be for example a positive function on the plane or the sphere (depending on the problem) or a greyscale image, which is seen as a piecewise affine function. First the support of the source density ρ is approached by a triangulation T and it is assumed that the density ρ: T⁺ is affine on each triangle. Then p is normalized by dividing it by the total integral ∫_T ρ(x)dx.

In a similar manner, the target light intensity function can also be any discrete probability measure. If the user provides an image, it can be transformed into a discrete measure on the form σ=Σ_iσ_iδ_yi by making use of Lloyd's algorithm or more simply, by taking one Dirac measure per pixel, the latter approach being adopted in this instance. The target measure is also normalized by dividing by the discrete integral σ=Σ_iσ_i. Furthermore, min_iσ_i>0 is needed for the damped Newton algorithm, however this is not a restriction: if σ_i=0, this simply leads to removal of the corresponding Dirac measure δ_yi, thus ensuring that no light is sent to y_i.

Selection of the Initial Family of Weights ψ⁰: As previously mentioned here above, it is necessary to ensure that at each iteration all Visibility cells have non-empty interiors. In particular, it is necessary to choose a set of initial weights ψ⁰=(ψ_i⁰)_1≤i≤N such that the initial Visibility cells are not empty.

For the collimated light sources cases (with mirror or lens), it is noted that if ψ_i^0=∥p_i∥²/2 is chosen, then ω_i=0, where p_i is obtained by using the formulas of Section 4 (Visibility and Power Cells). The Visibility diagram then becomes a Voronoi diagram, and consequently p_i∈V_i(ψ^o).

For the point light source mirror case, a calculation shows that if one were to choose ψ_i⁰=1, then −y_i∈V_i(ψ^o).

For the point light source lens case, it can be shown that if one were to also choose ψ_i⁰=1, then y_i∈V_i(ψ^o).

It should be noted that the previous expressions for ψ^o ensure that G_i(ψ⁰)=ρ(V_i(ψ⁰))>0 only when the support X_ρ of the light source is sufficiently large. By way of example in the case of a point source mirror, if y_i∉X_ρ, then G_i(ψ⁰)=0 may be obtained. In order to deal with this difficulty, use is made of a linear interpolation between ρ and a constant density supported on a set that contains the (−y_i)'s. This strategy also works for the collimated source lens case, the point source lens and collimated source lens cases.

Damped Newton Algorithm: When the light source is collimated (that is X=²×{0}), the problem is known to be an optimal transport problem in the plane for the quadratic cost, the function G is the gradient of a concave function, its Jacobian matrix DG is symmetric and DG≤0. Moreover, if G_i(ψ)>0 for all i and if X_ρ is connected, then the kernel of DG spans over ψ=cst. This ensures the convergence of the damped Newton algorithm presented in algorithm 2 here below, where A⁺ denotes the pseudo-inverse of the matrix A. In practice, taking the pseudo-inverse matrix of D{tilde over (G)}() ensures that the mean of the remains constant. Still in practice, one row and one column are removed from the matrix in order to make it full rank.

When the light source is a point source, a change is effected in respect of the variables {tilde over (ψ)}=ln (ψ) and {tilde over (G)}=G·exp, such that G(ψ)=σ. This change in variable transforms the optical component design problem into an optimal transport problem, ensuring that {tilde over (G)} is the gradient of a concave function and that D{tilde over (G)} is symmetric negative, and therefore easily invertible. In the point light source mirror problem with convex mirrors, the damped Newton algorithm has been proven to be converging.

Computation of G and DG: According to Section 4, the Visibility cells V_i(ψ) may be computed by the intersecting of a certain 3D power diagram with a triangulation T of the support X_ρ of ρ. Such an intersection may for example be computed by the algorithm described in “A numerical algorithm for L² semi-discrete optimal transport in 3D”, Bruno Lévy, arXiv preprint arXiv: 1409.1279 (2014). Then G_i(ψ)=∫_Vi(ψ) ρ(x)dx may be computed by using first order quadrature formulas. The computation of DG is done by using forward-mode automatic differentiation, where the gradient of G_i(ψ) is stored as a sparse vector. It should be noted that this works quite efficiently since all the numbers that occur in the computation of G_i(ψ) depend only on the values of ψ_j where j is such that (i, j) are neighbours in the Visibility diagram, i.e. V_i(ψ)∩V_j(ψ)≠Ø.

Linear system: Given that D{tilde over (G)} is sparse and symmetric negative, the solving of the linear systems is done using the preconditioned conjugate gradient.

Surface construction: In the last step of Algorithm 1, a triangulation of the mirror or lens surface is constructed. The input is a family of weights solving Equation (2) and the parameterization function R_ψ whose formula is given here above and depends on the different cases. Each Visibility cell is triangulated by taking the convex envelope of the vertices of its boundary. A vertex of the triangulation will belong to at least one Visibility cell. For each vertex, it is possible to compute exactly the normal to the continuous surface using the Snell-Descartes law since the incident ray and the corresponding reflected/refracted direction y_i are known.

Algorithms 1 and 2 are described here below:

Algorithm 1: Mirror/Lens Construction
Input:  A light source intensity function ρin
        A target light intensity function σin
        A tolerance η
Output: A triangulation RT of a mirror or a lens
Step 1: Initialization
        T, ρ ← DISCRETISATION_SOURCE (ρin)
        Y, σ ← DISCRETISATION_TARGET (σin)
        ψ0 ← INITIAL_VECTORS (Y)
Step 2: Solve Equation (2): G(ψ) = σ
        ψ ← DAMPED_NEWTON (T, ρ, Y, σ, ψ0, η)
Step 3: Construct a triangulation RT of R
        RT ← SURFACE_CONSTRUCTION (ψ, Rψ)

Algorithm 2: Damped Newton Method for G(ψ) = σ
Input: The source ρ : T   + and a target σ = Σiσiδyi; an initial vector
ψ0 and a tolerance η > 0.
Step 1: Transformation to an Optimal Transport Problem
IF X = 2 × {0} then   = ψ0 (et {tilde over (G)} = G)
IF X = 2, then   = (ln(ψi0))1≤i≤N (and {tilde over (G)} = (Gi ∘ exp)1≤i≤N)
Step 2: Solve the equation {tilde over (G)}({tilde over (ψ)}) = σ
Initialization: ε0 : = [mini Gi(ψ0), miniσi] > 0, k: = 0
While ||{tilde over (G)}(   ) − σ||∞ ≥ η
- Compute dk = −D{tilde over (G)}(   ) + ({tilde over (G)}(   ) − σ)
- Find the smallest   ∈   provided that {tilde over (ψ)}k,l : = {tilde over (ψ)}k + 2−ldk satisfies:
{mini⁢G~i⁡(ψ~k,ℓ)≥ɛ0G~⁡(ψ~k,ℓ)-σ∞≥(1-2-(ℓ+1))⁢G~⁡(ψ~k)-σ∞
- Set {tilde over (ψ)}k +1 = {tilde over (ψ)}k + 2−ldk and k ← k + 1
Return ψ := ({tilde over (ψ)}ik)1≤i≤N si X = 2 × {0} or
ψ := (exp({tilde over (ψ)}ik))1≤i≤N si X = 2

Although this value is given only by way of an example, in practice, it is possible to choose η=10⁻⁸.

The foregoing sections therefore serve to explain one of the methods for solving far-field non-imaging (anidolic) optics problems, that is to say, constructing a lens or a mirror, whether concave or convex, from a given source of light, which is either a collimated or point source, in order to achieve a given illumination on a target at infinity.

Although this is not the only method that it is possible to use in the context of the invention, the method described here above can be used in the context of the step (d) of the method of the invention described here below.

The problem solved in the far field by the above method can also be solved in the near field.

The method of the invention for solving the near-field problem is detailed here below, in conjunction with FIGS. 1-3.

In the method according to the invention, there is an illumination target T and an illumination light source S, the illumination target T being in the near field and therefore at a finite distance from the illumination light source S.

From this illumination light source S, it is sought to obtain a given illumination on the target T, by incidence of the light coming from the illumination light source S on an optical component Cnf to be designed.

The illumination on the target T is characterized by a quantity of light received alpha_i at any point i of the target T.

According to the method, the following steps are carried out:

• • (a) positioning an origin point O between the illumination light source S and the illumination target T;
  • (b) for each point i of the illumination target T, computing a direction dir_i which corresponds to the direction of the vector connecting the origin point O and the point i on the illumination target T (FIG. 1);
  • (c) positioning a reference point A between the origin point O and the illumination target T;
  • (d) creating the optical component Cff whose surface passes through the reference point A and which, when the illumination originating from an illumination light source S is incident on the optical component Cff, generates an illumination comprised of all the illuminations of direction dir_i and quantity of light alpha_i (FIG. 2);
  • (e) for each point z_i of the optical component Cff that generates the directional illumination dir_i and quantity of light alpha_i, computing a corrected direction dirc_i which corresponds to the direction of the vector connecting the point z_i and the point i on the illumination target T (FIG. 3);
  • (f) determining whether or not, for each point i, the difference between the direction dir_i and the corrected direction dirc_i satisfies a predetermined criterion; (g) if, for each point i, the difference between the direction dir_i and the corrected direction dirc_i satisfies the predetermined criterion, creating the optical component C that corresponds to the optical component Cff,
  • (h) if, for each point i, the difference between the direction dir_i and the corrected direction dirc_i does not satisfy the predetermined criterion, reiterating the steps (d) to (f) by substituting dir_i with dirc_i.

The method of the invention here above is applicable to any non-imaging (anidolic) optics problem, in particular for a fabricated optical component that is a concave or convex mirror or a lens.

The optical component created in the step (g) is created according to the conventional means for creating an optical component, whether this be a mirror or a lens, for example by milling on a 3-axis numerical control (CNC) machine tool after prior milling of the blank, this being preferably computer assisted or computer controlled so as to improve the finish of the optical component thus produced, optionally followed by sanding and/or polishing. For a lens, it is possible to use, for example, though not exclusively, poly(methyl methacrylate) (PMMA), while for a mirror, aluminum may be used.

The method of the invention may be translated into algorithmic form as in the Algorithm 3 here below, in order to provide for a better understanding thereof and possibly an automated implementation of the same, in a manner analogous to the method set out in the preamble to the detailed description for a far field target.

The method described here above, with a target illumination σ=Σ_i=1^Nσ_iδ_zi supported on a set of points Z={z₁, . . . , x_N)⊂³, is, as indicated in the method, tantamount to iteratively solving a far-field problem, namely Equation (1), which rapidly converges to the near field solution.

In the Algorithm 3, SOLVE_FF (T, ρ, Y^k, σ, η) denotes an algorithm that solves the far field problem between a source ρ: T⁺ and a target σ=Σ_iσ_iδ_yi supported on Y⊂² for a numerical error η. The Step 2 of Algorithm 1 may for example be implemented in order to solve this problem.

Algorithm 3: Optical Component Design for a Near-Field Target
Input: The source ρ : T   + and a target σ = Σi=1Nσiδzi ; an initial vector ψ0 and two
tolerances η, ηNF > 0.
Initialization: ∀i, ci0 = 0
While ∥ck+1 − ck∥1/N > ηNF
- Compute vik = Rψk(cik)
- Set yik = (zi − vik)/∥zi − vik∥
- Solve ψk+1 ← SOLVE_FF (T, ρ, Yk, σ, η)
- Update ck+1 to be the barycenter of Vi(ψk+1).

It is clear that when a fixed point is reached in Algorithm 3, which corresponds algorithmically to the method of the invention, the corresponding weight vector ψ is a near-field solution.

The Applicant has observed in practice that the process converges very quickly. Over several attempts, an error η_NF of less than 10⁻⁶ was obtained after only 6 iterations.

The convergence of the method of the invention is illustrated in FIG. 4, which represents, on the example of the image of the train on the first line, a rendering of the target image, the first iteration, the second iteration and the sixth iteration, for a collimated source concave lens configuration; while the second line represents a rendering of the target image, the first iteration, the second iteration and the sixth iteration, for a collimated source convex mirror configuration. It can therefore be observed that after six iterations, the image is almost identical to the target image.

The method according to the invention may be applied to any non-imaging optics problem, regardless of whether the light source is a point or collimated source, whether the optical component is a mirror or a lens, whether the geometry of the optical component is concave or convex. The method of the invention works equally well even if the light source is not uniform.

Claims

1. Method for fabricating an optical component that is configured so as to generate on a near field illumination target an illumination that has a determined pattern according to which each point (i) of the illumination target receives a quantity of light (alpha_i) via an illumination originating from an illumination light source which is incident on the optical component placed between the illumination light source and the illumination target, wherein the method includes the following steps: (a) positioning an origin point O between the illumination light source and the illumination target;(b) for each point (i) of the illumination target, computing a direction (dir_i) which corresponds to the direction of the vector connecting the origin point O and the point (i) on the illumination target;(c) positioning a reference point A between the origin point O and the illumination target;(d) creating the optical component (Cff) whose surface passes through the reference point A and which, when the illumination originating from an illumination light source is incident on the optical component (Cff), generates an illumination comprised of the all the illuminations of direction (dir_i) and quantity of light (alpha_i);(e) for each point (z_i) of the optical component (Cff) that generates the illumination of direction (dir_i) and quantity of light (alpha_i), computing a corrected direction (dirc_i) which corresponds to the direction of the vector connecting the point (z_i) and the point (i) on the illumination target;(f) determining whether or not, for each point (i), the difference between the direction (dir_i) and the corrected direction (dirc_i) satisfies a predetermined criterion;(g) if, for each point (i), the difference between the direction (dir_) and the corrected direction (dirc_i) satisfies the predetermined criterion, fabricating the optical component that corresponds to the optical component (Cff),(h) if, for each point (i), the difference between the direction (dir_i) and the corrected direction (dirc_i) does not satisfy the predetermined criterion, reiterating the steps (d) to (f) by substituting (dir_i) with (dirc_i).

2. The method according to claim 1, wherein the step (f) further comprises determining whether on each point (i) the mean of the norms of the differences between the direction (dir_i) and the corrected direction (dirc_i) is less than a predetermined value.

3. The method according to claim 1, wherein the optical component is one of a concave optical component and a convex optical component.

4. The method according to claim 1, wherein the optical component is one of a mirror and a lens.

5. The method according to claim 2, wherein the optical component is one of a concave optical component and a convex optical component.

6. The method according to claim 2, wherein the optical component is one of a mirror and a lens.

7. The method according to claim 3, wherein the optical component is one of a mirror and a lens.

8. The method according to claim 5, wherein the optical component is one of a mirror and a lens.