POLARIMETRIC AND MULTISPECTRAL IMAGING, PARTICULARLY FOR FORMING A CELESTIAL COMPASS

Abstract

The imager comprises: - a spectral filter matrix formed of a repeated pattern of spectral filters; - a polarizer matrix; - a photosensor matrix; and - a processing device designed to, for each of the different spectral bands, form at least one image of pixels having positions in the image corresponding respectively to predefined positions (i, j) of a mask (M) in the photosensor matrix, each of the pixels being calculated from the light intensities measured by the photosensors of the spectral band considered contained in the mask (M) at the position (i,j) of the pixel.The spectral filter pattern is designed so that, for each of the different spectral bands and for all predefined positions (i, j) of the mask (M), the photosensitive cells of the spectral band in question have an isobarycenter located in a central virtual cell (PVC) of the mask (M) of the same size as cells in the photosensitive cell array, and each of the predefined positions (i, j) of the mask (M) is offset by another predefined position (i, j) by one column or one row of the photosensitive cell array. Figure for the abbreviation: Fig. 7.

Description

Title of the invention: POLARIMETRIC AND MULTISPECTRAL IMAGING DEVICE, IN PARTICULAR FOR FORMING A CELESTIAL COMPASS Technical field of the invention

[0001] The present invention relates to a polarimetric and multispectral imager designed to determine its orientation (azimuth) relative to a luminous celestial body, generally the Sun or the Moon. Such an imager is particularly useful for forming a celestial compass designed to determine its orientation (azimuth) relative to a given direction, for example, true north, by reference to the ephemeris of the celestial body.

[0002] Thus, the invention is particularly applicable in the aeronautical field, where a celestial compass can be installed in an aircraft to determine its heading relative to a given direction. However, a celestial compass according to the invention can also be fitted to any type of vehicle, such as a boat or a land vehicle like a motor vehicle. Technological background

[0003] Prior art is known of an imager of the type comprising: - a spectral filter matrix formed from a repeating pattern of spectral filters of at least three different spectral bands, of dimension NxN, with N an integer greater than or equal to three; - a polarizer matrix of at least three different polarizations; - a matrix of photosensors, each designed to receive light passed through one of the spectral filters and one of the polarizers and to measure the luminous intensity of the received light; and - a processing device designed to, for each of the different spectral bands, form at least one image of pixels having positions in the image corresponding respectively to predefined positions of a mask in the photosensor matrix, each of the pixels being calculated from the light intensities measured by the photosensors of the spectral band considered contained in the mask at the position of the pixel.

[0004] Such an imager is generally used in digital cameras, where the spectral filter pattern is most often an RGB pattern. The latter thus comprises four filters for the color red, grouped in the upper left of the pattern, four filters for the color blue, grouped in the lower right of the pattern, and eight filters for the color green, half of which are grouped in the lower left of the pattern and the other half in the in the upper right corner of the pattern. The mask is then successively positioned on each sub-matrix of the photosensor array. The drawback of this is that the output image has a resolution four times lower than that of the photosensor array.

[0005] It may therefore be desirable to provide an imager which makes it possible to overcome at least some of the aforementioned problems and constraints. Summary of the invention

[0006] An imager of the aforementioned type is therefore proposed, characterized in that the spectral filter pattern is designed so that, for each of the different spectral bands and for all predefined positions of the mask, the photosensors of the spectral band considered have an isobarycenter located in a central virtual cell of the mask of the same size as the cells of the photosensor matrix, and in that each of the predefined positions of the mask is offset by another predefined position of a column or a row of the photosensor matrix.

[0007] Thus, thanks to the invention, to move from one position to another, the mask is only moved by one row or column of the photosensor matrix, which makes it possible to obtain a resolution almost equal to that of the photosensor matrix and this while maintaining good measurement quality because, thanks to the judicious choice of the spectral filter pattern, for each spectral band, the barycenter of the photosensors contained in the mask and receiving this spectral band is located in the middle of the mask, regardless of the position of the latter.

[0008] The invention may further include one or more of the following optional features, according to any technically possible combination.

[0009] Optionally, the spectral filter pattern presents, for each of the different spectral bands, a unique spectral filter of the spectral band considered in each row and each column of the spectral filter pattern.

[0010] Optionally, the spectral filter pattern is as follows: C1 C3 C2, where C1, C2, C3 are spectral filters of three respectively C2 C3 Cl C2 C3 Cl different spectral bands, or the following: Cl C3 C4 C2 C2 C4 C3 Cl C3 Cl C2 C4 C4, where Cl, C2, C3, C2 Cl C3 C4 are spectral filters of four different spectral bands respectively, or the following: Cl C2 C4 C3 C2 C5 C3 C4 C3 Cl C5 C2 C4 C3 Cl C5 C5 C4 C2 C l C5, where Cl, C2, C3, C4, C5 are filters Cl C4 C2 C3 spectral data from five different spectral bands respectively.

[0011] Optionally also, at least one of the different spectral bands is an ultraviolet spectral band or an infrared spectral band.

[0012] Optionally also, at least one of the different spectral bands is one of: a spectral band described as red color from 60 nm to 80 nm included in the interval [625 nm ; 740 nm], a spectral band described as green color from 60 nm to 80 nm included in the interval [520 nm ; 565 nm], and a spectral band described as blue color from 60 nm to 80 nm included in the interval [446 nm ; 520 nm].

[0013] Optionally also, the polarizer matrix is ​​formed of a repeated pattern of polarizers of at least three different polarizations, of dimension NxN, the polarizer patterns being respectively aligned in the stack with the spectral filter patterns.

[0014] Optionally, the polarizer pattern is also as follows: P1 P2 P3, P2 P3 PI P3 PI P2 where PI, P2, P3 are three polarizers following three polarizations respectively different predefined options, or the following: PI P2 PI P2 or the following: P3 P4 P3 P4 PI P2 PI P2 P3 P4 P3 P4 PI PI PI PI, where PI, P2, P3, P4 are the polarizers of four respectively P2 P2 P2 P2 P3 P3 P3 P3 P4 p4 p4 P4 different polarizations, or the following: PI P3 P4 P2 P3 or P2 P2 PI PI P4 P4 P2 PI P5 P5 P2 P5 P3 P4 PI P5 P3 P4 P5 P3 Next: PI P2 P3 P4 P5, where P1, P2, P3, P4, P5 are five following polarizers PI P2 P3 P4 P5 PI P2 P3 P4 P5 PI P2 P3 P4 P5 PI P2 P3 P4 P5 respectively five different predefined polarizations.

[0015] Optionally also, the different predefined polarizations include at least one of the following polarizations: 0°, 45°, 90° and 135°.

[0016] Optionally also, the imager further comprises a microlens array, so that the light received by each photosensor has passed through one of the microlenses.

[0017] Optionally also, the processing device is designed to form, for each of the different spectral bands, an image, called a local polarization angle map, in which each pixel is the local polarization angle calculated from the light intensities measured by the photosensors of the spectral band considered contained in the mask at the position of the pixel.

[0018] Optionally, the processing device is also designed to form, for each of the different spectral bands: - an image, called a Stokes parameter map S0, in which each pixel is the Stokes parameter S0 calculated from the light intensities measured by the photosensors of the spectral band considered contained in the mask at the position of the pixel; - an image, called a Stokes parameter map (SI), in which each pixel is the Stokes parameter (SI) calculated from the light intensities measured by some of the photosensors in the spectral band under consideration contained in the mask at the pixel's position; and - an image, called a Stokes parameter map S2, in which each pixel is the Stokes parameter S2 calculated from the light intensities measured by some of the photosensors of the spectral band considered contained in the mask at the pixel position; and wherein the processing device is designed to form the local polarization angle map from the three Stokes parameter maps S0, S1, S2.

[0019] Optionally, the processing device is also designed to form a meridian proximity map of a luminous celestial body by combining the local polarization angle maps of the different spectral bands, by means of a reading function that gives, for each predefined position, a sum over the different spectral bands of terms, each term being a function of the local polarization angle at that position for the spectral band considered, increasing by 0° to 90° and decreasing from 90° to 180° depending on the absolute value of the local polarization angle.

[0020] Optionally also, each term includes a peak function of the local polarization angle, increasing from 0° to 90° and decreasing from 90° to 180° as a function of the absolute value of the local polarization angle, normalized to be one at 90° and raised to a power greater than or equal to one.

[0021] Optionally, the peak function is also given by: a-cos(2-APLS) Xs “ 2

[0022] Optionally, the power is also given by: with L ... J the floor function and flstat(Xs) a polynomial function of a statistical moment, such as the mean or the standard deviation, or of a quantile, such as the median or a quartile, of the peak function, for example f Istat(Xs) = 9(Xs).

[0023] Optionally also, the processing device is designed to form, for each of the different spectral bands, an image, called a linear polarization degree map, in which each pixel is the linear polarization degree calculated from the light intensities measured by the photosensors of the spectral band considered contained in the mask at the pixel position, and each term includes a sky-likeness function of the linear polarization degree at that position for the spectral band considered, multiplied by the peak function raised to the power.

[0024] Optionally, the sky-like resemblance function is also given by: | 2-oiXO | with Ks “min{max( 1, DPL), 0) and f2stat(Ys) a polynomial function of a statistical moment, such as the mean or standard deviation, or of a quantile, such as the median or a quartile, of the function, for example f2stat(Xs) = a(Ys).

[0025] Optionally also, the reading function given by: rz- VS vz- \Ys(ijyXnfMYs) with S being the different spectral bands.

[0026] Optionally also, the processing device is designed to analyze the proximity map to determine an azimuth of a luminous celestial body with respect to a direction attached to the imager.

[0027] A celestial compass comprising: is also proposed - an imager according to the invention; - a positioning device designed to provide a position for the imager; - a clock designed to provide the current time; - an ephemeris giving the azimuth of the luminous celestial body relative to a predefined direction, for example true north, depending on the time and location; and - a heading determination device designed to provide an angle, called heading, between the direction attached to the imager and the predefined direction, from the azimuth provided by the ephemeris and the azimuth determined by the imager.

[0028] An aircraft comprising an imager according to the invention or a celestial compass according to the invention is also proposed. Brief description of the figures

[0029] The invention will be better understood with the aid of the following description, given solely by way of example and made with reference to the accompanying drawings in which: - [Fig. 1] is a schematic view of an aircraft equipped with a celestial compass fitted with an imager according to the invention, - [Fig.2] is a side view of the imager, - [Fig. 3] is a top view of a spectral filter matrix of the imager, - Figure [Fig. 4] is a top view of a polarizer array of the imager, - Figure 5 is a top view of a photosensor array of the imager, - [Fig. 6] is an exploded three-dimensional view of patterns from the matrices of filters and polarizers and sub-array of the photosensor array and a microlens array, - [Fig.7] is a top view of the spectral filter array and the photosensor array, stacked one on top of the other, with a mask in the first position in the photosensor array, - [Fig.8] is a view similar to that of [Fig.7], with the mask in a second position in the photosensor array, - [Fig.9] is a view similar to those in figures 7 and 8, with the mask in a third position in the photosensor array, - Figure 10 is a view similar to those in Figures 7 to 9, with the mask in a fourth position in the photosensor array. - Figure 11 is a block diagram illustrating the steps of a data processing procedure implemented by an imager processing device, and - [Fig. 12] is a simplified view of an example implementation of the treatment device. Detailed description of the invention

[0030] With reference to [Fig.1], an aircraft 100, such as an airplane, equipped with a celestial compass 102 implementing the invention, will now be described.

[0031] The celestial compass 102 comprises an imager 104 according to the invention, for example fixed to the aircraft 100, arranged so that, when the aircraft 100 is substantially horizontal, the imager 104 is oriented along the terrestrial vertical Z direction. In the following description, the imager 104 will be described with reference to this vertical Z direction. In particular, positioning terms such as "up," "down," "above," and "below" will refer to the vertical Z direction.

[0032] The imager 104 is in particular designed to provide a measurement of an azimuth A of a luminous celestial body 106 (i.e. a star), such as the Sun or the Moon, with respect to a predefined direction X attached to the aircraft 100.

[0033] The celestial compass 102 further includes a positioning device 108 designed to provide a position of the celestial compass 102 (and therefore of the aircraft 100). The positioning device 108 includes, for example, a GPS (Global Positioning System).

[0034] The celestial compass 102 further comprises a clock 110 designed to provide the current time, as well as an ephemeris 112, for example in the form of a database, giving an azimuth S of the luminous celestial body 106 with respect to a predefined direction NN, for example, true north, as a function of the time and position. The azimuth S is equal to the angle between the predefined direction NN and the azimuthal direction D of the celestial body 106. The celestial compass 102 further comprises a heading determination device 114 designed to provide a heading C of the aircraft 100, this heading C being the angle between the direction X attached to the aircraft 100 and the predefined direction NN. For this purpose, the heading determination device 114 is designed to receive the current time provided by the clock 110 and the position at this current time provided by the positioning device 108, in order to obtain from the ephemeris 112 the azimuth S corresponding to the current time and the position received.The heading determination device 112 is further designed to calculate the heading C from the azimuth S and the azimuth A determined by the imager 104, in particular by the following formula: C = A + S. .

[0035] Furthermore, it is possible to define a meridian MD of the star 106 as a circle on the celestial sphere passing through the star 106. This circle thus delimits a disk passing through the star 106, the imager 104 and containing the vertical axis Z and the azimuthal direction D.

[0036] With reference to [Fig.2], the imager 104 first of all comprises a top light inlet 202.

[0037] Under the upper light inlet 202, the imager 104 comprises a stack 204, along the vertical direction Z, of matrices 206, 208, 210, 212.

[0038] More specifically, the stack 204 comprises, from top to bottom, a matrix 206 of spectral filters, a matrix 208 (optional) of microlenses, a matrix 210 of polarizers and a matrix 212 of photosensors.

[0039] The 212 matrix of photosensors is always at the base of the 204 stack, but the order of the other matrices 206, 208, 210 in the 204 stack may be different.

[0040] For example, the polarizer array 210 can be made on the photosensor array 212 by oblique angle deposition. The filter array 206 and the microlens array 208 can then be attached on top.

[0041] Each photosensor is designed to measure a received light intensity, these measured light intensities being designated in [Fig.2] by the general reference IL. The efficiency of the photosensors is increased by the presence of microlenses, which serve to concentrate the light flux and orient it parallel to the stacking direction Z.

[0042] The imager 104 further includes a processing device 214 designed to receive the measured light intensities IL and to analyze them as will be described later.

[0043] With reference to [Fig. 3], the matrix 206 is formed from an MC pattern of spectral filters, repeated several times. In the illustrated example, the MC pattern is repeated four times to form the matrix 206. The MC pattern has the form of a matrix of cells of the same size, in each of which there is a spectral filter, the matrix being of dimension NxN, with N an integer greater than or equal to three.

[0044] Each spectral filter in the MC pattern is designed to allow only a predefined spectral band to pass through. More precisely, each spectral filter has a certain frequency spectrum of sensitivity. This spectrum has a peak (i.e., a maximum value) and a full width at half maximum (FWHM) of the peak. The spectral band of the spectral filter is, by definition, equal to this FWHM. The MC pattern thus comprises several groups of identical spectral filters, that is, in particular, filters having the same spectral band. In other words, the spectral filters in the same group have the same spectral band, the spectral bands being disjoint from one group to another. The number of different spectral bands, and therefore of different groups of spectral filters, is denoted S.

[0045] When the spectral bands extend into visible light, they correspond to colors and spectral filters can be described as filters of color. In this case, matrix 206 can be described as a color filter array (CFA). However, preferably, at least one of the S spectral bands lies outside the visible spectrum, for example, in the ultraviolet (UV) or near-infrared (nIR) range. This allows for the discernment of sky polarization patterns despite degraded atmospheric conditions. For example, at least one of the S spectral bands is chosen from: - an ultraviolet spectral band UV: spectral band included in the interval [350 nm ; 380 nm]; - a spectral band in the near-infrared nIR: spectral band included in the interval [700 nm ; 900 nm]; - a spectral band described as red color: spectral band from 60 nm to 80 nm included in the interval [625 nm ; 740 nm], with for example the peak at 630 nm; - a spectral band described as green: a spectral band from 60 nm to 80 nm included in the interval [520 nm; 565 nm], with, for example, the peak at 550 nm; and - a spectral band described as blue color: spectral band from 60 nm to 80 nm included in the interval [446 nm ; 520 nm], with for example the peak at 450 nm.

[0046] Preferably, the MC motif comprises N different spectral filters (S = N).

[0047] It is possible to define, in the middle of the MC pattern, a central virtual PVC pixel, of the The size of the MC pattern cells. The spectral filters of the MC pattern are arranged so that, for each of the S different spectral bands, the spectral filters of the considered spectral band have an isobarycenter in the central virtual pixel PVC. The isobarycenter is the average point of the center points of the cells in which the spectral filters of the considered spectral band are located.

[0048] For example, as illustrated in Figure 3, when N and S are equal to four, the MC pattern can be as follows: Cl C2 C3 C4, where Cl, C2, C3, C4 are the four C3 C4 Cl C2 C4 C3 C2 Cl C2 Cl C4 C3 Spectral filters for the four predefined spectral bands. In this MC pattern, for each of the four spectral bands, the four spectral filters have an isobarycenter located at a center C of the MC pattern, and therefore in the central virtual pixel PVC. More precisely, the four spectral filters C1 have an isobarycenter located at point C, as do the four spectral filters C2, C3, and C4.

[0049]

[0050]

[0051]

[0052] Alternatively, when N and S are equal to three, the MC pattern can be as follows: Cl C3 C2, where Cl, C2, C3 are the three spectral filters for C2 Cl C3 C3 C2 Cl respectively three predefined spectral bands. Alternatively, when N and S are equal to five, the MC pattern can be as follows: Cl C2 C4 C3 C5, where Cl, C2, C3, C4, C5 are the five spectral filters for C2 C5 C3 C4 Cl C3 Cl C5 C2 C4 C4 C3 Cl C5 C2 C5 C4 C2 Cl C3 respectively the five predefined spectral bands. Referring to [Fig. 4], the polarizer matrix 210 is formed by a repeated MP pattern of polarizers. The MP pattern is a matrix of cells the same size as the cells of the MC pattern. Each cell of the MP pattern contains a polarizer. The MP pattern has the same dimensions NxN as the MC pattern, and this MP pattern is repeated several times in the same way as the MC pattern. Thus, in the illustrated example, the MP pattern is repeated four times. For example, when N equals 4 and the MP pattern has four polarizers different, the MP pattern can be the following: PI P3 PI P3 P2 PA P2 PA PI P3 PI P3 P2, where P1, P2, P3, P4 PA P2 PA

[0053] These are four polarizers, each with a predefined polarization, for example: 0°, 45°, 90°, and 135°. In this example, it is preferable that the MP pattern itself be composed of several identical blocks (four identical 2x2 blocks). However, this is not generally required. Thus, for each spectral band of the MC pattern, the spectral filters are aligned, along the Z stacking direction, with polarizers of different polarizations. More precisely, in the illustrated example, one of the Cl spectral filters is aligned with a 0° polarizer, another Cl spectral filter is aligned with a 45° polarizer, another Cl spectral filter is aligned with a 90° polarizer, and the last Cl spectral filter is aligned with a 135° polarizer. The same is achieved for the C2, C3, and C4 spectral filters.

[0054] Alternatively, the MP pattern could be: PI PI PI PI, where PI, P2 P2 P2 P2 P3 P3 P3 P3 P4 P4 P4 P4

[0055] P2, P3, P4 are four polarizers following respectively the four predefined polarizations, for example: 0°, 45°, 90° and 135°. Alternatively, when N equals 3 and the MP pattern has three different polarizers, the MP pattern could be: PI P2 P3, where PI, P2, P3 are three P2 P3 P3 PI PI P2 polarizers following three different predefined polarizations respectively.

[0056] Alternatively, when N is equal to 5 and the MP pattern comprises five polarizers different, the MP pattern could be the following: PI P3 P4 P2 P3 or P2 P2 PI PI P4 P4 P2 PI P5 P5 P2 P5 P3 P4 PI P5 P3 P4 P5 P3 Next: PI P2 P3 P4 P5, where P1, P2, P3, P4, P5 are five following polarizers PI P2 P3 P4 P5 PI P2 P3 P4 P5 PI P2 P3 P4 P5 PI P2 P3 P4 P5 respectively five different predefined polarizations.

[0057] With reference to [Fig. 5], the photosensor matrix 212 is divided into sub-matrices SM of photosensor cells of the same size as the cells of the MC and MP patterns. The SM sub-matrices are the same size as the MC and MP patterns, i.e., 4x4 in the illustrated example. The microlens matrix 208, when present, is divided in the same way.

[0058] With reference to [Fig.6], the MC patterns of filters, the MP patterns of polarizers, the microlens sub-matrices and the SM sub-matrices of photosensors are aligned in the stacking direction Z. Thus, the light received by each photosensor passes through the spectral band filter and the polarizer located above that photosensor in the stacking direction Z, so as to be in one of the S spectral bands and according to one of the polarizations.

[0059] With reference to Figures 7 to 10, with the 212 photosensor matrix, it is possible to form at least one image of pixels having positions (i,j) in the image corresponding respectively to predefined positions (i, j) of a mask M in the 212 photosensor matrix. Each of the pixels can thus be calculated from the light intensities IL measured by the photosensors of the spectral band considered contained in the mask M at the position (i,j) of the pixel, but not from the other photosensors contained in the mask M.

[0060] The MC pattern is designed such that, for each of the different spectral bands and for all predefined positions (i, j) of the mask (M), the photosensitive cells of the spectral band in question have an isobarycenter located in a central virtual cell PVC of the mask M, the same size as the cells of the 212-cell matrix of photosensitive cells. Furthermore, each of the predefined positions (i, j) of the mask M is offset by another predefined position (i, j) by one column or one row of the 212-cell matrix of photosensitive cells, as illustrated in Figures 7 to 10.

[0061] Thus, the image 1002 has dimensions I = K-3 x J = L-3, where K and L are the dimensions of the 212 photosensor matrix. The pixels of the image 1002 are thus identified by the indices i (from 1 to I) and j (from 1 to J), these indices (i,j) equivalently representing the predefined positions of the mask M in the 212 photosensor matrix.

[0062] With reference to [Fig. 11], an example of a method 1100 for operating the imager 104 will now be described.

[0063] During a step 1102, the processing device 214 receives the light intensities IL measured by the photosensors, these light intensities coming from a celestial scene (part of the sky) observed by the imager 104.

[0064] During a step 1104, for each position (i,j), the processing device 214 forms, for each of the S spectral bands, a Stokes parameter map S0 (C so), a Stokes parameter map SI (Cs i) and a Stokes parameter map S2 (Cs 2)-

[0065] The value of each pixel in the Stokes parameter map S0 is the Stokes parameter S0 calculated from the light intensities IL measured by the photosensors of the spectral band considered, contained in the mask M at position i,j of the pixel. More precisely, in the illustrated case with four different polarizations, the Stokes parameter S0 is given by: S0 ^0^90^45^135 where Io, I45,190,1135 are the light intensities of the four photosensors for the spectral band considered, contained in the mask M and respectively associated with the polarizations 0°, 45°, 90° and 135°.

[0066] The value of each pixel in the Stokes parameter map S1 is the Stokes parameter SI calculated from the light intensities IL measured by some of the photosensors in the considered spectral band contained in the mask M at position i,j of the pixel. More precisely, in the illustrated case with four different polarizations, the Stokes parameter SI is given by:

[0067] The value of each pixel in the Stokes parameter map S2 is the Stokes parameter SI calculated from the light intensities IL measured by some of the photosensors in the spectral band considered, contained in the mask M at position i,j of the pixel. More precisely, in the illustrated case with four different polarizations, the Stokes parameter SI is given by: S2 = / 45-Zl35

[0068] In a step 1106, the processing device 214 calculates, for each of the S spectral bands and for each position (i,j), a degree of linear polarization DPL from the Stokes parameters SI, S2 for that spectral band and that position, in order to form a CDPL map of the degrees of linear polarization DPL. The degree of linear polarization DPL is given, for example, by: DPL

[0069] In a step 1108, the processing device 214 calculates, for each predefined spectral band and for each position (i,j), the local Stokes parameters S0, SU, S21 from the Stokes parameters S0, SI, S2 for that spectral band and that position. The local Stokes parameters S0, SU, S21 are given, for example, by: S0 = 50, 51 / — S1 ■ COS ( 2 • « jj ) + S2 • sin ( 2 • ij ) and S21 = - SI • sin ( 2 • « ij ) + 52 • COS ( 2 • « ij ). where a'7 is the angle of the pixel (i,j) taking a midpoint of the image 1002 as the origin.

[0070] During a step 1110, the processing device 214 calculates, for each predefined spectral band and for each position (i,j), a local polarization angle APL from the local Stokes parameters SOI, SU, S21 for that spectral band and that position, in order to form a Capl map of local polarization angles APL. The local polarization angle APL is, for example, given by: APL = jongle (SU + i • S21 ) ' with 1 the imaginary number.

[0071] In a step 1112, the processing device 214 calculates a proximity map CPROx at the meridian MD of the major illuminating body of the scene observed by the imager, i.e., the luminous celestial body 106, by combining the CDPL maps of linear polarization degrees DPL and the CAPl maps of local polarization angles APL for all predefined colors. To do this, the processing device 214 uses a readout function I(i,j) which, at each position (i,j), provides a scalar number derived from the linear polarization degrees DPL and the local polarization angles APL at that position (i,j), for all predefined colors.

[0072] Preferably, the reading function I(i,j) is given by: i J) = Lt A(ÎJŸ- Fs(4 J) V S V / • ■ \ # lsfa " y «>j II ^L^Cw) 2 • | "mît I with _ (^os(2APL)) y_; = min ( max ( 1, DPL ), 0 ), f 1 stat(Xs) and f2stat(Ys) being polynomial functions of a statistical moment (such as the mean or the standard deviation) or of a quantile (such as the median or the quartile) of Xs, respectively Ys. For example, / \stat ( Xs ) = 9 (Xs ) and f 2stat ( Xs ) = a ( Yç ).

[0073] In general, all other things being equal (i.e., with all other parameters constant), each Ts term increases from 0° to 90° and decreases from 90° to 180° as a function of the absolute value of the local polarization angle APL. Thus, each Ts term is higher when the local polarization angle APL is close to 90°, indicating the MD meridian of the luminous celestial body 106. Thus, all spectral bands are used, with the one having the most information on the MD meridian (i.e., the closest APL value to 90°) being amplified.

[0074] For example, each term Ts includes the function Xs, called the peak function, of the local polarization angle APL. This peak function Xs is increasing from 0° to 90° and decreasing from 90° to 180° as a function of the absolute value of the local polarization angle APL, normalized to be one at 90° (and preferably zero at 0° and 180°) and raised to the power p greater than or equal to one. In this way, the values ​​of T APL close to 90° are close to one, while the values ​​far from 90° quickly become very small (almost zero), especially the further they are from 90°. This makes it possible to highlight the meridian MD more precisely than a simple thresholding around 90°.

[0075] The function Fs is a sky-likeness map and expresses, at each position (i,j), whether an undesirable polarizing object (i.e., other than the sky) is located at that position. Such an undesirable object is, for example: a building or metallic or glazed building materials, a partial occlusion, and masking by vegetation cover (leaves, branches, etc.). For example, the function Fs(i,j) is normalized between 0 and 1, with values ​​close to 0 corresponding to the presence of an undesirable object and values ​​close to 1 corresponding to the sky. In general, the function Fs can be omitted (i.e., taken as 1) or take different forms than the one indicated above.

[0076] During a step 1114, the processing device 214 analyzes the proximity map CPROx the azimuth A of the luminous celestial body 106 with respect to the direction X attached to the imager 104. For this purpose, step 1114 includes for example the following steps.

[0077] During a step 1114-2, the processing device 214 first analyzes the CPROx proximity map to determine the direction of the MD meridian.

[0078] In a first example, the processing device 214 can use the Hough transform given by: ImgHough(p, 8) = fl ( x, j).5(x.cos(8) + ysin(8) -p)dxdy with x~^Center and y - j-jmtre and 5 defined as 5( 0) = 1,5(x * 0) =0. UHU i

[0079] The Hough transform gives, at each point (p, 8), the quantity of lines present in I(i,j) oriented by +gj and passing at a distance P from the center of the image I(i,j). Thus, the processing device 214 then calculates r, which corresponds to the number of lines oriented at Lr(8) = JImgHough(p, 8)dp -r (0 + ^) of I(i,j) passing at a distance less than r from the center of I(i,j).

[0080] The value 8 for which Lr(8) is maximum therefore indicates the orientation verified by the most lines of I(i,j) passing near the center of the image.

[0081] Ideally, on noise-free and pollution-free images, the orientation of the MD meridian would therefore be + with & for which L,_o( 8 ) is maximum.

[0082] In practice, the images are noisy and polluted. Thus, the processing device 214 works on a first iteration with a high r (on the order of a few pixels, for example between two pixels and 10% of the image width), so as to obtain a curve Lr>0(8) which is a "blurred" or "smoothed" version of the curve 8), which removes some of the noise present in the image of the reading function. From this first iteration Lr>0(0), the processing device 214 calculates 8nutx such that maxt^L^) = Lr>0(8max). This is a first approximation of the orientation of the MD meridian. The processing device 214 then determines a study interval around &nutx, @2] with ^>0(^2) = = ThresholdInterval^L^Ç dmax). The processing device 214 then studies the curve only on this interval, to find the orientation at j of the meridian MD which is at the maximum, or close to the maximum due to noise, of Lr=0( 0). The first iteration avoids concentrating around a peak of Lo( 6*) which would not be the desired orientation but a peak due to noise.

[0083] In this case, the processing device 214 first performs, for example, a thresholding of I(i,j) to keep only the points whose grey level / proximity level is greater than the chosen / predefined threshold proximity level.

[0084] The processing device 214 calculates the direction 8^ of the MD meridian by: ^mer — ~ T JL^ffjdü

[0085] In a second example, the processing device 214 performs a Ridgelet transform of the proximity map (radon transform followed by a wavelet transform). Then the processing device 214 determines the maximum of the Ridgelet transform to determine the direction of the MD meridian.

[0086] In a third example, the processing device 214 compares the CPROx proximity map to predefined proximity maps for each of which the position of the sun is known. These predefined proximity maps preferably correspond to different cloud cover and field of view obstruction conditions. These predefined proximity maps are, for example, generated from simulations of scenes under different cloud cover and vegetation or urban obstruction conditions, with a common feature: a fixed azimuth of the luminous celestial body 106, for example at 0°.

[0087] Preferably, the predefined proximity maps are expressed in polar coordinates. In this case, the processing device 214 performs, before comparison, a polar coordinate transform of the CPROx proximity map-

[0088] The comparison is carried out, for example, by calculating a correlation of the proximity map CPROx with each of the predefined proximity maps. In this case, the processing device 214 then calculates the average (weighted or unweighted) of the positions of the maxima of the convolutions, these maxima corresponding to the angular difference between the orientation of the meridian MD sought and the known orientations of the MD meridians in the predefined proximity maps (for example, 0°).

[0089] The direction of the MD meridian corresponds to the direction of the luminous celestial body 106 or to the opposite direction to the luminous celestial body 106.

[0090] Thus, during a step 1114-4, the processing device 214 removes this ambiguity to obtain the direction of the sun.

[0091] To this end, the processing device 214 analyzes, for example, the overall trend of the gradient of the ratio between S0 in the UV and S0 in the visible along the direction of the MD meridian. The direction along which this ratio tends to decrease is considered to be the direction of the luminous celestial body 106. Indeed, the portion of the MD meridian where the ratio / c is the lowest corresponds to the portion of the MD meridian pointing towards the luminous celestial body 106. Conversely, the portion of the MD meridian where the ratio s(y;v A, is the strongest corresponds to the portion of the MD meridian pointing at 180° from the luminous celestial body 106.

[0092] With reference to [Fig.

[12] , the processing device 214 includes, for example, a computer system comprising a data processing unit 1202 (such as a microprocessor) and a main memory 1204 (such as RAM, from the English "Random Access Memory") accessible by the processing unit 1202. The computer system further includes, for example, a network interface and / or a computer-readable medium, such as a local medium 1206 (such as a local hard disk) or a remote medium (such as a remote hard disk accessible via the network interface through a communication network) or a removable medium (such as a USB key, from the English "Universal Serial Bus", or a CD, from the English "Compact Disc" or a DVD, from the English "Digital Versatile Disc") readable by means of an appropriate reader of the computer system (such as a USB port or a CD and / or DVD disc drive).A computer program 1208 containing instructions for the processing unit 1202 is stored on the medium 1206 and / or downloadable via the network interface. This computer program 1208 is, for example, intended to be loaded into the main memory 1204, so that the processing unit 1202 executes its instructions to carry out the steps detailed previously.

[0093] Alternatively, all or part of the steps could be implemented in the form of hardware modules, i.e. in the form of an electronic circuit, for example micro-wired, not involving a computer program.

[0094] In conclusion, it should be noted that the invention is not limited to the embodiments described above. It will indeed be apparent to those skilled in the art that various modifications can be made to the embodiments described above, in light of the information just disclosed to them.

[0095] In the detailed presentation of the invention given above, the terms used shall not be interpreted as limiting the invention to the embodiments set forth in this description, but shall be interpreted as including all equivalents which can be foreseen by a person skilled in the art by applying their general knowledge to the implementation of the teaching which has just been disclosed to them.

Claims

Demands

1. Imager (104) comprising: - a spectral filter array (206) formed of a repeating pattern (MC) of spectral filters of at least three different spectral bands, of dimension NxN, with N an integer greater than or equal to three; - a polarizer array (210) of at least three different polarizations; - a photosensor array (212) each designed to receive light passed through one of the spectral filters and one of the polarizers and to measure a luminous intensity (IL) of the received light;and - a processing device (214) designed to, for each of the different spectral bands, form at least one image (CSo, Capl, Cdpl) of pixels having positions in the image (CSo, Capl, Cdpl) corresponding respectively to predefined positions (i, j) of a mask (M) in the matrix (212) of photosensors, each of the pixels being calculated from the light intensities (IL) measured by the photosensors of the spectral band considered contained in the mask (M) at the position (i,j) of the pixel;characterized in that the pattern (MC) of spectral filters is designed so that, for each of the different spectral bands and for all predefined positions (i, j) of the mask (M), the photosensors of the spectral band considered have an isobarycenter located in a central virtual cell (PVC) of the mask (M) of the same size as cells of the matrix (212) of photosensors, and in that each of the predefined positions (i, j) of the mask (M) is offset by another predefined position (i, j) of one column or one row of the matrix (212) of photosensors.;

2. Imager (104) according to claim 1, wherein the pattern (MC) of spectral filters presents, for each of the different spectral bands, a unique spectral filter of the spectral band considered in each row and each column of the pattern (MC) of spectral filters.

3. Imager (104) according to claim 2, wherein the pattern (MC) of spectral filters is the following: C1 C3 C2, where Cl, C2, C3 C2 Cl C3 C3 C2 Cl are spectral filters of three different spectral bands respectively, or the following: Cl C2 C3 C4, where Cl, C2, C3 C4 Cl C2 C4 C3 C2 Cl C2 Cl C4 C3 C3, C4 are spectral filters of four different spectral bands respectively, or the following: Cl C2 C4 C3 C5, C2 C5 C3 C4 Cl C3 Cl C5 Cl C4 C4 C3 Cl C5 Cl C5 C4 Cl Cl C3 where Cl, C2, C3, C4, C5 are spectral filters of five different spectral bands respectively.

4. Imager (104) according to any one of claims 1 to 3, wherein the polarizer matrix (210) is formed of a repeating pattern (MP) of polarizers of at least three different polarizations, of dimension NxN, the polarizer patterns being respectively aligned in the stack (204) with the patterns (MC) of spectral filters.

5. Imager (104) according to claim 4, wherein the polarizer pattern (MP) is as follows: P1 PI P3, where PI, P2, P3 are three PI P3 PI P3 PI PI polarizers following three different predefined polarizations respectively, or the following: PI P2 PI P2 or the following: PI PI PI PI PI, where PI, P2, P3, P4 are the polarizers of PI PI PI PI P3 P3 P3 P3 P4 P4 P4 P4 respectively four different polarizations, or the following: PI P3 P4 P2 P3 or the following: PI P2 P3 P4 P5, P2 P2 PI PI P4 PI P2 P3 P4 P5 P4 P2 PI P5 P5 PI P2 P3 P4 P5 P2 P5 P3 P4 PI PI P2 P3 P4 P5 P5 P3 P4 P5 P3 PI P2 P3 P4 P5

6.

7.

8. where PI, P2, P3, P4, P5 are five polarizers following five different predefined polarizations respectively. Imager (104) according to any one of claims 1 to 5, wherein the processing device (214) is designed to form, for each of the different spectral bands, an image, called a local polarization angle map (Capl), in which each pixel is the local polarization angle calculated from the light intensities (IL) measured by the photosensors of the spectral band considered contained in the mask (M) at the position (i,j) of the pixel. Imager (104) according to claim 6, wherein the processing device (214) is designed to form, for each of the different spectral bands: - an image, called a Stokes parameter map S0 (CSo), in which each pixel is the Stokes parameter S0 calculated from the light intensities (IL) measured by the photosensors of the spectral band considered contained in the mask (M) at the position (i,j) of the pixel; - an image, called a Stokes parameter map SI (Cs i), in which each pixel is the Stokes parameter S1 calculated from the light intensities (IL) measured by certain photosensors of the spectral band considered, contained in the mask (M) at position (i,j) of the pixel; and - an image, called a Stokes parameter map S2 (Cs 2), in which each pixel is the Stokes parameter S2 calculated from the light intensities (IL) measured by certain of the photosensors of the spectral band considered contained in the mask (M) at the position (i,j) of the pixel; and in which the processing device (214) is designed to form the local polarization angle map (Capl) from the three Stokes parameter maps S0, SI, S2 (CSo, CSi, CS2). Imager (104) according to claim 6 or 7, wherein the processing device (214) is designed to form a proximity map (Cprox) at the meridian (MD) of a luminous celestial body (106), by combining the maps of local polarization angles (CApl) of the different spectral bands, by means of a reading function (I) giving, for each predefined position (i,j), a sum over the different spectral bands, of terms (Ts), each term (Ts) being a function of the local polarization angle (APL) at this position (i,j) for the spectral band considered, increasing from 0° to 90° and decreasing from 90° to 180° as a function of the absolute value of the local polarization angle (APL).

9. Imager (104) according to claim 8, wherein each term (Ts) comprises a peak function (Xs) of the local polarization angle (LPA), increasing from 0° to 90° and decreasing from 90° to 180° as a function of the absolute value of the local polarization angle (LPA(i,j)), normalized to be one at 90° and raised to a power greater than or equal to one.

10. Imager (104) according to claim 9, wherein the peak function (Xs) is given by: _ g-cos(2 APLy) Xs — 2

11. Imager (104) according to claim 9 or 10, wherein the power is given by: 21 / 1^¾)] with [ ... J the floor function and flstat(Xs) a polynomial function of a statistical moment, such as the mean or the standard deviation, or of a quantile, such as the median or a quartile, of the peak function (Xs), for example f Istat ( Xs ) — 9 ( Xs ).

12. Imager (104) according to any one of claims 9 to 11, wherein the processing device (214) is designed to form, for each of the different spectral bands, an image, called a degree of linear polarization map (DPL), wherein each pixel is the degree of linear polarization calculated from the light intensities (IL) measured by the photosensors of the spectral band under consideration contained in the mask (M) at the position (i,j) of the pixel, and wherein each term (Ts) comprises a sky-likeness function (Fs) of the degree of linear polarization (DPL) at this position (i,j) for the spectral band under consideration, multiplied by the peak function (Xs) raised to the power.

13. Imager (104) according to any one of claims 8 to 12, wherein the processing device (214) is designed to analyze the proximity map (CPROx) to determine an azimuth (A) of a luminous celestial body (106) with respect to a direction (X) attached to the imager (104).

14. Celestial compass (102) comprising: - an imager (104) according to claim 13; - a positioning device (108) designed to provide a position of the imager (104); - a clock (110) designed to provide a current time; - an ephemeris (112) giving an azimuth (S) of the luminous celestial body (106) with respect to a predefined direction (NN), for example geographic North, as a function of the time and position; and - a heading determination device (114) designed to provide an angle, called a heading (C), between the direction (X) attached to the imager (104) and the predefined direction (N), from the azimuth (S) provided by the ephemeris (112) and the azimuth (A) determined by the imager (104).

15. Aircraft comprising an imager (104) according to any one of claims 1 to 13 or a celestial compass (102) according to claim 14.

Images