Polarimetric and multispectral imager, in particular for forming a celestial compass

Abstract

The invention relates to an imager (1) comprising: - an array of spectral filters formed of a repeating pattern of spectral filters; - an array of polarisers; - an array of photosensors; and - a processing device designed so that, for each of the different spectral bands, it forms at least one image of pixels having positions in the image that correspond, respectively, to predefined positions (i, j) of a mask (M) in the array of photosensors, wherein each of the pixels is calculated on the basis of the brightnesses measured by the photosensors for the stated spectral band contained in the mask (M) at the position (i, j) of the pixel. The pattern of spectral filters is designed such that, for each of the different spectral bands and for all of the predefined positions (i, j) of the mask (M), the photosensors for the stated spectral band have an isobarycentre located in a central virtual box (PVC) of the mask (M) having the same size as the boxes of the array of photosensors, and each of the predefined positions (i, j) of the mask (M) is offset from another predefined position (i, j) by a column or row of the array of photosensors.

Description

Description TITLE: POLARIMETRIC AND MULTISPECTRAL IMAGING DEVICE, PARTICULARLY 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 creating a celestial compass designed to determine its orientation (azimuth) relative to a given direction, for example, true north, by referring to the celestial body's ephemeris.

[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] We know from the prior art an imager of the type comprising: - a spectral filter matrix formed of a repeated 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 photosensor matrix each designed to receive light passed through one of the spectral filters and one of the polarizers and to measure a 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 luminous 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 commonly used in digital cameras, where the spectral filter pattern is most often an RGB pattern. This pattern typically consists of four filters for red, grouped in the upper left corner; four filters for blue, grouped in the lower right corner; and eight filters for green, half grouped in the lower left corner and the other half in the upper right corner. The mask is then successively positioned on each sub-array 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 that makes it possible to overcome at least some of the aforementioned problems and constraints.

[0006] Furthermore, US patent 2022 / 236463 A1 describes an imager comprising a spectral filter array, a polarizer array, and a photosensor array, stacked one on top of the other. US patent 2019 / 331762 A1 describes a method for obtaining a polarization map from an imager. Summary of the invention

[0007] 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 one column or one row of the photosensor matrix.

[0008] 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.

[0009] The invention may further include one or more of the following optional features, in any technically feasible combination.

[0010] Optionally, the spectral filter pattern presents, for each of the different spectral bands, a single spectral filter of the spectral band considered in each row and each column of the spectral filter pattern.

[0011] Optionally, the spectral filter pattern is also as follows: 1 3 2 2 1 3 , where C1, C2, C3 are spectral filters of three bands respectively 3 2 1 1 2 3 4 different spectral types, or the following: 2 4 3 2 1 where C1, C2, C3, C4 are 2 1 4 3 spectral filters of four different spectral bands respectively, or 1 2 4 3 5 2 5 3 4 1 the following: 3 1 5 2 4 , where C1, C2, C3, C4, C5 are spectral filters 4 3 1 5 2 5 4 2 1 3 of five different spectral bands respectively.

[0012] Optionally, at least one of the different spectral bands is either an ultraviolet spectral band or an infrared spectral band.

[0013] Optionally, at least one of the different spectral bands is one of: a spectral band described as red from 60 nm to 80 nm included in the interval [625 nm; 740 nm], a spectral band described as green from 60 nm to 80 nm included in the interval [520 nm; 565 nm], and a spectral band described as blue from 60 nm to 80 nm included in the interval [446 nm; 520 nm].

[0014] Optionally, the polarizer matrix is ​​formed from a repeating 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.

[0015] Optionally, the polarizer pattern is also as follows: !1 !2 !3 !2 !3 !1 where P1, P2, P3 are three polarizers following three respectively !3 !1 !2!1 !2 !1 !2 different predefined polarizations, or the following: !3 !4 !3 !4 !1 !2 !1 !2 or the following !3 !4 !3 !4 !1 !1 !1 !1: !2 !2 !2 !2 !3 !3 !3 !3, where P1, P2, P3, P4 are the polarizers of respectively !4 !4 !4 !4 !1 !3 !4 !2 !3 !2 !2 !1 !1 !4 four different polarizations, or the following: !4 !2 !1 !5 !5 or the following !2 !5 !3 !4 !1 !5 !3 !4 !5 !3 !1 !2 !3 !4 !5 !1 !2 !3 !4 !5: !1 !2 !3 !4 !5, where P1, P2, P3, P4, P5 are five polarizers following !1 !2 !3 !4 !5 !1 !2 !3 !4 !5 respectively five different predefined polarizations.

[0016] Optionally, the different predefined polarizations also include at least one of the following polarizations: 0°, 45°, 90° and 135°.

[0017] Optionally, the imager also includes a microlens array, so that the light received by each photosensor has passed through one of the microlenses.

[0018] Optionally, the processing device is also 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 in question contained in the mask at the pixel position.

[0019] Optionally, the processing device is also designed to form, for each of the different spectral bands: - an image, called the 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 in question contained in the mask at the pixel's position; - an image, called the Stokes parameter map S1, in which each pixel is the Stokes parameter S1 calculated from the light intensities measured by some of the photosensors of the spectral band in question contained in the mask at the pixel's position; and - an image, called the Stokes parameter map S2, in which each pixel is the Stokes parameter S2 calculated from the intensities luminous measured by some of the photosensors of the spectral band considered contained in the mask at the pixel position; and in which the processing device is designed to form the local polarization angle map from the three Stokes parameter maps S0, S1, S2.

[0020] Optionally, the processing device is also designed to form a proximity map to the meridian of a luminous celestial body, by combining the maps of local polarization angles of the different spectral bands, by means of a reading function giving, 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 from 0° to 90° and decreasing from 90° to 180° depending on the absolute value of the local polarization angle.

[0021] Optionally, each term also 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.

[0022] Optionally, the peak function is also given by: also, the power is given by: 2⌊7%#898(:;)⌋with ⌊… ⌋ the floor function and f1stat(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 >1?@A@("#) = 9〈"#〉.

[0024] Optionally, the processing device is also designed to form, for each of the different spectral bands, an image, called a degree of linear polarization map, in which each pixel is the degree of linear polarization calculated from the light intensities measured by the photosensors of the spectral band in question contained in the mask at the pixel position, and each term includes a sky-likeness function of the degree of linear polarization at that position for the spectral band in question, multiplied by the peak function raised to the power.

[0025] Optionally, the function of resemblance to the sky is also given by: with L# = MNO(MAP(1, Q!R), 0) and f2stat(Ys) 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 function, for example >2?@A@("#) = T(L#).

[0026] Optionally, the reading function provided by: with S being the different spectral bands.

[0027] Optionally, the processing device is also designed to analyze the proximity map to determine an azimuth of a luminous celestial body relative to a direction attached to the imager.

[0028] A celestial compass is also proposed comprising: - an imager according to the invention; - a positioning device designed to provide a position of the imager; - a clock designed to provide a current time; - an ephemeris giving an azimuth of the luminous celestial body with respect to a predefined direction, for example geographic North, as a function of the time and position; and - a heading determination device designed to provide an angle, called a 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.

[0029] Also proposed is an aircraft comprising an imager according to the invention or a celestial compass according to the invention. Brief description of the figures

[0030] 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: - Figure 1 is a schematic view of an aircraft equipped with a celestial compass fitted with an imager according to the invention, - Figure 2 is a side view of the imager, - Figure 3 is a top view of a spectral filter array of the imager, - Figure 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, - Figure 6 is a three-dimensional exploded view of the patterns of the filter and polarizer arrays and the sub-array of the photosensor array and a microlens array, - Figure 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, - Figure 8 is a view similar to Figure 7, with the mask in the second position in the photosensor array, - Figure 9 is a view similar to those of Figures 7 and 8, with the mask to a third position in the photosensor matrix,- 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 method implemented by an imager processing device; and - Figure 12 is a simplified view of an example embodiment of the processing device. Detailed description of the invention,

[0031] With reference to Figure 1, an aircraft 100, such as an airplane, equipped with a celestial compass 102 implementing the invention, will now be described.

[0032] The celestial compass 102 includes 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 direction Z. In the following description, the imager 104 will be described with reference to this vertical direction Z. In particular, positioning terms such as "Up", "down", "above", "below", will refer to the vertical Z direction.

[0033] The imager 104 is specifically 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.

[0034] The celestial compass 102 also 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).

[0035] The celestial compass 102 further includes 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 includes 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. .

[0036] 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.

[0037] With reference to Figure 2, the imager 104 first includes a top light inlet 202.

[0038] Below the upper light inlet 202, the imager 104 has a stack 204, along the vertical direction Z, of matrices 206, 208, 210, 212.

[0039] More specifically, the 204 stack comprises, from top to bottom, a 206 matrix of spectral filters, a 208 matrix (optional) of microlenses, a 210 matrix of polarizers and a 212 matrix of photosensors.

[0040] 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.

[0041] For example, the 210 polarizer array can be fabricated on the 212 photosensor array by oblique angle deposition. The 206 filter array and the 208 microlens array can then be applied on top.

[0042] Each photosensor is designed to measure a received light intensity, these measured light intensities being designated in Figure 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.

[0043] The imager 104 also includes a processing device 214 designed to receive the measured light intensities IL and to analyze them as will be described later.

[0044] Referring to Figure 3, the 206 matrix 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 206 matrix. 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, where N is an integer greater than or equal to three.

[0045] 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 specific frequency spectrum of sensitivity. This spectrum has a peak (i.e., a maximum value) and a full width at half maximum (FWHM). 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, filters with the same spectral band. In other words, the spectral filters in the same group have the same spectral band, with 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 by S.

[0046] When the spectral bands extend into the visible light spectrum, they correspond to colors, and spectral filters can be called color filters. In this case, matrix 206 can be described as a Color Filter Array (CFA). However, preferably, at least one of the spectral bands lies outside the visible spectrum, for example, in the ultraviolet (UV) or near-infrared (nIR) range. This allows for the discernment of polarization patterns in the sky even under degraded atmospheric conditions.For example, at least one of the S spectral bands is taken from: - an ultraviolet (UV) spectral band: spectral band included in the interval [350 nm; 380 nm]; - a near-infrared (nIR) spectral band: spectral band included in the interval [700 nm; 900 nm]; - a spectral band described as red: 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: 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: spectral band from 60 nm to 80 nm included in the interval [446 nm; 520 nm], with, for example, the peak at 450 nm.

[0047] Preferably, the MC pattern includes N different spectral filters (S = N).

[0048] It is possible to define, in the middle of the MC pattern, a central virtual pixel PVC, the size of the squares of the MC pattern. The spectral filters of the MC pattern are arranged such 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 squares in which the spectral filters of the considered spectral band are located.

[0049] For example, as illustrated in Figure 3, when N and S are equal to 1 2 3 4 Fourth, the MC pattern could be the following: where C1, C2, C3, C4 are 2 1 4 3 the four spectral filters for the four spectral bands respectively predefined. 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, the four spectral filters C3, and the four spectral filters C4.

[0050] Alternatively, when N and S are equal to three, the MC pattern can be the 1 3 2 following : 2 1 3 , where C1, C2, C3 are the three spectral filters for 3 2 1 respectively the three predefined spectral bands.

[0051] Alternatively, when N and S are equal to five, the MC pattern can be the 1 2 4 3 5 2 5 3 4 1 following : 3 1 5 2 4 , where C1, C2, C3, C4, C5 are the five spectral filters 4 3 1 5 2 5 4 2 1 3 for the five predefined spectral bands respectively.

[0052] Referring to Figure 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.

[0053] For example, when N equals 4 and the MP pattern comprises four different polarizers, the MP pattern can be as follows: P1, P2, P3, P4, and P4 are four polarizers with four predefined polarizations, 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.

[0054] 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 spectral filters C1 is aligned with a 0° polarizer, another of the filters Spectral C1 is aligned with a 45° polarizer, another C1 spectral filter is aligned with a 90° polarizer, and the last C1 spectral filter is aligned with a 135° polarizer. The same is achieved for spectral C2, C3, and C4 filters.

[0055] Alternatively, the MP pattern could be the following: !2 !2 !2 !2 !3 !3 !3 !3, where P1, !4 !4 !4 !4P2, P3, P4 are four polarizers following respectively the four predefined polarizations, for example: 0°, 45°, 90° and 135°.

[0056] Alternatively, when N equals 3 and the pattern MP contains three! 1 !2 !3 With different polarizers, the MP pattern could be as follows: !2 !3 !1 , where P1, P2, ! 3 !1 !2 P3 are three polarizers following three different predefined polarizations respectively.

[0057] Alternatively, when N equals 5 and the pattern MP has five different polarizers, the pattern MP could be: 4 2 1 5 5 or 2 5 3 4 1 5 3 4 5 3 1 2 3 4 5 1 2 3 4 5 or 1 2 3 4 5, where P1, P2, P3, P4, P5 are five polarizers following five different predefined polarizations respectively.

[0058] Referring to Figure 5, the 212 photosensor matrix is ​​divided into SM sub-matrices 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 208 microlens matrix, when present, is divided in the same way.

[0059] Referring to Figure 6, the MC filter patterns, MP polarizer patterns, microlens sub-matrices, and SM photosensor sub-matrices are aligned in the Z stacking direction. Thus, the light received by each photosensor passes through the spectral band filter and the polarizers located above this photosensor in the stacking direction Z, so as to be in one of the S spectral bands and according to one of the polarizations.

[0060] Referring 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 pixel can thus be calculated from the light intensities IL measured by the photosensors of the considered spectral band contained in the mask M at position (i,j) of the pixel, but not from the other photosensors contained in the mask M.

[0061] The MC 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, 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.

[0062] Thus, 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 image 1002 are therefore identified by the indices i (ranging from 1 to I) and j (ranging from 1 to J), these indices (i,j) equivalently representing the predefined positions of the mask M in the 212 photosensor matrix.

[0063] With reference to Figure 11, an example of method 1100 of operation of the imager 104 will now be described.

[0064] 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.

[0065] 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 (CS0), a Stokes parameter map S1 (CS1) and a Stokes parameter map S2 (CS2).

[0066] 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 the pixel position i,j. More precisely, in the illustrated case with four different polarizations, the Stokes parameter S0 is given by: where I0, I45, I90, I135 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°.

[0067] The value of each pixel in the Stokes parameter map S1 is the Stokes parameter S1 calculated from the light intensities IL measured by some of the photodetectors in the considered spectral band contained in the mask M at position i,j of the pixel. More precisely, in the case illustrated with four different polarizations, the Stokes parameter S1 is given by: 1 = Ug − Uig

[0068] The value of each pixel in the Stokes parameter map S2 is the Stokes parameter S1 calculated from the light intensities IL measured by some of the photodetectors in the considered spectral band contained in the mask M at position i,j of the pixel. More precisely, in the case illustrated with four different polarizations, the Stokes parameter S1 is given by: S2 = Ujk − U%lk

[0069] In 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 S1, 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, for example, given by:

[0070] In step 1108, the processing device 214 calculates, for each predefined spectral band and for each position (i,j), the local Stokes parameters S0l, S1l, S2l from the Stokes parameters S0, S1, S2 for that spectral band and that position. The local Stokes parameters S0l, S1l, S2l are, for example, given by is the angle of pixel (i,j) taking a midpoint of image 1002 as the origin.

[0071] In a step 1110, the processing device 214 calculates, for each predefined spectral band and for each position (i,j), a local APL polarization angle from the local Stokes parameters S0l, S1l, S2l for that spectral band and that position, in order to form a map C APL of local polarization angles APL. The local polarization angle APL is for example given by: u!R = %* AOvpw(_1p + x ∙ _2p), with x the imaginary number.

[0072] During step 1112, the processing device 214 calculates a proximity map C PROX at the meridian MD of the major illuminating star of the scene observed by the imager, that is to say of the luminous celestial body 106, by combining the C maps DPL degrees of linear polarization DPL and C cards APL local polarization angles APL for all predefined colors. For this, the processing device 214 uses a reading function I(i,j) which gives, at each position (i,j), a scalar number from the degrees of linear polarization DPL and the local polarization angles APL at that position (i,j), for all predefined colors.

[0073] Preferably, the reading function I(i,j) is given by: étantpolynomial 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, >1?@A@("#) = 9〈"#〉 and >2?@A@("#) = T(L#).

[0074] 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 meridian MD of the luminous celestial body 106. Therefore, all spectral bands are used, with the one containing the most information about the MD meridian (i.e., the closest APL value to 90°) being amplified.

[0075] For example, each Ts term includes the Xs function, called the peak function, of the local polarization angle APL. This peak function XS is increasing from 0° to 90° and The value of the local polarization angle (LPA) decreases from 90° to 180° as a function of the absolute value of the LPA, 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, LPA values ​​near 90° are close to one, while values ​​far from 90° quickly become very small (almost zero), especially the further they are from 90°. This allows the MD meridian to be highlighted more precisely than a simple thresholding around 90°.

[0076] The function F Sis 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 F S (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 F S can be omitted (i.e., taken as 1) or take different forms than the one indicated above.

[0077] 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 X direction attached to the imager 104. For this, step 1114 includes for example the following steps.

[0078] During a step 1114-2, the processing device 214 first analyzes the CPROX proximity map to determine the direction of the MD meridian.

[0079] In a first example, the processing device 214 can use the Hough transform given by: ^(P. cos(^) +^. sin(^) − ~)^P^^ with P = N − N'^^8^^ and ^ = V − V'^^8^^ and ^ defined as ^(0) = 1 ,^(P ≠ 0) = 0.

[0080] La transformée de Hough est donne en chaque point (~, ^) la quantité de lines present in I(i,j) oriented by (^ and passing at a distance ~ from the center of the image I(i,j). Thus, the processing device 214 then calculates R^(^) =∫^ UMv{r|vh(~, ^)^~ which corresponds to the number of lines oriented at (^ +^ & ^ *) of I(i,j) passing at a distance less than r from the center of I(i,j).

[0081] The value ^ for which R ^ (^) is maximal therefore indicates the orientation verified by the most lines of I(i,j) passing close to the center of the image.

[0082] Ideally, on noise-free and pollution-free images, the orientation of the MD meridian would therefore be (^ + ^*) with ^ for which R^^g(^) is maximum.

[0083] In practice, the images are noisy and polluted. Therefore, the 214 processing device 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), in order to obtain a curve R ^^g (^) which is a "fuzzy" or "smoothed" version of the R curve ^^g ( ^ ) This removes some of the noise present in the image of the read function. From this first iteration R ^^g(^), the processing device 214 calculates ^^9^ such that MAP(R^^g) = R^^g(^^9^). This is a first approximation of the orientation of the MD meridian. The processing device 214 then determines a study interval around ^^9^, [^%, ^*] with R^^g(^*) = R^^g(^%) =_w|NpUO@w^^Appw ∗ R^^g(^^9^). The processing device 214 then studies the curve R ^^g (^) only on this interval, to find the orientation at of the MD meridian which is at its maximum, or close to its maximum due to noise, of R ^^g (^). The first iteration avoids concentrating around a peak of R g ( ^ ) which would not be the desired direction but a peak due to noise.

[0084] In this case, the processing device 214 first performs, for example, thresholding of I(i,j) to keep only the points whose gray level / proximity level is greater than the chosen / predefined threshold proximity level. Processing 214 calculates the direction ^ ^^^ of the MD meridian by:

[0086] In a second example, the processing device 214 performs a Ridgelet transform on 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.

[0087] In a third example, the processing device 214 compares the CPROX proximity map to predefined proximity maps, each for which the sun's position is known. These predefined proximity maps preferably correspond to different cloud cover and field-of-view obstruction conditions. For example, these predefined proximity maps are generated from simulations of scenes under various cloud cover and vegetation obstruction conditions. or urban, with the common point of a fixed constant azimuth of the luminous celestial body 106, for example at 0°.

[0088] Preferably, predefined proximity maps are expressed in polar coordinates. In this case, the processing device 214 performs, before comparison, a polar coordinate transform of the proximity map C PROX .

[0089] The comparison is, for example, carried out 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 maximum of the convolutions, these maximums corresponding to the angular difference between the orientation of the sought-after MD meridian and the known orientation of the MD meridians in the predefined proximity maps (for example 0°).

[0090] 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.

[0091] Thus, during a step 1114-4, the processing device 214 removes this ambiguity to obtain the direction of the sun.

[0092] 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 in 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 Sg^^ / Sg is the lowest corresponds to the portion of the ¡¢¡£¤¥ meridian MD pointing towards the luminous celestial body 106. Conversely, the portion of the meridian MD where the ratio Sg^^¦ is the strongest corresponds to the portion of the MD meridian pointing at 180° from the luminous celestial body 106.

[0093] With reference to Figure 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, or 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 drive) or a remote medium (such as a remote hard drive accessible via the network interface through a network of communication) or a removable storage medium (such as a USB flash drive, a CD, a Compact Disc, or a DVD, a Digital Versatile Disc) readable by means of a suitable drive on the computer system (such as a USB port or a CD and / or DVD drive). A computer program 1208 containing instructions for the processing unit 1202 is stored on the storage medium 1206 and / or downloadable via the network interface. This computer program 1208 is, for example, intended to be loaded into main memory 1204, so that the processing unit 1202 executes its instructions to perform the steps detailed above.

[0094] 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.

[0095] In conclusion, it should be noted that the invention is not limited to the embodiments described above. Indeed, it will 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 provided.

[0096] 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

Claims [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 passing through one of the spectral filters and one of the polarizers and to measure a luminous intensity (LI) of the received light; and - a processing device (214) designed to, for each of the different spectral bands, form at least one image (C S0 , C APL , C DPL ) of pixels having positions in the image (C S0 , C APL , C DPL) 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 a column or a row of the matrix (212) of photosensors.[2] Imager (104) according to claim 1, wherein the pattern (MC) of spectral filters has, 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. Imager (104) according to claim 2, wherein the pattern (MC) of filters 1. 3 2 spectral is as follows: 2 1 3 , where C1, C2, C3 are spectral filters of 3 2 1 respectively three different spectral bands, or the following: 1 2 3 4 4 1 2 3 2 1 , where C1, C2, C3, C4 are spectral filters of respectively 2 1 4 3 1 2 4 3 5 2 5 3 4 1 four different spectral bands, or the following: 3 1 5 2 4 , where 4 3 1 5 2 5 4 2 1 3C1, 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 array (210) is formed of a repeating pattern (MP) of polarizers of at least three different polarizations, of dimension NxN, the polarizer patterns being aligned in the stack (204) with the patterns (MC) of spectral filters, respectively. [5] Imager (104) according to claim 4, wherein the polarizer pattern (MP)! 1 !2 !3 is as follows: !2 !3 !1 where P1, P2, P3 are three following polarizers! 3 !1 !2 respectively three different predefined polarizations, or the following: !2 !5 !3 !4 !1 !1 !2 !3 !4 !5!5 !3 !4 !5 !3 !1 !2 !3 !4 !5P5 are five polarizers following respectively five different predefined polarizations. [6] 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 (LI) measured by the photosensors of the spectral band considered contained in the mask (M) at the position (i,j) of the pixel. [7] 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 (C S0), 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 position (i,j) of the pixel; - an image, called the Stokes parameter map S1 (C S1 ), in which each pixel is the Stokes parameter S1 calculated from the light intensities (IL) measured by some of the photosensors of the spectral band under consideration contained in the mask (M) at position (i,j) of the pixel; and - an image, called the Stokes parameter map S2 (C S2), in which each pixel is the Stokes parameter S2 calculated from the light intensities (IL) measured by some 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, S1, S2 (CS0, CS1, CS2).[8] 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 local polarization angle maps (CAPL) of the different spectral bands, by means of a readout 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 that 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: according to claim 9 or 10, wherein the power is 2⌊7%#898(:;)⌋with ⌊… ⌋ the floor function and f1stat(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 >1?@A@("#) = 9〈"#〉.[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 (DLPL), wherein each pixel is the degree of linear polarization 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, and wherein each term (Ts) comprises a sky-likeness function (Fs) of the degree of linear polarization (DLP) at this position (i,j) for the spectral band considered, 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 the current time; - an ephemeris (112) 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; and - a heading determination device (114) designed to provide an angle, called heading (C), between the direction (X) attached to the imager (104) and the direction. predefined (N), based on 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