Multispectral imaging device

Abstract

A multispectral imaging device (1) comprising: - an imaging system (3), - a measurement system (5) comprising: -- a filter array (21) arranged in a plane of an exit pupil (14) of the imaging system, -- a lens array (7) located in an image plane (9) of the imaging system, and -- a pixel array (11), each lens (7a, 7b) of the lens array being configured to image each filter (21a, 21b) on one pixel (11a, 11b, 11c, 11d), each filter being configured to be imaged by each lens on one pixel, the pixel array comprising a number of pixels greater than or equal to a product of the number of filters in the filter array and the number of lenses in the lens array. Figure to be published for the abstract: Figure 1

Description

Title of the invention: Multispectral imaging device DOMAIN

[0001] The invention relates to the field of multispectral imaging and in particular in the infrared field. STATE OF THE ART

[0002] Multispectral imaging is a technique for acquiring images of the same scene in several spectral bands, that is, several ranges of wavelengths. The acquisition of images in the different spectral bands can be simultaneous or staggered over time.

[0003] Multispectral imaging devices are known which are based on the use of an optical lens array and a pixel array, each optical lens being configured to transmit light to a group of pixels.

[0004] Filtering can be achieved by placing spectral filters on the pixels. The light arriving at a pixel is spectrally filtered compared to all the light collected by the device. It corresponds to a particular spectral band. These filters are most often permanently attached to the pixels, making it difficult to modify the desired spectral filtering without changing the detector. Furthermore, manufacturing these filters requires relatively complex technologies, such as the deposition of dielectric layers, or technologies that lead to unsatisfactory filter performance, such as the etching of nanostructures. These operations are particularly complex if a sensitive imaging device in the infrared wavelengths is desired.Finally, imaging devices incorporating this type of filter have drawbacks regarding the multispectral nature of the resulting images: some of the light intended for one pixel may actually be transmitted to a neighboring pixel, thus reducing the quality of the spectral filtering. To address this, a lens array can be placed between the filter array and the pixel array to improve the collection of light seen by each pixel.

[0005] Filtering can also be achieved by placing different spectral filters in different areas of the imaging device's lens, located upstream of the actual detection system. It is possible to configure the entire device, called a plenoptic camera, so that a group of pixels receives light passing through only one of these lens areas.

[0006] A plenoptic camera is understood to be a system which, by adding a component (lens array or prism array or a combination of both) upstream or downstream of a diaphragm of the system, allows the diaphragm to be subdivided into several zones. For each zone of the diaphragm, the light rays passing through the zone They illuminate downstream only a portion of the pixels in the imaging system. In this way, each pixel receives only the light rays that have passed through a single area of ​​the diaphragm.

[0007] These devices tend either to subdivide the diaphragm too much, thus reducing the number of pixels associated with each zone, which is problematic in the infrared range due to the smaller number of pixels per detection array, or to require a modification of the focal length of the imaging system. Furthermore, these devices comprise a lens and a detection system with a lens array or a prism array or a combination of both, the sizing parameters and configurations of the lens and the detection system being closely related. It is therefore always necessary to adapt the lens to the detection system and to the lens array or prism array or combination of the two.

[0008] There is therefore a need for a multispectral imaging device which allows the use of a range of standard lenses, whose spectral filtering can be easily modified and whose manufacture is simple, especially if a device sensitive in infrared wavelengths is desired. EXPOSED

[0009] One aim of the present presentation is to propose a multispectral imaging device that is simpler than in the prior art.

[0010] The goal is achieved by means of a multispectral imaging device comprising:

[0011] - an imaging system configured to form an image of an object,

[0012] - a measuring system comprising:

[0013] — a filter matrix arranged in a plane of an output pupil of the system imaging,

[0014] — a lens array located in an image plane of the imaging system, and

[0015] — a pixel matrix, and

[0016] each lens of the lens matrix being configured to image each filter on a pixel, so that each lens is associated with a first number of pixels equal to a first number of filters in the filter matrix, each filter being configured to be imaged by each lens on a pixel, so that each filter is associated with a second number of pixels equal to a number of lenses in the lens matrix, the pixel matrix comprising a number of pixels greater than or equal to a product of the first number and the second number.

[0017] Such a device is advantageously and optionally complemented by the following various features, taken alone or in combination: - the number of filters in the filter matrix is ​​equal to four or nine; - a substrate occupying a volume separating the lens matrix and the pixel matrix, the substrate having an optical index greater than or equal to 1, preferably greater than or equal to 1.5 and even more preferably greater than or equal to 2; - for each lens, the pixels on which the lens is configured to image the filters of the filter array form a group of pixels, the lens being offset with the group of pixels by a gap A measured orthogonally to an optical axis of the multispectral imaging device, the gap A being equal to the ratio between a product of a distance measured along the optical axis separating the lens and the group of pixels by a distance from a center of the lens to the optical axis, and a product of an optical index of the medium separating the lens and the group of pixels by a distance measured along the optical axis separating the lens array from the exit pupil, the lens being closer to the optical axis than the group of pixels; - a hood surrounding the measurement system, the measurement system comprising a cryostat configured to cool the pixel matrix and the lens matrix, the measurement system comprising a screen located between the pixel matrix and the hood, the cryostat being configured to cool the screen, the screen defining an optical diaphragm upstream of the lens matrix, the filter matrix being arranged in the optical diaphragm; - for each filter in the filter matrix a filter mask placed against the filter, the filter mask being configured to optically mask a peripheral portion of the filter; - for each filter in the filter matrix, the filter is stopped down by the filter mask so that an image of the filter stopped down by a lens on a pixel is located at a minimum distance from each edge of the pixel, the minimum distance being greater than or equal to a scattering length of a photon in the pixel; and - for at least one filter in the filter matrix, the filter mask is off-center with respect to the filter, such that a spacing between the center of the filter and a center of the exit pupil is greater than a spacing between the center of the mask and the center of the exit pupil.

[0018] The presentation also relates to a multispectral imaging method for an object comprising: - a step of constructing an image of the object using an imaging system on a lens array and light from the object,

[0019] - a spectral filtering step of light by a filter matrix arranged in a plan of an exit pupil of the imaging system,

[0020] - an optical conjugation step of a plane of the filter matrix and a plane from a pixel matrix to a lens matrix,

[0021] - a light measurement step using the pixel matrix,

[0022] each lens of the lens matrix being configured to image each filter on a pixel, so that each lens is associated with a first number of pixels equal to a first number of filters in the filter matrix, each filter being configured to be imaged by each lens on a pixel, so that each filter is associated with a second number of pixels equal to a number of lenses in the lens matrix, the pixel matrix comprising a number of pixels greater than or equal to a product of the first number and the second number. DESCRIPTION OF THE FIGURES

[0023] Other features and advantages will become apparent from the following description, which is purely illustrative and not limiting, and should be read in conjunction with the accompanying drawings on which:

[0024] [Fig.1] [Fig.1] is a schematic representation of an example of a multispectral imaging device;

[0025] [Fig.2] [Fig.2] is a schematic representation of a detail of an example of a multispectral imaging device;

[0026] [Fig.3]

[0027] [Fig. 4] Figures 3 and 4 are schematic representations of the image of the exit pupil by a microlens in the plane of a photodetector array of an example of a multispectral imaging device; and

[0028] [Fig.5] [Fig.5] is a schematic representation of an example of a multispectral imaging process. DETAILED DESCRIPTION OF THE INVENTION

[0029] With reference to [Fig. 1], a multispectral imaging device 1 comprises an imaging system 3 configured to form an image of an object 2. The imaging system 3 collects a portion of the radiation, and in particular the infrared radiation, emitted by the object 2. The imaging system forms an image of the object 2 in an image plane 9 of the imaging system 3. The imaging system is centered on the optical axis X, an optical axis that can be oriented in the direction of light propagation, i.e., from the object 2 towards the imaging system 3 or from the imaging system 3 towards the image plane 9. The imaging system 3 has an exit pupil in a pupillary plane 14 which is upstream of the image plane 9, i.e., the exit pupil is located between the imaging system 3 and the image plane 9.

[0030] The exit pupil of the imaging system 3 is understood here as the image of the aperture diaphragm of the imaging system 3 by all the optics of the system The imaging system 3 is located downstream of the aperture diaphragm, or, where applicable, the aperture diaphragm itself when it is a single diaphragm positioned downstream of all the optics. The diaphragm is the mechanical element of the imaging system 3 that determines the maximum angle of the cone of rays reaching a point on the image plane 9 and originating from a point in the field and passing through the imaging system 3. The maximum angle allows the calculation of a working aperture number, designated by WF#.

[0031] The imaging system may include, in particular, a photographic lens.

[0032] The multispectral imaging device 1 also includes a measurement system 5 located downstream of the imaging system 3 with respect to the direction of light propagation. In other words, the imaging system 3 is placed between the object 2 and the measurement system 5.

[0033] The measuring system 5 is centered on the optical axis X.

[0034] The measuring system 5 comprises a filter matrix 21 arranged in the plane The pupillary 14 output of the imaging system 3. The filter array comprises a plurality of filters 21a, 21b, etc., arranged in rows and columns, the rows and columns being orthogonal to each other and to the optical axis X. The filter array 21 is centered on the optical axis. Each filter in the filter array 21 is configured to spectrally filter a particular spectral band of the incident light, that is, to allow only a specific wavelength range of the incident light to pass downstream. The filters can be chosen so that their wavelength ranges do not overlap.

[0035] The filters can be square and joined end to end to form a filter matrix. In particular, they can be made separately and then butted together, that is, fixed end to end, glued to one another, or assembled in a mount. If a filter is identified as defective, it is then simpler to replace than in the prior art.

[0036] Alternatively, the filters can also be made on the same optical plate after a succession of thin film deposition steps or photolithography steps.

[0037] This fixing can in particular be removable, that is to say it is easy to separate one of the filters from the rest of the matrix and replace it with another filter.

[0038] The number of filters in the filter matrix 21 can be either four or nine. When the matrix contains four filters, these are arranged in four squares forming a larger square, such that the matrix is ​​centered on a junction between the four filters. When the matrix contains nine filters, these are arranged in nine squares forming a larger square, such that the matrix is ​​centered on a central filter surrounded by the other eight filters.

[0039] The measurement system 5 includes a lens matrix 7 located in the image plane 9 of the imaging system 3. The lens matrix comprises a plurality of lenses 7a, 7b, 7c distributed in rows and columns, the rows and columns being orthogonal to each other and orthogonal to the optical axis X. The lens matrix 7 is centered on the optical axis.

[0040] The lenses may in particular be microlenses, that is to say lenses having a diameter less than or equal to 100pm.

[0041] Since the lens matrix 7 is located in the image plane 9, each lens receives light from a specific portion of the object plane 2.

[0042] Each lens has an aperture number which is designated by F#pL. All lenses having the same aperture number.

[0043] The measurement system 5 comprises a pixel array 11 located downstream of the lens array 7, i.e., the lens array 7 is located between the imaging system 3 and the pixel array 11. For example, the pixel array is a photosensitive array of a photographic sensor based on a charge-transfer device. The pixels are sensitive to infrared radiation, the sensor being configured to produce images in this spectral range.

[0044] The photodetector matrix 11 comprises a plurality of pixels 1la, 11b, 1le distributed in rows and columns, the rows and columns being orthogonal to each other and orthogonal to the optical axis X. The optical axis X passes through the pixel matrix 11.

[0045] Each lens 7a, 7b, of the lens array 7 is configured to image each filter 21a, 21b on a pixel 1la, 11b, or on a pixel 1le, 1Id, such that each lens is associated with a first number of pixels equal to a first number of filters in the filter array. More precisely, lens 7a is configured to image filter 21a on pixel 1la and filter 21b on pixel 11b. Lens 7b is configured to image filter 21a on pixel 1Id and filter 21b on pixel 1le. The filter array contains the first number of filters. For each lens in the lens array, the lens images the first number of filters on a first group of pixels. This first group of pixels comprises pixels adjacent to each other. Two pixels are adjacent if they are directly neighbors, that is, there cannot be a pixel located between two adjacent pixels.Each pixel in the first group of pixels receives light from only one of the filters in the filter array and the lens.

[0046] When the filter matrix comprises four filters, each lens is associated with a group of four pixels. This first group can be designated as a 'quad-pixel'.

[0047] When the filter matrix comprises nine filters, each lens is associated with a group of nine pixels. This first group can be designated as a 'nona-pixel'.

[0048] Each filter is configured to be imaged by each lens on a pixel, so that each filter is associated with a second number of pixels equal to a number of lenses in the lens matrix.

[0049] The lens matrix 7 comprises the second number of lenses. For each filter in the filter matrix, each lens forms an image of the filter on a pixel of the pixel matrix, all these pixels being different from one another. There are as many pixels receiving light from the filter and as many images of the filter constructed on the pixel matrix as there are lenses in the matrix; that is, the number of images of the filter is equal to the second number, or in other words, the number of pixels receiving light from the filter is equal to the second number. This second group of pixels associated with the filter comprises pixels that are not adjacent to each other. Each pixel associated with a filter belongs to a first group of pixels. The pixels associated with a filter all belong to a first group of pixels that differs from one pixel to another.

[0050] The pixel matrix comprises a number of pixels greater than or equal to a product of the first number and the second number. In other words, the number of pixels in the matrix is ​​at least equal to the product of the number of filters in the filter matrix and the number of lenses in the lens matrix.

[0051] Optionally, the multispectral imaging device 1 includes a substrate 15 occupying a volume separating the lens array 7 and the pixel array 11, the substrate 15 having an optical index greater than or equal to 1, preferably greater than or equal to 1.5, and even more preferably greater than or equal to 2. By choosing a high-index material separating the lens array 7 and the photodetector array 11, it is easier to obtain a very low aperture number for the lens. This is an advantage for an infrared configuration compared to the visible range. Furthermore, when the lenses are microlenses, their diffractive effects can be limited by configuring the system to define the highest possible generalized Fresnel number (FN) thanks to the material between the microlens and the detector having a high refractive index.The description further includes a mathematical relationship defining the generalized Fresnel number.

[0052] In operation, the imaging system 3 constructs an image of the object 2 on the lens array 7 such that each lens receives light from a region of the object. Before reaching the lens, the light from the object 2 passes through the filter array so that downstream of the filter array, each filter only allows the portion of light corresponding to its spectral bandwidth to pass through. Each lens is associated with a first group of pixels and constructs the image of a filter on one of the pixels of the first group. Each pixel of the first group receives only light light is emitted from the lens and one of the filters in the filter array. Thus, each pixel receives light from the object originating from a specific area of ​​the object, this light corresponding to the spectral band associated with the filter. The device therefore performs multispectral imaging by determining, for each area of ​​the imaged object, its light emission in different spectral ranges.

[0053] In relation to [Fig. 1], lens 7a is associated with pixels 1a and 11b, which are part of a quad-pixel associated with lens 7a. The filter matrix 21 includes, in particular, filters 21a and 21b.

[0054] The light beam from filter 21a arriving at lens 7a is defined between rays 27 and 28. Rays 27 and 28 are identified by double arrows oriented according to the direction of light propagation. This entire light beam is directed by lens 7a towards pixel lia.

[0055] The light beam from filter 21b arriving at lens 7a is defined between rays 29 and 30. Rays 29 and 30 are marked by triple arrows oriented according to the direction of light propagation. This entire light beam is directed by lens 7a towards pixel 11b.

[0056] The flux received by each pixel of a first group of pixels (quad-pixel or nona-pixel, for example) comes from the entire associated lens. This avoids blind spots in the sampling of the object to be imaged, unlike a device comprising a Bayer matrix.

[0057] Each lens behaves like a "macro-pixel" and it is the lenses of the lens matrix that sample the object space.

[0058] The filter array is positioned upstream of the lens array and the pixel array. The filters are then no longer attached to the pixels: their manufacture can be simplified, and it is possible to change the filters without completely modifying the device, in particular the lens or the pixel array. The filters are shared with their associated pixels, and therefore there is not a filter in front of each pixel. This results in fewer filters to manufacture, and they are larger. The filters are thus easier to produce, and more complex filtering patterns can be considered.

[0059] Furthermore, by arranging a filter array in an exit pupil plane of the imaging system and a lens array in an image plane of the imaging system, the spectral distribution function is performed downstream of the imaging system. In this way, it is possible to use a standard imaging system, such as a standard lens, to collect and shape the light. A standard imaging system is understood to mean, in particular, a system compatible with a cryogenic infrared detection block. It should be noted that such a system, standard in the infrared, is not standard in the visible spectrum because it is not common to have an accessible exit pupil when working in the visible spectrum. When working In the infrared, such standard systems make it possible to design a whole range of lenses that can be interchanged without having to adapt the detection block.

[0060] The filters can be made in a standard way (for example, by stacking dielectric layers), and they can be made separately and then either assembled on a mechanical mount or butted together to form a filter matrix. The efficiencies are significantly higher than those of prior art techniques.

[0061] By separating the filter part from the lens matrix part, the realization of the two components on the detector is technologically less delicate.

[0062] The lens array can be produced in a limited number of steps, which improves efficiency. A mask array placed in the pupillary plane 14 limits chromatic overlap between pixels. The multispectral imaging device allows a pixel to be associated with a filter and all pixels of the pixel array to be used.

[0063] The measurement system 5 can be chosen as generic in the sense that it is compatible with any imaging system 3 or lens having an exit pupil coinciding with the diaphragm of the measurement system (lenses of different focal lengths, zoom...).

[0064] It should be noted that in the prior art, plenoptic multispectral cameras have been proposed in the visible spectrum by placing a filter array at the diaphragm of a lens. However, the operation of these cameras relies on the use of lens arrays or pinhole arrays decoupled from the detector, which presents 3D imaging challenges. The image of a filter on the pixel array is then formed across several pixels. It is possible, for example, to sum the signals produced by these pixels associated with the same filter to form a single chromatic signal associated with the filter's spectral range. This type of configuration requires a large number of pixels in the pixel array and is not compatible with infrared multispectral imaging because infrared-sensitive pixel arrays contain far fewer pixels than visible-sensitive pixel arrays.This approach is therefore not valid in the infrared range, which is limited in the number of pixels. Furthermore, such a prior art system presents several difficulties: there is significant spectral overlap between the pixels, and since the filter array is located at the aperture of the optical system, its dimensional characteristics are linked to the objective lens, making it impossible to use a generic imaging system. The multispectral imaging device described above is significantly more efficient and better suited to multispectral imaging in the infrared range than these prior art systems.

[0065] As mentioned previously, each lens is associated with a first group of pixels, each pixel receiving only light from the lens. These are the pixels on which the lens is configured to image the filters of the filter array. Advantageously, the lens can be offset from the group of pixels by a gap A measured orthogonally to an optical axis of the multispectral imaging device, the gap A being equal to the ratio between a product of the distance measured along the optical axis separating the lens from the group of pixels by the distance from the center of the lens to the optical axis, and a product of an optical index of the medium separating the lens and the group of pixels by a distance measured along the optical axis separating the lens array from the exit pupil, the lens being closer to the optical axis than the group of pixels.

[0066] This condition can be noted in mathematical form, based on the notations used in [Fig.2]: A = (H2-H1) =(Hl*L2) / (n*Ll).

[0067] The length H2 separates the center of the pixel group associated with the lens 7c from the optical axis X, the length H1 separates the center of the lens 7c from the optical axis X, n is the optical index of the medium separating the lens and the pixels, L1 is the distance separating the lens matrix from the exit pupil.

[0068] More specifically the distance L2 separating the lens from the pixel group is defined between the vertex of the lens and the pixel group, the vertex of the lens being the vertex of the lens surface furthest from the pixels, the vertex being centered with respect to the lens.

[0069] Figure 2 illustrates this situation and in particular a light ray 43 passing through the center of a lens 7c and the center of the exit pupil, that is to say the intersection between the exit pupil plane 14 and the optical axis X. Since the lens 7c is not centered on the optical axis X, the light ray 43 arrives at a lens at a non-zero incidence i with the direction of the optical axis X.

[0070] The center of the lens 7c is separated from the optical axis X by a length referenced H1 in [Fig.2]. The length L1 separates the exit pupil 14 and the lens array 7.

[0071] Downstream of lens 7c, the light ray 43 defines an exit angle r (or angle of refraction r) with the direction of the optical axis X. The exit angle r may be different from the angle of incidence i, in particular if the refractive index ni of the medium upstream of lens 7c is different from the refractive index n of the medium 15 downstream of lens 7c. In this case, the angles of incidence i and refraction r are related by Snell's law nl*sin(i)=n*sin(r).

[0072] The light ray impacts the pixel matrix after traveling a length L2. This distance separates the lens matrix 7 and the pixel matrix 11. Figure 2 represents the case where the filter matrix comprises nine filters such that each lens is associated with a nona-pixel. The light ray 43 then arrives at the central pixel 1 of the nona-pixel associated with lens 7c.

[0073] The pixel group associated with lens 7c (here the nona-pixel centered on pixel 1 le) is not centered on the optical axis X, so that the center of the pixel group is separated from the optical axis X by a length referenced H2 in [Fig.2].

[0074] There is a difference A between the length H2 separating the center of the pixel group associated with lens 7c from the optical axis X and the length H1 separating lens 7c from the optical axis X. H2 is greater than HL

[0075] This deviation A can be determined according to the parameters of the measurement system 5: A = (H2-H1) =(Hl*L2) / (n*Ll).

[0076] In other words, each lens is offset from the center of its associated pixel group by a gap A measured orthogonally to the optical axis X, the gap A being equal to the ratio between a product of the distance measured along the optical axis separating the lens from the pixel group by a distance from a center of the lens to the optical axis, and a product of an optical index of the medium separating the lens and the pixel group by a distance measured along the optical axis separating the lens matrix from the exit pupil, the lens being closer to the optical axis than its main photodetector.

[0077] This amounts to choosing a spatial period in the lens matrix 7 that is not a multiple of the spatial period in the pixel matrix 11. These two periods are not in an integer ratio.

[0078] The technical effect associated with this particular choice of periods is to ensure that the image of a filter by the lens is located within the surface of a pixel.

[0079] It should be noted that in more specific cases, the light rays arriving at the lenses may have an incidence of zero or sufficiently low to be neglected. In these cases, each lens is aligned with the center of its associated pixel group along a direction parallel to the optical axis. This so-called telecentric embodiment corresponds to the case where the average angles of the light cones from the imaging system 3 arrive at the measurement system 5 with very small angles of incidence. It is as if the exit pupil 14 were very far from the lens array.

[0080] The multispectral imaging device 1 may advantageously include a cooling device for all the elements integrated within the area defined by the wall 17, and in particular the pixel array, the lens array, the filter array, and the wall 17 itself. The cooling device may, for example, include a cryostat 13 configured to cool the pixel array 11.

[0081] With reference to [Fig. 1], the multispectral imaging device may include a hood 45 surrounding the measurement system 5. The hood 45 may, in particular, allow to fix a window 23 located between the imaging system 3 and the filter matrix. The window 23, which is included in the measuring system 5, is sufficiently transparent to allow the light flux from the imaging system 3 to pass through without significant absorption.

[0082] The hood 45 is not cooled by the cryostat 13.

[0083] Still with reference to [Fig.1], the multispectral imaging device may include a screen 17 located between the pixel matrix 11 and the hood 45, the cooling device represented in part by the cold table 13 being configured to cool the screen 17.

[0084] The screen 17 is a cold screen that blocks a portion of the radiation from the hood 45 so that the pixels do not receive light emitted by the hood 45. Similarly, the screen 17 also blocks rays from the object plane known as 'grazing' rays, that is, rays with a very high angle of incidence with the optical axis and which are not part of the field of view of the imaging system 13 or do not pass through the imaging system 13. These are, for example, rays passing through a pixel associated with a first lens and a second lens adjacent to the first lens. The screen 17 thus protects the main photodetector from these high-angle rays and improves the quality of the images produced.

[0085] The screen 17 can advantageously define an optical diaphragm upstream of the lens array, the filter array being positioned at the optical diaphragm, for example, within the optical diaphragm itself or within a ring or band that represents the diaphragm. The filter array can be positioned slightly upstream or downstream of the optical diaphragm relative to the optical axis. The screen 17 thus serves to define the exit pupil of the imaging system. In this case, the screen 17 also serves as a support for the filter array.

[0086] Optionally, the multispectral imaging device may include for each filter in the filter array a filter mask disposed against the filter, the filter mask being configured to optically mask a peripheral portion of the filter.

[0087] A peripheral portion of the filter extends from the outer perimeter of the filter towards the inside of the filter in a plane orthogonal to the axis. In other words, the peripheral portion extends from an edge of the filter or from edges of the filter towards the center of the filter such that a central portion of the filter allows light to pass through while the peripheral portion, optically masked by the mask, does not allow light to pass through. The central and peripheral portions are complementary and together constitute the surface of the filter.

[0088] The central portion of the filter may include the center of the filter.

[0089] Such a filter with a mask is a diaphragmed filter whose image through the lens is reduced and does not illuminate the entire pixel. This ensures that the image The structure created by the lens is precisely localized to a single pixel, limiting the scattering of charge carriers to adjacent pixels. This improves the quality of the multispectral imaging performed.

[0090] A shape of the mask can be chosen so that the central portion has the shape of a disk, such as a disk centered on the center of the filter or of a square, such as a square centered on the center of the filter. When the filter matrix comprises four square filters arranged in a larger square, each filter can be masked by two strips, each strip extending along an edge adjacent to a neighboring filter. The filter matrix behaves as if it were masked by a cross passing through its center and following the edges of the filters.

[0091] The size of the central portions defined in the filters in the pupillary plane 14 is related to the size of the images of the diaphragmed filter on the pixel according to the following relation:

[0092]

[0093] With 0 the size of the central portion in the pupillary plane 14, t the size of the image of the diaphragmed filter at the pixel level, n the optical index of the medium separating the lens and the group of pixels, L2 is the distance between the lens and the photodetector matrix and L1 designates the distance between the pupillary plane 14 and the lens matrix 7.

[0094] Each mask can be part of the frame of its filter. Each mask can be placed directly onto its filter.

[0095] The filter array and associated masks are located within the diaphragm defined by the screen. They are situated where, in the prior art, a single cold bandpass filter is placed. This does not significantly alter the mass to be cooled compared to the prior art, so the temperature descent time of the multispectral imaging device is not significantly modified.

[0096] It can be imposed that, when the filter is diaphragmed by a mask, the image of the filter by a lens on a pixel is located at a minimum distance from each edge of the pixel, the minimum distance being greater than or equal to a scattering length of a photon in the pixel.

[0097] In a more general case, we can consider the situation where the image of the filter is separated from an inner edge of the pixel by a distance greater than the scattering length of a photon in the pixel.

[0098] More advantageously, it can be further imposed that the minimum distance is greater than or equal to the sum of a scattering length of a photon in the pixel and a characteristic diffraction length due to the lens in a plane of the pixel.

[0099] By requiring that the minimum distance be greater than the scattering length of a photon in the pixel, the chromatic overlap between pixels due to the Charge carrier scattering between pixels. Each pixel receives photons from a specific filter. If a photon scatters from its destination pixel to a neighboring pixel, the multispectral quality of the resulting images is degraded: the destination pixel will count one less photon for its first associated spectral range, and the neighboring pixel will count one more photon for its second associated spectral range. The first spectral range is thus undervalued, and the second spectral range is overvalued.

[0100] This minimum distance condition, greater than the scattering length of a photon in the pixel, allows us to work in a configuration where the aperture number of a lens F#pL is less than the working aperture number of the imaging system WF#, i.e., F#pL < WF#. Reducing the lens aperture number is beneficial for limiting the diffractive effects of the lens, particularly when the lens is a microlens.

[0101] In the case of microlenses, the size of the microlenses (such as the diameter λ) can be sufficiently small and close to the wavelength λ, so that diffractive effects appear and predominate over the refractive property. To determine the principal mode of operation, the generalized Fresnel number can be defined by the following relation:

[0102]

[0103] with the diameter of the lens, n the optical index of the medium separating the lens and the pixel group, X the wavelength of the light and L2 the distance between the lens and the photodetector array.

[0104] If FN is significantly greater than 1 (typically FN > 10), the laws of refraction apply and ray tracing allows the operation of the optical component to be described. If FN is less than 1 or close to 1, diffraction phenomena must be taken into account to describe the operation of the component.

[0105] The condition F#pL < WF# is favorable for increasing the generalized Fresnel number FN. This condition is particularly valid for circular pupils.

[0106] By requiring that the minimum distance be greater than the characteristic diffraction length due to the lens in a plane of the pixel, the chromatic overlap between pixels due to diffraction is limited. If the optical geometry between the lens and the pixel is such that diffraction effects occur, then the lens will transmit photons onto a larger area centered on its destination pixel than if the diffraction effects were absent. There is then a risk that a pixel near the destination pixel will receive the photon, thus degrading the multispectral nature of the images produced.

[0107] The characteristic diffraction length r can be given by the product of a proportionality coefficient, a wavelength of the light and an aperture number of the lens, the proportionality coefficient preferably being equal to 1.22. In other words r= 1.22XF#pL.

[0108] In the example of a quad-pixel comprising 2x2 pixels of size T = 15 pm illuminated by mid-infrared (MWIR) light (corresponding to a wavelength range between 3.7 pm and 4.8 pm) from microlenses having an aperture number F#pL of 1 and an index n of 2.75, the focal length for the microlens is denoted L2 and is L2 = 1 x F#pL x 2T = 82.5 pm. The distance L1 between the pupil plane 14 and the lens array 7 is 20 mm. The scattering length is 3 pm. The minimum diffraction spot has a radius denoted r, which is r = 1.222 - F#pL, or 5 pm at the wavelength 4 pm, in the case where r > Ld. By choosing a distance separating two adjacent images of a stopped-down filter equal to 2xLd, we want images of the filter stopped down according to square patterns with sides of 9 µm and centered in each pixel. In the pupillary plane 14, we place square filters with sides of 6 mm. The pupil inscribing the patterns has a diameter of 22.6 mm.The number of openings of the circle that inscribes the filter matrix is ​​therefore 0.88.

[0109] This number of apertures, lower than that of the lens, is related to the shape of the pupil consisting of 4 square apertures compared to a more standard circular aperture.

[0110] When the device includes a filter mask for each filter in the filter matrix, it is possible to adjust the position of at least one filter mask in the filter matrix 21 to offset it relative to the filter, so that a spacing between the center of the filter and a center of the exit pupil is greater than a spacing between the center of the mask and the center of the exit pupil.

[0111] The aim here is to move the mask within the filter so as to bring the central portion of the filter defined by the mask closer to the center of the filter matrix. The center of the filter matrix corresponds to the center of the exit pupil. By making this move, the area of ​​the exit pupil that needs to be illuminated decreases in width. This makes it possible to work with a large f-number of the imaging system, so that it is easier to meet the condition F#pL < WF# and also easier to limit the complexity of the imaging system associated with an excessively small f-number. The image of each filter stopped down on the detector is convolved by the image spot of the lens, particularly when it is a microlens. It is therefore advantageous to use large f-numbers to limit the widening of the image of the diaphragm.Therefore, it is preferable that r <Ld, r étant le rayon de la tache de diffraction de la lentille et Ld la longueur de diffusion d’un photon dans le pixel. Les effets diffractifs sur le recouvrement chromatique entre . This allows the pixels to be reduced, which improves the quality of the multispectral imaging performed.

[0112] It is even more advantageous to move each of the mask images from the center of their respective pixel and / or towards the center of the pixel group associated with the lens. This makes it possible to increase the aperture number of the imaging system while maintaining a low lens aperture number (F#pL < WF#): the imaging system is simpler to design and the lens diffraction spot is reduced.

[0113] With reference to [Fig. 3], we present the case of the image of a filter matrix comprising four diaphragmed filters. Each diaphragmed filter image has been offset from the center of its respective pixel. This case corresponds to a "quad-pixel" configuration. [Fig. 3] represents the four pixels 11e, 11f, 11g, and 11h of a quad pixel 35 of the pixel matrix. On each pixel, the lens associated with quad pixel 35 forms the image of one of the diaphragmed filters by its mask: the image 25e on pixel 11e, the image 25f on pixel 11f, the image 25g on pixel 11g, and finally the image 25h on pixel 11h. Each mask defines a square portion of the filter, so that each image of a filter is also square. Each image is a square with side length C. Each pixel is a square of size T.We consider the situation where the image of a stopped-down filter is separated from an inner edge of the pixel (that is, an edge adjacent to another pixel of the quad pixel) by a distance Ld corresponding to the scattering length of a photon in the pixel. Since side C has a value less than T-2Ld, the center of the image of the stopped-down filter is shifted from the center of the corresponding pixel and is brought closer to the center 33 of the quad pixel.

[0114] The exit pupil of the imaging system must be larger than the size of the diaphragmed filter array so as to encompass them all. This imposes the following relationship:

[0116] With the size of the exit pupil of the imaging system in the pupillary plane 14, n the refractive index of the medium separating the lens and the pixel group, L2 is the distance between the lens and the photodetector array, L1 designates the distance between the pupillary plane 14 and the lens array 7.

[0117] In this case, we can determine the expression for the working aperture number WF# and the expression for the aperture number F#ul of the lenses: [0H8] WF#=y

[0119] F# y

[0120] Where N is the number of pixels along an axis in a group of pixels associated with a lens: N=2 in the case of the quad pixel and N=3 in the case of the nona-pixel.

[0121] In the case of a quad pixel T \ 21 WF# = +Ld)

[0122] In the case where the lens is at the diffraction limit, the number of lens apertures required for the radius r of the smallest image spot to be equal to Ld, the scattering length of a photon in the pixel, is obtained by the relation:

[0123] t1 H fiL~ 1.22À

[0124] We can then deduce the side C of the image of a filter stopped down on the pixel:

[0125] r__lJ__I__ ^\T 1.22ÀWF#

[0126] In a configuration where Ld = 3 pm, for a wavelength of 4 pm, WF# = 1, and T = 15 pm, we obtain C = 3.5 pm. The microlens has an aperture number F#pL = 0.6, a focal length L2 = 51 pm, a radius of curvature of 32 pm, and a generalized Fresnel number FN of 3. The distance L1 between the pupillary plane 14 and the lens array 7 is 20 mm. The portion of the filter defined by the mask is a square with sides of 3.8 mm and has an aperture number approximately equal to 5 (circle inscribed in the square). F#pL = 0.6 is an extremely low aperture number that is viable because the refractive index of the substrate between the microlens and the pixel array is greater than 2.

[0127] In a configuration where Ld=4.5 pm, for the wavelength 4 pm, WF#=1, T=15 pm, we obtain C=5.3 pm. The microlens has an aperture number F#pL = 0.9, a focal length L2=76 pm, a radius of curvature of 48 pm, the generalized Fresnel number FN is 2. The distance L1 between the pupillary plane 14 and the lens matrix 7 is 20 mm, the portion of the filter defined by the mask is a square with sides of 3.8 mm and has an aperture number approximately equal to 5 (circle inscribed in the square).

[0128] In relation to [Fig. 4], we present the case of a filter matrix comprising nine filters. Each image of the eight peripheral filters to the central filter has been decentered from the center of its pixel. This case corresponds to a nona-pixel configuration. [Fig. 4] represents the 9 pixels 1 li, 1 Ij, 11k, 111, 11m, lin, llo, 1 Ip and 1 Iq of a nona-pixel 37 of the pixel matrix. On each pixel, the lens associated with the nona-pixel 37 forms the image of one of the filters stopped down by its mask: image 25i on pixel 1li, image 25j on pixel 1Ij, image 25k on pixel 11k, image 251 on pixel 111, image 25m on pixel 11m, image 25n on pixel 1In, image 25o on pixel 1lo, image 25p on pixel 1Ip, and finally image 25q on pixel 1Iq. Each mask defines a square portion of the filter, so that each image of a central portion is also square. Each image is a square of side C. The images of the eight peripheral filters are off-center with respect to the center of their respective pixels and are shifted towards the center 39 of the nona-pixel 37. Each pixel is a square of size T. We consider the situation where the image of a peripheral filter is separated from an inner edge of the pixel (that is, an edge adjacent to another pixel of the nona-pixel) by a distance Ld corresponding to the scattering length of a photon in the pixel. In a more general case, we can consider the situation where the image of a peripheral filter is separated from an inner edge of the pixel by a distance greater than the scattering length of a photon in the pixel. On the central pixel 11m, the lens associated with the nona-pixel 37 forms, at the center of the central pixel 11, the image of the central filter stopped down by its mask. This image is separated from the edge of the central filter 11m by a distance U greater than Ld.The distance U can be formulated as a function of the sizes C and T: .

[0129] U=(TC) / 2

[0130] Following the same calculation steps as for the quad-pixel, we obtain the following relationship for the nona-pixel: 101311 r]

[0132] In a configuration where Ld=3pm, for wavelength 4pm, WF#=1, T=15pm, we obtain C=-0.7pm<0. There is therefore no solution in this configuration.

[0133] In a configuration where Ld = 5 pm, therefore greater than 3 pm, for the wavelength 4 pm, WF# = 1, T = 15 pm, we obtain C = 3.8 pm. The microlens has an aperture number F#pL = 1, a focal length L2 = 126 pm, a radius of curvature of 80 pm, the generalized Fresnel number FN is 2.7. The distance L1 between the pupillary plane 14 and the lens matrix 7 is 20 mm, the portion of the filter defined by the mask is a square with sides of 1.6 mm and has an aperture number approximately equal to 12 (circle inscribed in the square).

[0134] In another configuration, the active area of ​​the pixels will be a microdiode the size of the images of the central portion of the filter defined by the mask on the detector and centered on these images.

[0135] It should also be noted that these dimensioning rules are taken here in the thin optics approximation and in the geometric optics regime. Diffractive effects can modify the dimensioning rules. The notions of the geometric regime and the diffractive regime have been discussed with the generalized Fresnel number. Adjustments may be necessary to take into account the effects of microlenses exhibiting significant sag. The sag of a lens is understood to be the height between the apex of the lens and its edge, or the maximum distance separating a surface of the lens and a plane tangent to that surface, the tangent plane. being normal to the optical axis. The approach used here remains relevant, however, particularly for preliminary sizing of the optical system.

[0136] In the case where an application requires bispectral vision (i.e., requiring a filter matrix comprising only two filters), it is possible to implement a 2x2 filter matrix with only two different filter types. A spectral image is obtained by summing the pixels receiving the same spectral band.

[0137] Advantageously, 3D information can be obtained by exploiting the parallax at the pupil generated by images obtained with different pixels of the same optical path, that is, with different pixels associated with the same lens. Multispectral imaging method

[0138] Using the multispectral imaging device as just described, it is possible to implement a multispectral imaging method P. With reference to [Fig. 5], the method P comprises the following steps.

[0139] In a first step SI, an image of the object 2 is constructed (or formed) by the imaging system 3 on the lens array 7 using the light from the object 2. The imaging system 3, placed between the object 2 and the lens array 7, allows, for example, this first step SL to be carried out.

[0140] In a second step S2, the light is spectrally filtered by the filter matrix 21 arranged in the plane 14 of the exit pupil of the imaging system 3,

[0141] In a third step S3, an optical conjugation of a plane of the filter matrix and a plane of the pixel matrix of the filter matrix 21 is carried out by the lens matrix 7 on the pixel matrix 11. This allows, for example, the construction of an image of the filter matrix on the pixel matrix.

[0142] In a fourth step S4, the light received by the photodetector array 11 is measured. This makes it possible to produce an image of the object 2 in different spectral bands.

Claims

Demands

1. A multispectral imaging device (1) comprising: - an imaging system (3) configured to form an image of an object (2), - a measurement system (5) comprising: — a filter array (21) arranged in a plane of an exit pupil (14) of the imaging system, — a lens array (7) located in an image plane (9) of the imaging system, and — a pixel array (11), and each lens (7a, 7b) of the lens array being configured to image each filter (21a, 21b) on one pixel (1a, 11b, 1a, 1d), such that each lens is associated with a first number of pixels equal to a first number of filters in the filter array, each filter being configured to be imaged by each lens on one pixel, such that each filter is associated with a second number of pixels equal to a number of lenses in the lens array,the pixel matrix comprising a number of pixels greater than or equal to a product of the first number and the second number.

2. Device according to claim 1 wherein the number of filters in the filter matrix is ​​equal to four or nine.

3. Device according to any one of claims 1 and 2 further comprising a substrate (15) occupying a volume separating the lens matrix and the pixel matrix, the substrate having an optical index greater than or equal to 1, preferably greater than or equal to 1.5 and even more preferably greater than or equal to 2.

4. A device according to any one of claims 1 to 3, wherein for each lens, the pixels on which the lens is configured to image the filters of the filter array form a pixel group, the lens being offset from the pixel group by a gap (A) measured orthogonally to an optical axis (X) of the multispectral imaging device, the gap (A) being equal to the ratio of a product of a distance (L2) measured along the optical axis separating the lens and the pixel group by a distance (H1) from a center of the lens to the optical axis, and a product of an optical index of the medium separating the lens and the pixel group by a distance (Ll) measured along the optical axis separating the lens matrix from the exit pupil, the lens being closer to the optical axis than the pixel group.

5. Device according to any one of claims 1 to 4 comprising a hood (45) surrounding the measurement system, the measurement system comprising: - a cryostat (13) configured to cool the pixel matrix and the lens matrix, - a screen (17) located between the pixel matrix and the hood, the cryostat being configured to cool the screen, the screen defining an optical diaphragm upstream of the lens matrix, the filter matrix being disposed in the optical diaphragm.

6. Device according to any one of claims 1 to 5 comprising for each filter of the filter matrix a filter mask disposed against the filter, the filter mask being configured to optically mask a peripheral portion of the filter.

7. Device according to claim 6 wherein for each filter in the filter matrix, the filter is diaphragmed by the filter mask such that an image of the filter diaphragmed by a lens on a pixel is located at a minimum distance from each edge of the pixel, the minimum distance being greater than or equal to a scattering length of a photon in the pixel.

8. Device according to any one of claims 6 to 7 wherein, for at least one filter of the filter matrix, the filter mask is off-center with respect to the filter, such that a spacing between the center of the filter and a center of the exit pupil is greater than a spacing between the center of the mask and the center of the exit pupil.

9. A multispectral imaging method for an object comprising: - a construction step (S1) of an image of the object by an imaging system on a lens array using light from the object, - a spectral filtering step (S2) of the light by a filter array arranged in a plane of an exit pupil of the imaging system, - an optical conjugation step (S3) of a plane of the filter array and a plane of a pixel array by the lens array, - a light measurement step (S4) by the pixel array, each lens in the lens matrix being configured to image each filter on one pixel, so that each lens is associated with a first number of pixels equal to a first number of filters in the filter matrix, each filter being configured to be imaged by each lens on one pixel, so that each filter is associated with a second number of pixels equal to a number of lenses in the lens matrix, the pixel matrix comprising a number of pixels greater than or equal to a product of the first number and the second number.

Images