WO2016146830A1 - Polarimetric image sensor - Google Patents

Abstract

The invention relates to an imaging device for collecting data of images of a target area, designed to be mounted on an aircraft or satellite (2), said device (1) comprising an image sensor (10) having a capture surface (101), designed to record images of the target area (Z), and at least two polarising filters (21, 22), wherein the linear polarising filters (21, 22) have respectively different transmission axes, with different polarisation directions, and each polarising filter (21, 22) covers part of the capture surface (101).

Description

 Polarimetric image sensor

GENERAL TECHNICAL FIELD

 The invention relates to the field of ground imagery by an overflight aircraft or a satellite.

 More specifically, the invention relates to a sensor for imaging an aircraft (such as an aircraft, drone, aerostat, etc.), or a satellite, with high spatial resolution exploiting the polarized nature of the light. STATE OF THE ART

 One of the advances in high-resolution Earth observation is, of course, to seek to increase spatial resolution, which implies the development of increasingly large telescopes and improved image quality. intrinsic (active optics for example).

 Another way is to increase the number of spectral bands and observed spectral coverage (infrared, etc.), or to improve the notions of temporal resolution (revisit, observation of dynamic phenomena with matrix shots, etc.).

 All these evolutions pursue the same goal: the increase of information acquired by the information system.

 Terrestrial imaging systems that use polarimetric imaging are already known.

 These systems are of low resolution type and are not suitable for high resolution satellite imagery.

 Embedded systems are especially known in aircraft that use several cameras, each equipped with polarization filters, the optical flow being divided into as many paths as cameras.

Such an arrangement is for example not transposable for a satellite, as soon as the high spatial resolution is sought, because of the congestion of the system. In addition, the POLDER system (for "Polarization and Directionality of the Earth's Reflections", polarization and directionality of the reflections of the Earth), launched by CNES in 1996, implemented polarimetric shooting from space to observe solar radiation reflected by the Earth and the atmosphere.

 However, it was not high-resolution imaging (resolution of the order of a kilometer) and the embedded technology used a filter wheel, ie polarizing filters arranged on a wheel rotating in front of the lens of the sensor.

 Such a device does not make it possible to respond to imaging problems of observing the terrestrial surface in high resolution (metric or even centimeter scale), where, because of the orbital movement, such an observation using a sequential shooting by mechanical filter exchange is not operative.

 The time taken to mechanically switch the polarizing filters is indeed too long compared to the time spent shooting. In addition, moving elements such as a filter wheel can cause disturbances due to vibration of the components.

PRESENTATION OF THE INVENTION

A general object of the invention is to provide a sensor for high-resolution overflight imaging that allows acquisitions of better qualities than those known to date in the state of the art.

 For this purpose, the invention notably proposes an imaging device for collecting image data of a target zone, adapted to be mounted on an aircraft or a satellite, said device comprising:

an image sensor having a capture surface, configured for taking images of the target zone, at least two polarizing filters,

 wherein the linear polarizing filters have different polarization directions and each polarizing filter covers a portion of the capture surface.

 The sensor and polarizing filters provide high resolution data that has been difficult to obtain up to now: the polarimetry of aerial data. Thanks to the latter, elements that are not visible in traditional imagery become visible (cables, vehicles, vegetation, etc.). Indeed, spectropolarimetric signatures are now accessible and bring new information.

 This polarimetric observation from Earth orbit and / or the atmosphere allows the improvement of the quality of certain images (complementarity with the current observation modes) and opens the way to the development of new applications (object detection, nature surfaces, state of vegetation, etc.).

 The invention is adapted to be mounted on an aircraft or a satellite configured to make the acquisition of ground image overflight.

 Aircraft means an airplane, a drone, an aerostat, a balloon (stratospheric balloon or balloon probe), etc.

PRESENTATION OF THE FIGURES Other characteristics, objects and advantages of the invention will emerge from the description which follows, which is purely illustrative and nonlimiting, and which should be read with reference to the appended drawings, in which:

 FIGS. 1a, 1b, 1c represent a polarized wave respectively rectilinear, elliptical and circular.

FIG. 2 represents an embodiment of a device according to the invention, FIG. 3a represents a satellite comprising the device 1 and the target zone,

 FIG. 3b represents an exemplary diagram of an optical path inside a satellite,

 Figure 3c shows an objective and a focal plane.

 Figures 4, 5 and 6 show other embodiments of the invention,

 FIG. 7 represents two successive images taken,

 FIG. 8 represents a device in push-broom mode,

 FIG. 9 represents a device with polarizing and chromatic filters, according to one embodiment of the invention,

 FIG. 10 represents a type of sensor, referred to as Bayer,

 FIG. 11 represents the dead zones between several sensors simply positioned in staggered rows,

 FIG. 12 represents a plurality of sensors arranged in staggered rows with partial overlap,

 FIG. 13 represents steps of a "drizzle" algorithm. FIG. 14 represents the positioning of the compensation mechanism in the optical path diagram.

 Figures 15a, 15b and 16 show the compensation mechanism and its operating principle.

DETAILED DESCRIPTION General Reminders on Polarizing Filters

Polarization is a fundamental characteristic of an electromagnetic wave, just like its wavelength (spectral dimension), its direction of propagation (spatial dimension) and its amplitude (radiometric dimension). The electric field of the wave and the magnetic field oscillate simultaneously in different directions. The polarization of light describes the vibration of the electric field: it can oscillate in one direction (rectilinear polarization or linear - Figure la) or around the axis of propagation of the wave (elliptical or circular polarization - Figures lb, the).

 A non-polarized wave, that is to say without privileged orientation, is associated with a so-called "natural" light, like that coming from the Sun.

 The polarization is caused by the asymmetries that the light encounters during its journey (for example a reflection on a liquid). The polarization rate can then reach 100% under certain circumstances (totally polarized state). Phenomena such as reflection or scattering change the polarization rate. In addition, the nature of the object and many physical parameters are involved, including the physico-chemical composition, the specularity, the optical permeability of the optical layer, and so on.

 Since from space or atmosphere we observe the light backscattered by the Earth, the polarimetric signature is ubiquitous in the flux entering the imaging device.

 There are polarizing optical filters allowing the transmission of the wave in a different way depending on its polarization state. In the case of filters sensitive to linear polarization, with a coefficient, they are fully passable for incident light whose polarization plane is parallel to a preferred axis of said filter.

The imaging device 1 will be described for a satellite 2, but the acquisition of images by any other type of aircraft configured to make the image acquisition of the Earth overflight is equivalent. In a nonlimiting manner, the aircraft include aircraft, drones, aerostats, balloons (stratospheric balloons or balloons-probes in particular).

Imaging device

The imaging device 1 shown in FIG. 2 is intended to be embarked on an aircraft or a satellite 2, and to collect data from a target zone Z. The target zone Z is typically a part of the earth's surface. The imaging device 1 comprises:

 a sensor 10, the sensor having a capture surface 101, configured for taking images of the target zone Z,

 at least two polarizing filters 21, 22, the polarizing filters being sensitive to linear polarization and each having distinct transmission axes, relative to the sensor 10.

 In use, the sensor 10 takes images of the target zone Z in a direction of movement X-X '. The target zone Z is therefore a band, called swath. According to the reference system in which one is placed, it can be considered that the target zone Z scrolls on the sensor 10 (see Figure 3a).

 The sensor 10 is typically located at the end of an optical path comprising a set of mirrors M making it possible to obtain a clear image of the target zone Z in the focal plane of the sensor 10, for example a telescope system (see FIG. 3b). .

 More generally, the optical path comprises an objective O (represented as a thin lens in FIG. 3c), focusing in the focal plane of the sensor 10.

 The sensor 10 is a two-dimensional sensor, of matrix type, of the CCD or CMOS type. It allows the image shooting in high or very high resolution. CMOS technology enables high acquisition speeds, while integrating a fairly efficient electronic shutter function. Thus, thanks to short integration times, the "spun" effect of the satellite or micro-vibrations is partially frozen: this is the principle of snapshot or snapshot in English.

The sensor 10 preferably has a rectangular shape of capture height h _c and capture width l _c .

Each polarizing filter 21, 22 covers a part of the sensor surface 101. Preferably, the filters have a rectangular shape, the width _f of substantially equal to the width l _c of the sensor, or at least equal to the width l _c of the sensor and a height h _f less than the height h of the sensor _c. They are arranged side by side with a small interval between them in the direction of the capture height h _c , so that the accumulated heights of the filters 21, 22 is less than or equal to the height of the capture h _c .

 Between two filters, a space G (for "gap"), corresponding to the connection, is not usable on the image taking.

 The principle of a linear polarizing filter being to let pass only the waves having a privileged direction, it is necessary to have several filters 21, 22 with different directions of polarization to recover the polarimetric information contained in the target zone Z.

 For example, if the axes of four polarizing filters are oriented at four angles Θ to 45 ° from each other (0, 45, 90, 145 °), the signal measured by the sensor is set for S (9):

I = _^1/2 (S (0 °) + S (45 °) + S (90 °) + S (135 °)),

 Q = S (0 °) - S (90 °),

 U = S (45 °) - S (135 °).

 I, Q and U are called Stokes parameters and their determination makes it possible to have access to the information on the state of rectilinear polarization at any point of the captured image (orientation of the polarization vector, polarization rate) in more traditional radiometric information.

 The knowledge of the values of S (0 °), S (45 °), S (90 °), and S (135 °) makes it possible to know the parameters of Stokes but it can be demonstrated that by knowing only three on all four, the parameters are known. Finally, by means of a first-order approximation of the angle of polarization of light (which can be known), only two measurements can suffice: indeed, the plane of polarization (defined by sun-earth-satellite) is known.

 The number of polarizing filters can be two with different polarization directions; from four filters, there is redundancy of information.

In one embodiment, a portion of the capture surface 101 is covered by the first filter 21 and adjacent to the first filter 21, i.e., side by side, is the filter 22 (see FIG. ). A portion 102 of the capture surface may remain uncovered, so as to capture traditional images or the surface 101 may be fully covered by the filters 21, 22 (Figure 4);

In another embodiment, the invention comprises three or four different polarization polarizing filters 21, 22, 23, 24 of polarization angle, arranged in a similar manner to the previous embodiment, i.e. each filter is adjacent (or adjoining, side by side) to another in the direction of the capture height h _c . A portion 102 of the capture surface 101 may remain uncovered to capture traditional images (Figures 5, 6).

 In a preferred manner, the values of the polarization direction angles of the filters are included:

 • in 0, 45, 90, 135 ° for four or fewer filters,

• in 0, 120, 240 ° for three or fewer filters. In practice, any combination of angles making it possible to retrieve the parameters of Storing (see above) associated with the angular polarization are possible, that is to say at least two different angles by means of a first order approximation and minus three different angles otherwise.

Referring to Figure 7, in which are presented two shots k and k + 1 (also called frames) at times t close. We denote by F the oversampling factor, defined by the equation F = h _f / S, with S the displacement ("step") of the target zone Z between two images, which corresponds approximately to the number of images to be taken for that each point is observed through the following filter 21, 22, 23, 24.

In order for the polarimetric information to be complete, it is necessary that each point of the target zone has been observed at least once via each of the filters 22, 23, 24, 25 of the device 1. In this case, the oversampling factor is equal to 1 ("step" equal to the height h _f of the filters), which limits the volume of data to be transmitted and processed. In other words, each point of the target zone Z will be "photographed" through each of the filters 21, 22, 23, 24 successively.

 Alternatively, if each point is observed several times, there is redundancy of the information, which allows oversampling and thus a better fineness, either in terms of image resolution, or in terms of cosmetic dressing, or in terms of signal to noise ratio. The oversampling factor is thus greater than 1. If each point appears y times in each filter, then the oversampling factor is y. In polarization, it is common to have a high factor F, greater than ten.

 In use, during the scrolling of the target zone Z, the target zone Z traverses all the filters 21, 22, 23, 24 in succession, with at least one shot through each of the filters.

 Next, algorithms for processing and reconstructing polarimetric images are applied. They reconstruct in particular an image from each point taken by the filters at different times.

 These algorithms are known to those skilled in the art and will not be explained.

 The interest of keeping an area 102 of the capture surface 101 not covered with a polarizing filter 21, 22, 23, 24 makes it possible to obtain conventional imaging information, in addition to the polarimetric images. Push-broom imaging

According to the so-called "push-broom" mode of photography known to those skilled in the field of space imaging in particular, the polarizing filters 21, 22, 23 may be arranged facing detector strips 103 operating in pure push-broom mode (the bar is composed of a single line of photosensitive pixels) or in TDI mode (or "Time Delay Integration", with a bar composed of lines that are adjacent). Each bar 103 is covered by a single filter. The arrows of the frame xxx3 in Figure 8 show an example of the axis of polarization of the filters. The image of the Earth moves along the X-X 'axis.

Chromatic imaging

 According to one embodiment (see FIG. 9), the invention furthermore comprises several color filters, arranged in a manner similar to the polarizing filters, that is to say that they cover the capture surface and are juxtaposed in the direction the height of the sensor hc.

 The color filters 25, 26 are arranged on a portion of the sensor 10 which is not covered by a polarizing filter 21, 22, 23, 24.

 It is necessary to have at least two filters 25, 26 of different colors, so as to obtain information in two different frequency domains. When the target zone Z scrolls, each point will thus go through all the color filters and polarizing filters, which offers a greater amount of information than that provided by the existing devices.

 For example, it is possible to provide three filters 25 (R for red for example), 26 (J for yellow for example), 27 (B for blue for example) or four filters including a red, a green, a blue and a yellow.

 Such a device 1 incorporating both polarizing and chromatic filters makes it possible simultaneously to obtain data relating to polarimetry and color shooting. Matrix of Baver

 According to one embodiment (see FIG. 10), the sensor 10 is a Bayer 104 matrix, that is to say a matrix of colored filters

RGB (red green blue) is placed in front of the sensor (between the sensor 10 and the polarizing filters 21, 22, 23, 24).

In this way, it is possible to obtain the polarimetric information in each color and thus obtain a greater wealth in terms of information. Arrangement of staggered matrices

The size of the currently existing sensors 10 generally does not allow to enter the entire target zone Z. For this, we assemble several devices 1 mosaic in a staggered arrangement. If the assembly is mechanical, there are generally Zm inter-sensor dead zones, which are difficult to accept (see FIG. 11) in the direction of travel XX 'of the target zone Z. Indeed, the dimensions of the sensor 1 are very often greater than that of the capture surface 101 and a simple staggered arrangement is not enough. To remedy this, the sensors are arranged in staggered staggered matrices so that in the direction of the capture width l _c , no point of the target zone Z can be missed after scrolling and capture images (see Figure 12) . In other words, the swath is covered by a slight geometric overlap of sensor 10 with sensor 10.

 It will be understood that each device 1 forming the staggered assembly comprises a sensor 10 and the polarizing filters 21, 22, 23, 24 and, if appropriate, the color filters 25, 26, 27.

 With such an assembly, it is necessary to produce, in the same way as for a single device 1, successive images at a regular time interval. The operation is similar to a pushbroom (a one-dimensional sensor that snapshots at a very short interval a direction axis) but with a higher time step. Method of improving the signal-to-noise ratio

Dithering

If the number of images is sufficient, it is possible to apply a generalized supermode type algorithm known as dithering. For this, it is necessary to know the exact offset between two images k, k + 1, in number of pixels. This information is evaluated from the very content of the images by calculating the intercorrelation between the successive frames k, k + 1 and in the zones common between these images. This image processing of averaging shots is valuable in polarimetry, where the signal-to-noise ratio is, at equivalent parameters, worse than for traditional imaging. The method is known to those skilled in the art, the results fairly accurate, but the cost in calculation is quite high.

One of the most efficient processing algorithms works in the image space (drizzle algorithm). On the other hand, it requires at least a tenfold increase in information.

 The algorithm artificially manufactures a virtual Gv grid with P pixels smaller than the pixel size of the sensor (by a factor of 0.5 to 0.7) while keeping the inter-pixel pitch. This virtual grid is then projected in the final image I from a geometric transformation calculated elsewhere (see Figure 13). The surface fraction of the reduced pixel is multiplied by the pixel luminance (intensity of the light source divided by ease) and then added to the intensity already found in the final image. The process is repeated for each image.

 With this method, it is possible to improve the compactness of the optical system (mirrors M for example), such as the telescopic assembly.

Yarn Compensation Strategy (or Blur)

During a succession of frames k, k + 1, in particular with a lower number of images per second (so as to fill the gap G between the filters) and / or with a longer exposure time ( in order to improve the signal-to-noise ratio), it may be necessary to set up a yarn compensation strategy. The gain of the signal-to-noise ratio passes through the effective increase of the integration time per frame, that is to say the exposure time. In order not to impair the quality of the image, it is imperative to freeze the image on the pixel grid of the sensor (s) 10 (in the case of the staggered arrangement) during the acquisition period. The fixity of the image during the pose therefore amounts to compensating the movement of the earth (or satellite, according to the reference system in which one is positioned) during the integration time.

 For this, the device may provide a compensation mechanism 3, preferably at the end of the optical path, so as not to affect the rest of the optical combination (see Figure 14).

 It comprises a pivoting mirror 31 pivoting 32, low angular amplitude, actuated by a piezoelectric motor 33 (see Figures 15a, 15b) which is synchronized to the read frame frequency (a few tens of Hertz in general).

 The amplitude of the displacement of the end of the tilting mirror 31 is of the order of one micrometer, which makes it possible to consider that said end performs a translation movement and not a rotation. In this way, the piezoelectric motor 33, in contact with said end, makes it move back and forth (see Figure 16) with reset position and tracking the movement.

 A redundant actuator may be provided, such as a piezoelectric motor provided on the other end of the mirror.

 The rotation of the mirror 32 will follow the displacement of the target zone Z (integration time T1), so that the image remains fixed in the focal plane of the sensor 10 of the device 1 or the staggered arrangement of a set of devices 1, then the mirror 31 returns to its original position to maintain fixed the next image (time T2 rallying) By the amplitudes involved, the distortions of images are negligible.

 This mechanism is advantageously suited to the device 1 insofar as the stabilization needs to be performed in only one direction X-X ', that of the scrolling of the target zone Z.

Claims

 claims An imaging device for collecting image data of a target area (Z), adapted to be mounted on an aircraft or a satellite, said device (1) comprising:  an image sensor (10) having a capture surface (101) configured to capture images of the target area (Z), at least two polarizing filters (21, 22), wherein the linear polarizing filters (21, 22) have separate transmission axes from each other and each polarizing filter (21, 22) covers a portion of the capture surface (101). An imaging device according to claim 1, wherein: the capture surface (101) of the sensor (10) has a capture width (l _c ) and a capture height (h _c ), the filters have the same width (l _f ) and the same height (l _h ), said width of the filters (l _f ) being substantially equal to the capture width (l _c ) and the cumulative heights of the filters (h _f ) being less than or equal to the catch height (h _c ), the filters being juxtaposed according to the height of the sensor (h _c ). Imaging device according to one of the preceding claims, wherein the sensor is a sensor type 2D CCD or CMOS. An imaging device according to one of claims 1 to 3, wherein the image sensor (10) comprises a plurality of sensor arrays (103), and each polarizing filter (21, 22, 23) covers a single sensor array (103). Imaging device according to one of the preceding claims, further comprising at least two color filters (25, 26), said color filters covering a portion of the sensor (10) not being covered by the polarizing filters (21, 22, 23, 24). An imaging device according to one of claims 1 to 3, comprising a Bayer mosaic (104) placed on the sensor (10), between the sensor (10) and the polarizing filters (21, 22, 23, 24). 7. Imaging device according to one of the preceding claims, wherein the filters each have a linear polarization direction to find the Stokes parameters associated with the angular polarization. An imaging system comprising a plurality of imaging devices according to claim 1, wherein the imaging devices (1) form a staggered array. 9. Imaging system according to the preceding claim, wherein the staggered arrangement allows a geometric partial overlap sensor sensor. 10. System according to one of claims 8 to 9, further comprising a compensation system (3), using a mirror (32) for maintaining an image of the fixed target area on the sensor when the system is moving relative to the target area (Z). 11. A method of collecting data of a target area, comprising an image acquisition step using a device (1) or a system according to any one of claims 1 to 10. A method of acquiring polarimetric image data according to claim 11, wherein the device is mounted on an aircraft or a satellite (2), said aircraft or satellite (2) being movement relative to the target areas in one direction (X-X '), and the capture surface of the sensor has a capture width (l _c ) and a capture height (h _c ), the direction of the capture height (h _c ) being parallel to said direction (Χ-Χ '), the filters have the same width (l _f ) and the same height (l _h ), said width of the filters being substantially equal to the capture width and the cumulative heights of the filters being less than or equal to the capture height, the filters being juxtaposed according to the height of the sensor (h _c ), said method comprising the successive steps of:  - Acquisition of a first shot,  - Acquisition of a second shot,  wherein the displacement of the target area on a sensor (10) between the two acquisitions is less than or equal to the height of the polarizing filters (21, 22, 23, 24), so that each point of the target area (Z) was acquired through each filter, so as to obtain information on the entire target area (Z). 13. The method of claim 12, wherein said displacement is less than or equal to one quarter of the height of the polarizing filters, in order to average the images and to obtain a high redundancy of the data by oversampling. 14. The method of claim 13, wherein said displacement is less than or equal to one quarter of the height of the polarizing filters, and further comprising a data processing step, called "drizzle".