System and method for acquiring a hyperspectral image of a landscape under a real sky.
The system addresses the whitish veil and blooming issues in panoramic imaging by using a CCD hyperspectral camera with a hypergonal lens and diaphragm, achieving up to 90% reduction in the whitish veil and minimal light blur, ensuring high-quality hyperspectral images.
Patent Information
- Application Number
- FR2022012517
- Authority / Receiving Office
- FR · FR
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2022-11-29
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2042-11-29
AI Technical Summary
Current panoramic imaging systems with rotating photographic devices and hypergonal lenses suffer from a whitish veil and 'blooming' effect when capturing images in sunlight, leading to unrealistic light reflections and loss of information.
A CCD hyperspectral camera with a hypergonal lens oriented perpendicular to the astronomical zenith, combined with a diaphragm featuring a narrow vertical orifice parallel to the zenith, reduces light refraction and diffusion during 360-degree rotation, accompanied by a data processing method to correct pixel columns affected by blooming.
The solution significantly reduces the spatial extension of the whitish veil by up to 90% and minimizes light blur with limited information loss, resulting in more realistic and accurate hyperspectral images.
Smart Images

Figure 00000016_0000 
Figure 00000017_0000 
Figure 00000018_0000
Abstract
Description
Title of the invention: System and method for acquiring a hyperspectral image of a landscape under real sky. Technical field
[0001] The present invention relates to a system and a method for acquiring a hyperspectral image of a landscape under real sky. Technical background
[0002] In various industrial fields, recent advances in computer-aided simulation make it possible to create realistic virtual previews of articles even before they are manufactured or renovated. These previews can provide decision-making assistance when choosing between different articles under consideration depending on whether or not they meet certain technical and / or aesthetic requirements.
[0003] There is, nowadays, a strong and ever-increasing demand from customers to provide them with realistic predictions of future results of complex renovation operations on which they can confidently base their decision to start or continue these operations.
[0004] For example, in the automotive and construction industries, it is now common practice to use simulation tools, such as physico-realistic rendering or ray tracing methods, in order to generate three-dimensional scenes as anticipatory digital representations of physical or functional characteristics of a motor vehicle or a building. In these digital representations, the objects represented, for example all or part of a vehicle or a building, evolve in an environment or a landscape whose digital simulation can be carried out using an environment map projected as a texture on the surface of a geometric object, such as a sphere, a cube or a half-sphere, used as a support for a virtual sky, and / or projected onto the reflective surfaces of certain elements or objects of the simulated three-dimensional scene.
[0005] The images used to generate the environment map may be manually or computer-generated drawings or graphics, or real-world photographs acquired using a photographic device. In the latter case, the environment maps may be acquired using a panoramic photographic device that is adapted for acquiring a complete landscape, including the sky, along a solid angle defined vertically by an angle of 180° from the zenith to the horizon, and horizontally by an angle of 360° to the horizon. Examples of such devices are 360° rotating photographic devices, photographic devices with a “wide-angle” lens, or photographic devices with fisheye lens.
[0006] CN 207442995 U, SHENZHEN REAL SCENE TECH CO LTD [CN] 01.06.2018 discloses a hybrid panoramic imaging system comprising several conventional cameras and a hypergon lens camera. The conventional cameras are oriented in a direction perpendicular to the astronomical zenith. The hypergon lens camera is oriented along the astronomical zenith axis.
[0007] JP 2007028272 A, DOMUS INTERNAT LTD [JP] 01.02.2007 describes a panoramic imaging system comprising a camera equipped with a CCD photographic sensor and a hypergon lens. The camera is mounted on a rotating movable stage and is oriented in a direction perpendicular to the astronomical zenith direction. Opposite the hypergon lens is arranged an occulting diaphragm equipped with a vertical slit so that the camera acquires images along a vertical arc of 180° from the zenith to the horizon. A complete panoramic photograph is reconstructed by assembling all the vertical images acquired during the rotation of the rotating movable stage. Summary of the invention Technical problem
[0008] A first disadvantage of current panoramic imaging systems comprising a rotating camera equipped with a hypergon lens is that, when an image is acquired in full sunlight or in full light using this type of device, certain regions of the image thus acquired may be affected by a whitish veil. This whitish veil is particularly visible when the sun itself is photographed, in particular in the peripheral regions of the image of the sun.
[0009] Concretely, the white veil materializes by a whitening of certain regions of the image which obscures certain elements of the landscape such as thin clouds, trees or buildings. Also, when the images are used for the design of an environment map in a rendering engine, a disadvantage is that these whitened regions induce unrealistic reflections of light, in particular with regard to brightness and colors, and, consequently, a lack of representativeness of the real landscape.
[0010] Another disadvantage is what is commonly referred to as "blooming or bloom effect". This light blur originates from the glare of the CCD sensors of photographic devices by direct sunlight. In concrete terms, it appears on panoramic images as a more or less thick vertical line of pixels whose color shade is at its maximum. These pixels may be in a state of saturation. It is obvious that such a defect can lead to negative consequences when the images are used as an environment map in a rendering engine. This type of defect can be accidentally eliminated by some digital image processing or other tricks, but it always results in a loss of information in the form of aesthetic discontinuities in the image. Solution to the technical problem
[0011] According to a first aspect of the invention, there is provided a system (1000) for acquiring a hyperspectral image of a landscape under real sky, the system comprises: - a CCD hyperspectral camera (1001) equipped with a hypergonal lens (1002) whose optical axis (AO) is oriented in a direction substantially perpendicular to the astronomical zenith vertical direction (ZA); - a rotating mobile plate (1002) configured to ensure the rotation of said hyperspectral camera according to an angle of 360° in the reference plane of said hyperspectral camera with, as axis of rotation, the astronomical zenith vertical direction (ZA); - a diaphragm (2001) provided with a vertical orifice (2002) and arranged on said hypergone lens (1002); wherein the width of said vertical orifice (2002) is less than or equal to 5mm, preferably less than or equal to 4mm, preferably less than or equal to 3mm, and wherein the vertical axis (AV) of said vertical orifice (2002) is substantially parallel to the astronomical zenith vertical direction (ZA).
[0012] Advantageous embodiments are described below.
[0013] According to a second aspect of the invention, a method is provided for acquiring a hyperspectral image of a landscape under real sky, said method comprising the following steps: (a) the provision of a CCD hyperspectral camera equipped with a hypergone lens whose optical axis is oriented in a direction substantially perpendicular to the astronomical zenith vertical direction, and a diaphragm arranged on said hypergone lens, said vertical orifice having a width less than or equal to 5mm, preferably less than or equal to 4mm, preferably less than or equal to 3mm, and a vertical axis substantially parallel to the astronomical zenith vertical direction; (b) the acquisition of a panoramic image of a landscape under real sky by rotating said hyperspectral camera through an angle of 360° in the reference plane of said hyperspectral camera with, as axis of rotation, the astronomical zenith vertical direction
[0014] Advantageous embodiments are described below. Advantages of the invention
[0015] A first remarkable advantage of the invention is a reduction in the extension spatial reduction of the whitish veil. This reduction can reach 90% or more.
[0016] For some embodiments, a second advantage is the elimination or reduction of light blur with limited loss of information. Brief description of the drawings
[0017] [Fig.l] is a schematic representation of a conventional system for acquiring a panoramic hyperspectral image.
[0018] [Fig.2] is a schematic representation of the diaphragm of a system according to the first aspect of the invention.
[0019] [Fig.3] schematic representation of a front view of the diaphragm of [Fig.2].
[0020] [Fig.4] is a schematic representation of a side view of the diaphragm of the [Fig.2],
[0021] [Fig.5] is a schematic representation of a particular embodiment of the diaphragm of [Fig.2].
[0022] [Fig.6] is a schematic representation of the matrix structure of a hyperspectral image.
[0023] [Fig.7] is a graphical representation of the luminance variations on a line of pixels for a channel of a panoramic hyperspectral image acquired according to three examples in accordance with the invention and a counter-example CE not in accordance with the invention.
[0024] [Fig.8] is an example of a black and white panoramic hyperspectral image of a real sky with a light blur phenomenon in the sun region.
[0025] [Fig.9] is a graphical representation of the variations in luminance along a vertical profile in the light blur area of [Fig.9], before and after processing using a system according to certain embodiments of the first aspect of the invention.
[0026] [Fig. 10] is the image of [Fig.9] after processing using a system according to certain embodiments of the first aspect of the invention. Detailed description of the embodiments
[0027] By "substantially perpendicular direction", respectively "substantially parallel direction", with respect to another direction, is meant a direction approximately or exactly perpendicular, respectively parallel, with a tolerance of at most 10°, preferably at most 5° with respect to said other direction. Otherwise, a direction is considered perpendicular, respectively parallel, to another direction, when it forms an angle between 80° and 100°, preferably between 85° and 95°, respectively between -10° and 10°, preferably between -5° and 5°, with respect to said other direction.
[0028] By "hyperspectral imaging" is meant hyperspectral imaging such as defined in the technical field of imaging. It includes in particular the techniques, devices and / or systems for constructing an image in which each of the individual elements, called pixels, is associated with a reflectance spectrum. Simply put, hyperspectral imaging allows exploration of the electromagnetic spectrum beyond the RGB channels of conventional imaging systems. Thus, a hyperspectral image is formed from an assembly of several dozen, or even hundreds, of contiguous spectral bands acquired for each of the pixels. It generally has a matrix structure and can be represented in the form of a data cube whose first two dimensions represent the spatial dimensions of the image and the third its spectral dimension.
[0029] According to a first aspect of the invention, with reference to [Fig.l], [Fig.2], [Fig.3] and [Fig.4], there is provided a system 1000 for acquiring a hyperspectral image of a landscape under real sky, the system comprises: - a CCD hyperspectral camera 1001 equipped with a hypergone lens 1002 whose optical axis (AO) is oriented in a direction substantially perpendicular to the astronomical zenith vertical direction (ZA); - a rotating mobile plate 1002 configured to ensure the rotation of said hyperspectral camera according to an angle of 360° in the reference plane of said hyperspectral camera with, as axis of rotation, the astronomical zenith vertical direction (ZA); - a diaphragm 2001 provided with a vertical orifice 2002 and arranged on said hypergone lens 1002; wherein the width of said vertical orifice 2002 is less than or equal to 5mm, preferably less than or equal to 4mm, preferably less than or equal to 3mm, and wherein the vertical axis (AV) of said vertical orifice 2002 is substantially parallel to the astronomical zenith vertical direction (ZA).
[0030] By "reference plane of the hyperspectral camera", is meant the plane of the camera 1001 comprising the optical axis (AO) of the hypergonal objective 1002 and perpendicular to the astronomical zenith vertical direction (ZA).
[0031] A system 1000 according to the first aspect of the invention operates in the following manner. Once the system is installed in a landscape under a real sky, the hyperspectral camera 1001 acquires a vertical hyperspectral image continuously through the vertical orifice 2002 of the diaphragm 2001, preferably at a vertical angle of at least 170°, while the movable plate 1002 rotates said camera 1001 at an angle of 360° in its reference plane. A complete panoramic hyperspectral image can then be reconstructed by assembling all of the acquired vertical hyperspectral images.
[0032] A system 1000 according to the first aspect of the invention allows a reduction of the spatial extent of the whitish veil likely to appear on a panoramic image in the regions of said image including the sun. Without the present disclosure necessarily being bound by any theory, it is suspected that, thanks to diaphragm 2001 provided with a vertical orifice 2002 as described, the refraction and / or diffusion of light in the hypergon lens 1002 during rotation of the system is considerably reduced, without detriment to the quality of the image. A reduction in the extent of the whitish veil of approximately 90% or even more is generally observed.
[0033] With reference to [Fig.2], the diaphragm 2001 is provided with a vertical orifice 2002, or slit, the vertical axis (AV) of which is substantially parallel to the astronomical zenith vertical direction (ZA) and, preferably, is perpendicular to the optical axis (AO) of the hypergone objective 1002. Thus, only the light coming from the vertical orifice 2002, that is to say according to a vertical solid angle, enters the hypergone objective 1002.
[0034] The diaphragm 2001 is opaque by its function. It can be of any suitable color, for example black. It is also of any suitable material, flexible or rigid. It can be made of rubber in particular.
[0035] The hypergon lens 1002 is of any suitable type. Examples of hypergon lens 1002 are described in JP 2005292280 A [FUJINON CORP] 20.10.2005, WO 2009084842 A2 [NANOPHOTONICS CO LTD [KR]] 09.07.2009, JP 2009 128654 A [SONY CORP] 11.06.2009, WO 2010 077050 A2 [NANOPHOTONICS CO LTD [KR] 08.07.2010. The surface of this type of lens generally has a certain surface curvature.
[0036] Also, according to certain embodiments, the curvature of the surface of the diaphragm 2001 may correspond to the curvature of the surface of the hypergone lens 1002. A correspondence between the curvature of the diaphragm 2001, on the one hand, and the curvature of the hypergone lens 1002, on the other hand, allows better fixing of the diaphragm 2001 on the hypergone lens 1002 and avoids the refraction and / or diffusion of any stray light in the lens, that is to say any light which would not come from the vertical orifice.
[0037] According to certain embodiments, with reference to [Fig.5], the vertical orifice 2002 of the diaphragm 2001 can be offset relative to the optical axis (AO) of said hyperspectral camera 1001. It is surprisingly found that such an offset, i.e. the non-coincidence of said vertical orifice 2002 with the optical center of the hyperspectral camera 1001, further reduces the spatial extension of the whitish veil. According to certain advantageous embodiments, the offset of the vertical orifice 2002 relative to the optical axis (AO) of said hyperspectral camera 1001 can be at least 0.1 mm, preferably at least 0.5 mm.
[0038] According to certain advantageous embodiments, the system 1000 further comprises a data processing device (not shown) configured for the digital processing of the panoramic hyperspectral images acquired by said hyperspectral camera 1001, such that said data processing device comprises means for executing a computer-implemented method and said computer-implemented method takes, as input data, a panoramic image acquired by said hyperspectral camera, provides, as output data, a corrected image of said image, and comprises the following steps: (a) detecting and selecting, in the region of the panoramic image including the sun, one or more columns C[j] of pixels crossing the sun, the majority of whose pixels have a maximum value; (b) applying a corrective model to the columns C[j] of pixels selected in step (a), such that said corrective model is based on one or more pixels of the adjacent columns, C[j-1], C[j+1], to said columns of pixels selected.
[0039] The data processing device may be any device adapted to execute sequences of arithmetic or logical operations automatically in order to perform tasks or actions. This device may comprise one or more central processing units (CPUs) and at least one control device which are adapted to perform these operations.
[0040] The device may also include other electronic components such as input / output interfaces, non-volatile or volatile storage devices, and communication buses for transferring data between electronic components inside and outside said device. One of the input / output devices may be a user interface for human-machine interaction, for example a graphical user interface for displaying human-understandable information.
[0041] A first example of a data processing device may be a computer. A second example may be an embedded system, for example a low-power computer card such as a single-board computer.
[0042] The method implemented by the data processing device may be a computer program comprising instructions which, when the program is executed by said device, causes said device to implement said method.
[0043] Any type of programming language, compiled or interpreted, can be used to implement the steps of said method. The computer program may be part of a software solution, i.e. a collection of executable instructions, codes, scripts or the like and / or databases.
[0044] In some embodiments, the computer program may be stored on a non-transitory computer-readable medium. Accordingly, this non-transitory computer-readable medium includes instructions that, when executed by a computer- diner, make the computer execute said method.
[0045] Generally, a matrix image can be represented, for each channel or spectral band, in the form of a table of lines L[l]-L[m] and columns C[l]-C[n] of pixels Px[l,l]-Px[m,n], as illustrated in [Fig. 6]. In the context of the present invention, the columns are oriented along the vertical axis AV of the vertical orifice 2002.
[0046] The columns C[j] of pixels in which the majority of pixels have a maximum value correspond to the columns comprising at least 50%, or even all of their pixels, whose color shade value is maximum in the encoding scale of the panoramic image.
[0047] As a purely illustrative and educational example, for a conventional, non-hyperspectral image, encoded on 24 bits in the RGB color space, i.e. 8 bits or 1 byte for each R, G and B channel, the pixels of maximum value are those which have the greatest color shade value between 0 and 255 in each R, G and B channel compared to the other adjacent columns of pixels. In the case of a hyperspectral image, generally encoded in a floating point format, for example 32 or 64 floating point bits, the pixels of maximum value are those which have the greatest reflectance value in all the spectral bands, or channels, of said image according to the sensitivity of the detector.
[0048] It has been found that a corrective model based on one or more pixels Px[i,j-1], Px[i,j+1] of the adjacent columns C[i-1], C[i+1] to the columns C[i] of pixels whose value is maximum in the region of the panoramic image including the sun allows an acceptable reconstruction of said selected columns. It allows an elimination or, at least, a drastic reduction of the light blur ("blooming", "bloom effect") with a very limited loss of information relating to said panoramic image.
[0049] According to certain embodiments, with reference to [Fig. 6], the corrective model may replace the values of the pixels Px[i,j] in the selected columns with the average value of the pixel values Px[i,j-1], Px[i,j+1] in the adjacent columns C[i-1], C[i+1] to said selected columns C[i] and located on the same line L[i] in the panoramic image. The number of adjacent columns to be taken into account in calculating the average value depends on the nature of the objects represented in the image. It is preferable that the adjacent columns do not include pixels whose value would be too different from that which could be expected for the pixels of the selected columns if there were no light blur. In other words, adjacent columns comprising objects of a different nature to those of the selected columns should not be taken into account or in a limited manner.
[0050] According to certain embodiments, the corrective model comprises a first step subtraction of the pixel values of one of the adjacent columns C[i-1], C[i+1] from the pixel values of the selected columns C[i], a second step of modeling the profile obtained after this subtraction by a model function, then a third step of subtraction of the modeled profile from the pixel values of the selected columns C[i].
[0051] According to certain preferred embodiments, the corrective model is a trained adversarial generative network in which the generator of said adversarial generative network comprises an artificial convolutional neural network trained on the pixels Px[i,j-1], Px[i,j+1] of the adjacent columns C[i-1], C[i+1] to the selected columns C[i]. The advantage of an adversarial generative network is the possibility of a reconstruction more faithful to the expected reality of the selected columns. The loss of information is minimal and the blurring effect is considerably reduced.
[0052] Preferably, the corrective model can be applied to all the spectral channels or bands of the hyperspectral image. Spectral continuity in each channel can thus be preserved.
[0053] According to certain preferred embodiments, the hyperspectral camera has a spectral range between 400nm and 1000nm, a resolution between 2nm and 4nm, a vertical field angle of at least 170°, an exposure time between 21ps and 10s, and, preferably, a standard resolution of at least 900x1600 pixels.
[0054] Advantageously, the hyperspectral image may be an image acquired according to at least three different exposure times. The hyperspectral image may then be an average of the images acquired according to these three exposure times, for example 21ps, 1ms and 100ms. Alternatively, it is possible to merge the hyperspectral images thus acquired to construct a high dynamic range image, called a GGD image (High Dynamic Range or HDR), encoded in a floating point format, for example 64 floating point bits or higher. The construction of the GGD image may be carried out using any suitable algorithm, for example those described in Kumar et al., 'Camera Calibration Curves for Luminance Data Acquisition using Matlab', INTERNATIONAL JOURNAL OF ENGINEERING RESEARCH & TECHNOLOGY, vol. 03, n°103, 2014.
[0055] A system according to any one of the embodiments of the first aspect of the invention can be advantageously used to generate an environment map for the digital simulation of images, in particular, for the physico-realistic rendering of images or the simulation by ray tracing. The better quality of the hyper-spectral images capable of being obtained by said system makes it possible to significantly improve the simulation of digital representations of buildings and / or vehicles, in a landscape under real sky. The digital representations are more realistic and less likely to be subject to simulation artifacts, particularly those originating from the environment map.
[0056] According to a second aspect of the invention, a method is provided for acquiring a hyperspectral image of a landscape under real sky, said method comprising the following steps: (a) the provision of a CCD hyperspectral camera equipped with a hypergone lens whose optical axis is oriented in a direction substantially perpendicular to the astronomical zenith vertical direction, and a diaphragm arranged on said hypergone lens, said vertical orifice having a width less than or equal to 5mm, preferably less than or equal to 4mm, preferably less than or equal to 3mm, and a vertical axis substantially parallel to the astronomical zenith vertical direction; (b) the acquisition of a panoramic image of a landscape under real sky by rotating said hyperspectral camera through an angle of 360° in the reference plane of said hyperspectral camera with, as axis of rotation, the astronomical zenith vertical direction.
[0057] According to certain advantageous embodiments, the method may further comprise the following steps: (c) detecting and selecting, using a data processing system, within said panoramic image, one or more columns of pixels whose value is maximum in the region of said image including the sun, (d) applying, using said data processing system, a corrective model to the columns of pixels selected in step (c), such that said corrective model is based on one or more pixels adjacent to said selected columns of pixels.
[0058] According to certain embodiments, the hyperspectral image is acquired according to three different exposure times.
[0059] The embodiments described in the context of a system according to the first aspect of the invention are applicable to the method according to the second aspect of the invention. In particular, the method according to the second aspect of the invention may be implemented using a system according to any of the embodiments of the first aspect of the invention. Example
[0060] In [Fig.7] are represented the variations of luminance on a line of pixels for a channel of a panoramic hyperspectral image acquired according to three examples, E1, E2 and E3 of system or method in accordance with the invention, and a counter-example CE not in accordance with the invention.
[0061] For the three examples and the counter-example, the hyperspectral images of a sunny landscape were acquired over a horizontal angle of 360° using a hyperspectral camera having a spectral range between 400nm and 1000nm, a resolution between 2nm and 4nm, a maximum horizontal field angle of 360°, a vertical field angle of at least 170°, an exposure time between 21ps and 10s, and, preferably, a standard resolution of at least 900x1600 pixels.
[0062] In the three examples E1, E2 and E3, the hypergonal lens of the hyperspectral camera is provided with a diaphragm comprising a vertical opening with a width of 3 mm, 4 mm and 5 mm respectively. In the counter-example, the hypergonal lens is devoid of a diaphragm.
[0063] In [Fig.7], all the luminance curves have a maximum around pixel 710. This sudden variation corresponds to the image of the sun.
[0064] The luminance curve (solid line) of the CE counterexample shows, around this maximum, a very rapid variation between 500 and 850 pixels. The width at half-height in this region between 500 and 850 pixels corresponds to the extension of a whitish veil which affects the hyperspectral image in the form of a whitening of the areas of the image adjacent to the sun.
[0065] The width at half-height in this region of 500 - 850 px is significantly reduced for the three examples (3mm, 4mm and 5mm) represented by the dashed curves. This reduction corresponds to a reduction in the extension of the whitish veil and allows the appearance of finer structures, for example, around 620 pixels, almost invisible on the curve corresponding to the counter-example CE. The advantage of the present invention in reducing the whitish veil is clearly demonstrated here.
[0066] [Fig.8] is an example of a black and white panoramic hyperspectral image 8000 of a ground 8001 and a real sky 8002. A light blur phenomenon 8004 appears in the region 8003 of the sun. In the corresponding color hyperspectral image, the area of the real sky 8002 is azure blue with some cirrus clouds, the region 8003 of the sun is a very bright halo and the light blur 8004 materializes in the form of a vertical white line of saturated pixels crossing the image from one side to the other.
[0067] The variation in luminance, L, expressed in logarithmic scale, in the 8004 light blur according to the L1 profile of [Fig.8] is shown in [Fig.9] (dotted curve). Also shown in [Fig.9] (dashed curve) is the variation in luminance according to the L2 profile outside the 8004 light blur. On both curves, the intensity peak around the pixel, Px, 300 corresponds to the 8003 region of the sun.
[0068] According to certain embodiments of the system according to the first aspect of the invention, a corrective model of the light blur 8004 of the column L1 of [Fig.9], based on the values of the pixels of the adjacent column L2. In the present example, the corrective model comprises a first step of subtracting the values of the pixels of the adjacent column L2 to the pixel values of column L1, a second step of modeling the profile obtained after this subtraction by a model function, then a third step of subtraction of the modeled profile to the pixel values of column L1. The model function used in the example is of the type y = A*lxlB where y is the luminance, x the pixel number and A, B are adjustable parameters. The result of this subtraction is represented by the solid line curve in [Fig.9]. Thanks to this operation, the intensity of the peak around 300x is reduced, and allows to reveal details, for example in the region of pixels 400 to 600, which did not appear on the other curves.
[0069] [Fig. 10] is an example of image 10000 obtained after application of the corrective model on [Fig.9]. The light blur 8004, materialized by the vertical line of saturated pixels on [Fig.8], is eliminated. List of references Patent literature
[0070] JP 2005292280 A [FUJINON CORP] 20.10.2005.
[0071] JP 2007028272 A, DOMUS INTERNAT LTD [JP] 01.02.2007.
[0072] JP 2009 128654 A [SONY CORP] 11.06.2009.
[0073] WO 2009084842 A2 [NANOPHOTONICS CO LTD [KR]] 09.07.2009.
[0074] WO 2010 077050 A2 [NANOPHOTONICS CO LTD [KR] 08.07.2010.
[0075] CN 207442995 U, SHENZHEN REAL SCENE TECH CO LTD [CN] 01.06.2018. Non-patent literature
[0076] Kumar et al., 'Camera Calibration Curves for Luminance Data Acquisition using Matlab', INTERNATIONAL JOURNAL OF ENGINEERING RESEARCH & TECHNOLOGY, vol. 03, n°103, 2014
Claims
Claims
1. System (1000) for acquiring a hyperspectral image of a landscape under a real sky, the system comprises: - a CCD hyperspectral camera (1001) provided with a hypergone lens (1002) whose optical axis (AO) is oriented in a direction substantially perpendicular to the astronomical zenith vertical direction (ZA); - a rotating mobile stage (1002) configured to ensure the rotation of said hyperspectral camera through an angle of 360° in the reference plane of said hyperspectral camera with, as axis of rotation, the astronomical zenith vertical direction (ZA); - a diaphragm (2001) provided with a vertical orifice (2002) and arranged on said hypergone lens (1002);wherein the width of said vertical orifice (2002) is less than or equal to 5mm, preferably less than or equal to 4mm, preferably less than or equal to 3mm, and wherein the vertical axis (AV) of said vertical orifice (2002) is substantially parallel to the astronomical zenith vertical direction (ZA).;
2. System (1000) according to claim 1, such that the vertical orifice (2002) of said diaphragm is offset relative to the optical axis (AO) of said hyperspectral camera.
3. System (1000) according to claim 2, such that the offset of the vertical orifice (2002) relative to the optical axis (AO) of the hyperspectral camera (1001) is at least 0.1 mm.
4. System (1000) according to any one of claims 1 to 4, such that the curvature of the surface of the diaphragm (2001) corresponds to the curvature of the surface of the hypergonal objective (1002).
5. A system (1000) according to any one of claims 1 to 4, such that said system (1000) further comprises a data processing device configured for digital processing of panoramic hyperspectral images acquired by said hyperspectral camera (1001), such that said data processing device comprises means for executing a computer-implemented method and said computer-implemented method takes, as input data, a panoramic image acquired by said hyperspectral camera, provides, as output data, a corrected image of said image, and comprises the following steps: (a) detecting and selecting, in the region of the panoramic image comprising the sun, one or more columns C[j] of pixels crossing the sun, the majority of whose pixels have a maximum value; (b) applying a corrective model to the columns C[j] of pixels selected in step (a), such that said corrective model is based on one or more pixels of the adjacent columns C[j-1], C[j+1], to said selected columns of pixels.
6. System (1000) according to claim 6, such that the corrective model can replace the values of the pixels Px[i,j] in the selected columns by the average value of the values of pixels Px[i,j-1], Px[i,j+1] in the adjacent columns C[i-1], C[i+1] to said selected columns C[i] and located on the same line L[i] in the panoramic image.
7. System (1000) according to claim 6, such that the corrective model is a trained adversarial generative network in which the generator of said adversarial generative network comprises an artificial convolutional neural network trained on the pixels Px[i,j-1], Px[i,j+1] of the adjacent columns C[i-1], C[i+1] to the selected columns C[i].
8. System (1000) according to any one of claims 6 to 8, such that the corrective model is applied to all the channels of the hyperspectral image.
9. System (1000) according to any one of claims 1 to 9, such that the hyperspectral camera has a spectral range between 400nm and 1000nm, a resolution between 2nm and 4nm, a maximum horizontal field angle of 360°, a vertical field angle of at least 170°, an exposure time between 21ps and 10s, and, preferably, a standard resolution of at least 900x1600 pixels.
10. System (1000) according to any one of claims 1 to 9, such that the hyperspectral image is an image acquired according to at least three different exposure times.
11. Use of a system (1000) according to any one of claims 1 to 10 for generating an environment map for digital image simulation, in particular, for physicorealistic image rendering or ray tracing simulation.
12. Method for acquiring a hyperspectral image of a landscape under real sky, said method comprising the following steps: (a) providing a CCD hyperspectral camera equipped with a hypergone lens whose optical axis is oriented in a direction substantially perpendicular to the astronomical zenith vertical direction, and a diaphragm arranged on said hypergonal objective, said vertical orifice having a width less than or equal to 5mm, preferably less than or equal to 4mm, preferably less than or equal to 3mm, and a vertical axis substantially parallel to the astronomical zenith vertical direction; (b) the acquisition of a panoramic image of a landscape under real sky by rotating said hyperspectral camera through an angle of 360° in the reference plane of said hyperspectral camera with, as axis of rotation, the astronomical zenith vertical direction.
13. A method according to claim 12, such that said method further comprises the following steps: (c) detecting and selecting, using a data processing system, within said panoramic image, one or more columns of pixels whose value is maximum in the region of said image including the sun, (d) applying, using said data processing system, a corrective model to the columns of pixels selected in step (c), such that said corrective model is based on one or more pixels adjacent to said selected columns of pixels.
14. Method according to any one of claims 12 to 13, such that the hyperspectral image is acquired according to three different exposure times.
