• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    The invention relates to a system for acquiring a hyperspectral image (x), comprising: a sensor (101) for grayscale images; and a scattering and dispersive element (105) placed on the optical path between the sensor (101) and a scene (103), which element (105) comprises an array of individually controllable liquid crystal cells, wherein each cell can receive a control voltage selected from a series of at least three distinct control voltages.
    公开号:FR3021127A1
    申请号:FR1454453
    申请日:2014-05-19
    公开日:2015-11-20
    发明作者:Timothe Laforest；Antoine Dupret；William Guicquero；Arnaud Verdant
    申请人:Commissariat a lEnergie Atomique CEA；Commissariat a lEnergie Atomique et aux Energies Alternatives CEA；
    IPC主号:
    专利说明:

    [0001] TECHNICAL FIELD The present application relates to a system and method for acquiring hyperspectral or multispectral images, and the calibration of such a system or method. BACKGROUND OF THE INVENTION
    [0002] BACKGROUND OF THE PRIOR ART Generally, a scene is called a hyperspectral or multispectral image of a plurality of two-dimensional elementary images of the scene, each image representing a component of the scene in a restricted band of wavelengths. Each elementary image generally corresponds to the integration of the luminous intensity in a specific spectral band. In the following, the expression "hyperspectral image" is considered to be equivalent to the expression "multispectral image", and denotes a series of at least two elementary images of the same scene in bands of distinct wavelengths. . To acquire a hyperspectral image, one method consists in successively acquiring the different images of the series using an image sensor, disposing at each acquisition, between the scene and the sensor, an optical filter passes band tape. narrow pass, for example from a few nanometers to a few tens of nanometers, corresponding to B13227 - DD14758ST 2 one of the components of the hyperspectral image. With each new acquisition, a new filter is placed between the scene and the sensor, so as to successively select the different wavelength bands of the hyperspectral image.
    [0003] A major disadvantage of this method lies in the need to scroll successively different optical filters between the scene and the sensor. This results in relatively bulky acquisition systems and relatively long acquisition times. Once the acquisition is complete, the information contained in each elementary image corresponds to the integration of the luminous intensity into the bandwidth of the corresponding filter. Another method is to use, instead of the filter, a prism, to spatially disperse the different wavelengths of the scene. A matrix system of programmable shutters, of the type commonly designated in the art by the acronym SLM (of the English "Spatial Light Modulator" - spatial light modulator) can be placed between the scene and the prism, upstream of the sensor images. The SLM is for example constituted by a matrix of micro-mirrors or by a matrix of liquid crystal cells. Depending on the spatial and spectral resolutions sought, an acquisition or a small number (that is, less than the total number of spectral bands in the hyperspectral image) of acquisitions may be sufficient to reconstruct the hyperspectral image. This type of method is generally referred to in the art by the acronym CASSI (Coded Aperture Snapshot Spectral Imaging). A disadvantage of this method however lies in the complexity of the optical system necessary to achieve the acquisition, and in the relatively low resolution, elementary images that can be acquired simultaneously by the sensor. Other methods have been proposed, for example in the paper "Compressive sensing spectroscopy based on B13227 - DD14758ST 3 liquid crystal devices" by Yitzhak August et al., Using theories of compressive acquisition to reconstruct the different components. extended spectrum from a small number of acquisitions. In the aforementioned article, an electrically controllable liquid crystal filter is disposed between the scene and an image sensor. During an acquisition, a control voltage is applied to the filter, so that the latter selects not a narrow bandwidth, as in traditional methods, but a wide bandwidth of irregular shape, corresponding to a juxtaposition of the lengths of the filter. waves of the different components of the hyperspectral image to which weights are assigned. Several acquisitions are successively carried out, each time modifying the control of the liquid crystal filter, which amounts to modifying the response of the filter and therefore the weighting coefficients assigned to the components of the hyperspectral image. The authors have shown that by using the mathematical models of compressive acquisition, it is possible to reconstruct a series of elementary components of the extended spectrum from an acquisition number smaller than the total number of spectral bands of the reconstructed series. . A disadvantage of this method however lies in the complexity of the calibration of the liquid crystal filters, and in the reduced possibilities in terms of compression or super-resolution. There is a need for a system and method for acquiring hyperspectral images, overcoming all or part of the disadvantages of known solutions. Thus, an embodiment provides a hyperspectral image acquisition system, comprising: a grayscale image sensor; and a diffusing and dispersive element placed on the optical path between the sensor and a scene, this element comprising an individually controllable liquid crystal cell array, wherein each cell can receive a control voltage selected from a series at least three separate control voltages. According to one embodiment, the system further comprises a control module adapted to control the sensor and the element for successively acquiring an integer M greater than 1 of grayscale elementary images of the scene, by applying, when each acquisition, a command specific to the element. According to one embodiment, for each command applied to the element, the element has a point spreading function which depends on the wavelength of the rays it receives. According to one embodiment, the M commands applied to the diffusive and dispersive element are such that the MxK point spreading functions of the element, corresponding to the M commands applied and K spectral bands of the hyperspectral image, where K is greater than 1, they are irregular and all are different from each other. According to one embodiment, the system further comprises a processing module adapted to reconstruct the hyperspectral image from the M elementary images acquired by the sensor. Another embodiment provides a control method of the aforementioned system, comprising an acquisition phase in which the sensor and the element are controlled to successively acquire an integer M greater than 1 of elementary images in levels of gray of the scene, applying, at each acquisition, a command specific to the element. According to one embodiment, the control method 30 further comprises a phase of reconstruction of the hyperspectral image from the M elementary images acquired during the acquisition phase. According to one embodiment, the control method further comprises a prior calibration phase during which the MxK point spreading functions of the element, B13227-DD14758ST corresponding to the M commands applied and K spectral bands of the hyperspectral image, with K integer greater than 1, are determined. According to one embodiment, the calibration phase 5 comprises the acquisition successively by the sensor of MxK images of spots resulting from the diffusion, by the element, for the M commands of the element and for K light spectral bands of the hyperspectral image of a wavelength-adjustable point light source.
    [0004] According to one embodiment, the MxK images acquired during the calibration phase are put in correspondence with a theoretical behavioral model of the element. BRIEF DESCRIPTION OF THE DRAWINGS These and other features and advantages will be set forth in detail in the following description of particular embodiments in a non-limiting manner with reference to the accompanying figures in which: Figure 1 schematically shows and partial an example of an embodiment of a hyperspectral image acquisition system; FIG. 2 is a diagram illustrating the operating principle of the hyperspectral image acquisition system of FIG. 1; and Figure 3 schematically and partially shows the hyperspectral image acquisition system of Figure 1 during a calibration phase. DETAILED DESCRIPTION For the sake of clarity, the same elements have been designated with the same references in the various figures. Figure 1 schematically and partially shows an example of an embodiment of a hyperspectral image acquisition system. The system of FIG. 1 comprises an image sensor 101, for example a photodetector array, adapted to provide grayscale images. The image sensor 101 is a wideband sensor, ie it is sensitive to all the wavelengths of the hyperspectral images that it is desired to acquire. The system of FIG. 1 further comprises, on the optical path between the sensor 101 and an object or a scene 103 (OBJ) whose hyperspectral image is to be acquired, a programmable diffusive and dispersive element 105, comprising a matrix of cells with individually controllable liquid crystals. The liquid crystals of the element 105 are liquid crystals having a wavelength-dependent refractive index, and preferably liquid crystals whose wavelength dispersion varies as a function of the bias voltage applied, by example of the liquid crystal of the type marketed by MERCK under the reference TL216 or under the reference E44. By way of nonlimiting example, each elementary cell may have dimensions of the order of 1 to 10 gm for applications in the visible range. In practice, the element 105 may comprise a continuous layer of liquid crystals whose first face is coated by a continuous common electrode and a second face opposite the first face is covered by a matrix array of individual controllable discrete elementary electrodes. In this case, an elementary cell of the element 105 is constituted by the portion of the liquid crystal layer disposed between an elementary electrode and the common electrode. In this example, the object 103 which one wishes to acquire a hyperspectral image is supposed to be located at infinity. If necessary, a complementary optical system, not shown, can be arranged between the object and the element 105 so that the object 103 appears at infinity from the point of view of the element 105. The system of the FIG. 1 may further comprise one or more memories (not shown) adapted to store one or more digital images acquired by the sensor 101, a calculation module (not shown), for example a microprocessor, B13227-DD14758ST 7 adapted to process images acquired by the sensor 101, and / or a control module of the sensor 101 and the diffusive and dispersive element 105. The operation of the system of FIG. 1 will now be described. During a phase of acquisition of a hyperspectral image of the object 103, the sensor 101 successively acquires M images ym of the object 103 seen through the scattering and dispersive element 105, where M is a higher integer at 1 and m is an integer in the range from 1 to M. When acquiring an image ym, the scattering and scattering element 105 is modulated by a command Cm, i.e. say a set of control signals of the different liquid crystal cells of the element 105. Each control set makes it possible to generate a specific map of refractive indices on the liquid crystal matrix. In the embodiments described, the cells are not controlled in shutter / opening as in an SLM of a CASSI type system, but each cell can receive a control voltage selected from among a series of at least three distinct voltages. and preferably from a series of at least ten different voltages. Preferably, the element 105 is such that, for a given command Cm of the element 105, there is no discontinuity of refractive index at the boundary between two neighboring cells of the matrix, so that the The liquid crystal layer of element 105 is perceived as a non-discrete (i.e., continuous or analog) element. The M commands Cm applied to the element 105 during the M acquisitions of images by the sensor 105 are all different from each other. At each command Cm of the element 105, the element 105 has a Point Spread Function (PSF) which varies according to the wavelength of the rays received by the element 105 (dispersive nature of the element 105). For a given command Cm of the element 105 and for a given wavelength, the PSF of the element 105 forms a diffusion spot, preferably irregular, which can spread over a plurality of photodetectors. of the sensor 101 and preferably on the major part of the surface of the sensor 101 (diffusing characteristic of the element 105). If we define the PSF of the element 105, at a given wavelength, by the image of the task, in the plane of the sensor 101, resulting from the diffusion by the element 105 of a beam originating from a point light source at this wavelength, the commands Cm are preferably chosen so that the MxK PSF of the element 105 (where K is an integer greater than 1 corresponding to the number of spectral bands of the hyperspectral image that one wishes to acquire) are sufficiently decorrelated from each other. By way of example, a group of M commands of the element 105 may be selected from a group of commands generated on the basis of Zernike polynomials so as to minimize the maximum cross correlation and the maximum autocorrelation of the PSFs. By way of example, a group of commands adapted for a given wavelength can be selected, and then a bias adapted to minimize the correlation of PSFs in the spectral axis can be determined for each control of the group. Preferably, for a given command, the spectral deviation of the PSF between two successive spectral bands of the hyperspectral image to be acquired is less than or equal to one pixel of the sensor, so as to avoid certain spectral artefacts. From the point of view of the compressive acquisition, one can further seek to minimize the mutual coherence between the acquisition matrix (relative to PSF) and the base in which the hyperspectral images that one wishes to acquire are parsimonious. In addition, to define commands for generating orthogonal PSFs for a given wavelength, a non-negative principal component analysis of a group of PSFs generated from commands derived from Zernike polynomials can be performed. It is then possible to return to the commands associated with the orthogonal PSF with the help of an algorithm making it possible to find the wave fronts corresponding to the generated PSFs.
    [0005] The commands Cm are preferably chosen so that, for a given command Cm and for a given wavelength, the PSF of the element 105 is substantially the same at every point of the element 105. If this is not the case, a correction of the distortion can be incorporated in the algorithm for reconstruction of the hyperspectral image. At each command Cm of the element 105, different liquid crystal cells of the element 105 are controlled to present different refractive indices (at constant wavelength), so as to create phase shifts of the front of the element. wave output of the element 105, to generate irregular diffusion spots or PSF. By way of example, the commands Cm correspond to commands making it possible to generate known PSFs, for example PSFs derived from aberrations of the wavefront corresponding to the Zernike polynomials or to a combination of these polynomials. Note in particular, by way of non-limiting example, the case of aberrations of astigmatism or spherical aberrations coupled to a bias or tilt. As will be explained in more detail below, the inventors have determined that, from the M gray level fuzzy images acquired by the sensor 101 during the acquisition phase, it is possible to reconstruct a hyperspectral image. comprising K elementary images xxk corresponding to the components of the image x in the different spectral bands Xk of the image x (with k being an integer ranging from 1 to K). The calculations making it possible to reconstruct the hyperspectral image x from the M images ym will be detailed below. These calculations can be implemented either using a calculation module integrated in the acquisition system, or by an external processing unit. Here, hm, xk denotes the convolution core corresponding to the effect of the PSF of the element 105 on the light of the spectral band Xk, when the element 105 receives the command Cm. If we consider that the elementary images xxk sought have a resolution PxQ greater than a factor s at the resolution B13227 - DD14758ST 10 PsxQs of the sensor 101, with P, Q, s, Ps and Qs integers, Ps = P / s and 45 = Q / s, and if we consider that the MxK PSF of the element 105 can be defined with a level of granularity or no discretization identical to that of the images xxk, each image ym acquired by the sensor 101 can be expressed analytically by the following formula: yin - Ds * X2, k), where k = 1 where * represents the (non-cyclic) convolution operator, and where Ds denotes a decimation operator by the factor s, for example a corresponding averaging operator to the spatial decimation, by the sensor 101, of the hyperspectral image x. Figure 2 shows a schematic modeling of this equation. In a first step, it is considered that one seeks to obtain a hyperspectral image x at the same spatial resolution PsxQs as the sensor 101, and that the MxK PSF of the element 105 can be defined with a level of granularity corresponding to the Ps x Qs resolution of the sensor. The operator Ds can be omitted (case s = 1). Spatial operations in the space domain correspond to point-to-point multiplications or Hadamard products in the Fourier domain. The above equation can then be approximated as follows: Yin - k = 1 where '.' is the product of Hadamard, Ym is the Fourier transform of the image ym, Hm, xk is the Fourier transform of the convolution kernel hm, xk (possibly supplemented by zero if the kernel dimensions hm, xk are smaller than dimensions of the image xxk), and Xxk is the Fourier transform of the elementary image xxk. It should be noted that the edge effects can, if necessary, be compensated by using zero-padding methods, that is to say by increasing the size of the matrix and supplementing with zeros. For each pixel of spatial coordinates p, q of the hyperspectral image x, where p is an integer from 1 to P and q 5 is an integer from 1 to Q, this equation can be expressed as a simple matrix multiplication as follows: Y (p, g) = G (p, q) X (p, g), where Y (p, q) is a vector of dimension M, containing the M values of the pixel of coordinates p, in the M Fourier transform images Ym, X (p, q) is a K-dimensional vector containing the K values of the pixel of coordinates p, q in the K Fourier transform images Xxk, and G (p, q) is a matrix of dimensions MxK containing the MxK values of the pixels of coordinates p, q in the transformed MxKs of Fourier Hm, xk convolution nuclei 15 hm, u. Consider the case where the number M of acquisitions made by the sensor 101 is equal to the number K of elementary images of the hyperspectral image x sought. The matrices G (p, q) are then square matrices. If the MxK PSFs of element 105 are sufficiently decorrelated from each other, the square matrices G (p, q) can be inverted or pseudo-reversed. The M commands Cm of the element 105 are preferably chosen so that the square matrices G (p, q) are invertible. Each vector X (p, q) of the hyperspectral Fourier transform image can be directly calculated by a simple matrix multiplication according to the following formula: X (p, g) = G (p, q) -1-Y ( p, g), where G (p, q) -1 denotes the inverted or pseudo-inverted matrix G (p, q). The different elementary images xxk of the hyperspectral image 30 x can then be reconstructed by simple inverse Fourier transform operations. An advantage of the hyperspectral image acquisition system described above is that it is simple and compact. In particular, it is not necessary to provide a mechanism for mechanically changing the optical filter between the sensor and the scene between the different acquisitions. In addition, the embodiment of the element 105 is relatively simple, since the element 105 must not perform a specific traditional optical filtering function, but only introduce a blur in the image, the only constraint being that the MxK PSF of element 105 have a good level of decorrelation with respect to each other.
    [0006] Note that the matrices G (p, q) -1 may be predetermined, and stored in a memory of a reconstruction module of the hyperspectral image. The matrices G (p, q) -1 may further optionally be preconditioned to avoid certain errors related to edge effects.
    [0007] More generally, if the number of pixel values actually acquired is greater than the number of pixel values to be reconstructed, for example if the number M of acquisitions made by the sensor 101 is less than the number K of elementary images of the image hyperspectral x that we seek to reconstruct, and / or if the spatial resolution PxQ of the hyperspectral image x is greater than the resolution pxq of the sensor 101 (case s> 1), the image x can be reconstructed at the using an iterative regularization method of the type used in the field of compressive acquisition, corresponding to the resolution of a minimization problem formulated as follows: MK 2 argminx 1 (x) + y Fshirt, 2.1, * x2A, - Yrn m = 1 k = 1 2 / where J (x) corresponds to a regularization operator allowing to exacerbate the internal structure of the hyperspectral image and y is a scalar parameter of regularization. By way of nonlimiting example, the operator J (x) is based on five gradient operators V1, V2, V3, V4 and VS, corresponding to constraints that apply to the cube of dimensions PxQxK of the hyperspectral image. x, respectively in the spatial directions B13227 - DD14758ST 13 vertical (P), horizontal (H) and diagonal and in the spectral direction (K), expressed as follows: Xp + 1, q, k Xp, q, k, p <P, iX) p, q, k = to = 13, (v2x) p ,,,, k = r0p, q + 1, k Xp, q, k, q <Q, = (2, rp + 1 , q + 1, k Xp, q, k, P <Pnq <Q, (173x) p, q, k =, p = PUq = Q,, p <Pnq> 1, 4X) p, q, k =, p = Puq = 1, IXpa, k + 1-Xp, q, k, k <K,, k = K, (V 5x) p, q, k = 0 The operator J (x) can for example be defined by the following formula: ## EQU1 ## where IF denotes a chosen two-dimensional wavelet transform operator. Alternatively, the operator J (x) may take other forms. For example, the operator J (x) may be an operator allowing, by using the so-called nuclear standard, to minimize the number of different spectral signatures in the reconstructed data. Moreover, rather than reconstructing the complete hyperspectral image (a value of intensity per pixel and per spectral band), it is possible to reconstruct data directly exploitable by the application in question, for example a map of materials, or indications as to the presence or absence of certain phenomena or objects in the hyperspectral image.
    [0008] An advantage of the proposed acquisition system and method is that they enable the acquisition of hyperspectral images having a spatial resolution greater than that of the sensor and / or having a number of distinct spectral bands greater than the number of acquisitions made by the sensor, particularly simple. In particular, the acquisition of a hyperspectral image having a resolution greater than that of the sensor can be performed by a system having no prism or coded aperture matrix closure device. Another major advantage lies in the fact that, during the acquisition, no spectral filtering is performed, which implies that the choice of the reconstructed bands does not depend on the acquisition but only results from a choice made at the time of the acquisition. 15 reconstruction. It is thus possible, from the same series of M images ym of a scene acquired by the sensor 101, to choose the number and position of the reconstructed bands. A compromise is of course often to be made between the size of the spectral bands, their number, the spatial resolution of the reconstructed image, and the quality / reliability of the reconstructed image. Whatever the method chosen to reconstruct the hyperspectral image x from the M images ym acquired by the sensor 101, it is necessary to know the different PSF hm, u of the element 105. The PSF hm, u may for example be determined during a calibration phase of the system, or be determined by simulation from theoretical response models of the element 105, or be determined by a method combining calibration and simulation. FIG. 3 schematically and partially illustrates the hyperspectral image acquisition system of FIG. 1 during a calibration phase making it possible to determine the different PSF's hm, u of the element 105. In this example, for the calibration the object 103 of FIG. 1 is replaced by a wavelength-adjustable point light source 303 that can successively scan the K spectral bands Xk of the hyperspectral images that it is desired to acquire. The source 303 is for example a wavelength-adjustable laser source. Alternatively, the source 303 may comprise a broadband source coupled to a plurality of individually passable optical bandpass filters corresponding to the different spectral bands of the hyperspectral images that one wishes to acquire. During a calibration phase, it is possible, in a first step, to apply to the element 105 the control C1 and to control the source 303 to transmit only in the spectral band Xi. An image can then be acquired by the sensor 101, defining the PSF hi, x1 of the element 105 for the control C1 and for the wavelength Xi. The operation can be repeated K times without modifying the control of the element 105, but each time modifying the emission wavelength of the source, so as to scan the K spectral bands of the hyperspectral images that are wish to acquire. The entire operation can then be repeated M times by modifying each time the command applied to the element 105 so as to scan the M commands Cm likely to be applied to the element 105 and thus obtain the M x K PSF hm, xk of the element 105. This method makes it possible to obtain PSF hm, xk having a level of granularity or no discretization identical to that of the sensor 101, and having a noise level corresponding to that of the sensor. However, it may be advantageous to define the PSF hm, xk with a discretization pitch smaller than that of the sensor and / or with a noise level lower than that of the sensor (especially if it is desired to reconstruct a hyperspectral image having a higher spatial resolution to that of the sensor or having a number of elementary images xxk greater than the number of images ym acquired by the sensor). For this, the PSFs can be determined by simulation, using a theoretical behavioral model of the element 105.
    [0009] B13227 - DD14758ST 16 Another particularly advantageous method of determining PSF hm, u, combining calibration and simulation, will now be described. At first, MxK images hsm, xk corresponding to MxK PSF of the element 105 to the resolution of the sensor 101, can be acquired using a wavelength-adjustable source according to the aforementioned calibration method. It is then expected to pool the information collected during this acquisition phase to reduce noise and possibly increase the resolution of the PSF acquired. For each band of wavelengths Xk of the hyperspectral images that one wishes to acquire, a theoretical behavioral model hu (1-2) can be defined, where S2 denotes a set of a restricted number of parameters of the element 105. The theoretical model hxk (1-2) may, for example, have a smaller discretization step than that of the sensor 101. For each command Cm of the element 105, the measurements performed hsm, xk are matched with the theoretical behavioral model. hxk so as to determine a set of 20 S.2m parameters defining the actual behavior of the element 105 for the command Cm. The set S.2m can be determined by means of an error minimization algorithm, for example an iterative algorithm, solving the following minimization problem (in the case where it is desired to reduce the effect of a Gaussian noise type): ç n = ar ar ar ar (C C C,,,,,,,,,,,,,,,, k = 1 where Ds is a decimation operator corresponding to the averaging performed by the sensor 101 during the acquisition, compared to an image having the same resolution as the theoretical model hxk, where y is a scalar parameter of regularization, and where C (S2) is a constraint operator on a set of parameters S) . Each of the MxK PSF hm, xk of the element 105 can then be determined by the following formula: B13227 - DD14758ST 17 hrn ,, Ak - hAk (nrii) - This makes it possible to take into account any imperfections of the acquisition system can not be modeled using theoretical assumptions. Advantageously, an interpolation and smoothing function I can be applied. PSF hm, xk of element 105 can then be determined by the following formula: = hAk (nm) - (hsm, Ak - D shAk (flm)) - Particular embodiments have been described.
    [0010] Various variations and modifications will be apparent to those skilled in the art. In particular, a new solution for acquiring hyperspectral images has been proposed. From the teaching described above, those skilled in the art will be able to adapt certain aspects of the described method and system according to the intended application. For example, although only the exemplary embodiments in which the diffusing and dispersive element 105 operates in transmission have been shown in the figures, the described embodiments are not limited to this particular case. Thus, we can provide an acquisition system of the type described in relation to Figures 1 to 3, wherein the element 105 operates in reflection.
    权利要求:
    Claims (10)
    [0001]
    REVENDICATIONS1. A hyperspectral image acquisition system (x), comprising: a grayscale image sensor (101); and a scattering and dispersive element (105) placed on the optical path between the sensor (101) and a scene (103), this element (105) comprising an individually controllable liquid crystal matrix, in which each cell can receive a control voltage selected from a series of at least three distinct control voltages. 10
    [0002]
    2. The system of claim 1, further comprising a control module adapted to control the sensor (101) and the element (105) to successively acquire an integer M greater than 1 of elementary images (ym) in levels of gray of the scene (103), applying, at each acquisition, a command (Cm) specific to the element (105).
    [0003]
    3. System according to claim 2, wherein, for each control (Cm) applied to the element (105), the element (105) has a point spread function which depends on the wavelength of the rays. he receives. 20
    [0004]
    The system of claim 3, wherein the M commands applied to the diffusive and dispersive element (105) are such that the MxK point spreading functions (hm, xk) of the element (105), corresponding to the M commands applied and K spectral bands of the hyperspectral image (x), with K integer greater than 1, are irregular and are all different from each other.
    [0005]
    5. System according to any one of claims 2 to 3, further comprising a processing module adapted to reconstruct the hyperspectral image (x) from the M elementary images (ym) acquired by the sensor (101).
    [0006]
    6. A method of controlling a system according to any one of claims 1 to 5, comprising an acquisition phase during which the sensor (101) and the element (105) are controlled to successively acquire an integerB13227 - DD14758ST 19 M greater than 1 of elementary images (ym) in grayscale of the scene (103), applying, at each acquisition, a command (Cm) specific to the element (105).
    [0007]
    7. The method according to claim 6, further comprising a phase of reconstruction of the hyperspectral image (x) from the M elementary images (ym) acquired during the acquisition phase.
    [0008]
    The method of claim 6 or 7, further comprising a prior calibration phase in which the MxK point spread functions (hm, u) of the element (105), corresponding to the M commands applied and at K spectral bands of the hyperspectral image (x), with K integer greater than 1, are determined.
    [0009]
    9. A method according to claim 8, wherein the calibration phase comprises the acquisition successively by the sensor (101) of MxK images of spots resulting from the diffusion, by the element (105), for the M commands (Cm ) of the element (105) and for the K light spectral bands of the hyperspectral image, a light source (303) point adjustable wavelength.
    [0010]
    The method of claim 9, wherein the MxK images acquired during the calibration phase are mapped to a theoretical behavioral model of the element (105).
    类似技术:
    公开号 | 公开日 | 专利标题
    EP2947433A1|2015-11-25|System and method for acquiring hyperspectral images
    CA2592293C|2015-04-14|Method for processing images using automatic georeferencing of images derived from a pair of images captured in the same focal plane
    EP2567537B1|2015-11-25|Polychromatic imaging method
    EP2909671B1|2016-12-07|Method for designing a single-path imager able to estimate the depth of field
    EP1012549B1|2004-11-17|Method and device for wavefront optical analysis
    EP3290891A1|2018-03-07|Method and device for characterizing optical aberrations of an optical system
    Toivonen et al.2021|Snapshot hyperspectral imaging using wide dilation networks
    FR2994007B1|2019-08-23|METHOD FOR MERGING MULTI-RESOLUTION IMAGES
    FR2645976A1|1990-10-19|ELECTRONIC HOLOGRAPHY APPARATUS
    EP2176633B1|2017-08-30|Method of estimating at least one deformation of the wave front of an optical system or of an object observed by the optical system and associated device
    EP3072081B1|2018-10-10|Determination of the image depth map of a scene
    US20200256732A1|2020-08-13|Holographic mode filter for super-resolution imaging
    JP2016090290A|2016-05-23|Imaging apparatus, spectroscopy system, and spectroscopy method
    FR2974899A1|2012-11-09|MULTISPECTRAL SCANNING TELESCOPE HAVING WAVEN FRONT ANALYSIS MEANS
    Block2005|A sensitivity study of a polychromatic sparse-aperture system
    FR3066271B1|2019-07-12|MOTION SENSOR AND IMAGE SENSOR
    EP2002226B1|2013-10-09|Spectrum-forming device on an optical sensor with spatial rejection
    EP3824269A2|2021-05-26|Method and systems for the non-invasive optical characterization of a heterogeneous medium
    Abi-Rizk et al.2020|Non-Stationary Hyperspectral Forward Model and High-Resolution
    EP2839254B1|2021-06-30|Method and device for reconstructive multi-spectral imaging
    FR3059453A1|2018-06-01|METHOD AND DEVICE FOR DECONVOLUTION OF DIGITAL IMAGES
    WO2014174195A1|2014-10-30|Image capture with addition of accumulation signals for adjacent photodetectors
    EP3278407A1|2018-02-07|Device and method for characterising a femtosecond laser pulse
    FR2692060A1|1993-12-10|Contour extractor type image processing device and earth sensor comprising such an extractor.
    FR2974898A1|2012-11-09|Image analysis device for scanning telescope, has inter-bar region including active optical component to introduce optical aberration on image portion, where optical thickness of optical component varies in accordance with applied command
    同族专利:
    公开号 | 公开日
    EP2947433A1|2015-11-25|
    US9875407B2|2018-01-23|
    US20150332081A1|2015-11-19|
    FR3021127B1|2017-10-13|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    US20060038705A1|2004-07-20|2006-02-23|Brady David J|Compressive sampling and signal inference|CN106679807A|2016-11-01|2017-05-17|北京理工大学|Image compression and reconstruction method based on LCTFhyperspectral imaging system|WO2008149677A1|2007-05-31|2008-12-11|Nikon Corporation|Tunable filter, light source device and spectrum distribution measuring device|
    DE112012004100T5|2011-09-30|2014-07-10|Los Alamos National Security, Llc|Programmable full-frame hyperspectral imaging device|US9581543B2|2014-11-10|2017-02-28|Ci SystemsLtd.|Infrared detection and imaging device with no moving parts|
    US9395293B1|2015-01-12|2016-07-19|Verily Life Sciences Llc|High-throughput hyperspectral imaging with superior resolution and optical sectioning|
    US10605660B2|2015-07-30|2020-03-31|Technology Innovation Momentum FundLimited Partnership|Spectral imaging method and system|
    DE202016101941U1|2016-04-13|2017-07-17|Hans Sasserath Gmbh & Co Kg|Reducer assembly|
    法律状态:
    2015-05-22| PLFP| Fee payment|Year of fee payment: 2 |
    2015-11-20| PLSC| Publication of the preliminary search report|Effective date: 20151120 |
    2016-05-24| PLFP| Fee payment|Year of fee payment: 3 |
    2017-05-30| PLFP| Fee payment|Year of fee payment: 4 |
    2018-05-28| PLFP| Fee payment|Year of fee payment: 5 |
    2019-05-31| PLFP| Fee payment|Year of fee payment: 6 |
    2020-05-30| PLFP| Fee payment|Year of fee payment: 7 |
    2021-05-31| PLFP| Fee payment|Year of fee payment: 8 |
    优先权:
    申请号 | 申请日 | 专利标题
    FR1454453A|FR3021127B1|2014-05-19|2014-05-19|SYSTEM AND METHOD FOR ACQUIRING HYPERSPECTRAL IMAGES|FR1454453A| FR3021127B1|2014-05-19|2014-05-19|SYSTEM AND METHOD FOR ACQUIRING HYPERSPECTRAL IMAGES|
    EP15168023.8A| EP2947433A1|2014-05-19|2015-05-18|System and method for acquiring hyperspectral images|
    US14/715,413| US9875407B2|2014-05-19|2015-05-18|Hyperspectral image acquisition system and method|
    [返回顶部]