WO2019000659A1 - 一种强大气散射条件下的成像装置及方法 - Google Patents

一种强大气散射条件下的成像装置及方法 Download PDF

Info

Publication number
WO2019000659A1
WO2019000659A1 PCT/CN2017/102512 CN2017102512W WO2019000659A1 WO 2019000659 A1 WO2019000659 A1 WO 2019000659A1 CN 2017102512 W CN2017102512 W CN 2017102512W WO 2019000659 A1 WO2019000659 A1 WO 2019000659A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
function
frequency
dimensional
laser
Prior art date
Application number
PCT/CN2017/102512
Other languages
English (en)
French (fr)
Inventor
程雪岷
张临风
郝群
Original Assignee
清华大学深圳研究生院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 清华大学深圳研究生院 filed Critical 清华大学深圳研究生院
Publication of WO2019000659A1 publication Critical patent/WO2019000659A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
    • G01S17/003—Bistatic lidar systems; Multistatic lidar systems
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
    • G01S17/88—Lidar systems specially adapted for specific applications
    • G01S17/89—Lidar systems specially adapted for specific applications for mapping or imaging
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
    • G01S7/48—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S17/00
    • G01S7/481—Constructional features, e.g. arrangements of optical elements

Definitions

  • the present invention relates to the field of image imaging technology, and in particular, to an imaging apparatus and method under strong air scattering conditions.
  • the imaging method in the existing common strong scattering medium is the near-infrared laser active illumination imaging, which utilizes the transmittance of the specific wavelength of light to achieve good imaging effect, and the method essentially uses the near-infrared laser as the light source.
  • the traditional optical imaging principle but the active illumination imaging method under the strong gas scattering conditions, the imaging quality is greatly reduced; improving the imaging quality under the condition of strong gas scattering is the direction that the person skilled in the art is working hard.
  • the present invention proposes an imaging device and method under strong gas scattering conditions.
  • the invention discloses an imaging device under strong gas scattering conditions, which is used for imaging a target through a scattering medium, including a laser, a spatial light modulator, a first lens, a second lens and an image sensor,
  • the spatial light modulator comprises a plurality of reversible micromirrors
  • the spatial light modulator is disposed on a laser light path emitted by the laser, and the laser light path passes through the micromirror of the spatial light modulator
  • the first lens is transmitted and transmitted through the scattering medium to the target, and then the laser light path is reflected by the target and transmitted through the scattering medium, and then passes through the first Two
  • the lens is transmitted and incident on the image sensor to image the target.
  • the laser uses a laser source having a wavelength of 720 to 904 nm.
  • the spatial light modulator comprises M x N matrix arrays of reversible micromirrors.
  • the invention also discloses an imaging method under the condition of strong gas scattering, which is imaged by using the above imaging device, comprising the following steps:
  • S2 inputting 3 sets of M ⁇ N measurement matrices including 0 and 1 to the spatial light modulator, and generating, by using the three sets of the measurement matrix and the corresponding light intensity information received by the image sensor, a second image, wherein 0 in the measurement matrix indicates that the corresponding micromirror in the spatial light modulator is inverted to reflect the laser light path emitted by the laser onto the target;
  • step S1 further comprises: performing filtering processing on the first image to generate a filtered first image, and the first image in step S3 is the filtered first image.
  • step S2 restoring and generating the second image by using the three sets of the measurement matrix and the light intensity information received by the corresponding image sensor specifically includes: adopting the following formula:
  • x is a one-dimensional image original information
  • y is the total intensity of the reflected light received by the image sensor in m samples
  • ⁇ is a measurement matrix set
  • m is a matrix of a set of the measurement matrix
  • n M ⁇ N; by the above formula, x can be reconstructed according to ⁇ and y reconstruction, that is, the second image is restored.
  • step S2 specifically includes the following steps:
  • step S23 sampling and reconstructing the useful pixels in step S22 by using a second set of measurement matrices with a resolution of (N/2) ⁇ (M/2) to obtain a partial image of the medium resolution;
  • step S25 sampling and reconstructing the useful pixels in step S24 using a third set of measurement matrices with a resolution of (N) ⁇ (M) to obtain a high-resolution detail image;
  • the algorithm in which the reconstruction is performed employs an OMP algorithm.
  • step S3 specifically includes:
  • S31 Perform a Fourier transform on the first image and the second image respectively to obtain a frequency domain function
  • step S32
  • F is the highest frequency.
  • step S3 specifically includes:
  • S31 Perform a Fourier transform on the first image and the second image respectively to obtain a frequency domain function
  • step S32 is
  • the corresponding weight When the frequency in the first two-dimensional Gaussian function is less than the first predetermined value, the corresponding weight is 0, and when the frequency is greater than the second predetermined value, the corresponding weight is 1, and the frequency is at the first predetermined value and the second predetermined value. When the frequency is higher, the corresponding weight is larger;
  • the corresponding weight When the frequency in the second two-dimensional Gaussian function is less than the first predetermined value, the corresponding weight is 1, and when the frequency is greater than the second predetermined value, the corresponding weight is 0, and the frequency is at the first predetermined value and the second predetermined value. When the frequency is higher, the corresponding weight is smaller.
  • the invention has the beneficial effects that the imaging device under the powerful gas scattering condition proposed by the invention can simultaneously realize imaging methods of two different principles, including an active illumination imaging method and a compressed sensing ghost imaging method, so that The first image obtained by the active illumination imaging method and the second image obtained by the compressed perceptual ghost imaging method can be simultaneously obtained by the imaging device, so that the first image and the second image can be further processed to obtain a better image.
  • the integrated image greatly improves the image quality under strong air scattering conditions.
  • the laser uses a laser source having a wavelength of 720 nm to 904, so that the laser beam path emitted by the laser has better penetration into the scattering medium in the atmosphere, and the diffraction effect is not maintained.
  • the spatial light modulator comprises M ⁇ N matrix-arranged reversible micromirrors, so that the M ⁇ N measurement matrix can be input to the spatial light modulator to act as a modulation light source by passing three sets of randomly generated measurement matrices.
  • Performing a corresponding test to generate a second image reduces the number of samples for generating the second image; further, the imaging method innovatively designs a layered reconstruction method to reduce the number of sampling times at high resolution by distributing sampling, The amplitude reduces the sampling time and shortens the reconstruction time.
  • a Gaussian function or a specific piecewise function can be used as a weight function of the first image and the second image, and the frequency is performed.
  • the method of domain weight addition adds that the final composite image is superior to the first image and the second image.
  • FIG. 1 is a schematic view of an image forming apparatus under strong gas scattering conditions in accordance with a preferred embodiment of the present invention
  • Figure 2a is a spectrogram of the original image of the target
  • 2b and 2c are spectrograms of the first image and the second image at low scattering coefficients
  • 2d and 2e are spectrograms of the first image and the second image at high scattering coefficients
  • FIG. 3 is a schematic diagram of a second two-dimensional piecewise function in some embodiments of the present invention.
  • 4a and 4b are schematic views respectively showing a composite image in which a Gaussian function and a piecewise function are treated as a weight function;
  • FIG. 5a is a schematic diagram of a first two-dimensional Gaussian function according to Embodiment 1 of the present invention.
  • FIG. 5b is a schematic diagram of a second two-dimensional Gaussian function according to Embodiment 1 of the present invention.
  • FIG. 6a is a schematic diagram showing a result of multiplying a first two-dimensional Gaussian function by a frequency domain function of a first image according to Embodiment 1 of the present invention
  • 6b is a schematic diagram showing a result of multiplying a second two-dimensional Gaussian function of the first embodiment of the invention with a frequency domain function of the second image;
  • Figure 6c is the result of the addition of Figures 6a and 6b;
  • FIG. 7a is a schematic diagram of a first image obtained according to Embodiment 1 of the present invention.
  • FIG. 7b is a schematic diagram of a second image obtained according to Embodiment 1 of the present invention.
  • Figure 7c is a schematic diagram of a composite image obtained in Embodiment 1 of the present invention.
  • FIG. 8a is a schematic diagram of an original image of a target of Embodiment 2 of the present invention.
  • FIG. 8b is a schematic diagram of a first image obtained according to Embodiment 2 of the present invention.
  • Figure 8c is a schematic diagram of the filtered first image obtained by Gaussian filtering of Figure 8b;
  • Figure 8d is a schematic diagram of a second image obtained in the second embodiment of the present invention.
  • Embodiment 8e is a schematic diagram of a comprehensive image obtained in Embodiment 2 of the present invention.
  • 9a is a schematic diagram of an original image of a third object of the embodiment of the present invention.
  • FIG. 9b is a schematic diagram of a first image obtained in Embodiment 3 of the present invention.
  • Figure 9c is a schematic diagram of the filtered first image obtained by Gaussian filtering of Figure 9b;
  • Figure 9d is a schematic diagram of a second image obtained in the third embodiment of the present invention.
  • Figure 9e is a schematic diagram of a composite image obtained in the third embodiment of the present invention.
  • an imaging apparatus under a strong gas scattering condition of a preferred embodiment of the present invention includes a laser 10, a spatial light modulator 20, a first lens 30, a second lens 40, and an image sensor 50, through which the target device is targeted
  • the object 60 is imaged with a scattering medium 70 between the imaging device and the target 60.
  • the main structure of the imaging device is that the spatial light modulator 20 includes a plurality of reversible micromirrors, the spatial light modulator 20 is disposed on the laser light path emitted by the laser 10, and the laser light path is reflected by the micromirror of the spatial light modulator 20.
  • the laser 10 After being transmitted through the first lens 30 and passing through the scattering medium 70, the laser light path is reflected by the object 60 and transmitted through the scattering medium 70, and then transmitted through the second lens 40 to be incident on the image sensor 50.
  • the target 60 is imaged.
  • the laser 10 uses a laser light source having a wavelength of 720 to 904 nm
  • the spatial light modulator 20 includes M x N matrix-arranged reversible micromirrors.
  • the laser 10 of the imaging device uses a near-infrared laser light source with a wavelength of 808 nm, which has better permeability to rain and fog in the atmosphere, and the spatial light modulator 20 includes M ⁇ N matrix arrays.
  • Reversible micromirror. Imaging the target by the imaging device includes the following steps:
  • a measurement matrix of M ⁇ N of all 1 is input to the spatial light modulator 20, and a first image is generated on the image sensor 50, wherein 1 in the measurement matrix represents a corresponding micromirror in the spatial light modulator 20.
  • 1 in the measurement matrix represents a corresponding micromirror in the spatial light modulator 20.
  • the spatial light modulator 20 reflects all the light, and the entire optical path is a laser active illumination imaging optical path, and a two-dimensional image of the target object is received on the image surface of the image sensor 50;
  • the first image is also subjected to filtering processing to generate a filtered first image, wherein the filtering process may adopt a Gaussian filtering method.
  • S2 Input three sets of M ⁇ N measurement matrices (which may be randomly generated) including 0 and 1 to the spatial light modulator 20, and integrate the light intensity information received by the three sets of measurement matrices and the corresponding image sensor 50. Reducing to generate a second image, wherein 0 in the measurement matrix means that the corresponding micromirror in the spatial light modulator 20 is inverted to reflect the laser light path not emitted by the laser 1 onto the target 60;
  • the spatial light modulator 20 is composed of M ⁇ N reversible micromirrors, and by inputting a specific measurement matrix, a part of the micromirrors of the surface can be flipped, so that light of a specific spatial position is obtained.
  • the modulation of the light source is realized; whether the rotation of the micromirror array determines whether the light in the region is reflected to the target 60, thereby determining whether the corresponding region of the surface of the target 60 is illuminated, that is, each measurement
  • the matrix actually corresponds to the area where the surface of the object is illuminated; through the plurality of measurement matrices in the three groups and the light intensity information received by the corresponding image sensor, the second image is restored, and the following formula is specifically adopted:
  • x is the one-dimensional image original information
  • y is the total intensity of the reflected light received by the m-sampled image sensor 50
  • ⁇ is the measurement matrix set
  • the constructed algorithm can use the OMP algorithm (orthogonal matching pursuit algorithm).
  • step S2 specifically includes the following steps:
  • step S23 sampling and reconstructing the useful pixels in step S22 by using a second set of measurement matrices with a resolution of (N/2) ⁇ (M/2) to obtain a partial image of the medium resolution;
  • step S25 sampling and reconstructing the useful pixels in step S24 using a third set of measurement matrices with a resolution of (N) ⁇ (M) to obtain a high-resolution detail image;
  • the spectrum of the original image is shown in Fig. 2a (the abscissa is the frequency, the vertical The coordinates are the magnitude of the frequency domain function after Fourier transform), under the low scattering coefficient (3.5),
  • the spectrum of the first image obtained by the active illumination imaging method is shown in Fig. 2b
  • the spectrum of the second image obtained by the compressed perceptual ghost imaging method is shown in Fig. 2c.
  • the dominant interval of the method the interval of more than one tenth of the highest frequency (5-45) can be considered as the absolute interval; under the high scattering coefficient (6.5), the spectrum of the first image obtained by the active illumination imaging method is shown in Figure 2d. As shown in the figure, the spectrum of the second image obtained by the compressed perceptual ghost imaging method is shown in Fig. 2e. It can be seen from the comparison that the high frequency band should still be based on the active illumination imaging result, but the highest frequency is one-twentieth. Below (within 2.5 in the figure), the compressed-chase ghost imaging method shows an advantage.
  • the first two-dimensional piecewise function and the second two-dimensional piecewise function may be employed as weights of the first image and the second image, respectively.
  • the function performs frequency domain weighting, and directly uses the spectrum of the dominant method in the absolute dominant range of the two, and smoothly transitions in the middle region, as follows:
  • F is the highest frequency, which depends on the size of the image of the target, and the actual value is half the diagonal length of the image of the target.
  • step S3 may specifically include:
  • S31 Perform a Fourier transform on the first image and the second image respectively to obtain a frequency domain function
  • FIG. 4a is a comprehensive image processed by the piecewise function
  • SSIM is 0.76053
  • FIG. 4b is a comprehensive image processed by Gaussian function.
  • the SSIM is 0.78426.
  • the first two-dimensional piecewise function and the second two-dimensional piecewise function in the above step S32 may also take the first two-dimensional Gaussian function and the second two-dimensional Gaussian function respectively.
  • step S3 may further include:
  • S31 Perform a Fourier transform on the first image and the second image respectively to obtain a frequency domain function
  • S32 using a first two-dimensional Gaussian function and a second two-dimensional Gaussian function as weight functions of the first image and the second image, respectively, wherein the first two-dimensional Gaussian function and the second two-dimensional Gaussian function are respectively normalized, and The sum of the first two-dimensional Gaussian function and the second two-dimensional Gaussian function is one;
  • the corresponding weight when the frequency in the first two-dimensional Gaussian function is less than the first predetermined value, the corresponding weight is 0, and when the frequency is greater than the second predetermined value, the corresponding weight is 1, and the frequency is at the first predetermined value and the second predetermined value. When the frequency is higher, the corresponding weight is larger;
  • the corresponding weight When the frequency in the second two-dimensional Gaussian function is less than the first predetermined value, the corresponding weight is 1, and when the frequency is greater than the second predetermined value, the corresponding weight is 0, and the frequency is between the first predetermined value and the second predetermined value. When the frequency is higher, the corresponding weight is smaller.
  • Embodiment 1 is a diagrammatic representation of Embodiment 1:
  • FIG. 5a A schematic diagram of the first two-dimensional Gaussian function is shown in FIG. 5a, and a schematic diagram of the second two-dimensional Gaussian function is shown in FIG. 5b, and the sum of the two is 1, and the two are respectively subjected to the first image and the second image.
  • Multiplying the frequency domain function obtained by the Fourier transform (the first two-dimensional Gaussian function is multiplied by the frequency domain function obtained by the Fourier transform of the first image as shown in Fig. 6a, the second two-dimensional Gaussian function and the second The image is multiplied by the frequency domain function obtained by Fourier transform as shown in Fig. 6b), and then added to obtain a comprehensive frequency domain function.
  • FIG. 7c after the Fourier transform, the integrated image of the final target is generated as shown in FIG. 7c, wherein the first image and the second image obtained by the steps S1 and S2 respectively are as shown in FIG. 7a and FIG. 7b, respectively.
  • FIG. 7c As shown, comparing Fig. 7c with Figs. 7a and 7b, respectively, it can be seen that the effect of the integrated image is superior to the first image and the second image.
  • Embodiment 2 is a diagrammatic representation of Embodiment 1:
  • the original image of the target is shown in Fig. 8a.
  • structural similarity SSIM
  • the first image obtained according to step S1 is as shown in FIG. 8b
  • the SSIM is 0.7564
  • the first image after Gaussian filtering is as shown in FIG. 8c
  • the SSIM is 0.89081
  • the second image obtained according to step S2 is as shown in FIG. 8d.
  • the SSIM is 0.60799; according to step S3, the first two-dimensional Gaussian function and the second two-dimensional Gaussian function as shown in FIG. 5a and FIG. 5b are respectively added with the frequency domain weighting of the first image and the second image respectively to obtain a composite image.
  • the SSIM is 0.88449.
  • Embodiment 3 is a diagrammatic representation of Embodiment 3
  • the original image of the target is shown in Fig. 9a.
  • structural similarity SSIM
  • the first image obtained according to step S1 is as shown in FIG. 9b
  • the SSIM is 0.20654
  • the first image after Gaussian filtering is as shown in FIG. 9c
  • the SSIM is 0.43112
  • the second image obtained according to step S2 is as shown in FIG. 9d.
  • SSIM is 0.37954; according to step S3, the first two-dimensional Gaussian function and the second two-dimensional Gaussian function of FIG. 5a and FIG. 5b are respectively used to perform frequency domain weighted addition with the first image and the second image, respectively, to obtain a composite image.
  • the SSIM is 0.54104.
  • the second embodiment and the third embodiment are the comprehensive reconstruction effects under different scattering coefficients. It can be seen that by combining the two imaging methods, the difference between the two methods can be very close to the better imaging effect. When it's not good, you can get an image that works better than both.
  • the first image is an image obtained according to an active illumination imaging method
  • the second image is an image obtained according to a compressed perceptual ghost imaging method, wherein the active illumination imaging method and the compressed perceptual ghost imaging method are in principle large
  • the compressed perceptual ghost imaging method is not sensitive to random noise, but sensitive to the overall light field fluctuation, and the active illumination imaging method is exactly the opposite; the compressed perceptual ghost imaging method can better preserve the low frequency information of the image.
  • the active illumination imaging method preserves the medium and high frequency information well within a certain range of scattering coefficients.
  • the image obtained by the active illumination imaging method is also subjected to filtering processing, such as Gaussian filtering, in a preferred embodiment of the present invention.
  • the present invention firstly combines the two methods.
  • the frequency domain characteristics resulting in results that are superior to both methods.
  • the present invention overcomes the prejudice researched in the prior art, and integrates two imaging modes into the same imaging device, so that the imaging device can realize imaging methods of two different principles, and is better than two by the above specific algorithm.
  • the images obtained by the respective images greatly improve the imaging quality under the condition of strong gas scattering.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • General Physics & Mathematics (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Electromagnetism (AREA)
  • Image Processing (AREA)
  • Investigating Or Analysing Materials By Optical Means (AREA)

Abstract

一种强大气散射条件下的成像装置，用于透过散射介质(70)对目标物(60)进行成像，包括激光器(10)、空间光调制器(20)、第一透镜(30)、第二透镜(40)和图像传感器(50)，其中，空间光调制器(20)包括多个可翻转的微镜，空间光调制器(20)设置在激光器(10)射出的激光光路上，激光光路经空间光调制器(20)的微镜反射后再经过第一透镜(30)透射后并透过散射介质射(70)在目标物(60)上，然后激光光路经目标物(60)反射并透过散射介质(70)后再经过第二透镜(40)透射后射在图像传感器上(50)以对目标物(60)进行成像。一种强大气散射条件下的成像方法。强大气散射条件下的成像装置及方法，大大提高了在强大气散射条件下的成像质量。

Description

一种强大气散射条件下的成像装置及方法 技术领域
本发明涉及图像成像技术领域，尤其涉及一种强大气散射条件下的成像装置及方法。
背景技术
航空、航海和公路交通等行业对在雾霾、雨雾等强散射介质中的成像有着广泛的需求。现有常见的强散射介质中的成像方法为近红外激光主动照明成像，利用特定波长的光对大气散射的穿透性实现较好的成像效果，该方法本质上就是使用近红外激光作为光源的传统的光学成像原理，但是主动照明成像方法在强大气散射条件下，成像质量大大降低；提高在强大气散射条件下的成像质量，是本领域技术人员正在努力的方向。
以上背景技术内容的公开仅用于辅助理解本发明的构思及技术方案，其并不必然属于本专利申请的现有技术，在没有明确的证据表明上述内容在本专利申请的申请日已经公开的情况下，上述背景技术不应当用于评价本申请的新颖性和创造性。
发明内容
为了提高在强大气散射条件下的成像质量，本发明提出一种强大气散射条件下的成像装置及方法。
为了达到上述目的，本发明采用以下技术方案：
本发明公开了一种强大气散射条件下的成像装置，所述成像装置用于透过散射介质对目标物进行成像，包括激光器、空间光调制器、第一透镜、第二透镜和图像传感器，其中，所述空间光调制器包括多个可翻转的微镜，所述空间光调制器设置在所述激光器射出的激光光路上，所述激光光路经所述空间光调制器的所述微镜反射后再经过所述第一透镜透射后并透过所述散射介质射在所述目标物上，然后所述激光光路经所述目标物反射并透过所述散射介质后再经过所述第二 透镜透射后射在所述图像传感器上以对所述目标物进行成像。
优选地，所述激光器采用波长为720～904nm的激光光源。
优选地，所述空间光调制器包括M×N个矩阵排列的可翻转的所述微镜。
本发明还公开了一种强大气散射条件下的成像方法，采用上述的成像装置进行成像，包括以下步骤：
S1：将一个全为1的M×N的测量矩阵输入到所述空间光调制器，在所述图像传感器上生成第一图像，其中所述测量矩阵中的1表示将所述空间光调制器中对应的所述微镜翻转至将所述激光器射出的激光光路反射到所述目标物上；
S2：将3组包含0和1的M×N的测量矩阵输入到所述空间光调制器，通过该3组所述测量矩阵和相应的所述图像传感器接收的光强信息，综合还原生成第二图像，其中所述测量矩阵中的0表示将所述空间光调制器中对应的所述微镜翻转至不将所述激光器射出的激光光路反射到所述目标物上；
S3：采用频域加权的方式将所述第一图像和所述第二图像进行加权相加，生成最终的所述目标物的综合图像。
优选地，步骤S1还包括，对所述第一图像进行滤波处理，生成滤波后的第一图像，步骤S3中的所述第一图像为滤波后的所述第一图像。
优选地，步骤S2中通过该3组所述测量矩阵和相应的所述图像传感器接收的光强信息，还原生成第二图像具体包括：采用以下计算公式：
y＝Φx
Figure PCTCN2017102512-appb-000001
其中，x是一维化的图像原始信息，y是m次采样所述图像传感器接收到的反射光总强度，Φ是测量矩阵集，m是一组所述测量矩阵的矩阵数量，n＝M×N；通过上述公式可以根据Φ和y重构生成x，即还原生成所述第二图像。
优选地，步骤S2具体包括以下步骤：
S21：使用分辨率为(N/4)×(M/4)的第一组测量矩阵，对图像进行采样和重构，得到低分辨率的完整图像；
S22：根据低分辨率图像的梯度排序，选取前50％的像素作为有用像素；
S23：使用分辨率为(N/2)×(M/2)的第二组测量矩阵，对步骤S22中的有用像素进行采样和重构，得到中分辨率的局部图像；
S24：根据中分辨率图像的梯度排序，选取前50％的像素作为有用像素；
S25：使用分辨率为(N)×(M)的第三组测量矩阵，对步骤S24中的有用像素进行采样和重构，得到高分辨率的细节图像；
S26：将上述图像组合，得到50％区域低分辨率，25％区域中分辨率和25％区域高分辨率的整体图像，即为第二图像。
优选地，其中重构的算法采用OMP算法。
优选地，步骤S3具体包括：
S31：将所述第一图像和所述第二图像分别经过傅里叶变换得到频域函数；
S32：采用第一二维分段函数和第二二维分段函数分别作为所述第一图像和所述第二图像的权重函数；
S33：将所述第一二维分段函数和所述第二二维分段函数分别与所述第一图像和所述第二图像经过傅里叶变换得到的频域函数相乘，然后相加，得到综合的频域函数，再进行反傅里叶变换，即生成最终的所述目标物的综合图像；
其中在步骤S32中：
所述第一二维分段函数w1与频率f的关系如下：
Figure PCTCN2017102512-appb-000002
所述第二二维分段函数w2与频率f的关系如下：
Figure PCTCN2017102512-appb-000003
其中，F为最高频率。
优选地，步骤S3具体包括：
S31：将所述第一图像和所述第二图像分别经过傅里叶变换得到频域函数；
S32：采用第一二维高斯函数和第二二维高斯函数分别作为所述第一图像和 所述第二图像的权重函数，其中所述第一二维高斯函数和所述第二二维高斯函数分别经过归一化，且所述第一二维高斯函数和所述第二二维高斯函数之和为1；
S33：将所述第一二维高斯函数和所述第二二维高斯函数分别与所述第一图像和所述第二图像经过傅里叶变换得到的频域函数相乘，然后相加，得到综合的频域函数，再进行反傅里叶变换，即生成最终的所述目标物的综合图像。
优选地，步骤S32中：
所述第一二维高斯函数中频率小于第一预定值时，对应的权重为0，频率大于第二预定值时，对应的权重为1，频率在第一预定值和所述第二预定值之间时，频率越大，对应的权重越大；
所述第二二维高斯函数中频率小于第一预定值时，对应的权重为1，频率大于第二预定值时，对应的权重为0，频率在第一预定值和所述第二预定值之间时，频率越大，对应的权重越小。
与现有技术相比，本发明的有益效果在于：本发明提出的强大气散射条件下的成像装置可以同时实现两种不同原理的成像方式，包括主动照明成像方法和压缩感知鬼成像方法，使得通过该成像装置可以同时得到采用主动照明成像方法得到的第一图像以及采用压缩感知鬼成像方法得到的第二图像，从而可以进一步将第一图像和第二图像进行综合处理，以得到较优的综合图像，从而大大提高了在强大气散射条件下的成像质量。
在进一步的方案中，激光器采用波长为720nm～904的激光光源，使得激光器发出的激光光路对大气中的散射介质具有更好的穿透性，且保持不会发生衍射效应。空间光调制器包括M×N个矩阵排列的可翻转的微镜，从而可以将M×N的测量矩阵输入到空间光调制器，起到调制光源的作用，通过将3组随机产生的测量矩阵进行相应的测试生成第二图像，减少了产生第二图像的采样次数；进一步，该成像方法创新性地设计了分层重构的方式，通过分布采样，减少高分辨率下的采样次数，大幅度减少采样时间，缩短重构时间。
在更进一步的方案中，结合主动照明成像方法和压缩感知鬼成像方法的成像特性，本发明中可以通过高斯函数或者特定的分段函数作为第一图像和第二图像的权重函数，并进行频域加权相加的方法，得到最终的综合图像均优于第一图像和第二图像。
附图说明
图1是本发明优选实施例的强大气散射条件下的成像装置的示意图；
图2a是目标物的原图的频谱图；
图2b和图2c是在低散射系数下第一图像和第二图像的频谱图；
图2d和图2e是在高散射系数下第一图像和第二图像的频谱图；
图3是本发明一些实施例中的第二二维分段函数的示意图；
图4a和图4b分别是将高斯函数和分段函数作为权重函数处理的综合图像的示意图；
图5a是本发明实施例一的第一二维高斯函数的示意图；
图5b是本发明实施例一的第二二维高斯函数的示意图；
图6a是本发明实施例一的第一二维高斯函数与第一图像的频域函数相乘的结果示意图；
图6b是发明实施例一的第二二维高斯函数与第二图像的频域函数相乘的结果示意图；
图6c是图6a和图6b的相加的结果；
图7a是本发明实施例一得到的第一图像的示意图；
图7b是本发明实施例一得到的第二图像的示意图；
图7c是本发明实施例一得到的综合图像的示意图；
图8a是本发明实施例二目标物的原图图像的示意图；
图8b是本发明实施例二得到的第一图像的示意图；
图8c是图8b经过高斯滤波得到的滤波后的第一图像的示意图；
图8d是本发明实施例二得到的第二图像的示意图；
图8e是本发明实施例二得到的综合图像的示意图；
图9a是本发明实施例三目标物的原图图像的示意图；
图9b是本发明实施例三得到的第一图像的示意图；
图9c是图9b经过高斯滤波得到的滤波后的第一图像的示意图；
图9d是本发明实施例三得到的第二图像的示意图；
图9e是本发明实施例三得到的综合图像的示意图。
具体实施方式
下面对照附图并结合优选的实施方式对本发明作进一步说明。
如图1所示，本发明优选实施例的强大气散射条件下的成像装置包括激光器10、空间光调制器20、第一透镜30、第二透镜40和图像传感器50，通过该成像装置对目标物60进行成像，其中在该成像装置和目标物60之间存在散射介质70。其中该成像装置的主要结构为：空间光调制器20包括多个可翻转的微镜，空间光调制器20设置在激光器10射出的激光光路上，激光光路经空间光调制器20的微镜反射后再经过第一透镜30透射后并透过散射介质70射在目标物60上，然后激光光路再经过目标物60反射并透过散射介质70后经过第二透镜40透射后射在图像传感器50上以对目标物60进行成像。其中，在部分实施例中激光器10采用波长为720～904nm的激光光源，空间光调制器20包括M×N个矩阵排列的可翻转的微镜。
在本发明具体实施例中，该成像装置的激光器10采用波长为808nm的近红外激光光源，对大气中的雨雾具有比较好的穿透性，空间光调制器20包括M×N个矩阵排列的可翻转的微镜。通过该成像装置对目标物进行成像，包括以下步骤：
S1：将一个全为1的M×N的测量矩阵输入到空间光调制器20，在图像传感器50上生成第一图像，其中测量矩阵中的1表示将空间光调制器20中对应的微镜翻转至将激光器10射出的激光光路反射到目标物60上；
此时，空间光调制器20反射所有光，整个光路就是一个激光主动照明成像光路，在图像传感器50的像面上接收到的就是目标物的二维图像；
在一些实施例中，还对第一图像进行滤波处理，生成滤波后的第一图像，其中滤波处理可以采用高斯滤波方法。
S2：将3组包含0和1的M×N的测量矩阵(可以是随机产生的)输入到空间光调制器20，通过该3组测量矩阵和对应的图像传感器50接收的光强信息，综合还原生成第二图像，其中测量矩阵中的0表示将空间光调制器20中对应的微镜翻转至不将激光器1射出的激光光路反射到目标物60上；
其中，在本实施例中，空间光调制器20由M×N个可翻转的微镜组成，通过输入特定的测量矩阵，可以让其表面的部分微镜翻转，使得特定空间位置的光 才能被反射，实现对光源的调制；微镜阵列的翻转与否决定了该区域的光是否向目标物60反射，进而决定了目标物60表面的对应区域是否被照亮，也即每一个测量矩阵实际对应了物体表面被照亮的区域；通过3组中多个测量矩阵和对应的图像传感器接收的光强信息，还原生成第二图像，具体采用以下计算公式：
y＝Φx
Figure PCTCN2017102512-appb-000004
其中，x是一维化的图像原始信息，y是m次采样图像传感器50接收到的反射光总强度，Φ是测量矩阵集，m是一组测量矩阵的矩阵数量，n＝M×N，也即测量矩阵集中的每一行即对应一次采样中空间光调制器的一组编码(对应一个测量矩阵)；通过上述公式可以根据Φ和y重构生成x，即还原生成第二图像；其中重构的算法可以采用OMP算法(正交匹配追踪算法)。
在一些实施例中，步骤S2具体包括以下步骤：
S21：使用分辨率为(N/4)×(M/4)的第一组测量矩阵，对图像进行采样和重构，得到低分辨率的完整图像；
S22：根据低分辨率图像的梯度排序，选取前50％的像素作为有用像素；
S23：使用分辨率为(N/2)×(M/2)的第二组测量矩阵，对步骤S22中的有用像素进行采样和重构，得到中分辨率的局部图像；
S24：根据中分辨率图像的梯度排序，选取前50％的像素作为有用像素；
S25：使用分辨率为(N)×(M)的第三组测量矩阵，对步骤S24中的有用像素进行采样和重构，得到高分辨率的细节图像；
S26：将上述图像组合，得到50％区域低分辨率，25％区域中分辨率和25％区域高分辨率的整体图像，即为第二图像。
S3：采用频域加权的方式将第一图像和第二图像进行加权相加，生成最终的目标物的综合图像。
通过对不同散射系数条件下主动照明成像方法和压缩感知鬼成像方法的频谱图分别与目标物的原图的频谱图作比较，原图的频谱图如图2a所示(横坐标为频率，纵坐标为傅里叶变换后的频域函数的幅值)，在低散射系数(3.5)下， 主动照明成像方法得到的第一图像的频谱图如图2b所示，压缩感知鬼成像方法得到的第二图像的频谱图如图2c所示，通过比较可以看出整个波段几乎都是主动照明成像方法的优势区间，最高频率的十分之一以上(5-45)这个区间可以认为是绝对区间；在高散射系数(6.5)下，主动照明成像方法得到的第一图像的频谱图如图2d所示，压缩感知鬼成像方法得到的第二图像的频谱图如图2e所示，通过比较可以看出在高频段仍然应该以主动照明成像结果为基准，但在最高频率为二十分之一以下(图中的2.5以内)，压缩感知鬼成像方法显现出了优势。
因此根据主动照明成像方法和压缩感知鬼成像方法的该特性，在一些实施例中，可以采用第一二维分段函数和第二二维分段函数分别作为第一图像和第二图像的权重函数来进行频域加权，在两者的绝对优势区间直接使用优势方法的频谱，在中间区域平滑过渡，具体如下：
第一二维分段函数w1与频率f的关系如下：
Figure PCTCN2017102512-appb-000005
第二二维分段函数w2与频率f的关系如下，如图3所示：
Figure PCTCN2017102512-appb-000006
其中，F为最高频率，该值取决于目标物的图像的尺寸，实际值为目标物的图像的对角线长度的一半。
进一步地，步骤S3具体可以包括：
S31：将第一图像和第二图像分别经过傅里叶变换得到频域函数；
S32：采用第一二维分段函数和第二二维分段函数分别作为第一图像和第二图像的权重函数；
S33：将第一二维分段函数和第二二维分段函数分别与第一图像和第二图像经过傅里叶变换得到的频域函数相乘，然后相加，得到的综合的频域函数，再进行反傅里叶变换，即生成最终的目标物的综合图像。
从图3所示的分段函数的示意图可以看出，该函数类似于高斯函数，因此，在本发明另一些实施例中，还可采用高斯函数作为权重函数来进行频域加权，如图4a和图4b所示，将分段函数和高斯函数分别作为权重函数处理进行比较，图4a为通过分段函数综合处理的综合图像，SSIM为0.76053，图4b为通过高斯函数综合处理的综合图像，SSIM为0.78426，可以看出分段函数综合处理和高斯函数综合的效果几乎没有区别，也即通过分段函数和高斯函数作为权重函数均得到了较优的效果，甚至高斯函数的效果可以更佳。
因此，在另一种实施例中，上述步骤S32中的第一二维分段函数和第二二维分段函数还可以分别采用第一二维高斯函数和第二二维高斯函数来取值，即步骤S3具体还可以包括：
S31：将第一图像和第二图像分别经过傅里叶变换得到频域函数；
S32：采用第一二维高斯函数和第二二维高斯函数分别作为第一图像和第二图像的权重函数，其中第一二维高斯函数和第二二维高斯函数分别经过归一化，且第一二维高斯函数和第二二维高斯函数之和为1；
S33：将第一二维高斯函数和第二二维高斯函数分别与第一图像和第二图像经过傅里叶变换得到的频域函数相乘，然后相加，得到的综合的频域函数，再进行反傅里叶变换，即生成最终的目标物的综合图像。
其中，第一二维高斯函数中频率小于第一预定值时，对应的权重为0，频率大于第二预定值时，对应的权重为1，频率在第一预定值和所述第二预定值之间时，频率越大，对应的权重越大；
第二二维高斯函数中频率小于第一预定值时，对应的权重为1，频率大于第二预定值时，对应的权重为0，频率在第一预定值和所述第二预定值之间时，频率越大，对应的权重越小。
实施例一：
第一二维高斯函数的示意图如图5a所示，第二二维高斯函数的示意图如图5b所示，两者相加的和为1，两者分别与第一图像和第二图像经过傅里叶变换得到的频域函数相乘(第一二维高斯函数与第一图像经过经过傅里叶变换得到的频域函数相乘结果如图6a所示，第二二维高斯函数与第二图像经过经过傅里叶变换得到的频域函数相乘结果如图6b所示)，然后相加，得到综合的频域函数如图 6c所示，再经过返傅里叶变换，即生成最终的目标物的综合图像如图7c所示，其中通过步骤S1和S2分别得到的第一图像和第二图像分别如图7a和图7b所示，将图7c分别与图7a和图7b作比较，可以看出，综合图像的效果优于第一图像和第二图像。
实施例二：
目标物的原图图像如图8a所示，在散射介质的前向散射系数bd为4.5的条件下，以结构相似性(SSIM)作为图像质量的评价标准。根据步骤S1得到的第一图像如图8b所示，SSIM为0.7564，经过高斯滤波后的第一图像如图8c所示，SSIM为0.89081；根据步骤S2得到的第二图像如图8d所示，SSIM为0.60799；根据步骤S3，采用如图5a和图5b的第一二维高斯函数和第二二维高斯函数分别与第一图像和第二图像进行频域加权相加，得到综合图像如图8e所示，SSIM为0.88449。
实施例三：
目标物的原图图像如图9a所示，在散射介质的前向散射系数bd为5.5的条件下，以结构相似性(SSIM)作为图像质量的评价标准。根据步骤S1得到的第一图像如图9b所示，SSIM为0.20654，经过高斯滤波后的第一图像如图9c所示，SSIM为0.43112；根据步骤S2得到的第二图像如图9d所示，SSIM为0.37954；根据步骤S3，采用如图5a和图5b的第一二维高斯函数和第二二维高斯函数分别与第一图像和第二图像进行频域加权相加，得到综合图像如图9e所示，SSIM为0.54104。
实施例二和实施例三分别是在不同散射系数下的综合重构效果，可以看出，通过结合两种成像方法，在两种方法差距悬殊时能够非常接近较好的成像效果，在两者都不好时，可以得到效果同时优于两者的图像。
在本发明中，第一图像是根据主动照明成像方法得到的图像，第二图像是根据压缩感知鬼成像方法得到的图像，其中，主动照明成像方法和压缩感知鬼成像方法在原理上有很大区别，同时在成像特性上也有差异。通过申请人的研究发现，压缩感知鬼成像方法对随机噪声不敏感，但是对整体光场涨落较为敏感，而主动照明成像方法恰恰相反；压缩感知鬼成像方法能够较好地保留图像的低频信息，主动照明成像方法则在一定的散射系数范围内能够很好地保留中高频信息。为了 抑制噪声，本发明的优选实施例中也对主动照明成像方法得到的图像进行滤波处理，如高斯滤波。
在目前的图像成像技术领域，研究主要是集中在压缩感知的重构算法和实际运用，几乎没有人对其成像结果的诸如频域特性进行分析；另外，由于鬼成像往往都是采用透射式的结构，常规的主动照明成像则为反射式结构，在现有的技术中，也没有这两者分别和常规成像方式整合的系统。值得一提的是，通过本方案独创的分层采样方式，相比传统压缩感知算法，大幅减少了采样的次数和重构的时间，因此本方案并不是对于两种现有技术的简单相加。且在此前的研究中，往往是局限于用一种成像方法取代另一种，然而本领域技术人员没有注意到以上两者对于图像的信息有不同的侧重，本发明中首创地结合两种方法的频域特性，从而得到同时优于两种方法的结果。而本发明克服了现有技术中研究的偏见，将两种成像方式集成到同一套成像装置中，使得该成像装置可以实现两种不同原理的成像方式，并通过上述特定的算法得到优于两者分别得到的图像，大大提高了强大气散射条件下的成像质量。
以上内容是结合具体的优选实施方式对本发明所作的进一步详细说明，不能认定本发明的具体实施只局限于这些说明。对于本发明所属技术领域的技术人员来说，在不脱离本发明构思的前提下，还可以做出若干等同替代或明显变型，而且性能或用途相同，都应当视为属于本发明的保护范围。

Claims (10)

  1. 一种强大气散射条件下的成像装置，其特征在于，所述成像装置用于透过散射介质对目标物进行成像，包括激光器、空间光调制器、第一透镜、第二透镜和图像传感器，其中，所述空间光调制器包括多个可翻转的微镜，所述空间光调制器设置在所述激光器射出的激光光路上，所述激光光路经所述空间光调制器的所述微镜反射后再经过所述第一透镜透射后并透过所述散射介质射在所述目标物上，然后所述激光光路经所述目标物反射并透过所述散射介质后再经过所述第二透镜透射后射在所述图像传感器上以对所述目标物进行成像。
  2. 根据权利要求1所述的成像装置，其特征在于，所述激光器采用波长为720～904nm的激光光源。
  3. 根据权利要求1或2所述的成像装置，其特征在于，所述空间光调制器包括M×N个矩阵排列的可翻转的所述微镜。
  4. 一种强大气散射条件下的成像方法，其特征在于，采用权利要求3所述的成像装置进行成像，包括以下步骤：
    S1：将一个全为1的M×N的测量矩阵输入到所述空间光调制器，在所述图像传感器上生成第一图像，其中所述测量矩阵中的1表示将所述空间光调制器中对应的所述微镜翻转至将所述激光器射出的激光光路反射到所述目标物上；
    S2：将3组包含0和1的M×N的测量矩阵输入到所述空间光调制器，通过该3组所述测量矩阵和相应的所述图像传感器接收的光强信息，综合还原生成第二图像，其中所述测量矩阵中的0表示将所述空间光调制器中对应的所述微镜翻转至不将所述激光器射出的激光光路反射到所述目标物上；
    S3：采用频域加权的方式将所述第一图像和所述第二图像进行加权相加，生成最终的所述目标物的综合图像。
  5. 根据权利要求4所述的成像方法，其特征在于，步骤S1还包括，对所述第一图像进行滤波处理，生成滤波后的第一图像，步骤S3中的所述第一图像为滤波后的所述第一图像。
  6. 根据权利要求4所述的成像方法，其特征在于，步骤S2中通过该3组所述测量矩阵和相应的所述图像传感器接收的光强信息，还原生成第二图像具体包括：采用以下计算公式：
    y＝Φx
    Figure PCTCN2017102512-appb-100001
    其中，x是一维化的图像原始信息，y是m次采样所述图像传感器接收到的反射光总强度，Φ是测量矩阵集，m是一组所述测量矩阵的矩阵数量，n＝M×N；通过上述公式可以根据Φ和y重构生成x，即还原生成所述第二图像。
  7. 根据权利要求4所述的成像方法，其特征在于，步骤S2具体包括以下步骤：
    S21：使用分辨率为(N/4)×(M/4)的第一组测量矩阵，对图像进行采样和重构，得到低分辨率的完整图像；
    S22：根据低分辨率图像的梯度排序，选取前50％的像素作为有用像素；
    S23：使用分辨率为(N/2)×(M/2)的第二组测量矩阵，对步骤S22中的有用像素进行采样和重构，得到中分辨率的局部图像；
    S24：根据中分辨率图像的梯度排序，选取前50％的像素作为有用像素；
    S25：使用分辨率为(N)×(M)的第三组测量矩阵，对步骤S24中的有用像素进行采样和重构，得到高分辨率的细节图像；
    S26：将上述图像组合，得到50％区域低分辨率，25％区域中分辨率和25％区域高分辨率的整体图像，即为第二图像。
  8. 根据权利要求4至7任一项所述的成像方法，其特征在于，步骤S3具体包括：
    S31：将所述第一图像和所述第二图像分别经过傅里叶变换得到频域函数；
    S32：采用第一二维分段函数和第二二维分段函数分别作为所述第一图像和所述第二图像的权重函数；
    S33：将所述第一二维分段函数和所述第二二维分段函数分别与所述第一图像和所述第二图像经过傅里叶变换得到的频域函数相乘，然后相加，得到综合的频域函数，再进行反傅里叶变换，即生成最终的所述目标物的综合图像；
    其中在步骤S32中：
    所述第一二维分段函数w1与频率f的关系如下：
    Figure PCTCN2017102512-appb-100002
    所述第二二维分段函数w2与频率f的关系如下：
    Figure PCTCN2017102512-appb-100003
    其中，F为最高频率。
  9. 根据权利要求4至7任一项所述的成像方法，其特征在于，步骤S3具体包括：
    S31：将所述第一图像和所述第二图像分别经过傅里叶变换得到频域函数；
    S32：采用第一二维高斯函数和第二二维高斯函数分别作为所述第一图像和所述第二图像的权重函数，其中所述第一二维高斯函数和所述第二二维高斯函数分别经过归一化，且所述第一二维高斯函数和所述第二二维高斯函数之和为1；
    S33：将所述第一二维高斯函数和所述第二二维高斯函数分别与所述第一图像和所述第二图像经过傅里叶变换得到的频域函数相乘，然后相加，得到综合的频域函数，再进行反傅里叶变换，即生成最终的所述目标物的综合图像。
  10. 根据权利要求9所述的成像方法，其特征在于，步骤S32中：
    所述第一二维高斯函数中频率小于第一预定值时，对应的权重为0，频率大于第二预定值时，对应的权重为1，频率在第一预定值和所述第二预定值之间时，频率越大，对应的权重越大；
    所述第二二维高斯函数中频率小于第一预定值时，对应的权重为1，频率大于第二预定值时，对应的权重为0，频率在第一预定值和所述第二预定值之间时，频率越大，对应的权重越小。
PCT/CN2017/102512 2017-06-29 2017-09-20 一种强大气散射条件下的成像装置及方法 WO2019000659A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710517013.X 2017-06-29
CN201710517013.XA CN107315176B (zh) 2017-06-29 2017-06-29 一种强大气散射条件下的成像装置及方法

Publications (1)

Publication Number Publication Date
WO2019000659A1 true WO2019000659A1 (zh) 2019-01-03

Family

ID=60181279

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/102512 WO2019000659A1 (zh) 2017-06-29 2017-09-20 一种强大气散射条件下的成像装置及方法

Country Status (2)

Country Link
CN (1) CN107315176B (zh)
WO (1) WO2019000659A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111652925A (zh) * 2020-06-29 2020-09-11 中国科学院合肥物质科学研究院 利用单像素成像提取目标全局特征Hu不变矩的方法
CN115220061A (zh) * 2022-07-15 2022-10-21 哈工大机器人（合肥）国际创新研究院 一种基于正交归一化的深度学习偏振鬼成像方法和系统
CN116609794A (zh) * 2023-07-21 2023-08-18 中国人民解放军国防科技大学 基于径向切比雪夫光场的单像素成像方法、装置及设备

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108594429B (zh) * 2018-04-13 2021-06-08 中国科学院光电研究院 基于波前校正的透云雾成像方法
CN109520969B (zh) * 2018-10-26 2021-03-09 中国科学院国家空间科学中心 一种基于大气介质自调制的分布式散射成像方法
CN110132901B (zh) * 2019-05-21 2020-07-31 北京理工大学 合成孔径穿散射介质成像的系统和方法
CN110865391B (zh) * 2019-11-14 2021-09-21 清华大学 用于目标增强的毫米波太赫兹多极化成像方法及成像系统
CN111352126B (zh) * 2020-03-11 2022-03-08 中国科学院国家空间科学中心 一种基于大气散射介质调制的单像素成像方法
CN113992840B (zh) * 2021-09-15 2023-06-23 中国航天科工集团第二研究院 一种基于压缩感知的大视场高分辨成像方法及装置

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1474162A (zh) * 2003-08-07 2004-02-11 中国科学技术大学 不可见像的光学成像方法及光学成像装置
EP1480441B1 (en) * 2003-05-21 2008-11-12 Esko-Graphics Imaging GmbH Method and apparatus for multi-track imaging using single-mode beams and diffraction-limited optics
CN102192787A (zh) * 2010-03-04 2011-09-21 范冰清 一种红外成像探测系统
CN102486410A (zh) * 2010-12-06 2012-06-06 中国科学院微电子研究所 光学成像装置
US20130100525A1 (en) * 2011-10-19 2013-04-25 Su Yu CHIANG Optical imaging system using structured illumination
CN103363924A (zh) * 2013-07-15 2013-10-23 中国科学院空间科学与应用研究中心 一种压缩的三维计算鬼成像系统及方法
CN105223582A (zh) * 2015-09-01 2016-01-06 西安交通大学 一种基于压缩感知的激光雷达成像装置及成像方法

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1480441B1 (en) * 2003-05-21 2008-11-12 Esko-Graphics Imaging GmbH Method and apparatus for multi-track imaging using single-mode beams and diffraction-limited optics
CN1474162A (zh) * 2003-08-07 2004-02-11 中国科学技术大学 不可见像的光学成像方法及光学成像装置
CN102192787A (zh) * 2010-03-04 2011-09-21 范冰清 一种红外成像探测系统
CN102486410A (zh) * 2010-12-06 2012-06-06 中国科学院微电子研究所 光学成像装置
US20130100525A1 (en) * 2011-10-19 2013-04-25 Su Yu CHIANG Optical imaging system using structured illumination
CN103363924A (zh) * 2013-07-15 2013-10-23 中国科学院空间科学与应用研究中心 一种压缩的三维计算鬼成像系统及方法
CN105223582A (zh) * 2015-09-01 2016-01-06 西安交通大学 一种基于压缩感知的激光雷达成像装置及成像方法

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111652925A (zh) * 2020-06-29 2020-09-11 中国科学院合肥物质科学研究院 利用单像素成像提取目标全局特征Hu不变矩的方法
CN111652925B (zh) * 2020-06-29 2023-04-07 合肥中科迪宏自动化有限公司 利用单像素成像提取目标全局特征Hu不变矩的方法
CN115220061A (zh) * 2022-07-15 2022-10-21 哈工大机器人（合肥）国际创新研究院 一种基于正交归一化的深度学习偏振鬼成像方法和系统
CN115220061B (zh) * 2022-07-15 2024-05-10 哈工大机器人（合肥）国际创新研究院 一种基于正交归一化的深度学习偏振鬼成像方法和系统
CN116609794A (zh) * 2023-07-21 2023-08-18 中国人民解放军国防科技大学 基于径向切比雪夫光场的单像素成像方法、装置及设备
CN116609794B (zh) * 2023-07-21 2023-09-26 中国人民解放军国防科技大学 基于径向切比雪夫光场的单像素成像方法、装置及设备

Also Published As

Publication number Publication date
CN107315176B (zh) 2019-04-26
CN107315176A (zh) 2017-11-03

Similar Documents

Publication Publication Date Title
WO2019000659A1 (zh) 一种强大气散射条件下的成像装置及方法
JP5334574B2 (ja) 符号化開口画像システム
US9405960B2 (en) Face hallucination using convolutional neural networks
KR101161471B1 (ko) 비선형 및/또는 공간적으로 변하는 이미지 프로세싱을 이용한 광학 이미징 시스템 및 방법
Liang Punching holes in light: recent progress in single-shot coded-aperture optical imaging
Farid Image forgery detection
US20070216798A1 (en) Post processing of iris images to increase image quality
JP2022509034A (ja) ニューラルネットワークを使用した輝点除去
JP5665775B2 (ja) 拡張被写界深度を有する正確な画像化方法および装置
US7692709B2 (en) End-to-end design of electro-optic imaging systems with adjustable optical cutoff frequency
JP2013535931A (ja) 圧縮撮像装置の画像取得時間の減少
JP5640612B2 (ja) 二重モードの拡張された被写界深度の結像システム
KR20230027012A (ko) 디지털 광학 수차 보정 및 스펙트럼 이미징을 위한 시스템 및 방법
Oktem et al. High-resolution multi-spectral imaging with diffractive lenses and learned reconstruction
EP3143583B1 (en) System and method for improved computational imaging
US10446600B2 (en) Imaging system and imaging device having a random optical filter array
WO2021099761A1 (en) Imaging apparatus
US11758297B2 (en) Systems, methods, and media for high dynamic range imaging using single-photon and conventional image sensor data
KR102235127B1 (ko) 멀티모달 영상의 적응적 융합 방법 및 그 장치
Zhang et al. Fourier single-pixel imaging based on lateral inhibition for low-contrast scenes
Le Li et al. Compressive Multi-Mission Electro-Optical Sensor System
JP2020031327A (ja) レンズレス撮像装置
CN112950507B (zh) 基于深度学习提高散射环境下单像素彩色成像性能的方法
Sun End-to-end Optics Design for Computational Cameras
EP3826283A1 (en) Imaging apparatus comprising at least a light source and a plenoptic camera

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17916416

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 17916416

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 17916416

Country of ref document: EP

Kind code of ref document: A1