Extremely-weak light super-dynamic scattering imaging system and method

Abstract

The invention relates to an extremely-weak light super-dynamic scattering imaging system and method, and proposes target information reconstruction based on speckle high-order statistical correlation for the extremely-weak light high-dynamic scattering imaging requirement of a complex scene, thereby breaking through the limitation that a traditional area-array camera is difficult to overcome strong scattering for effective imaging. A scattered field point spread function is calculated through a camera speckle image and a reconstructed image, and target structure reconstruction is performed by directly utilizing deconvolution, so that the real-time performance and reliability of the system are improved.

Description

Technical Field

[0001] The present invention relates to the field of scattering imaging, and in particular to an extremely weak light ultra-dynamic scattering imaging system and method. Background Art

[0002] The current scattering imaging technology can be divided into ballistic light imaging and scattered light imaging based on the scattered light field information. Ballistic light imaging uses a very small number of non-scattered ballistic photons in weak scattering scenes to reconstruct the target structure information, and usually uses a gating method to filter the scattered light. It can be specifically divided into time gating (Time Gating) and space gating (Spatial Gating) technology; scattered light imaging is suitable for strong scattering scenes, and uses the high-order statistical characteristics of the scattered light field and light field measurement technology to restore the target light field information in the time and space domain, specifically including wavefront shaping (Wavefront Shaping), adaptive optics (Adaptive Optics), diffuse optical tomography (Diffuse Optical Tomography), optical transmission matrix (Optical Transmission Matrix) and optical memory effect (Optical Memory Effect) imaging and other series of technologies.

[0003] In extreme environments such as clouds, haze, and dust, the scattering field is usually a strong scattering medium. It is difficult for the gated screening method to distinguish between stray light and ballistic light, and it is necessary to use the scattered light field for imaging. Traditional detection imaging technology uses cameras and other equipment to collect light field information of the target scene. In extreme environments, it is difficult to effectively obtain important information about the target and scene, and there are problems such as unclear vision, inaccurate vision, and inability to see. Summary of the invention

[0004] The present invention proposes an extremely weak light ultra-dynamic scattering imaging system and method, which specifically includes two parts: target information reconstruction and real-time imaging optimization, breaking through the limitation that traditional array cameras are difficult to overcome strong scattering for effective imaging.

[0005] 1. Aiming at the strong scattering imaging requirements in extreme environments, an optical imaging system is built to obtain speckle images, and target information is reconstructed based on the high-order statistical correlation of speckles. The specific steps include:

[0006] (1) Construct an optical imaging system for collecting target light field information. The system consists of a light source module, an incoherent light generation module, a scattering module, and a speckle collection module. The light source module provides illumination for the entire optical path system; the incoherent light generation module removes the coherence of the laser light source and generates collimated parallel light; the scattering module scatters and modulates the light passing through the target object; the speckle collection module obtains the target scattered light field information and collects the speckle image.

[0007] (2) Calculate the Fourier domain amplitude information based on the detected speckle. In view of the uneven illumination of the actual imaging speckle image, the original speckle low-pass filter is optimized and smoothed by the Gaussian kernel, and then the autocorrelation image R is calculated.

[0008] (3) Using the iterative phase recovery algorithm to solve the spatial domain information. Using a random initial phase and taking the power spectrum as the amplitude information, the corrected image frequency domain information is obtained to solve the spatial domain image information.

[0009] (4) Reconstruction constraints are imposed based on the real and non-negative properties of the target pixel information in the spatial domain.

[0010] (5) Iterative reconstruction using the hybrid input-output (HIO) algorithm.

[0011] (6) Use the Error Reduction (ER) algorithm to optimize the target information.

[0012] 2. In real application environments, extremely high requirements are placed on the real-time performance of algorithms, so the corresponding algorithms are optimized to meet the real-time and reliability requirements. The specific steps include:

[0013] (1) According to the relevant scattering mechanism of strong scattering media, the propagation process from the target light field to the imaging light field is constructed.

[0014] (2) The target structure information is calculated based on the first frame scene data, and the camera speckle image I and the reconstructed image O are used to calculate the target structure information. r , the point spread function of the scattered field is calculated.

[0015] (3) Directly use deconvolution to reconstruct the target structure, and achieve target reconstruction in a single step to improve the real-time response capability of the entire imaging. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 The optical system 10 mainly includes: a light source module 101 , an incoherent light generating module 102 , a target image 103 , a scattering module 104 , and a speckle collecting module 105 . DETAILED DESCRIPTION

[0017] The implementation of the method is further described in detail below with reference to the accompanying drawings:

[0018] An extremely weak light ultra-dynamic scattering imaging system and method mainly include target information reconstruction and real-time imaging optimization. The specific implementation method is divided into the following two parts:

[0019] 1. Refer to the attached drawings in the specification Figure 1Build an optical imaging system to capture speckle images and reconstruct target information based on the high-order statistical correlation of speckles. The specific steps are as follows:

[0020] (1) Build Figure 1 The optical system 10 shown in FIG. 1 includes: a light source module 101; an incoherent light generation module 102 composed of a rotating ground glass 1021, a beam expander 1022, and an aperture stop 1023; a target image 103; a scattering module 104 composed of a ground glass 1042 and an aperture stop 1043; and a speckle collection module 105.

[0021] In practical applications, the light source 101 generates a 655nm laser; it enters the incoherent generation module 102, removes its coherence by rotating the frosted glass (220 sand) 1021, and is collimated by a beam expander (focal length 50mm) 1022, while an aperture diaphragm (diameter of about 5mm) 1023 is used to limit the illumination area on the target image 103; then, the target light field enters the scattering module 104 after propagating through the free space, and after being scattered by the frosted glass (220 sand) 1042, an aperture diaphragm (diameter of about 5mm) 1043 is used to limit the speckle size; finally, a CCD camera (resolution 4000*3000) is used as a speckle acquisition module 105 to collect the speckle.

[0022] (2) Calculate the Fourier domain amplitude information based on the detected speckle. In view of the uneven illumination of the actual imaging speckle image, the original speckle low-pass filter is optimized, and a Gaussian kernel of 0.5 to 2 pixels is used for smoothing, and then the autocorrelation image R is calculated. The specific method is to obtain the frequency domain amplitude in the image Fourier domain, where FT is the Fourier transform, FT -1 is the inverse Fourier transform. After that, the power spectrum of R is calculated, and W(x, y) is the window function.

[0023] R(x,y)=I(x,y)★I(x,y)=FT -1 {|FT{I(x,y)}| 2} (1)

[0024] S meas (k x , k y )=|FT{W(x, y)R(x, y)}| (2)

[0025] (3) Use the iterative phase recovery algorithm to solve the spatial domain information. Use a random initial phase and use the power spectrum as the amplitude information to obtain the corrected image frequency domain information and solve the spatial domain image information. The k-th result is as follows:

[0026] G k (k x , k y)=FTg k (x, y) (3)

[0027] θ k (kx,k y ) = ar g G k (k x , k y ) (4)

[0028]

[0029] g′ k (x, y) = FT -1 G′ k (k x , k y ) (6)

[0030] (4) In the k+1 iteration, the image is restored using the reconstruction constraint (real, non-negative), which includes two steps:

[0031] (5) First, the HIO algorithm is used for iteration, and its formula is:

[0032]

[0033] (6) Then use the ER algorithm for optimization, and its calculation formula is:

[0034]

[0035] 2. Optimize the target reconstruction algorithm to meet the real-time and reliability requirements. The specific steps include:

[0036] (1) For scattered field modeling, it is equivalent to using the translation-invariant point spread function to perform a convolution operation on the original target. Where I(θ) is the speckle intensity received by the detector, O(r) is the actual image of the object, S(r) represents the system translation-invariant point spread function, r is the detection surface coordinate, and θ is the incident angle. Then, according to the convolution calculation theorem, the convolution in the spatial domain is equal to the product in the frequency domain. Where g(x, y) is the speckle function, f(x, y) is the target function, h(x, y) is the point spread function, and G, F, and H are their corresponding Fourier domain (frequency domain) functions.

[0037]

[0038]

[0039] G(f x , f y )=F(f x , f y ) (11)

[0040] H(f x , f y )g(x,y)=FT -1 (G(f x , f y )) (12)

[0041] (2) The target structure information is obtained from the first frame scene data. Using the convolution theorem, the camera speckle image I and the reconstructed image O can be used to obtain the target structure information. r , the scattered field point spread function S is calculated.

[0042] S=FT -1 (FT(I) / FT(O r )) (13)

[0043] (3) Directly use deconvolution to reconstruct the target structure. Deconv is deconvolution, and Wiener deconvolution is used.

[0044] O r =deconv(I,S) (14).

Claims

1. A system and method for ultra-weak light ultra-dynamic scattering imaging, characterized in that: The system consists of a light source module, an incoherent light generation module, a scattering module, and a speckle acquisition module. It reconstructs target information based on the high-order statistical correlation of speckles, and optimizes the real-time and reliability of the target reconstruction algorithm. Its characteristics include the following steps: (1) Build a target optical imaging system to collect speckle images and reconstruct target information based on the high-order statistical correlation of speckles. The specific process is as follows: S1: Build a target optical imaging system, which consists of a light source module, an incoherent light generation module, a scattering module, and a speckle acquisition module; S2: The original speckle pattern is optimized by low-pass filtering, smoothed by Gaussian kernel, and then its autocorrelation image R is calculated; specifically, the Fourier domain amplitude information is calculated based on the detected speckle image, and then the power spectrum of R is calculated; FT stands for Fourier transform, FT -1 represents inverse Fourier transform, W(x,y) is the window function; R(x,y)=I(x,y)*I(x,y)=FT -1 {|FT{I(x,y)}| 2 } (1) S meas (k x ,k y )=|FT{W(x,y)R(x,y)}| (2) S3: Using iterative phase recovery algorithm to solve spatial domain information; using random initial phase, taking power spectrum as amplitude information, and obtaining corrected image frequency domain information, thereby solving spatial domain image information; G k (k x ,k y )=FTg k (x,y) (3) θ k (k x ,k y )=argG k (k x ,k y ) (4) g′ k (x,y)=fT -1 G′ k (k x ,k y ) (6) S4: Reconstruction constraints are performed based on the real and non-negative characteristics of the target pixel information in the spatial domain; S5: Iterative reconstruction using the Hybrid Input-Output (HIO) algorithm; S6: Optimization using the Error Reduction (ER) algorithm; (2) Optimize the target reconstruction algorithm to meet the real-time and reliability requirements; specifically, the following steps are included: S1: According to the relevant scattering mechanism of strong scattering media, the propagation process from the target light field to the imaging light field is constructed: G(f x ,f y )=F(f x ,f y ) (11) H(f x ,f y )g(x,y)=FT -1 (G(f x ,f y )) (12) Where I(θ) is the speckle intensity received by the detector, O(r) is the actual image of the object, S(r) represents the system translation-invariant point spread function, r is the detection surface coordinate, θ is the incident angle, g(x, y) is the speckle function, f(x, y) is the target function, h(x, y) is the point spread function, and G, F, and H are their corresponding Fourier domain (frequency domain) functions respectively; S2: Calculate the target structure information based on the first frame scene data, and reconstruct the image O according to the camera speckle image I r , the scattered field point spread function S is calculated; S=FT -1 (FT(I) / FT(O r )) (13) S3: Finally, deconvolution is used to reconstruct the target structure: O r =deconv(I,S) (14) Among them, deconv is deconvolution, which uses Wiener deconvolution.

Images