A real-time scattering imaging method, device and storage medium without prior multi-target

Abstract

The present invention discloses a real-time scattering imaging method for multiple targets without a priori, comprising the following steps: S1, detecting the number of target positions based on a scale vector, calculating the autocorrelation of a speckle area centered on a derived position, and reconstructing a static target using a phase recovery method; S2, estimating a local point spread function corresponding to each area according to the reconstruction result of each target by an autocorrelation method and the speckle of the corresponding area; S3, collecting a single-frame speckle image of unknown dynamic multiple objects, and using a deconvolution method to achieve real-time reconstruction of dynamic position targets in each speckle area by relying on the estimated point spread function; S4, placing the reconstruction results of the dynamic position targets in each speckle area at the corresponding position, and splicing them into a complete large-field-of-view reconstructed image, so as to achieve real-time deconvolution imaging of multiple targets with a large field of view.

Description

Technical Field

[0001] The present invention relates to the technical field of computer vision and digital image processing, and in particular to a real-time scattering imaging method, a device and a storage medium without priori multiple targets. Background Art

[0002] Due to the uneven refractive index distribution inside the medium (such as biological tissue, clouds, etc.) or on the surface (such as frosted glass, rough walls, etc.), light waves will be strongly scattered when they are transmitted through the medium or interact with its surface. Therefore, traditional optical imaging systems can only receive speckles formed by interference between scattered light, and cannot obtain target information hidden behind the scattering medium through direct observation. Although the application fields of scattering imaging technology are very wide, including medical diagnosis, driving in complex environments, fire rescue, deep-sea exploration, astronomical observation, satellite remote sensing images, etc., how to clearly image through scattering media has always been a research difficulty and hotspot in the imaging field.

[0003] Existing scattering imaging methods through strong scattering media mainly include the following three categories: wavefront shaping method, deconvolution method based on point spread function (PSF), and correlation imaging algorithm. Among them, the correlation imaging algorithm is widely used due to its simple device, non-invasive acquisition, fast imaging speed and other characteristics. However, this method has an obvious limitation, that is, the imaging field range is limited by the memory effect range. Chinese patent document CN111257287A discloses a "large field of view scattering imaging method and device based on non-prior target positioning", which includes: collecting speckles formed by multiple objects under a large field of view through a scattering medium, and establishing a scaling vector describing the scaling relationship between two layers of speckles through a block matching algorithm in two speckle images at different image distances; obtaining the position information of multiple objects through a non-prior target position extraction algorithm based on the length and direction information of the scaling vector; taking the obtained position information as a reference, taking a rectangular block on the speckle with each object position as the center, and obtaining the autocorrelation information of each target object to be reconstructed; using a phase recovery algorithm, reconstructing the spatial domain information of the object from the autocorrelation information, placing the reconstructed object at the extracted position of each object, forming a complete large field of view multi-object reconstruction result, and realizing non-prior large field of view multi-object scattering imaging. As a non-invasive non-a priori scattering imaging method, the present invention can greatly improve the imaging field of view, and at the same time obtain the position information of the target without a priori. However, the above technical solution consumes a lot of computing resources and time, and can only image a set of fixed static targets. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention proposes a real-time scattering imaging method, device and storage medium without prior knowledge of multiple targets, which can realize real-time deconvolution imaging of multiple targets with a large field of view.

[0005] In order to solve the above technical problems, the technical solution of the present invention is:

[0006] A real-time scattering imaging method without prior knowledge of multiple targets comprises the following steps:

[0007] S1, detect the number of target positions based on the scale vector, calculate the autocorrelation of the speckle area centered on the derived position, and reconstruct the static target using the phase recovery method;

[0008] S2, estimating the local point spread function corresponding to each area according to the autocorrelation reconstruction results of each target and the speckle of the corresponding area;

[0009] S3, collecting single-frame speckle images of unknown dynamic multi-objects, and using the estimated point spread function to achieve real-time reconstruction of dynamic position targets in each speckle area using the deconvolution method;

[0010] S4. Place the reconstruction results of the dynamic position targets in each speckle area at the corresponding position and stitch them into a complete large-field-of-view reconstructed image.

[0011] Preferably, the step S1 includes

[0012] S1-1, collect speckles formed by multiple objects in a large field of view through scattering media;

[0013] S1-2, in two speckle images at different image distances, a scaling vector describing the scaling relationship between two layers of speckle is established by using a block matching algorithm;

[0014] S1-3, obtaining the position information of multiple objects through a non-a priori target position extraction algorithm based on the length and direction information of the scaling vector;

[0015] S1-4, taking a rectangular block on the speckle with each object position as the center;

[0016] S1-5. Obtain the autocorrelation information of the corresponding target object to be reconstructed from each rectangular block, reconstruct the static target using a phase recovery algorithm, and form an autocorrelation reconstruction result of each target object.

[0017] Preferably, in step S2, a deconvolution algorithm is used to estimate the local point spread function of each area according to the autocorrelation reconstruction result of each target object and the speckle in the corresponding rectangular block area.

[0018] Preferably, in step S4, the reconstructed unknown dynamic multiple objects are placed at their respective corresponding positions according to the acquired positions of the unknown dynamic multiple objects, and finally a complete real-time reconstruction result of multiple objects with a large field of view is formed, thereby realizing dynamic scattering imaging within a limited dynamic range of multiple objects with a large field of view without a priori.

[0019] Preferably, in step S1-1, a sensor is used to collect images of multiple objects under incoherent light illumination. k , k=1,2,…,n After passing through the scattering medium, the speckles I and I′ are formed, where the image distances of I and I′ are different, namely d1 and d2, respectively, and both are in the Fresnel near field relative to the scattering medium; the position corresponding to each object Ok is (s k ,t k ); the restriction conditions for multiple targets are: under the premise that the distance from the object plane to the scattering medium is d0 and the memory effect range of the scattering medium is Δθ, the size of each object is smaller than d0·Δθ, and the distance between two adjacent objects is larger than d0·Δθ, and the joint field of view formed by multiple objects is larger than the memory effect range of the medium.

[0020] Preferably, in step S1-2, the block matching algorithm establishes a plurality of spatially symmetrically distributed scaling vectors from speckles at two different image distances in accordance with the following formula:

[0021]

[0022] Among them, M x,y represents the search block centered at (x, y) on the speckle I′; N x,y represents the search block centered at (x, y) on the speckle I; Corr(.) represents a function describing the correlation between two pixel blocks; using formula (1), we get the scaling vector with the starting point at the speckle I plane and the end point at the speckle I′ plane By traversing the entire speckle I and I′ planes at a certain interval, multiple scaling vectors with symmetrical distribution in space can be obtained.

[0023] Preferably, the step S1-3 includes

[0024] 1) The area where the length of the scaling vector is less than a certain threshold is regarded as the possible location of the target object. The threshold value can be adjusted according to the actual situation. The 8-connected connected domain algorithm is used to cluster the identified possible locations to obtain different connected domains. The number of connected domains obtained by clustering is used as the number of targets to be reconstructed. Each connected domain corresponds to an object position.

[0025] 2) Select the scaling vector in the connected domain The position with the shortest length is taken as the initial positioning position of the target object in the connected domain;

[0026] 3) Near the initial positioning position of the connected domain, use the direction information of the scaling vector to make slight adjustments, and finally determine the position of the object in the connected domain to ensure that it is located at as many scaling vectors as possible. On the extension line of

[0027] 4) Repeat steps 2) and 3) on each connected domain, and finally determine the positions of multiple objects (s k ,t k ),k=1,2,…,n and quantity n.

[0028] Preferably, in step S1-5, the object O k Corresponding position (s k ,t k ), take the block according to the following formula:

[0029] I k (x,y)={I(x,y)|||(x,y)-(s k ,t k )||∞≤L / 2} (2)

[0030] Among them I k Indicates the speckle I in (s k ,t k ) as the center and L as the side length; the choice of L value depends on the resolution of I, and L value is equal to 1 / 5 of the minimum resolution of speckle I.

[0031] Preferably, in step S2, the estimation algorithm of the local point spread function is:

[0032] (1) For the rectangular area in the selected speckle image I, the speckle autocorrelation method is used to obtain the corresponding target autocorrelation information. The formula is as follows:

[0033]

[0034] in represents the autocorrelation operation, O k represents the original target image, I k is the selected speckle rectangular block, (x, y) is any coordinate point in the selected area, and the HIO-ER phase recovery algorithm is used to reconstruct the spatial information of the target based on the target autocorrelation information. To obtain more accurate results, the phase recovery algorithm is executed at least 50 times under different initial phase conditions, and the reconstructed image with the minimum mean square error of the Fourier spectrum with the speckle autocorrelation is taken as the final reconstruction result of the target;

[0035] (2) The overscattering medium imaging system can be understood as the original target forming the corresponding speckle through the optical point spread function, which can be expressed as follows:

[0036] I k =O k *S k (4)

[0037] Where * represents the convolution operation, Ok represents the original target image, S k is the point spread function, I k The Richardson-Lucy deconvolution algorithm is performed on the selected rectangular block in the speckle I and its corresponding target reconstruction result to estimate the local point spread function S corresponding to the area. k ;

[0038] (3) Repeat steps (1) and (2) for each matrix block to obtain its corresponding local estimated point spread function S k , k=1,2,…,n.

[0039] Preferably, in step S3, the formula of the deconvolution method is as follows:

[0040]

[0041] Where F represents Fourier transform, F -1 represents the inverse Fourier transform, O represents the original target image, S is the point spread function, and IIM is the speckle intensity image.

[0042] After estimating the point spread function corresponding to each matrix block of the speckle image I, the sensor is used to collect a single frame of speckle I formed by dynamic multi-objects Ok, k = 1, 2, ..., n under incoherent light illumination after passing through the scattering medium at the same image distance d1 as the speckle I. t Take the matrix block with the same center position as the speckle I, perform deconvolution processing on each speckle rectangular block using the Richardson-Lucy deconvolution algorithm, and reconstruct the single target spatial information O t , and according to the acquired object positions, the reconstructed objects are placed in their corresponding positions. By utilizing the short time-consuming characteristics of deconvolution imaging, a complete real-time reconstruction result of multiple objects with a large field of view can be finally formed.

[0043] A real-time scattering imaging device without a priori multiple targets comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the real-time scattering imaging method without a priori multiple targets is realized.

[0044] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned real-time scattering imaging method without priori multiple targets.

[0045] The present invention has the following characteristics and beneficial effects:

[0046] By adopting the above technical scheme, the field of view of the imaging without prior scattering is expanded, the limitation of the memory effect range is broken, and the dynamic multi-object real-time scattering imaging within a limited moving range is realized. The idea of ​​the present invention is to first estimate the point spread function using a stationary training target: the stationary multi-objects under a large field of view form speckles after passing through the scattering medium and are collected by the sensor. In the two speckle images collected at different image distances, a scaling vector describing the scaling relationship between the two layers of speckles is established through a block matching algorithm. Then, by using the coupling relationship between the scaling vector and the position of the target object, the position information of the multi-object is obtained through a non-priori target position extraction algorithm based on the length and direction information of the scaling vector. Then, based on the obtained object position information, a rectangular block is taken on the speckle with the position of each object as the center, and its autocorrelation information is calculated, and then the phase recovery algorithm is used to reconstruct the spatial information of each object. According to the obtained object position, deconvolution processing is performed on the spatial information of each object and the speckle rectangular block at the corresponding position to estimate the point spread function of the local position. After estimating the point spread function, it can be used to perform real-time deconvolution imaging of unknown dynamic multiple objects to be observed: using the specific point spread function of each rectangular area, the speckle image of the object to be observed in the area is deconvolved, the spatial information of the object is reconstructed, and according to the acquired object position, the reconstructed object is placed in the corresponding position, and finally a complete large-field-of-view multi-object dynamic reconstruction result is formed, realizing real-time scattering imaging of multiple objects with a large field of view without prior knowledge. As a non-invasive scattering imaging method without prior knowledge, the present invention can achieve real-time scattering imaging of dynamic multiple objects within a limited moving range while obtaining a larger imaging field of view. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0048] Figure 1 It is a principle structure block diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0049] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0050] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and the like are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first", "second", and the like may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0051] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood by specific circumstances.

[0052] The present invention provides a real-time scattering imaging method without prior multi-targets, such as Figure 1 As shown, the following steps are included;

[0053] S1, detect the number of target positions based on the scale vector, calculate the autocorrelation of the speckle area centered on the derived position, and reconstruct the static target using the phase recovery method;

[0054] S2, estimating the local point spread function corresponding to each area according to the autocorrelation reconstruction results of each target and the speckle of the corresponding area;

[0055] S3, collecting single-frame speckle images of unknown dynamic multi-objects, and using the estimated point spread function to achieve real-time reconstruction of dynamic position targets in each speckle area using the deconvolution method;

[0056] S4. Place the reconstruction results of the dynamic position targets in each speckle area at the corresponding position and stitch them into a complete large-field-of-view reconstructed image.

[0057] In the above technical solution, the field of view of imaging without a priori scattering is expanded and the limitation of the memory effect range is broken through, while the real-time scattering imaging of dynamic multiple objects within a limited moving range is realized.

[0058] Specifically, step S1 includes

[0059] S1-1, collect speckles formed by multiple objects in a large field of view through scattering media;

[0060] S1-2, in two speckle images at different image distances, a scaling vector describing the scaling relationship between two layers of speckle is established by using a block matching algorithm;

[0061] S1-3, obtaining the position information of multiple objects through a non-a priori target position extraction algorithm based on the length and direction information of the scaling vector;

[0062] S1-4, taking a rectangular block on the speckle with each object position as the center;

[0063] S1-5. Obtain the autocorrelation information of the corresponding target object to be reconstructed from each rectangular block, reconstruct the static target using a phase recovery algorithm, and form an autocorrelation reconstruction result of each target object.

[0064] Furthermore, in step S2, a deconvolution algorithm is used to estimate the local point spread function of each area according to the autocorrelation reconstruction result of each target object and the speckle in the corresponding rectangular block area.

[0065] Furthermore, in step S4, the reconstructed unknown dynamic multiple objects are placed at their respective corresponding positions according to the acquired positions of the unknown dynamic multiple objects, and finally a complete real-time reconstruction result of multiple objects with a large field of view is formed, thereby realizing dynamic scattering imaging within a limited dynamic range of multiple objects with a large field of view without a priori.

[0066] Specifically, in step S1-1, a sensor is used to collect images of multiple objects under incoherent light illumination. k , k=1,2,…,n After passing through the scattering medium, the speckles I and I′ are formed, where the image distances of I and I′ are different, namely d1 and d2, respectively, and both are in the Fresnel near field relative to the scattering medium; the position corresponding to each object Ok is (s k ,t k ); the restriction conditions for multiple targets are: under the premise that the distance from the object plane to the scattering medium is d0 and the memory effect range of the scattering medium is Δθ, the size of each object is smaller than d0·Δθ, and the distance between two adjacent objects is larger than d0·Δθ, and the joint field of view formed by multiple objects is larger than the memory effect range of the medium.

[0067] In step S1-2, the block matching algorithm establishes multiple spatially symmetrically distributed scaling vectors from speckles at two different image distances according to the following formula:

[0068]

[0069] Among them, M x,yrepresents the search block centered at (x, y) on the speckle I′; N x,y represents the search block centered at (x, y) on the speckle I; Corr(.) represents a function describing the correlation between two pixel blocks; using formula (1), we get the scaling vector with the starting point at the speckle I plane and the end point at the speckle I′ plane By traversing the entire speckle I and I′ planes at a certain interval, multiple scaling vectors with symmetrical distribution in space can be obtained.

[0070] The step S1-3 includes

[0071] 1) The area where the length of the scaling vector is less than a certain threshold is regarded as the possible location of the target object. The threshold value can be adjusted according to the actual situation. The 8-connected connected domain algorithm is used to cluster the identified possible locations to obtain different connected domains. The number of connected domains obtained by clustering is used as the number of targets to be reconstructed. Each connected domain corresponds to an object position.

[0072] 2) Select the scaling vector in the connected domain The position with the shortest length is taken as the initial positioning position of the target object in the connected domain;

[0073] 3) Near the initial positioning position of the connected domain, use the direction information of the scaling vector to make slight adjustments, and finally determine the position of the object in the connected domain to ensure that it is located at as many scaling vectors as possible. On the extension line of

[0074] 4) Repeat steps 2) and 3) on each connected domain, and finally determine the positions of multiple objects (s k ,t k ),k=1,2,…,n and quantity n.

[0075] It should be noted that the 8-connected connected domain algorithm is a conventional algorithm, so it is not described in detail in this application.

[0076] In step S1-5, object O k Corresponding position (s k ,t k ), take the block according to the following formula:

[0077] I k (x,y)={I(x,y)|||(x,y)-(s k ,t k )||∞≤L / 2} (2)

[0078] Among them I k Indicates the speckle I in (s k ,tk ) as the center and L as the side length; the choice of L value depends on the resolution of I, and L value is equal to 1 / 5 of the minimum resolution of speckle I.

[0079] Furthermore, in step S2, due to the envelope characteristics of the speckle, the influence of the selection of a certain area on the weight of each imaging target in the speckle I is compounded by the following formula:

[0080] C(x-u1,y-v1)>>C(xu k ,yv k ),k=2,...,n,

[0081] Where (x, y) is the coordinate point of any point in the selected area. The influence of other targets except the corresponding target on a single speckle area can be ignored, and a non-invasive local point spread function estimation algorithm is adopted.

[0082] Specifically, the estimation algorithm of the local point spread function is:

[0083] (1) For the rectangular area in the selected speckle image I, the speckle autocorrelation method is used to obtain the corresponding target autocorrelation information. The formula is as follows:

[0084]

[0085] in represents the autocorrelation operation, O k represents the original target image, I k is the selected speckle rectangular block, (x, y) is any coordinate point in the selected area, and the HIO-ER phase recovery algorithm is used to reconstruct the spatial information of the target based on the target autocorrelation information. To obtain more accurate results, the phase recovery algorithm is executed at least 50 times under different initial phase conditions, and the reconstructed image with the minimum mean square error of the Fourier spectrum with the speckle autocorrelation is taken as the final reconstruction result of the target;

[0086] (2) The overscattering medium imaging system can be understood as the original target forming the corresponding speckle through the optical point spread function, which can be expressed as follows:

[0087] I k =O k *S k (4)

[0088] Where * represents the convolution operation, O k represents the original target image, S k is the point spread function, I kThe Richardson-Lucy deconvolution algorithm is performed on the selected rectangular block in the speckle I and its corresponding target reconstruction result to estimate the local point spread function S corresponding to the area. k ;

[0089] (3) Repeat steps (1) and (2) for each matrix block to obtain its corresponding local estimated point spread function S k , k=1,2,…,n.

[0090] Furthermore, in step S3, for the speckle image after the scattering medium, after the point spread function of the optical imaging system is known, the deconvolution method can be used to reconstruct the original target. The formula of the deconvolution method is as follows:

[0091]

[0092] Where F represents Fourier transform, F -1 represents the inverse Fourier transform, O represents the original target image, S is the point spread function, and IIM is the speckle intensity image.

[0093] In the above technical solution, after estimating the point spread function corresponding to each matrix block of the speckle image I, the sensor is used to collect the single-frame speckle I formed by the dynamic multi-object Ok, k = 1, 2, ..., n under incoherent light illumination after passing through the scattering medium at the same image distance d1 as the speckle I. t Take the matrix block with the same center position as the speckle I, perform deconvolution processing on each speckle rectangular block using the Richardson-Lucy deconvolution algorithm, and reconstruct the single target spatial information O t , and according to the acquired object positions, the reconstructed objects are placed in their respective corresponding positions, and the short time-consuming feature of deconvolution imaging is utilized to finally form a complete real-time reconstruction result of multiple objects with a large field of view. This is a non-invasive imaging method without prior scattering. The present invention can achieve real-time scattering imaging of dynamic multiple objects within a limited moving range while obtaining a larger imaging field of view.

[0094] The present invention also discloses a real-time scattering imaging device without a priori multiple targets, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the real-time scattering imaging method without a priori multiple targets is implemented.

[0095] The present invention also discloses a computer-readable storage medium on which a computer program is stored, characterized in that when the computer program is executed by a processor, the real-time scattering imaging method without priori multiple targets is implemented.

[0096] It should be noted that the specific preferred methods in the above embodiments (such as the Richardson-Lucy deconvolution algorithm, the phase recovery algorithm, etc.) are only listed for illustration, and the scope of the present invention is not limited to these listed methods.

[0097] The background section of the present invention may contain background information about the problem or environment of the present invention, but does not necessarily describe the prior art. Therefore, the content included in the background section is not an admission by the applicant that the prior art is available.

[0098] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions and variations of these embodiments including components are made without departing from the principles and spirit of the present invention, and still fall within the scope of protection of the present invention.

Claims

1. A real-time scattering imaging method without prior knowledge of multiple targets, characterized in that: The steps include: S1, detect the number of target positions based on the scale vector, calculate the autocorrelation of the speckle area centered on the derived position, and reconstruct the static target using the phase recovery method; S1-1, collect speckles formed by multiple objects in a large field of view through scattering media; S1-2, in two speckle images at different image distances, a scaling vector describing the scaling relationship between two layers of speckle is established by using a block matching algorithm; S1-3, obtaining the position information of multiple objects through a non-a priori target position extraction algorithm based on the length and direction information of the scaling vector; 1) The area where the length of the scaling vector is less than a certain threshold is regarded as the possible location of the target object. The threshold value can be adjusted according to the actual situation. The 8-connected connected domain algorithm is used to cluster the identified possible locations to obtain different connected domains. The number of connected domains obtained by clustering is used as the number of targets to be reconstructed. Each connected domain corresponds to an object position. 2) Select the scaling vector in the connected domain The position with the shortest length is taken as the initial positioning position of the target object in the connected domain; 3) Near the initial positioning position of the connected domain, use the direction information of the scaling vector to make slight adjustments, and finally determine the position of the object in the connected domain to ensure that it is located at as many scaling vectors as possible. On the extension line of 4) Repeat steps 2) and 3) on each connected domain, and finally determine the positions of multiple objects (s k ,t k ), k = 1, 2, ..., n and the number n; S1-4, taking a rectangular block on the speckle with each object position as the center; S1-5, obtaining the autocorrelation information of the corresponding target object to be reconstructed from each rectangular block, reconstructing the static target using a phase recovery algorithm, and forming an autocorrelation reconstruction result of each target object; S2, estimating the local point spread function corresponding to each area according to the autocorrelation reconstruction results of each target and the speckle of the corresponding area; The estimation algorithm of the local point spread function is: (1) For the rectangular area in the selected speckle image I, the speckle autocorrelation method is used to obtain the corresponding target autocorrelation information. The formula is as follows: in represents the autocorrelation operation, O k represents the original target image, I k is the selected speckle rectangular block, (x, y) is any coordinate point in the selected area, and the HIO-ER phase recovery algorithm is used to reconstruct the spatial information of the target based on the target autocorrelation information. To obtain more accurate results, the phase recovery algorithm is executed at least 50 times under different initial phase conditions, and the reconstructed image with the minimum mean square error of the Fourier spectrum with the speckle autocorrelation is taken as the final reconstruction result of the target; (2) The overscattering medium imaging system can be understood as the original target forming the corresponding speckle through the optical point spread function, which is expressed as follows: I k =O k *S k (2) Where * represents the convolution operation, O k represents the original target image, S k is the point spread function, I k The Richardson-Lucy deconvolution algorithm is performed on the selected rectangular block in the speckle I and its corresponding target reconstruction result to estimate the local point spread function S corresponding to the area. k ; (3) Repeat steps (1) and (2) for each matrix block to obtain its corresponding local estimated point spread function S k , k=1,2,…,n; S3, collecting single-frame speckle images of unknown dynamic multi-objects, and using the estimated point spread function to achieve real-time reconstruction of dynamic position targets in each speckle area using the deconvolution method; S4. Place the reconstruction results of the dynamic position targets in each speckle area at the corresponding position and stitch them into a complete large-field-of-view reconstructed image.

2. The real-time scattering imaging method without prior knowledge of multiple targets according to claim 1, characterized in that: In step S2, the local point spread function corresponding to each area is estimated by using a deconvolution algorithm according to the autocorrelation reconstruction result of each target object and the speckle in the corresponding rectangular block area.

3. The real-time scattering imaging method without prior knowledge of multiple targets according to claim 2, characterized in that: In step S4, the reconstructed unknown dynamic multiple objects are placed at their respective corresponding positions according to the acquired positions of the unknown dynamic multiple objects, and finally a complete real-time reconstruction result of the large field of view multiple objects is formed, thereby realizing dynamic scattering imaging within a limited dynamic range of the large field of view multiple objects without a priori.

4. The real-time scattering imaging method without prior knowledge of multiple targets according to claim 1, characterized in that: In step S1-1, a sensor is used to collect images of multiple objects under incoherent light illumination. k , k=1,2,…,n After passing through the scattering medium, the speckles I and I′ are formed, where the image distances of I and I′ are different, namely d1 and d2, respectively, and both are in the Fresnel near field relative to the scattering medium; the position corresponding to each object Ok is (s k ,t k ); the restriction conditions for multiple targets are: under the premise that the distance from the object plane to the scattering medium is d0 and the memory effect range of the scattering medium is Δθ, the size of each object is smaller than d0·Δθ, and the distance between two adjacent objects is larger than d0·Δθ, and the joint field of view formed by multiple objects is larger than the memory effect range of the medium.

5. The real-time scattering imaging method without prior knowledge of multiple targets according to claim 4, characterized in that: In step S1-2, the block matching algorithm establishes multiple spatially symmetrically distributed scaling vectors from speckles at two different image distances according to the following formula: Among them, M x,y represents the search block centered at (x, y) on the speckle I′; N x,y represents the search block centered at (x, y) on the speckle I; Corr(.) represents a function describing the correlation between two pixel blocks; using formula (3), we get the scaling vector with the starting point at the speckle I plane and the end point at the speckle I′ plane By traversing the entire speckle I and I′ planes at a certain interval, multiple scaling vectors with symmetrical distribution in space can be obtained.

6. The real-time scattering imaging method without prior knowledge of multiple targets according to claim 5, characterized in that: In step S1-5, object O k Corresponding position (s k ,t k ), take the block according to the following formula: I k (x,y)={I(x,y)|||(x,y)-(s k ,t k )||∞≤L / 2} (4) Among them I k Indicates the speckle I in (s k ,t k ) as the center and L as the side length; the choice of L value depends on the resolution of I, and L value is equal to 1 / 5 of the minimum resolution of speckle I.

7. The real-time scattering imaging method without prior knowledge of multiple targets according to claim 6, characterized in that: In step S3, the formula of the deconvolution method is as follows: Where F represents Fourier transform, F -1 represents the inverse Fourier transform, O represents the original target image, S is the point spread function, and IIM is the speckle intensity image.

8. A real-time scatter imaging device without prior knowledge of multiple targets, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is executed by a processor, the real-time scattering imaging method without a priori multiple targets described in any one of claims 1 to 7 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the real-time scattering imaging method without a priori multiple targets as described in any one of claims 1 to 7 is implemented.

Images