CN119418131A - Method, system, medium and device for target image classification based on complex background - Google Patents
Method, system, medium and device for target image classification based on complex background Download PDFInfo
- Publication number
- CN119418131A CN119418131A CN202411742333.1A CN202411742333A CN119418131A CN 119418131 A CN119418131 A CN 119418131A CN 202411742333 A CN202411742333 A CN 202411742333A CN 119418131 A CN119418131 A CN 119418131A
- Authority
- CN
- China
- Prior art keywords
- target image
- background
- space
- frequency domain
- spatial
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The embodiment of the invention discloses a target image classification method based on a complex background, which comprises the steps of obtaining continuous and stable monochromatic p-polarized light, collecting a target image, loading the target image to an amplitude modulation array, modulating the monochromatic p-polarized light by using the amplitude modulation array after loading the target image to obtain a space time domain light signal containing the target image of the complex background, converting the space time domain light signal containing the target image of the complex background into a space frequency domain signal which is a space frequency distribution diagram containing space information of the target image, removing background information of the space frequency domain signal to obtain a space frequency domain signal after removing the background, converting the space frequency domain signal after removing the background into a space time domain light signal after removing the background, wherein the space time domain light signal after removing the background is a space intensity distribution diagram containing space information of the target image, and classifying the target image corresponding to the space time domain light signal after removing the background.
Description
Technical Field
The invention relates to the technical field of image processing and recognition, in particular to a target image classification method, a system, a medium and equipment based on complex background.
Background
Traditional von neumann architectures have long been dominant in the computing field, however, limitations of the architecture are emerging as technology continues to advance and demand grows. Due to the design principle of computational separation, von neumann architecture faces the dual challenges of power consumption walls and memory walls, which limits the improvement of computational performance and slows down the acceleration of moore's law. To address these challenges, researchers are continually exploring new computing technologies and architectures in an effort to break through the limits of traditional microelectronic computing.
The diffraction optical neural network (DIFFRACTIVE OPTICAL NEURAL NETWORKS, DONN) is used as an emerging computing technology, and has wide application prospect in the fields of high-performance computing and artificial intelligence by virtue of the advantages of ultra-high speed, large bandwidth, multi-dimension and the like. DONN fully fuses advanced technologies such as high-speed optical communication, optical interconnection, optical integration, silicon-based photoelectron and the like, and provides a very competitive solution for breaking through the bottleneck of traditional microelectronic calculation in the post-molar age.
However, despite the advantages of DONN, some challenges remain in practical use. At present, when object identification is performed by DONN, simple objects which are processed in advance are often identified. When faced with targets in complex background situations, DONN often cannot be accurately identified, which greatly reduces its practicality in real-world scenarios. Therefore, how to improve DONN the target recognition capability in a complex background becomes a current urgent problem to be solved.
Disclosure of Invention
Based on this, it is necessary to address the above-described problems, and a target image classification method based on a complex background is proposed.
A method for classifying a target image based on a complex background, the method comprising the steps of:
obtaining continuous and stable monochromatic p polarized light;
Collecting a target image, loading the target image to an amplitude modulation array, and modulating the monochromatic p-polarized light by using the amplitude modulation array loaded with the target image to obtain a spatial time domain optical signal of the target image containing a complex background;
converting the spatial time domain optical signal of the target image containing the complex background into a spatial frequency domain signal, wherein the spatial frequency domain signal is a spatial frequency distribution diagram containing spatial information of the target image;
Removing background information of the space frequency domain signal, and obtaining a space frequency domain signal after removing the background;
the space frequency domain signal after removing the background is converted into a space time domain optical signal after removing the background, the space time domain optical signal after removing the background is a space intensity distribution diagram containing space information of a target graph, and the target image corresponding to the space time domain optical signal after removing the background is classified.
In the above scheme, the obtaining continuous and stable monochromatic p polarized light specifically includes:
the acquisition laser emits continuous and stable monochromatic linear polarized laser beams;
the monochromatic linear polarized laser beams pass through a beam expanding lens pair and are adjusted to obtain parallel beams after beam expansion;
carrying out polarization modulation on the parallel light beams after beam expansion according to a half-wave plate;
and controlling the angle of the half wave plate to obtain continuous and stable monochromatic p polarized light.
In the scheme, the target image is acquired and loaded to an amplitude modulation array, and the continuous stable monochromatic p-polarized light is subjected to polarization filtering through a polarization beam splitter PBS.
In the above scheme, the method for obtaining the spatial time domain optical signal of the target image comprising the complex background by modulating the monochromatic p-polarized light by using the amplitude modulation array after loading the target image further comprises the step of distributing the light field intensity of the reflected light according to the gray value of the loaded gray image to obtain the spatial time domain optical signal of the target image comprising the complex background.
In the above solution, the converting the spatial time domain optical signal of the target image including the complex background into the spatial frequency domain signal specifically includes:
The method comprises the steps that an amplitude modulation array of a reflective amplitude type Spatial Light Modulator (SLM) receives monochromatic p-polarized light reflected by a first polarization splitting Prism (PBS), and the monochromatic p-polarized light is reflected into monochromatic s-polarized light after amplitude modulation;
Reflecting the single-color s polarized light to the direction perpendicular to the incident light path according to the first polarization splitting prism PBS;
The reflected light beam passes through a first polarization beam splitting prism PBS from an amplitude modulation array, and the focal length f of the lens is determined according to the distance of the light beam in the propagation path when the light beam reaches the Fourier lens;
When the spatial time domain optical signal of the target image containing the complex background passes through the Fourier lens, the Fourier lens converts the spatial time domain optical signal into a spatial frequency domain to acquire a spatial frequency domain signal.
In the above solution, the converting the spatial time domain optical signal of the target image including the complex background into the spatial frequency domain signal specifically includes:
performing two-dimensional Fourier transform on the modulated s polarized light, and converting the spatial information of the target image into frequency information;
From the spatial frequency distribution map formed on the back focal plane of the Fourier lens, a spatial frequency domain signal of a target image containing a complex background is determined by the signals after Fourier transformation.
In the above solution, the filtering and nonlinear activation are performed on the spatial frequency domain signal to obtain a processed spatial frequency domain signal, which specifically includes:
filtering the spatial frequency domain signal according to a first diffractive optical neural network DONN;
And carrying out nonlinear activation on the spatial frequency domain signal after the filtering processing according to the transmission phase type Spatial Light Modulator (SLM) to acquire the spatial frequency domain signal after the processing.
The application also provides a target image classification system based on the complex background, which comprises a polarized light acquisition unit, a time domain signal acquisition unit, a signal processing unit and a classification unit;
the polarized light acquisition unit is used for acquiring continuous and stable monochromatic p polarized light;
The time domain signal acquisition unit is used for acquiring a target image, loading the target image to an amplitude modulation array, and modulating the monochromatic p-polarized light by using the amplitude modulation array loaded with the target image to acquire a spatial time domain optical signal containing the target image with a complex background;
the signal processing unit is used for converting the space time domain optical signal of the target image containing the complex background into a space frequency domain signal, wherein the space frequency domain signal is a space frequency distribution diagram containing space information of the target image, removing background information of the space frequency domain signal to obtain a space frequency domain signal after background removal, converting the space frequency domain signal after background removal into a space time domain optical signal after background removal, wherein the space time domain optical signal after background removal is a space intensity distribution diagram containing space information of the target image, and the classifying unit is used for classifying the target image corresponding to the space time domain optical signal after background removal.
The application also proposes a readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of:
obtaining continuous and stable monochromatic p polarized light;
Collecting a target image, loading the target image to an amplitude modulation array, and modulating the monochromatic p-polarized light by using the amplitude modulation array loaded with the target image to obtain a spatial time domain optical signal of the target image containing a complex background;
converting the spatial time domain optical signal of the target image containing the complex background into a spatial frequency domain signal, wherein the spatial frequency domain signal is a spatial frequency distribution diagram containing spatial information of the target image;
Removing background information of the space frequency domain signal, and obtaining a space frequency domain signal after removing the background;
the space frequency domain signal after removing the background is converted into a space time domain optical signal after removing the background, the space time domain optical signal after removing the background is a space intensity distribution diagram containing space information of a target graph, and the target image corresponding to the space time domain optical signal after removing the background is classified.
The application also proposes a computer device comprising a memory and a processor, said memory storing a computer program, said computer program being executed by said processor to:
obtaining continuous and stable monochromatic p polarized light;
Collecting a target image, loading the target image to an amplitude modulation array, and modulating the monochromatic p-polarized light by using the amplitude modulation array loaded with the target image to obtain a spatial time domain optical signal of the target image containing a complex background;
converting the spatial time domain optical signal of the target image containing the complex background into a spatial frequency domain signal, wherein the spatial frequency domain signal is a spatial frequency distribution diagram containing spatial information of the target image;
Removing background information of the space frequency domain signal, and obtaining a space frequency domain signal after removing the background;
the space frequency domain signal after removing the background is converted into a space time domain optical signal after removing the background, the space time domain optical signal after removing the background is a space intensity distribution diagram containing space information of a target graph, and the target image corresponding to the space time domain optical signal after removing the background is classified.
The embodiment of the invention has the advantages that continuous and stable single-color p polarized light is firstly acquired, a target image is acquired, the target image is loaded to an amplitude modulation array, the single-color p polarized light is modulated by the amplitude modulation array after the target image is loaded, a space time domain optical signal of the target image containing a complex background is acquired, the space time domain optical signal of the target image containing the complex background is converted into a space frequency domain signal, the space frequency domain signal is a space frequency distribution diagram containing space information of the target image, background information of the space frequency domain signal is removed, a space frequency domain signal after the background is removed is acquired, the space frequency domain signal after the background is removed is converted into a space time domain optical signal after the background is removed, the space time domain optical signal after the background is removed is a space intensity distribution diagram containing space information of the target image, and the target image corresponding to the space time domain optical signal after the background is removed is classified. The invention can extract the target image from the complex background and accurately classify the target image by processing the optical signal.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Wherein:
FIG. 1 is a flow chart of a method for classifying target images based on complex backgrounds in one embodiment;
FIG. 2 is a schematic diagram of a target image classification system based on a complex background in one embodiment.
Description of the reference numerals
1, A laser, 2, a beam expanding lens, 3, a HWP, 4, a PBS, 5, a reflective amplitude type SLM, 6, a target acquisition detector, 7, a processing system, 8, a Fourier lens, 9, a first DONN, 10, a transmission phase type SLM, 11, an inverse Fourier lens, 12, a second DONN, and 13, a classification identification detector.
Detailed Description
The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, but not all embodiments, and all other embodiments obtained by those skilled in the art without making creative efforts based on the embodiments of the present invention are included in the protection scope of the present invention.
In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the present invention, it will be apparent, however, to one skilled in the art that the present invention may be practiced without one or more of these details, in other instances, in order to avoid obscuring the present invention, it will be understood that this invention is to be carried out in a different manner and is not to be construed as limited to the embodiments set forth herein, and that, on the contrary, these embodiments are presented to provide a thorough and complete understanding of the present invention and to convey the scope of the invention to those skilled in the art.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise, and it is to be understood that the terms "comprises" and/or "comprising" when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. As used herein, the term "and/or" includes any and all combinations of the associated listed items.
To solve this problem, the present application proposes a new method of fourier transforming an optical signal acquired in real time in a spatial domain by a 4f system into an optical signal in a spatial frequency domain, and then filtering and non-linearly activating the optical signal in the spatial frequency domain using DONN and a spatial light Modulator (SPATIAL LIGHT Modulator, SLM). The method can effectively filter out the complex background and retain the target to be identified, thereby improving DONN the target identification capability in the complex background. Finally, another DONN is used to identify the object image after the background is filtered, so as to realize accurate object classification. The method not only improves the practicability of DONN, but also provides wider space for the application of DONN in daily life.
In order that the invention may be fully understood, a detailed structure will be presented in the following description to illustrate the presently contemplated embodiments of the invention, and in addition to the detailed description, alternative embodiments of the invention may be described in detail.
As shown in fig. 1, in one embodiment, there is provided a method for classifying a target image based on a complex background, which includes steps S101 to S106, as follows:
s101, obtaining continuous and stable monochromatic p polarized light;
In some embodiments, obtaining continuous stable monochromatic p-polarization specifically includes:
the acquisition laser emits continuous and stable monochromatic linear polarized laser beams;
The monochromatic linear polarized laser beam passes through a beam expanding lens pair and is adjusted to obtain a parallel beam after beam expansion;
carrying out polarization modulation on the parallel light beams after beam expansion according to the half-wave plate;
And controlling the angle of the half wave plate to obtain continuous and stable monochromatic p polarized light.
Preferably, the laser emits a continuous and stable monochromatic linearly polarized laser beam, the beam diameter of which is enlarged and adjusted into a parallel beam by the beam expander, the incident light wave at this time is a plane wave, and the complex amplitude U (r) thereof is expressed as:
U(r)=u0eikz
Wherein u 0 is the real part of the complex amplitude, which determines the intensity and phase function The wave vector k represents the wave propagation direction independent of x and y, and the expression is k=2pi/λ.
Since the SLM needs p-polarized light (the component of the electric vector vibrating parallel to the incident plane is called p-component, if the linear polarized light has only p-component, it is called p-polarized light) to be incident, and the HWP is used to perform polarization modulation on the beam after beam expansion, so as to obtain continuous and stable monochromatic p-polarized light.
The polarization modulation process is to rotate the half wave plate to enable the fast axis direction and the original polarization direction to form an angle theta, and the half wave plate is a slow axis direction component added with phase delay pi, so that linearly polarized light is still linearly polarized after passing through the half wave plate, and the azimuth of the vibration surface is converted into the angle 2 theta compared with incident light, and therefore the polarization direction of the linearly polarized light can be adjusted to be p polarized light only by changing the angle theta.
S102, acquiring a target image, loading the target image to an amplitude modulation array, and modulating monochromatic p polarized light by using the amplitude modulation array loaded with the target image to obtain a spatial time domain optical signal of the target image containing a complex background;
in some embodiments, capturing the target image and loading the target image into the amplitude modulation array further comprises polarization filtering the continuously stabilized monochromatic p-polarized light by a polarization splitting prism PBS.
Specifically, the PBS performs polarization filtration on the HWP modulated emergent beam, and the PBS can transmit p polarized light and reflect s polarized light, so that only the p polarized light is incident on an amplitude modulation array of the reflective amplitude type SLM, and simultaneously, according to actual use requirements, a target gray level image containing a complex background and acquired in real time by a target acquisition detector is loaded on the amplitude modulation array of the reflective amplitude type SLM through a processing system, so that the light field intensity of reflected light is distributed according to the gray level value of the loaded gray level image;
The PBS works on the principle that a right-angle prism made of two negative uniaxial crystals (n e<no) is glued along an inclined plane by adopting a proper adhesive (n e<n<no), the optical axes of two s crystals are parallel to the right-angle plane and the incident plane of the light, at the moment, the e light is p polarized light, the o light is s polarized light (the component of the electric vector vibrating perpendicular to the incident plane is called s component, and if the linearly polarized light has only s component, the s polarized light is called s polarized light);
The working principle of the amplitude type SLM is that when the electrode is not electrified, the arrangement of liquid crystals is gradually distorted by 90 degrees along the direction of a light incident surface, natural light passes through a polarizer p 1 (parallel to the direction of the incident surface) to obtain polarized light, namely p polarized light, with the vibration direction parallel to the incident surface, the light beam is reflected by a dielectric reflector after entering a light valve and then exits through the light valve, the polarization direction of the emergent light is not changed, and no light passes through an orthogonal analyzer p 2 (perpendicular to the direction of the incident surface). When the electrodes of the liquid crystal light valve are electrified, the arrangement distortion of the liquid crystal is 45 degrees, the vibration direction of the emergent light reflected by the dielectric reflector is perpendicular to the incident surface, and the emergent light can completely pass through the analyzer p 2;
According to the actual use requirement, the processing module loads the gray level image of the target acquired by the target acquisition detector on the amplitude modulation array, the amplitude modulation array changes the resistance of the photosensitive layer according to the gray level value written into the image pixel point, so that an electric field changing along with the space is formed on the liquid crystal layer, the change of the electric field distribution on the liquid crystal layer changes the distortion of the liquid crystal arrangement, the vibration direction of polarized light emitted by the light valve changes, the intensity of light emitted by the polarization analyzer is distributed along the space, and the aim of spatial light modulation is achieved, and at the moment, the reflected light field can be expressed as:
where U (x, y) is the gray value of the pixel point of the written image at (x, y), δ (x 0-x,y0 -y) is a pulse function, the light field at this time can be regarded as a linear combination of several point light sources, and its light field complex amplitude U (x 0,y0) is distributed according to the gray value U (x, y) of the loaded image.
The complex amplitude of the reflected light field of the amplitude-reflective SLM is then:
U(x0,y0)＝A(x0,y0)r(x0,y0)
wherein A (x 0,y0) is light wave irradiated onto the amplitude modulation array, and r (x 0,y0) is amplitude reflection coefficient, namely, target image gray scale distribution containing complex background which is horizontally reversed and loaded onto the amplitude modulation array;
further, since the light source irradiated onto the amplitude modulation array is planar light, and the amplitude thereof is assumed to be 1, the complex amplitude of the reflected light field of the reflection type amplitude is:
U(x0,y0)＝r(x0,y0)
Where r (x 0,y0) is the amplitude reflection coefficient.
In some embodiments, the method for obtaining the spatial time domain optical signal of the target image comprising the complex background further comprises the steps of obtaining the spatial time domain optical signal of the target image comprising the complex background by distributing the light field intensity of the reflected light according to the gray value of the loaded gray image by using the amplitude modulation array after the target image is loaded.
S103, converting a spatial time domain optical signal of a target image containing a complex background into a spatial frequency domain signal, wherein the spatial frequency domain signal is a spatial frequency distribution diagram containing spatial information of the target image;
In some embodiments, converting a spatial time domain optical signal containing a target image of a complex background into a spatial frequency domain signal specifically includes:
The method comprises the steps that an amplitude modulation array of a reflective amplitude type Spatial Light Modulator (SLM) receives monochromatic p-polarized light reflected by a first Polarization Beam Splitter (PBS), and the monochromatic p-polarized light is reflected into monochromatic s-polarized light after amplitude modulation;
reflecting the monochromatic s polarized light to the direction perpendicular to the incident light path according to the first polarization splitting prism PBS;
The method comprises the steps that a reflected light beam passes through a first polarization beam splitting prism PBS from an amplitude modulation array, and the focal length f of a lens is determined according to the distance of the light beam in a propagation path when the light beam reaches the Fourier lens;
when the space time domain optical signal of the target image containing the complex background passes through the Fourier lens, the Fourier lens converts the space time domain optical signal into a space frequency domain, and a space frequency domain signal is obtained.
Specifically, the incident p polarized light is reflected into s polarized light after being subjected to amplitude modulation of the reflective amplitude type SLM, the s polarized light is reflected to a direction perpendicular to an incident light path by the first PBS, and the distance from the reflected light beam imaged by the amplitude modulation array through the PBS to the Fourier lens is just the focal length f of the Fourier lens;
when the light beam passes through a plurality of optical elements or optical systems, the image of the former element or system is the object of the latter element or system, and when the light field reflected by the amplitude modulation array is reflected by the PBS, the light field is equivalent to passing through a parallel plate, at this time, the image formed by the target is positioned in front of the amplitude modulation array, and the distance from the amplitude modulation array is as follows:
Wherein L is the side length of the PBS, n is the refractive index of the PBS, and the distance from the reflected light field imaged by the PBS to the Fourier lens should be the focal length f of the Fourier lens to form a 4f system;
The relationship between PBS surface-to-Fourier lens distance L 1, PBS side length L, amplitude modulation array-to-PBS surface distance L 2, and Fourier lens focal length f can be found from the above:
l1+L+Δl=f
Bringing Deltal into the above
When the above relation is satisfied, a 4f system is formed;
The reflected light field is reflected by the PBS and then irradiated onto the front surface of the first DONN through the Fourier lens to generate Fresnel diffraction, and the complex amplitude on the front surface of the first DONN is obtained by a Fresnel diffraction formula and is as follows:
U(ξ,η)=c·FT{r(x0,y0)}=c·R(ξ,η)
Where c is a complex constant, ζ=x/λf, η=y/λf, corresponding to the positions of x and y in the spatial frequency domain, respectively, the transformation is called standard fourier transformation, where the optical signal in the target spatial domain with complex background is transformed onto the spatial frequency domain.
In some embodiments, converting a spatial time domain optical signal containing a target image of a complex background into a spatial frequency domain signal specifically includes:
performing two-dimensional Fourier transform on the modulated s polarized light, and converting the spatial information of the target image into frequency information;
from the spatial frequency distribution map formed on the focal plane, a spatial frequency domain signal of the target image containing the complex background is determined from the signals after fourier transformation.
S104, removing background information of the space frequency domain signal, and obtaining the space frequency domain signal after removing the background;
In some embodiments, removing background information of the spatial frequency domain signal, and obtaining the spatial frequency domain signal after removing the background specifically includes:
filtering the spatial frequency domain signal according to the first diffractive optical neural network DONN;
And carrying out nonlinear activation on the spatial frequency domain signal after the filtering processing according to the transmission phase type Spatial Light Modulator (SLM) to acquire the spatial frequency domain signal after the processing.
Specifically, the optical signal converted into the spatial frequency domain is filtered by the first DONN, the parameter of the first DONN is obtained by training the specific filtering task data set on a computer by using a deep learning method, and the complex amplitude of the optical field after the filtering by the first DONN is given by the filtering function of the first DONN in the spatial frequency domain as H (ζ, η):
U′(ξ,η)=U(ξ,η)H(ξ,η)
After filtering, part of the optical signal containing the background information is removed;
In particular, the H (ζ, η) may be formed by stacking a random matrix phase function of the vortex phase function and a fresnel lens phase function. The vortex phase function modulates the light field to enable the wave front phase to be distributed in a spiral mode, at the moment, the light beam carries certain orbital angular momentum, and the light intensity also shows circular distribution. For the target optical signal transferred into the fourier space, part of the fundamental frequency information located in the center of the spatial frequency domain is filtered out by the vortex phase function, so that the final output only retains the medium-high frequency information.
The random matrix phase function is randomly distributed in a two-dimensional space, the range of the random matrix phase function is between 0 and 2 pi, and the random matrix phase function is used for accurately modulating the space frequency domain information to filter out complex background so as to only retain the medium-high frequency information of a target, and the random matrix phase function is shown as a target shape in the space domain. The fresnel lens phase function is used for simulating the fresnel lens and focusing or diverging the light beam, and since part of energy is diffracted out of the field of view of the system when the light field passes through DONN, and the final imaging intensity is reduced, the fresnel lens phase function needs to be added to the first DONN to make the energy converge so as to ensure the final processing effect.
S105, converting the space frequency domain signal after removing the background into a space time domain optical signal after removing the background, wherein the space time domain optical signal after removing the background is a space intensity distribution diagram containing space information of a target graph;
Wherein, through nonlinear activation of the transmissive phase SLM, an optical signal containing target information is enhanced and the remaining optical signal containing background information is attenuated, at this time, the optical signal almost contains only target information, the basic principle of the nonlinear activation is to set a gamma correction curve of the transmissive phase SLM to a nonlinear activation function Φ (ζ, η) =e f(ξ,η), and when the optical field is nonlinear activated, its complex amplitude can be expressed as:
U′(ξ,η)=U(ξ,η)H(ξ,η)Φ(ξ,η)=c·ef(ξ,η)R(ξ,η)H(ξ,η)
where c is a complex constant and H (ζ, η) is the filter function.
The distance from the back surface of the transmissive phase SLM and the front surface of the second DONN to the inverse fourier lens is the focal length f of the inverse fourier lens, at this time, the optical signal in the spatial frequency domain will be fourier transformed again, if the coordinates are opposite, the optical signal after nonlinear activation is inverse fourier transformed into the spatial domain, at this time, the optical signal in the spatial domain almost only remains the target information, and the optical wave field can be expressed as:
Ui(xi,yi)＝FT-1{c·ef(ξ,η)R(ξ,η)H(ξ,η)}
where c is a complex constant and H (ζ, η) is the filter function.
S106, classifying the target images corresponding to the space time domain optical signals after the background is removed.
Specifically, the second DONN classifies the spatial domain optical signals after the background is removed, the parameters of the spatial domain optical signals are obtained by training a computer by using a deep learning method according to a specific classification task data set, and the classification recognition detector acquires the classified optical signals and transmits the classified optical signals to the processing system for recognition, so that real-time classification recognition processing of the target under the complex background is realized.
As shown in fig. 2, the present application further proposes a target image classification system based on a complex background, the system comprising:
the laser is used for providing a continuous and stable light source for the whole system, and the emitted light is monochromatic light and linearly polarized light with fixed polarization direction;
The beam expanding lens is used for diverging a monochromatic beam emitted by the laser, and simultaneously carrying out collimation and shaping on the beam, so that the beam finally passes through a subsequent system in parallel, and the beam energy is dispersed to avoid damaging other components due to overhigh power, and the beam irradiated on the reflective amplitude type SLM can be ensured to fill the whole amplitude modulation array;
HWP is used for changing polarization direction of polarized light of laser output line, and modulating polarization state of light beam on the incident amplitude modulation array into p polarized light according to operation requirement of reflective amplitude type SLM so as to ensure that the reflective amplitude type SLM can work normally;
The light source comprises a PBS, a light source and a light source, wherein the PBS is used for dividing incident light into two perpendicular linearly polarized light, p polarized light completely passes through, s polarized light is reflected at an angle of 45 degrees, an outgoing direction forms an angle of 90 degrees with the p polarized light, the p polarized light and the s polarized light are both linearly polarized light, and the polarization directions are perpendicular to each other;
The reflective amplitude type SLM is used for carrying out amplitude modulation on incident p polarized light, and comprises a sandwich structure formed by a photoelectric guide layer, a dielectric reflector, a liquid crystal layer, a glass substrate transparent conductive electrode (ITO) and the like, and a two-dimensional array liquid crystal panel formed by a plurality of basic independent units, wherein a processing module can control the rotation direction of voltage-regulating liquid crystal molecules through software to realize amplitude regulation and control on single pixel points, and the reflected light of a modulation array is s polarized light;
The target acquisition detector is used for acquiring a target image in an application scene in real time so as to facilitate the subsequent system to process;
the processing system is used for transmitting the target image acquired by the target acquisition detector to the reflective amplitude type SLM, transmitting the preset phase modulation parameter to the transmissive phase type SLM, and acquiring the classification identification signal acquired by the classification identification detector;
the Fourier lens is used for converting the optical signal in the space domain into the optical signal in the space frequency domain, so that the subsequent modulation of the optical signal is carried out in the space frequency domain, and the optical signal is identical to the focal length f of the Fourier lens to form a 4f system;
the first DONN is used for performing filtering processing modulation on the optical signal in the space frequency domain, and specific parameters of the optical signal are obtained through computer simulation training according to a dataset and are manufactured by technologies such as 3D printing or photoetching;
the transmission phase type SLM is used for performing nonlinear activation operation on the spatial frequency domain optical signal filtered by the first DONN;
The anti-Fourier lens is used for converting the optical signals in the space frequency domain into the optical signals in the space domain, so that the subsequent processing identification of the optical signals is carried out in the space domain, and the optical signals are identical to the focal length f of the Fourier lens to form a 4f system;
The second DONN is used for identifying the filtered spatial domain optical signal, and specific parameters of the second DONN are obtained through computer simulation training according to a dataset and are manufactured by technologies such as 3D printing or photoetching;
And the classification and identification detector is used for collecting the optical signals after the second DONN classification and identification and transmitting the optical signals to the processing system for identification and processing.
The application also proposes a readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of:
obtaining continuous and stable monochromatic p polarized light;
collecting a target image, loading the target image to an amplitude modulation array, and modulating monochromatic p polarized light by using the amplitude modulation array loaded with the target image to obtain a space time domain optical signal of the target image containing a complex background;
converting a spatial time domain optical signal of a target image containing a complex background into a spatial frequency domain signal, wherein the spatial frequency domain signal is a spatial frequency distribution diagram containing spatial information of the target image;
Removing background information of the space frequency domain signal, and obtaining the space frequency domain signal after removing the background;
converting the space frequency domain signal after removing the background into a space time domain optical signal after removing the background, wherein the space time domain optical signal after removing the background is a space intensity distribution diagram containing space information of a target graph;
and classifying the target images corresponding to the space time domain optical signals after the background is removed.
The application also proposes a computer device comprising a memory and a processor, the memory storing a computer program, the computer program being executed by the processor to:
obtaining continuous and stable monochromatic p polarized light;
collecting a target image, loading the target image to an amplitude modulation array, and modulating monochromatic p polarized light by using the amplitude modulation array loaded with the target image to obtain a space time domain optical signal of the target image containing a complex background;
converting a spatial time domain optical signal of a target image containing a complex background into a spatial frequency domain signal, wherein the spatial frequency domain signal is a spatial frequency distribution diagram containing spatial information of the target image;
Removing background information of the space frequency domain signal, and obtaining the space frequency domain signal after removing the background;
converting the space frequency domain signal after removing the background into a space time domain optical signal after removing the background, wherein the space time domain optical signal after removing the background is a space intensity distribution diagram containing space information of a target graph;
and classifying the target images corresponding to the space time domain optical signals after the background is removed.
Those skilled in the art will appreciate that implementing all or part of the above-described methods in accordance with the embodiments can be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, can comprise the steps of the above-described embodiments of the methods. Any reference to memory, storage, database, or other medium used in embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (SYNCHLINK) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
The technical features of the above embodiments can be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The foregoing examples illustrate only a few embodiments of the application and are described in detail herein without thereby limiting the scope of the application. It should be noted that it is possible for those skilled in the art to make several variations and modifications without departing from the spirit of the application, and these are all the preferred embodiments of the application, and it is needless to say that the scope of the claims of the application shall not be limited thereto, and therefore equivalent variations according to the claims of the application shall still fall within the scope of the application.
Claims (10)
1. A method for classifying target images based on complex backgrounds, the method comprising:
obtaining continuous and stable monochromatic p polarized light;
Collecting a target image, loading the target image to an amplitude modulation array, and modulating the monochromatic p-polarized light by using the amplitude modulation array loaded with the target image to obtain a spatial time domain optical signal of the target image containing a complex background;
converting the spatial time domain optical signal of the target image containing the complex background into a spatial frequency domain signal, wherein the spatial frequency domain signal is a spatial frequency distribution diagram containing spatial information of the target image;
Removing background information of the space frequency domain signal, and obtaining a space frequency domain signal after removing the background;
converting the space frequency domain signal after removing the background into a space time domain optical signal after removing the background, wherein the space time domain optical signal after removing the background is a space intensity distribution diagram containing space information of a target graph;
and classifying the target image corresponding to the space time domain optical signal after the background is removed.
2. The method for classifying a target image based on a complex background according to claim 1, wherein the step of obtaining continuous and stable monochromatic p-polarized light specifically comprises:
the acquisition laser emits continuous and stable monochromatic linear polarized laser beams;
the monochromatic linear polarized laser beams pass through a beam expanding lens pair and are adjusted to obtain parallel beams after beam expansion;
carrying out polarization modulation on the parallel light beams after beam expansion according to a half-wave plate;
and controlling the angle of the half wave plate to obtain continuous and stable monochromatic p polarized light.
3. The method of complex background-based object image classification as defined in claim 2, wherein said capturing an object image and loading said object image into an amplitude modulation array further comprises polarization filtering said continuous stable monochromatic p-polarized light through a polarization splitting prism PBS.
4. The method for classifying a target image based on a complex background according to claim 3, wherein the step of modulating the monochromatic p-polarized light by using an amplitude modulation array after loading the target image to obtain a spatial time domain optical signal of the target image containing the complex background further comprises the step of distributing the light field intensity of the reflected light according to the gray value of the loaded gray image to obtain the spatial time domain optical signal of the target image containing the complex background.
5. The method for classifying a target image based on a complex background according to claim 4, wherein the converting the spatial time domain optical signal of the target image containing the complex background into a spatial frequency domain signal specifically comprises:
The method comprises the steps that an amplitude modulation array of a reflective amplitude type Spatial Light Modulator (SLM) receives monochromatic p-polarized light reflected by a first polarization splitting Prism (PBS), and the monochromatic p-polarized light is reflected into monochromatic s-polarized light after amplitude modulation;
Reflecting the single-color s polarized light to the direction perpendicular to the incident light path according to the first polarization splitting prism PBS;
The reflected light beam passes through a first polarization beam splitting prism PBS from an amplitude modulation array, and the focal length f of the lens is determined according to the distance of the light beam in the propagation path when the light beam reaches the Fourier lens;
When the spatial time domain optical signal of the target image containing the complex background passes through the Fourier lens, the Fourier lens converts the spatial time domain optical signal into a spatial frequency domain to acquire a spatial frequency domain signal.
6. The method for classifying a target image based on a complex background according to claim 5, wherein the converting the spatial time domain optical signal of the target image containing the complex background into a spatial frequency domain signal specifically comprises:
performing two-dimensional Fourier transform on the modulated s polarized light, and converting the spatial information of the target image into frequency information;
from the spatial frequency distribution map formed on the focal plane, a spatial frequency domain signal of the target image containing the complex background is determined from the signals after fourier transformation.
7. The method for classifying a target image based on a complex background according to claim 6, wherein the removing the background information of the spatial frequency domain signal, and obtaining the spatial frequency domain signal after removing the background, specifically comprises:
filtering the spatial frequency domain signal according to a first diffractive optical neural network DONN;
And carrying out nonlinear activation on the spatial frequency domain signal after the filtering processing according to the transmission phase type Spatial Light Modulator (SLM) to acquire the spatial frequency domain signal after the processing.
8. The target image classification system based on the complex background is characterized by comprising a polarized light acquisition unit, a time domain signal acquisition unit, a signal processing unit and a classification unit;
the polarized light acquisition unit is used for acquiring continuous and stable monochromatic p polarized light;
The time domain signal acquisition unit is used for acquiring a target image, loading the target image to an amplitude modulation array, and modulating the monochromatic p-polarized light by using the amplitude modulation array loaded with the target image to acquire a spatial time domain optical signal containing the target image with a complex background;
the signal processing unit is used for converting the space time domain optical signal of the target image containing the complex background into a space frequency domain signal, wherein the space frequency domain signal is a space frequency distribution diagram containing space information of the target image, removing background information of the space frequency domain signal to obtain a space frequency domain signal after background removal, converting the space frequency domain signal after background removal into a space time domain optical signal after background removal, wherein the space time domain optical signal after background removal is a space intensity distribution diagram containing space information of the target image, and the classifying unit is used for classifying the target image corresponding to the space time domain optical signal after background removal.
9. A readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of the method of any one of claims 1 to 7.
10. A computer device comprising a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the method as claimed in any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411742333.1A CN119418131A (en) | 2024-11-29 | 2024-11-29 | Method, system, medium and device for target image classification based on complex background |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411742333.1A CN119418131A (en) | 2024-11-29 | 2024-11-29 | Method, system, medium and device for target image classification based on complex background |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN119418131A true CN119418131A (en) | 2025-02-11 |
Family
ID=94459847
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202411742333.1A Pending CN119418131A (en) | 2024-11-29 | 2024-11-29 | Method, system, medium and device for target image classification based on complex background |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN119418131A (en) |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080101657A1 (en) * | 2006-10-30 | 2008-05-01 | The Regents Of The University Of California | Method and apparatus for performing qualitative and quantitative analysis of produce (fruit, vegetables) using spatially structured illumination |
| CN104537634A (en) * | 2014-12-31 | 2015-04-22 | 中国科学院深圳先进技术研究院 | Method and system for removing raindrop influences in dynamic image |
| CN109271902A (en) * | 2018-08-31 | 2019-01-25 | 西安电子科技大学 | Method for detecting infrared puniness target based on time domain empirical mode decomposition under complex background |
| CN111221001A (en) * | 2020-01-16 | 2020-06-02 | 宁波飞芯电子科技有限公司 | Detection method, device, equipment and storage medium |
| CN111736334A (en) * | 2020-06-22 | 2020-10-02 | 武汉市艾玻睿光电科技有限公司 | Device for realizing complex amplitude light field regulation and control based on spatial light modulator |
| CN114355320A (en) * | 2021-12-31 | 2022-04-15 | 南京牧镭激光科技有限公司 | Device and method for calibrating signal emission time and intermediate frequency of Doppler wind lidar end face |
| CN115793348A (en) * | 2022-12-03 | 2023-03-14 | 哈尔滨理工大学 | A target enhanced up-conversion imaging method and system |
| CN115935236A (en) * | 2022-12-14 | 2023-04-07 | 汕头大学 | Optical fiber vibration event classification method based on dual data amplification and storage medium |
| CN116128037A (en) * | 2022-12-12 | 2023-05-16 | 北京空间机电研究所 | An efficient training method and system for 4f-ONN system |
| US20230281437A1 (en) * | 2022-03-01 | 2023-09-07 | Massachusetts Institute Of Technology | Radio-Frequency Photonic Architecture for Deep Neural Networks, Signal Processing, and Computing |
| CN116908841A (en) * | 2023-06-21 | 2023-10-20 | 山东大学 | A high-resolution ground penetrating radar transceiver system signal processing method and system |
| CN118916755A (en) * | 2024-10-11 | 2024-11-08 | 南京航天工业科技有限公司 | Same-frequency signal separation and identification method based on real-time spectrum analysis |
-
2024
- 2024-11-29 CN CN202411742333.1A patent/CN119418131A/en active Pending
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080101657A1 (en) * | 2006-10-30 | 2008-05-01 | The Regents Of The University Of California | Method and apparatus for performing qualitative and quantitative analysis of produce (fruit, vegetables) using spatially structured illumination |
| CN104537634A (en) * | 2014-12-31 | 2015-04-22 | 中国科学院深圳先进技术研究院 | Method and system for removing raindrop influences in dynamic image |
| CN109271902A (en) * | 2018-08-31 | 2019-01-25 | 西安电子科技大学 | Method for detecting infrared puniness target based on time domain empirical mode decomposition under complex background |
| CN111221001A (en) * | 2020-01-16 | 2020-06-02 | 宁波飞芯电子科技有限公司 | Detection method, device, equipment and storage medium |
| CN111736334A (en) * | 2020-06-22 | 2020-10-02 | 武汉市艾玻睿光电科技有限公司 | Device for realizing complex amplitude light field regulation and control based on spatial light modulator |
| CN114355320A (en) * | 2021-12-31 | 2022-04-15 | 南京牧镭激光科技有限公司 | Device and method for calibrating signal emission time and intermediate frequency of Doppler wind lidar end face |
| US20230281437A1 (en) * | 2022-03-01 | 2023-09-07 | Massachusetts Institute Of Technology | Radio-Frequency Photonic Architecture for Deep Neural Networks, Signal Processing, and Computing |
| CN115793348A (en) * | 2022-12-03 | 2023-03-14 | 哈尔滨理工大学 | A target enhanced up-conversion imaging method and system |
| CN116128037A (en) * | 2022-12-12 | 2023-05-16 | 北京空间机电研究所 | An efficient training method and system for 4f-ONN system |
| CN115935236A (en) * | 2022-12-14 | 2023-04-07 | 汕头大学 | Optical fiber vibration event classification method based on dual data amplification and storage medium |
| CN116908841A (en) * | 2023-06-21 | 2023-10-20 | 山东大学 | A high-resolution ground penetrating radar transceiver system signal processing method and system |
| CN118916755A (en) * | 2024-10-11 | 2024-11-08 | 南京航天工业科技有限公司 | Same-frequency signal separation and identification method based on real-time spectrum analysis |
Non-Patent Citations (3)
| Title |
|---|
| WANG CHANGMING ET AL.: "Spatio-temporal pattern analysis of single-trial EEG signals recorded during visual object recognition", 《SCIENCE CHINA》, vol. 54, no. 12, 31 December 2011 (2011-12-31), pages 2499 - 2507, XP019986537, DOI: 10.1007/s11432-011-4507-1 * |
| 刘峰 等: "时频谱图和数据增强的水声信号深度学习目标识别方法", 《应用声学》, vol. 40, no. 4, 31 July 2021 (2021-07-31), pages 518 - 524 * |
| 李佳 等: "时空域非线性滤波红外序列图像非均匀性校正", 《光子学报》, vol. 44, no. 4, 30 April 2015 (2015-04-30), pages 1 - 5 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11586144B2 (en) | Dynamic holography focused depth printing device | |
| CN106560738B (en) | A device and method for generating a perfect IG vortex beam | |
| CN205620619U (en) | Produce device of gauss's vortex light beam in ending | |
| EP4040239A1 (en) | Image projection | |
| CN114019690B (en) | Optical systems for generating arbitrary-order optical vortex arrays and finite optical lattices with defects | |
| US10802440B2 (en) | Dynamic holography non-scanning printing device | |
| CN109870890B (en) | Integer order vortex light beam phase mask plate with fractional order vortex contour and light path system | |
| CN113448077A (en) | Method, device and system for generating multi-parameter adjustable light field based on DMD | |
| US20190004476A1 (en) | Dynamic Holography Printing Device | |
| CN103425035B (en) | A kind of holographic stereogram print system adopting special-shaped mask plate | |
| US20240104918A1 (en) | Non-interferometric, non-iterative complex amplitude reading method and apparatus | |
| Skirnewskaja et al. | Accelerated augmented reality holographic 4K video projections based on Lidar Point Clouds for Automotive head‐up displays | |
| CN114092757B (en) | Image processing | |
| CN119418131A (en) | Method, system, medium and device for target image classification based on complex background | |
| CN113805347A (en) | Method for mapping space phase to space-time light field time-space domain phase | |
| CN104439723A (en) | Laser boring method and device | |
| US20190025757A1 (en) | Holographic System for Controlling Plasma | |
| Vaziri et al. | A simple method to prepare and characterize optical fork-shaped diffraction gratings for generation of orbital angular momentum beams | |
| CN112327557A (en) | Self-accelerating light generation method and device | |
| CN101566732B (en) | Method for generating column vector beam by utilizing computer generated hologram | |
| CN209266023U (en) | Minimize volume holographicstorage and identifying system | |
| KR102778506B1 (en) | Optical artificial neural network system | |
| Owechko et al. | Holographic neurocomputer utilizing laser diode light source | |
| US20250028219A1 (en) | Method and apparatus of performing spatial domain-based optical convolution operation | |
| Chen et al. | Multilayer optoelectronic hybrid convolutional neural network with an optical 4f-system recurrent structure |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |