CN113192154A - Underwater ghost imaging system based on edge calculation and deep learning image reconstruction method - Google Patents
Underwater ghost imaging system based on edge calculation and deep learning image reconstruction method Download PDFInfo
- Publication number
- CN113192154A CN113192154A CN202110594841.XA CN202110594841A CN113192154A CN 113192154 A CN113192154 A CN 113192154A CN 202110594841 A CN202110594841 A CN 202110594841A CN 113192154 A CN113192154 A CN 113192154A
- Authority
- CN
- China
- Prior art keywords
- light
- light intensity
- image
- target object
- modulation device
- 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.)
- Granted
Links
- 238000003384 imaging method Methods 0.000 title claims abstract description 57
- 238000004364 calculation method Methods 0.000 title claims abstract description 50
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000013135 deep learning Methods 0.000 title claims abstract description 31
- 230000003287 optical effect Effects 0.000 claims abstract description 17
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 10
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 7
- 238000012549 training Methods 0.000 claims description 25
- 238000003062 neural network model Methods 0.000 claims description 23
- 230000006870 function Effects 0.000 claims description 14
- 238000004891 communication Methods 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 9
- 238000001514 detection method Methods 0.000 claims description 6
- 238000004458 analytical method Methods 0.000 claims description 3
- 150000001875 compounds Chemical class 0.000 claims description 3
- 238000004519 manufacturing process Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 230000003044 adaptive effect Effects 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 description 15
- 238000012545 processing Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000012827 research and development Methods 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 238000009434 installation Methods 0.000 description 2
- 238000010521 absorption reaction Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000012821 model calculation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—2D [Two Dimensional] image generation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/30—Assessment of water resources
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The application discloses an underwater ghost imaging system based on edge calculation and a deep learning image reconstruction method, wherein the system comprises: light source: for emitting light to the active modulation area of the light modulation device; an optical modulation device: the system comprises a projection lens, speckles, a reflector and a reflector, wherein the speckles are used for modulating light and reflecting the modulated light so that the reflected light passes through the projection lens along the optical axis of the projection lens and then is emitted to a target object in the water body; the converging lens is used for: converging the light reflected by the target; the light intensity detector is used for: collecting light intensity information of the converged light; the first end of the edge calculation module is connected with the light modulation device and used for sending speckles to the light modulation device; and the second end of the edge calculation module is connected with the light intensity detector and is used for controlling the light intensity detector to collect light intensity information and acquiring an image of the target object through a deep learning image reconstruction algorithm according to the light intensity information. The problems of large calculated amount, low imaging resolution and poor imaging quality in the prior art are solved.
Description
Technical Field
The application relates to the technical field of underwater imaging, in particular to an underwater ghost imaging system based on edge calculation and a deep learning image reconstruction method.
Background
Underwater imaging technology is an important means of understanding, developing and utilizing the ocean. Due to the absorption and scattering of light by the water body, it becomes very difficult to acquire high-resolution and high-definition underwater target images. The ghost imaging technology utilizes the statistical correlation of light field intensity fluctuation to carry out imaging, and compared with the traditional CCD and CMOS imaging technologies, the ghost imaging technology can be used for imaging in a weak light environment and a scattering medium and has important application value in underwater imaging. However, the existing underwater ghost imaging technology has the problems of large calculation amount, low imaging resolution and poor imaging quality, and the development of the underwater ghost imaging technology is restricted.
In recent years, edge calculation and deep learning have been widely applied in the fields of video and image processing. Edge computing can provide network, computing, storage, application, etc. functions at the data source end. The processing process is placed at a local edge end for processing, and the data computing and processing efficiency is greatly improved by matching with a cloud computing technology. The scene high-resolution image can be quickly reconstructed by deep learning, and the image is detected, identified, classified and the like. The method is used for solving the problems of large calculation amount, low imaging resolution and poor imaging quality in underwater ghost imaging.
Disclosure of Invention
The embodiment of the application provides an underwater ghost imaging system based on edge calculation and a deep learning image reconstruction method, which are used for solving the technical problems of large calculation amount, low imaging resolution and poor imaging quality in the existing underwater ghost imaging.
In view of the above, a first aspect of the present application provides an underwater ghost imaging system based on edge calculation, the system comprising:
the device comprises a light source, a light modulation device, a projection lens, a convergent lens, a light intensity detector and an edge calculation module;
the first end of the edge calculation module is connected with the light modulation device and used for sending speckles to the light modulation device;
the light source is: the light modulation device is used for emitting light to an effective modulation area of the light modulation device;
the light modulation device: the system comprises a projection lens, speckles, a lens holder, a projection lens, a reflector and a reflector, wherein the speckle is used for modulating light rays and reflecting the modulated light rays so that the reflected light rays are emitted to a target object in the water body after passing through the projection lens along the optical axis of the projection lens;
the converging lens: for converging the light reflected by the target;
the light intensity detector is: the light intensity information acquisition module is used for acquiring the light intensity information of the converged light;
and the second end of the edge calculation module is connected with the light intensity detector and is used for controlling the light intensity detector to collect light intensity information and acquiring an image of the target object through a deep learning image reconstruction algorithm according to the light intensity information.
Optionally, the method further comprises: a cloud computing center; the cloud computing center is in communication connection with the edge computing module;
the cloud computing center: and the image of the target object is obtained through a deep learning image reconstruction algorithm according to the light intensity information when the pixel size of the image of the target object is larger than a preset size.
Optionally, the method further comprises: the cloud storage platform is in communication connection with the cloud computing center and the edge computing module respectively;
the cloud storage platform is used for: storing the image of the object.
Optionally, the method further comprises: the display terminal is in communication connection with the cloud storage platform;
the display terminal is used for: and receiving the image of the target object sent by the cloud storage platform, and displaying the image of the target object.
Optionally, the converging lens is installed right in front of the light intensity detector, so that the focus of the converging lens is in the center of the effective detection area of the light intensity detector.
The second aspect of the present application provides a deep learning image reconstruction method, which is applied to the underwater ghost imaging system based on edge calculation according to the first aspect, and includes:
loading the speckle sequence on a light modulation device to modulate light, and projecting the modulated light to a target object through a projection lens;
carrying out light intensity detection on the light rays converged to the light intensity detector by the convergent lens to obtain a light intensity sequence;
by data setsTraining a neural network model, and optimizing a loss function by using an ADAM (adaptive dynamic analysis of moving average) function until the training reaches preset times to obtain parameters of the trained neural network model;
inputting the light intensity sequence into a trained neural network model, and outputting an image of a target object
Wherein the neural network model is represented as:
Q(x,y)=Ψ(S,Φ);
the above formula represents that a mapping relation is established between a one-dimensional optical field intensity sequence S and an output image Q (x, y) by using a hidden function psi (·), and phi represents a neural network model parameter.
Optionally, the method further comprises: on-pair neural network modelAdding m noise level graphs in the training process, wherein the pixel values of the b-th noise level graph are all Ub,b=1,2,...,m,UbIs at (0, 75)]Randomly generated within the range.
200000 different gray images with 64 × 64 pixel size are taken as original target object images, and the jth original target object image G is subjected tojObtaining a one-dimensional light field intensity sequence by calculation
In the formula (I), the compound is shown in the specification,denotes the nth light field intensity value, N1, 2.
Optionally, the neural network model parameters are represented as:
in the formula, | | the luminance2Denotes 2-norm, λ ═ 0.6 is the regularization parameter, GiRepresenting the i-th original object image for training, DiIs corresponding to the object image GiThe number of training times is 30000, and the total amount of data in the data set is 200000.
Optionally, the loss function is expressed as:
in the formula, GiRepresenting the i-th original object image for training, DiIs corresponding to the object image GiThe one-dimensional light field intensity sequence of (a),i is the number of training sessions.
According to the technical scheme, the embodiment of the application has the following advantages:
in an embodiment of the present application, an underwater ghost imaging system based on edge calculation is provided, including: the method comprises the following steps: the device comprises a light source, a light modulation device, a projection lens, a convergent lens, a light intensity detector and an edge calculation module; the first end of the edge calculation module is connected with the light modulation device and used for sending speckles to the light modulation device; light source: for emitting light to the active modulation area of the light modulation device; an optical modulation device: the system comprises a projection lens, speckles, a lens holder, a projection lens, a reflector and a reflector, wherein the speckle is used for modulating light rays and reflecting the modulated light rays so that the reflected light rays are emitted to a target object in the water body after passing through the projection lens along the optical axis of the projection lens; a converging lens: for converging the light reflected by the target; a light intensity detector: the light intensity information acquisition module is used for acquiring the light intensity information of the converged light; and the second end of the edge calculation module is connected with the light intensity detector and is used for controlling the light intensity detector to acquire light intensity information and acquiring an image of the target object through a deep learning image reconstruction algorithm according to the light intensity information.
The underwater ghost imaging system based on the edge calculation is composed of a light source, a light modulation device, a projection lens, a converging lens, a light intensity detector and an edge calculation module, is compact in structure, convenient to install and simple to operate, and obtains images of a target object by utilizing light intensity information obtained by the system through a deep learning image reconstruction algorithm. In addition, the edge computing technology is adopted, and the computing capability of the system is greatly improved by matching with the base station and the cloud computing platform, so that the practicability is higher. In addition, due to the fact that Fourier sinusoidal speckles have orthogonality, an effective light intensity sequence capable of reconstructing a target image can be obtained under the condition that Nyquist sampling is lower. The method and the device are beneficial to research and development of an underwater imaging technology and a deep learning technology. Therefore, the technical problems of large calculated amount, low imaging resolution and poor imaging quality in the existing underwater ghost imaging are solved.
Drawings
FIG. 1 is an architecture diagram of an underwater ghost imaging system based on edge calculation provided in an embodiment of the present application;
fig. 2 is a schematic flowchart of a deep learning image reconstruction method provided in an embodiment of the present application;
FIG. 3 is a schematic diagram of a neural network model training process in the present application;
FIG. 4 is a schematic diagram of the components of the modules A, B, C, D, E, F, G, H, and I shown in FIG. 3;
fig. 5 is a schematic composition diagram of the dense block of fig. 4.
Reference numbers: 101. a light source; 102. a light modulation device; 103. a projection lens; 104. a converging lens; 105. a light intensity detector 106 and an edge calculation module; 107. and a base station.
Detailed Description
In order to make the technical solutions of the present application better understood, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Referring to fig. 1, an underwater ghost imaging system based on edge calculation provided in an embodiment of the present application includes: a light source 101, a light modulation device 102, a projection lens 103, a converging lens 104, a light intensity detector 105, and an edge calculation module 106;
a first end of the edge calculation module 106 is connected to the light modulation device 102, and is configured to send speckles to the light modulation device 102;
light source 101: for emitting light to the active modulation area of the light modulation device 102;
the light modulation device 102: the system comprises a projection lens 103, a speckle filter, a light source and a controller, wherein the light source is used for modulating light through the speckle and reflecting the modulated light, so that the reflected light is emitted to a target object in the water body after passing through the projection lens along the optical axis of the projection lens 103;
the converging lens 104: for converging the light reflected by the target;
the light intensity detector 105: the light intensity information acquisition module is used for acquiring the light intensity information of the converged light;
the second end of the edge calculation module 106 is connected to the light intensity detector 105, and is configured to control the light intensity detector 105 to acquire light intensity information, and acquire an image of the target object through a deep learning image reconstruction algorithm according to the light intensity information.
It should be noted that the fourier sinusoidal speckles for modulating the light source are emitted by the edge calculation module, and are modulated with the light source emitted by the light modulation device 102 and the light source 101. The light source is projected to a target object in the water body through the projection lens 103, so that the light source is emitted through the target object and reaches the light intensity detector 105 through the convergent lens 104 for collection, and a light field intensity sequence is obtained. The edge calculation module comprises an image model obtained by training the image network model through the light field intensity sequence sample, so that the image of the target object can be quickly output through the image model calculation only by inputting the obtained light field intensity sequence into the edge calculation module.
In a specific installation mode, firstly, the optical modulation device 102 is fixed, the projection lens 103 is installed at the right side of the optical modulation device 102, and the optical axis of the projection lens 103 passes through the center of the effective modulation area of the optical modulation device 102; then, a light source 101 is arranged at the lower right of the light modulation device 102, so that light emitted by the light source 101 can cover the effective modulation area of the light modulation device, and after the central light of the light source 101 is reflected by the light modulation device 102, the reflected light can pass through the projection lens 103 along the optical axis of the projection lens 103; then, the projected light source is directed at the target, and the converging lens 104 is installed right in front of the light intensity detector 105, so that the focus of the converging lens 104 is at the center of the effective detection area of the light intensity detector 105.
The setting of the positional relationship between the modules in the system of the above installation manner is only an example of the present embodiment, and a person skilled in the art may also perform the setting according to the actual situation, which is not limited herein.
The underwater ghost imaging system based on the edge calculation is composed of a light source, a light modulation device, a projection lens, a converging lens, a light intensity detector and an edge calculation module, is compact in structure, convenient to install and simple to operate, and obtains images of a target object by utilizing light intensity information obtained by the system through a deep learning image reconstruction algorithm. In addition, the edge computing technology is adopted, and the computing capability of the system is greatly improved by matching with the base station and the cloud computing platform, so that the practicability is higher. In addition, due to the fact that Fourier sinusoidal speckles have orthogonality, an effective light intensity sequence capable of reconstructing a target image can be obtained under the condition that Nyquist sampling is lower. The invention is beneficial to the research and development of underwater imaging technology and deep learning technology. Therefore, the technical problems of large calculated amount, low imaging resolution and poor imaging quality in underwater ghost imaging are solved.
In a specific embodiment, the underwater ghost imaging system based on edge calculation of the present application further includes: a cloud computing center; the cloud computing center is in communication connection with the edge computing module;
the cloud computing center: and the image of the target object is obtained through a deep learning image reconstruction algorithm according to the light intensity information when the pixel size of the image of the target object is larger than the preset size.
It should be noted that, as a result of experimental analysis by the applicant, when the pixel size of the processed image is greater than 256 × 256, the speed of generating the image by the edge calculation module has no obvious advantage over the prior art, so that in order to ensure the reliability of the underwater ghost imaging system based on the edge calculation, in consideration of the characteristic that the edge calculation module has no obvious advantage over the processing of a large-size image, the cloud calculation center is configured to generate the image of the target object by using the prior art, and therefore, a person skilled in the art can set the preset size according to actual situations.
In a specific embodiment, the underwater ghost imaging system based on edge calculation of the present application further includes: the cloud storage platform is in communication connection with the cloud computing center and the edge computing module respectively;
the cloud storage platform is used for: storing the image of the object.
It should be noted that, in order to ensure the speed and reliability of the generated target object image, the present embodiment further arranges the cloud storage platform for storing the target object image.
In a specific embodiment, the underwater ghost imaging system based on edge calculation of the present application further includes: the display terminal is in communication connection with the cloud storage platform;
the display terminal is used for: and receiving the image of the target object sent by the cloud storage platform, and displaying the image of the target object.
It should be noted that, in order to enable the user to visually and real-timely obtain the image of the target object and improve the practicability of the system, the embodiment further provides a display terminal for displaying the image of the target object, and the display terminal may be of various types, such as a mobile phone, a tablet computer, a PC, and the like.
In a specific embodiment, in the underwater ghost imaging system based on edge calculation of the present application, the converging lens is installed right in front of the light intensity detector, so that a focal point of the converging lens is at the center of an effective detection area of the light intensity detector.
The embodiment of the underwater ghost imaging system based on edge calculation provided by the embodiment of the present application is as follows.
Referring to fig. 2, 3, 4, and 5, in an embodiment of the present application, a method for reconstructing a deep learning image includes:
It should be noted that the present embodiment uses fourier sinusoidal speckle Wk=Wk(x, y) is loaded onto the light modulation device and the illumination beam is modulated by Fourier sinusoidal speckle. The modulated light beam is projected onto target object by projection lens, and the reflected light beam is converged by convergence lens to light intensity detector, and its light field intensity value SkRecorded by the light intensity detector, as follows:
Sk=∫∫Q(x,y)Wk(x,y)dxdy
in the above equation, Q (x, y) is the objective function, K is 1,2,.., K is 500, which is the total number of fourier sinusoidal speckles, and (x, y) is the pixel coordinate. The pixel size of fourier sinusoidal speckle is 64 × 64.
And 500 speckles constitute a speckle sequence W, denoted as:
W=[W1,W2,...,W500]
corresponding to the speckle sequence W, a one-dimensional light field intensity sequence S can be obtained:
S=[S1,S2,...,S500]。
and 102, detecting the light intensity of the light rays converged to the light intensity detector by the converging lens to obtain a light intensity sequence.
It should be noted that the light is the light reflected by the target object in step 101, and the light is converged onto the light intensity detector by the converging lens, and the light intensity detector collects the light intensity sequence of the light.
After 30000 times of training, the network model parameters after training can be obtainedThe training times can also be set by those skilled in the art according to practical situations, and are not limited herein.
Need to make sure thatIllustratively, the data setThe manufacturing process comprises the following steps:
200000 different gray images with 64 × 64 pixel size are taken as original target object images, and the jth original target object image G is subjected tojObtaining a one-dimensional light field intensity sequence by calculation
In the formula (I), the compound is shown in the specification,denotes the nth light field intensity value, N1, 2.
It should be noted that the neural network model is expressed as:
Q(x,y)=Ψ(S,Φ);
the above formula represents that a mapping relation is established between a one-dimensional optical field intensity sequence S and an output image Q (x, y) by using a hidden function psi (·), and phi represents a neural network model parameter.
Wherein the neural network model parameters are expressed as:
in the formula, | | the luminance2Denotes 2-norm, λ ═ 0.6 is the regularization parameter, GiRepresenting the i-th original object image for training, DiIs corresponding to the object image GiThe number of training times is 30000, and the total amount of data in the data set is 200000.
Note that the loss function is expressed as:
in the formula, GiRepresenting the ith original for trainingImage of an object, DiIs corresponding to the object image GiI is the training times.
For a given underwater target object, a one-dimensional light field intensity sequence S is calculatedexpThen, a high-resolution and high-quality target object image can be reconstructed by using the trained neural network modelNamely, it is
In an optional embodiment, the deep learning image reconstruction method of the present application further includes: adding m-1 noise level graphs in the process of training the neural network model, wherein the pixel values of the b-1 noise level graphs are all Ub,b=1,2,...,m,UbIs at (0, 75)]Randomly generated within the range.
It should be noted that, the noise level map is added to obtain a denoising neural network model, which is used for removing noise in an image and improving the resolution and quality of the image.
The deep learning image reconstruction method is applied to the underwater ghost imaging system based on edge calculation in the embodiment, and due to the fact that Fourier sine speckles have orthogonality, an effective light intensity sequence capable of reconstructing a target image can be obtained under the condition of being lower than Nyquist sampling, and therefore the deep learning image reconstruction method can rapidly reconstruct the target image, removes noise interference in the image, enhances image details and improves image quality. The method and the device are beneficial to research and development of an underwater imaging technology and a deep learning technology. Therefore, the technical problems of large calculated amount, low imaging resolution and poor imaging quality in underwater ghost imaging are solved.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working process of the method described above may refer to the corresponding process in the foregoing device embodiment, and is not described herein again.
The terms "first," "second," "third," "fourth," and the like in the description of the present application and in the above-described drawings are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the application described herein are, for example, capable of operation in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
It should be understood that in the present application, "at least one" means one or more, "a plurality" means two or more. "and/or" for describing an association relationship of associated objects, indicating that there may be three relationships, e.g., "a and/or B" may indicate: only A, only B and both A and B are present, wherein A and B may be singular or plural. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship. "at least one of the following" or similar expressions refer to any combination of these items, including any combination of single item(s) or plural items. For example, at least one (one) of a, b, or c, may represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", wherein a, b, c may be single or plural.
In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be substantially implemented or contributed to by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The above embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions in the embodiments of the present application.
Claims (10)
1. An underwater ghost imaging system based on edge calculation, comprising: the device comprises a light source, a light modulation device, a projection lens, a convergent lens, a light intensity detector and an edge calculation module;
the first end of the edge calculation module is connected with the light modulation device and used for sending speckles to the light modulation device;
the light source is: the light modulation device is used for emitting light to an effective modulation area of the light modulation device;
the light modulation device: the system comprises a projection lens, speckles, a lens holder, a projection lens, a reflector and a reflector, wherein the speckle is used for modulating light rays and reflecting the modulated light rays so that the reflected light rays are emitted to a target object in the water body after passing through the projection lens along the optical axis of the projection lens;
the converging lens: for converging the light reflected by the target;
the light intensity detector is: the light intensity information acquisition module is used for acquiring the light intensity information of the converged light;
and the second end of the edge calculation module is connected with the light intensity detector and is used for controlling the light intensity detector to collect light intensity information and acquiring an image of the target object through a deep learning image reconstruction algorithm according to the light intensity information.
2. The edge-computation-based underwater ghost imaging system of claim 1, further comprising: a cloud computing center; the cloud computing center is in communication connection with the edge computing module;
the cloud computing center: and the image of the target object is obtained through a deep learning image reconstruction algorithm according to the light intensity information when the pixel size of the image of the target object is larger than a preset size.
3. The edge-computation-based underwater ghost imaging system of claim 2, further comprising: a cloud storage platform; the cloud storage platform is respectively in communication connection with a cloud computing center and the edge computing module;
the cloud storage platform is used for: storing the image of the object.
4. An edge computation based underwater ghost imaging system according to claim 3, further comprising: the display terminal is in communication connection with the cloud storage platform;
the display terminal is used for: and receiving the image of the target object sent by the cloud storage platform, and displaying the image of the target object.
5. An edge computation based underwater ghost imaging system in accordance with claim 1, wherein said converging lens is mounted directly in front of said intensity detector such that a focal point of said converging lens is centered in an effective detection area of said intensity detector.
6. A deep learning image reconstruction method applied to the underwater ghost imaging system based on edge calculation according to any one of claims 1 to 5, comprising:
loading the speckle sequence on a light modulation device to modulate light, and projecting the modulated light to a target object through a projection lens;
carrying out light intensity detection on the light rays converged to the light intensity detector by the convergent lens to obtain a light intensity sequence;
by data setsTraining a neural network model, and optimizing a loss function by using an ADAM (adaptive dynamic analysis of moving average) function until the training reaches preset times to obtain parameters of the trained neural network model;
inputting the light intensity sequence into a trained neural network model, and outputting an image of a target object
Wherein the neural network model is represented as:
Q(x,y)=Ψ(S,Φ);
the above formula represents that a mapping relation is established between a one-dimensional optical field intensity sequence S and an output image Q (x, y) by using a hidden function psi (·), and phi represents a neural network model parameter.
7. The deep learning image reconstruction method according to claim 6, further comprising: adding m noise level graphs in the process of training the neural network model, wherein the pixel values of the b-th noise level graph are all Ub,b=1,2,...,m,UbIs at (0, 75)]Randomly generated within the range.
8. The deep learning image reconstruction method of claim 6, wherein the data setThe manufacturing process comprises the following steps:
200000 different gray images with 64 × 64 pixel size are taken as original target object images, and the jth original target object image G is subjected tojObtaining a one-dimensional light field intensity sequence by calculation
9. The deep learning image reconstruction method according to claim 6, wherein the neural network model parameters are expressed as:
in the formula, | · the luminance | |2Denotes 2-norm, λ ═ 0.6 is the regularization parameter, GiRepresenting the i-th original object image for training, DiIs corresponding to the object image GiThe number of training times is 30000, and the total amount of data in the data set is 200000.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110594841.XA CN113192154B (en) | 2021-05-28 | 2021-05-28 | Underwater ghost imaging system based on edge calculation and deep learning image reconstruction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110594841.XA CN113192154B (en) | 2021-05-28 | 2021-05-28 | Underwater ghost imaging system based on edge calculation and deep learning image reconstruction method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113192154A true CN113192154A (en) | 2021-07-30 |
CN113192154B CN113192154B (en) | 2023-05-23 |
Family
ID=76986314
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110594841.XA Active CN113192154B (en) | 2021-05-28 | 2021-05-28 | Underwater ghost imaging system based on edge calculation and deep learning image reconstruction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113192154B (en) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105807289A (en) * | 2016-05-04 | 2016-07-27 | 西安交通大学 | High-speed calculation correlated imaging system and method based on preset modulatable light source |
US20200311878A1 (en) * | 2019-04-01 | 2020-10-01 | Canon Medical Systems Corporation | Apparatus and method for image reconstruction using feature-aware deep learning |
CN111833265A (en) * | 2020-06-15 | 2020-10-27 | 北京邮电大学 | Ghost imaging image recovery scheme based on group sparse cyclic modulation |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111986118B (en) * | 2020-08-31 | 2023-06-16 | 广东工业大学 | Underwater calculation ghost imaging image denoising method and system with minimized weighted kernel norm |
-
2021
- 2021-05-28 CN CN202110594841.XA patent/CN113192154B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105807289A (en) * | 2016-05-04 | 2016-07-27 | 西安交通大学 | High-speed calculation correlated imaging system and method based on preset modulatable light source |
US20200311878A1 (en) * | 2019-04-01 | 2020-10-01 | Canon Medical Systems Corporation | Apparatus and method for image reconstruction using feature-aware deep learning |
CN111833265A (en) * | 2020-06-15 | 2020-10-27 | 北京邮电大学 | Ghost imaging image recovery scheme based on group sparse cyclic modulation |
Non-Patent Citations (4)
Title |
---|
HENG WU等: ""Sub-Nyquist computational ghost imaging with deep learning"", 《OPTICS EXPRESS》 * |
傅爽: ""基于定量相位成像和深度学习的相位体识别分析方法研究"", 《中国优秀博硕士学位论文全文数据库(硕士) 基础科学辑》 * |
冯维等: ""基于卷积神经网络的计算鬼成像方法研究"", 《光子学报》 * |
吴衡: ""机器视觉鬼成像理论与实验研究"", 《中国优秀博硕士学位论文全文数据库(博士) 信息科技辑》 * |
Also Published As
Publication number | Publication date |
---|---|
CN113192154B (en) | 2023-05-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102973242B (en) | Image processing equipment, image processing method, image processing system, program and recording medium | |
US20190319036A1 (en) | Method and system for object reconstruction | |
KR101893047B1 (en) | Image processing method and image processing device | |
CN110646810B (en) | Speckle optimization compressed sensing ghost imaging method and system | |
US20130335535A1 (en) | Digital 3d camera using periodic illumination | |
CN108895985B (en) | Object positioning method based on single-pixel detector | |
Singh et al. | Single image dehazing for a variety of haze scenarios using back projected pyramid network | |
CN106405546B (en) | Rapid correlation imaging system and method based on compressed speckles | |
JP2023521270A (en) | Learning lighting from various portraits | |
JP6867645B2 (en) | Image processing equipment, methods, and programs | |
AU2020408599A1 (en) | Light field reconstruction method and system using depth sampling | |
CN113393548A (en) | Non-vision field imaging method based on depth residual error neural network | |
JP2022524923A (en) | Imaging systems and methods via scattering media | |
CN113192154A (en) | Underwater ghost imaging system based on edge calculation and deep learning image reconstruction method | |
US11758297B2 (en) | Systems, methods, and media for high dynamic range imaging using single-photon and conventional image sensor data | |
CN111788824A (en) | Image processing apparatus, imaging apparatus, and image processing method | |
CN108981782A (en) | A method of it is realized using mobile phone and calculates relevance imaging | |
Czajkowski et al. | Two-edge-resolved three-dimensional non-line-of-sight imaging with an ordinary camera | |
BR et al. | Machine Learning Application for Evidence Image Enhancement | |
Li et al. | Omnidirectional Ring Structured Light Noise Filtering Based On DCGAN Network And Autoencoder | |
CN114627520A (en) | Living body detection model training method, system, equipment and storage medium | |
CN117876837B (en) | Near infrared single-pixel imaging method and system based on depth expansion network | |
CN109120918A (en) | Drop shadow effect's self-appraisal system and method based on robot | |
CN114119781B (en) | Non-visual field imaging method based on light intensity cross correlation | |
CN113284071B (en) | Periopsis imaging method and system based on human retina mechanism simulation |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |