Physics-driven scattering generalization imaging method and experimental method
An imaging method and a physical technology, applied in neural learning methods, design optimization/simulation, biological neural network models, etc., can solve problems such as limitations, limited recovery effect of scattering target information, limited imaging quality, and long time-consuming target recovery. The effect of expanding the scope of work
Active Publication Date: 2021-05-14
NANJING UNIV OF SCI & TECH
View PDF5 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Among them, the first three are computational imaging methods based on physical models. The recovery effect of the method based on wavefront shaping is also limited by the pixel size and number of modulation devices, and it is impossible to effectively restore and measure the scattering scene when it is not fixed; based on The specific target recovery of compressed sensing in the single-pixel imaging method is limited by the selected sparse base, and the target recovery takes a long time; the speckle correlation technology based on the memory effect can perform non-invasive imaging of the target behind the scattering medium, based on the optical The speckle correlation technology of the memory effect can restore the target behind the medium through a single frame of speckle information, but it is limited in principle by the scope of the optical memory effect, once the target cannot be effectively reconstructed, and the traditional phase recovery algorithm has no effect on the The recovery effect and imaging quality of scattering target information are also very limited
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0040] The present invention is described in further detail now in conjunction with accompanying drawing.
[0041] The physical-driven scattering generalization imaging method of the present invention, as shown in Figure 1(a)-Figure 1(c), specifically includes the following steps:
[0042] Step 1: The physical / data model built, see figure 2 ,
[0043] Step 1.1: Only use the traditional U-Net convolutional neural network structure and weaken the requirements for the network structure. The physical constraint network model consists of an autocorrelation physical constraint layer and a convolutional neural network layer. The physical constraint layer mainly includes speckle The autocorrelation calculation of I(x,y) is adjusted, and the speckle autocorrelation is obtained by the two-dimensional inverse Fourier transform of the energy spectrum
[0044] (3),
[0045] Equation (3) is the main operation of the physical constraint layer, which adjusts and reconstructs the speckle...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention relates to a physical-driven scattering generalization imaging method and an experimental method, which belong to the technical field of imaging analysis. The scattering generalization imaging method includes a built physical / data model, uses a speckle pattern collected by a camera, and uses a scattered The speckle correlation theory can be used as a universal physical principle to constrain and guide neural networks to perform generalized imaging in different scattering scenes. The experimental method includes theoretical analysis and system experiment, and this method is systematically discussed in building a specific system experiment. Improve generalization ability and generalization quality, and generalize more complex targets, etc. The organic combination of the physical model based on speckle correlation and the data model of deep learning greatly improves the generalized imaging effect and imaging range of the neural network model.
Description
technical field [0001] The invention relates to a physics-driven scattering generalization imaging method and an experimental method, belonging to the technical field of optical imaging analysis. Background technique [0002] Taking the wavelength of light as the unit of measurement, most objects and scenes in life are rough, so light scattering exists in all aspects of life. After the target information is modulated by a strong scattering medium, the original effective target information will be seriously degraded. Scattering degradation has brought many problems and troubles to real life and scientific research observation applications, such as traffic supervision in dense fog weather in life, effective monitoring of vehicles through changing scattering scenes, including car license plate number recognition, and even drivers Face information recovery is of great significance. At the same time, in order to obtain specific target information, scattering interference is als...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G06F30/20G06N3/04G06N3/08
CPCG06F30/20G06N3/08G06N3/045
Inventor 韩静柏连发郭恩来朱硕顾杰崔倩莹周晨寅
Owner NANJING UNIV OF SCI & TECH