Deep neural network-based scattering medium penetrating target positioning and reconstruction method

Abstract

The invention discloses a deep neural network-based scattering medium penetrating target positioning and reconstruction method, which is based on DINet and can be used for simultaneously predicting depth information and reconstructing a target image from a single speckle pattern. Image data of an experiment and a distance between an object and a scattering medium are collected by utilizing systemconfiguration; a speckle pattern statistical model system generated at different positions is configured to be used for collecting image data of an experiment and the distance between an object and ascattering medium; and the speckle pattern passes through the dual-channel network, generates a depth prediction value through the positioning channel network, and performs image restoration and reconstruction through the imaging channel network. According to the method, the multi-task challenge can be effectively solved, and multiple pieces of physical information can be obtained under the complex scattering condition. By utilizing a multi-task total loss function, the method has stronger constraint force on learning and training of the network, so that multi-task cooperative training has better performance in positioning tasks, and the positioning precision and the imaging quality are improved.

Description

technical field [0001] The invention belongs to the field of machine learning and image reconstruction, in particular to a target positioning and reconstruction method through a scattering medium based on a deep neural network. Background technique [0002] The detection of hidden objects through scattering media has broad application prospects in many fields such as atmospheric optics and biophotonics. However, scattering will cause interference and degradation to the original information of the observed object, which limits the imaging and measurement of the object. [0003] At present, some traditional physical methods have been proposed to solve the scattering imaging problem, but they only restore the speckle image and do not contain other physical information. [0004] Ranging and localization of targets in scattering environments is crucial for atmospheric or biological applications. So far, there are some techniques to obtain the depth information of hidden objects...

AI Technical Summary

Problems solved by technology

However, due to scattering interference and model capabilities, it has not been possible to use neural networks to effectively measure the spatial location information of hidden objects with or without prior information.

Images