An Encoding Optimization Method for Spectral Imaging Systems Based on Optimization-Inspired Neural Networks

Abstract

The invention discloses a coding optimization method of a spectral imaging system based on an optimized heuristic neural network, belonging to the field of computational photography. The coding optimization method of the spectral imaging system based on the optimized heuristic neural network comprises the following steps: establishing a forward propagation model of the spectral imaging system, implementing the forward propagation model by using a network, and constructing a coding aperture optimization network; constructing a hyperspectral image reconstruction network based on optimization heuristic and considering both the hyperspectral image spatial correlation and the spectral correlation; making a training set; configuring parameters required for hyperspectral image reconstruction network training; training the hyperspectral image reconstruction network; establishing a connection between the coding aperture optimization network and the hyperspectral image reconstruction network toconstruct a joint network; configuring the parameters required for joint network training; training the joint network; extracting a coding template obtained after training, and completing modulation from the hyperspectral image to a two-dimensional compressed image based on a CASSI system imaging process; and reconstructing the target hyperspectral image block by block using the trained hyperspectral image reconstruction network.

Description

technical field [0001] The invention relates to a coding optimization method for a spectral imaging system, in particular to a method capable of acquiring high-quality hyperspectral images, and belongs to the field of computational photography. Background technique [0002] Unlike traditional RGB imaging or panchromatic imaging, spectral imaging captures the scene as a three-dimensional tensor that more finely samples the spectral information at each pixel location of the scene in the spectral dimension. The hyperspectral images obtained by spectral imaging are rich in spectral information, which makes them more advantageous than traditional imaging techniques in remote sensing, medical imaging, visual inspection, sewage detection, vegetation research, atmospheric monitoring and other fields. are being used more and more widely. [0003] Since hyperspectral images are three-dimensional tensors, and currently used imaging sensors are two-dimensional, spectral information mus...

AI Technical Summary

Problems solved by technology

[0009] Aiming at the problem that the existing imaging method does not consider the imaging process and the computational reconstruction process at the same time, the technical problem to be solved by the coding optimization method of the spectral imaging system based on the optimization-inspired neural network disclosed by the present invention is: to improve the coding aperture by optimizing the coding aperture The hyperspectral image reconstruction quality of the Coded Aperture Snapshot Spectral Imager (CASSI) system, while ensuring high spatial resolution and hyperspectral fidelity of the reconstruction results, improves the efficiency of hyperspectral image reconstruction and expands the hyperspectral image scope of application

Images