Optical splitting method based on neural network and broadband filter modulation

Abstract

The invention discloses a light splitting method based on neural network and broadband filtering modulation and mainly solves the problem that a least square method cannot be used for accurate splitting when noisy in the prior art. The implement steps comprises: 1) designing a set of modulation sheets as a filter and mounting to detectors to form a detector matrix; 2) detecting a known incident light source by using the detector matrix to acquire the data detected by each detector, and preprocessing; 3) constructing a neural network; 4) training the neural network by using the known incident light source spectrum and the preprocessed detection data; and 5) calculating the spectral data of the unknown light source to complete splitting right after detecting an unknown light source by usingthe detector matrix, and inputting the detected unknown light source data after being preprocessed to the trained neural network model. According to the light splitting method based on neural networkand broadband filtering modulation, the spectral multiplexing characteristics of the broadband filtering and the learning ability of the neural network are fully fused, so that the splitting accuracyis improved, and the method can be applied to site detection of the spectra.

Description

technical field [0001] The invention belongs to the field of spectrum technology, and in particular relates to a spectroscopic method, which can be used for spectrum detection and substance detection. Background technique [0002] The spectrum of a substance is called the "fingerprint" of the substance, and it is widely used in resource exploration, crop yield estimation, disease diagnosis and other fields. However, the current spectroscopic method for obtaining spectra requires huge equipment and complex optical paths, which seriously restricts the application of spectroscopy in on-site detection. Therefore, it is urgent to study efficient and smart spectral spectroscopic methods. [0003] The existing spectroscopic techniques mainly include prism spectroscopic, grating spectroscopic and interferometer spectroscopic. in: [0004] Prism splitting is to use the difference in refractive index of different wavelengths of light in the medium to separate them in space. Since t...

AI Technical Summary

Problems solved by technology

However, this method has obvious shortcomings: first, the quantum dot colloid may evaporate, causing a change in transmittance, reducing accuracy, and resulting in a shorter service life; second, the transmittance needs to be given in the post-processing, and the transmittance The equations must have a solution, but errors will inevitably occur during installation, resulting in changes in the transmittance of the installed filters, greatly reducing the calculation accuracy; third, when the number of filters is less than the number of bands, it cannot be solved accurately

Images