Neural Network Control of Liquid Crystal Spatial Light Modulator for Spectral Synthesis of Light

Abstract

If illuminating light with expected spectral distribution can be synthesized, it will be a great help in many applications, such as computer vision, pattern recognition, industrial quality control, and remote sensing. A light source system containing a liquid crystal spatial light modultor (LCSLM) is proposed for this purpose. A supervised neural network with a single layer was used to control the system. After training the neural nrtwork with data acquired from the system, it gives the data which should be written to LCSLM in order to synthesize illuminating light having certain spectral distribution. Eight spectral distributions were synthesized to test the performance of the system, and the average synthesizing error was 12%.