All-Optical Physical Field Recognition Via Sparse Feature Extraction

Abstract

Optical computing has been proven to have the ability to process information with ultra-high speed. Here, an all-optical feature extraction system via sparse representation (AFE-SR) is introduced. The AFE-SR, which is achieved by multiple diffractive optical elements (DOEs), can realize the recognition of generated physical fields with the speed of light. The sparse representation simplifies the target and improves the recognition accuracy. With the mathematical analysis principle of sparse optical features extraction and optical integration, the identification accuracy of the generation of physical fields is 100% in 2100 frames of the experimental videos. The application field of optical computing systems is extended to state recognition.