Entropy analysis of optical fiber specklegram sensors

Abstract

Optical fiber specklegrams originate from the interference between several guided light modes. Since these granular patterns are sensitive to changes in the fiber status, it is possible to retrieve an external stimulus by analyzing their morphology and spatial distribution. This paper evaluates the first and second-order entropies as metrics to interrogate fiber specklegram sensors. Firstly, the refractive index surrounding a no-core fiber is varied to modulate the number and average size of speckles. Secondly, a microbending transducer driven with controlled displacements creates deviations in the output speckle field. Then, the entropy data for each case is compared with Law’s energy texture and the zero-mean cross-correlation analyses, respectively. Conversely to the Shannon entropy, the joint and relative entropies process the spatial information of the image and corroborate the theoretical and experimental speckle count, yielding ∼0.02 refractive index resolution. Moreover, the relative entropy of the ratio between test and reference speckle field images agrees with the correlation coefficient, achieving ∼0.5 µm resolution. The results show that the second-order entropy is effective for interrogating fiber specklegrams, figuring as a straightforward alternative to the widespread texture processing and deep learning approaches.