Evaluation of image matching techniques for optical fiber specklegram sensor analysis.

ABSTRACT

A quantitative study of image matching techniques applied to fiber specklegram sensor analysis is presented. The fiber status is modulated by a microbending transducer, so the output speckle field can be correlated to the input displacements. Once acquired and preprocessed, the specklegrams' variations were evaluated according to seven approaches. Although the average intensity did not provide reliable information per se, the correlation and sum of differences methods yielded ∼11 mm-1 and ∼14 mm-1 sensitivities, respectively, within a ∼0.06 mm range and low linearity and hysteresis errors, with enhancement possibility by intensity level cancellation. Moreover, the phase-only correlation and the mutual information metrics provided very high sensitivities (22 mm-1 and 120 mm-1, respectively) for a <0.02 mm range, making these techniques suitable for detecting subtle variations in the fiber status due to physical or chemical stimuli.