A fiber optic refractive index sensor with extremely high dynamic range utilizing deep learning-speckle pattern imaging

In this paper, a tapered optical fiber sensor with high precision measurement capability in a wide dynamic range has been proposed and experimentally demonstrated. Speckle pattern images obtained from the fiber end were trained with a convolutional neural network, and the measurement process was carried out through a deep learning algorithm. The sensing region of the optical fiber with a diameter of 125 mu m used as the sensor was reduced to a waist diameter of 10 mu m along a length of 2.25 cm. The sensor exhibits highly sensitive measurement capabilities with a resolution of up to 1.34x10-5 within the refractive index range of 1.3322-1.4214. By utilizing the deep learning-based sensor structure with real-time measurement capabilities, refractive index measurements can be made in many fields such as chemical, biosensor, space and aviation fields.