A low-cost fiber-optic temperature sensor utilizing integrated sensing and leaky mode specklegram analysis

Fiber-optic sensors are highly sensitive to physical, chemical, and biological variations, making them essential for precision measurements in complex environments. Achieving cost-effectiveness, miniaturization, and scalability simultaneously, however, remains challenging. To address this, an integrated fiber-optic sensing approach is presented. A tapered fiber segment is employed to generate leaky-mode speckle patterns, with geometric parameters and a thermosensitive coating optimized to modulate leakage intensity and spatial distribution in response to temperature. Experimental results confirm a direct correlation between specklegram features and temperature across 5 °C–80 °C. Temperature inversion is performed using a deep learning model based on the Xception convolutional neural network. The modular system design allows adaptation to other sensing modalities through material or detector modification while preserving core functionality. This approach overcomes traditional trade-offs between cost, size, and scalability, demonstrating robust and versatile performance for practical applications.