Insufficient light or strong light reflection in the tunnel can affect visual inspection (such as cameras and laser scanning), causing the crack data collected by such sensors to show discrete jumps and making it difficult to extract spatial continuous features, resulting in a low diagnostic accuracy rate. To this end, distributed optical fiber sensing is introduced and an intelligent diagnosis system for tunnel lining cracks is designed. The hardware part of this system adopts distributed optical fiber sensors to achieve high-precision strain field monitoring. It combines optical signal demodulation instruments and acquisition devices to construct an anti-interference signal chain. Real-time data processing is defined through the core processor pins, forming a full hardware closed loop from the sensing layer to the computing layer. The software part adopts a three-stage fusion denoising algorithm. It separates the high and low frequencies of the crack data collected by the distributed optical fiber sensor through multi-scale wavelet decomposition, and combines the low-frequency subband to reconstruct high-quality denoising data. Aiming at the problems of feature decoupling and spatial correlation that still exist in the denoised data, a complete characterization system of cracks is constructed from three dimensions: geometric features, mechanical features, and material deterioration features. This effectively distinguishes the cross-scale continuous features of real cracks from the discretization jumps caused by environmental interference, and extracts the spatial continuous features of crack data. Crack classification diagnosis and severity grading are realized by combining the dynamic weight support vector machine. The experimental results show that after denoising by the designed system, the maximum peak signal-to-noise ratio of the tunnel lining crack data reaches 78dB, the extracted spatial curvature index tends to be consistent with the actual spatial curvature index, and the classification diagnosis accuracy rate of tunnel lining cracks reaches 99.6%.