Liquid film plays a crucial role in void fraction, friction pressure drops, the momentum and heat transfer in the two-phases flow. The film thickness measurement experiments of annular flow at four pressure conditions have been conducted by near infrared (NIR) sensor. The signal is processed by variational mode decomposition (VMD), of which parameters are optimized by sparrow search algorithm (SSA). Envelope spectrum and Pearson correlation coefficient (PCC) judgment criteria is adopted for signal reconstruction, and the value of liquid film thickness is obtained. The effect (Such as flow rate, pressure, entrainment, etc) of liquid film thickness are analyzed theoretically. The characterization parameters Weg”, Wel, Nμl and Xmod have been extracted and optimized, and then a new average liquid film thickness correlation is proposed. The laboratory results indicate that mean absolute percentage error (MAPE) of the predictive correlation is 4.35% (current data) and 12.02% (literatures’ data) respectively.