In the gas-liquid two-phase slug flow, the wet-wall fraction and the tilt Angle of the bullet are the key to study the liquid holdup and the elastic characteristics of the gas-elastic section. In this paper, based on optical imaging technology, a vision sensor test system with high speed camera method is designed, and the gas elastic characteristic parameters of gas-liquid two-phase slug flow are measured. The machine learning model is used as a new method to predict the characteristic parameters of multiphase flow. The data modeling of the inclination Angle of the aeroelastic head is studied, and the apparent rough surface (MARS) model applicable to stratified smooth flow, wave flow and annular flow as well as the flow state mechanism in the aeroelastic and liquid film region of slug flow are analyzed. The parameters are optimized based on the MARS model. The method can be applied to the solution of wet wall fraction and liquid holdup at slug film. The results show that the mean absolute percentage error (MAPE) of the model prediction result of the polynomial model of warhead inclination α by data fitting is 8.87%, and the relative error of 94.6% is within the error range of ±20%. The mean absolute percentage error (MAPE) of the prediction results of wet wall fraction at the liquid film of the horizontal pipe slug flow proposed by the modified MARS model is 9.42%, and the relative error of 96.1% of the prediction results is within the error range of ±20%. The MAPE of the prediction model is about 8.04%, and the relative error of 93.4% is within the error range of ±25%.