This paper uses computer image recognition algorithm to predict the dynamic modulus of asphalt mixture. The main parameters of asphalt mixture porosity, effective asphalt content, asphalt viscosity, load frequency, aggregate sieve mass fraction (and aggregate pass rate on 0.075 mm sieve openings, were established based on gene expression programming algorithm). Asphalt mixture dynamic modulus prediction model. Taking dynamic stability as the evaluation index, comprehensively considering its residual stability, creep rate, friction factor, water seepage coefficient and porosity, etc., to propose a functional asphalt pavement mineral material ratio the best optimization plan. Compare the prediction model with Witczak 1999 model, computer vision image recognition prediction model and artificial neural network model).