In this study, an adaptive remaining useful life (RUL) prediction method is developed based on the expectation maximization (EM) algorithm to solve the problems of insufficient prior information and lack of historical data with respect to the newly developed photoelectric products. Currently, two common problems are associated with majority of the existing RUL prediction methods. First, there is an underlying assumption in degradation modeling that a random parameter estimated at the current time is exactly equal to the posterior estimation of the random parameter at the previous time. Second, the historical degradation data are assumed to be available for parameter estimation based on which the initial model parameters can be determined for multiple photoelectric products of the same type. The RUL prediction accuracy is limited by data availability. Herein, we construct a novel degradation model under the state space model framework and derive the analytical form of the RUL distribution. Subsequently, we propose an adaptive parameter prediction method based on the EM algorithm to overcome the problems of insufficient prior information and lack of historical data. Finally, we conduct an experimental study with respect to the actual degradation data of a GaAs laser and fiber-optic gyroscope to denote that the proposed method improves the RUL prediction accuracy and can be effectively applied to the newly developed photoelectric products.