An online optimization method for improving the performance of a range-gated imaging system is introduced by means of adaptively adjusting the operation parameters of modules. The eye pattern parameter is treated as a random variable varying with the tunable parameters of the system, and the Gaussian process regression method is applied for fitting this variation relationship. In addition, the parameter optimization procedure is also involved in the learning process. Thus the variable characteristic of a random process in the high dimensional parameter space is learnt rapidly, and simultaneously the system parameters are optimized. The experimental results show that the proposed method can be used to optimize the system parameter configuration online and improve the range resolution. The dependence of the range resolution on the configuration parameters of modules is illustrated, which provides a basis for the design of an imaging system.