The remaining useful life (RUL) prediction of rolling bearing is significant for proactive maintenance of equipment, and selecting the features which can accurately reflect the performance degradation process as the inputs of the life prediction model is the premise of accurate RUL prediction. A novel method based on kernel principal component analysis (KPCA) and Weibull proportional hazard model (WPHM), is proposed to assess the reliability and predict the RUL of the rolling bearing. High relative feature set is constructed by selecting the effective features through extracting the time domain, frequency domain and time-frequency domain features of lifetime bearing. The kernel principal components (KPCs) which can accurately reflect the performance degradation process are obtained by KPCA. Then the KPCs are used as the covariates of WPHM to assess the reliability and predict the RUL. An example of bearing test is provided to demonstrate that this method can accurately assess the reliability and predict the RUL to provide timely maintenance resolution. Meanwhile, as the relative features are extracted, the differences in manufacturing, installation and working condition of the same type bearings are reduced, which enhances the practicability and stability of the method.