In reset control systems, the reset law determines reset values at reset times. This paper considers the observer-based reset law design for uncertain systems with both nonlinear uncertainties and time-varying parameter uncertainties. A model predictive strategy is proposed to design the reset law by minimizing a quadratic performance index. The optimization problem is transformed into a linear matrix inequalities (LMIs) formulation, and the reset law is solved by using an LMI technique. The proposed approach is applied to a typical continuous stirred-tank reactor system to show the effectiveness of the obtained results.