In order to achieve rapid and accurate estimation of stress relaxation parameters of tomato in grasping process, a method of estimating stress relaxation parameters of tomato based on BP neural network and genetic algorithm (GA) was proposed, which optimized the initial weights and thresholds of BP neural network. A three–element generalized Maxwell model, was used to characterize the stress relaxation characteristics of tomato and the sample data set was obtained by fitting. Then the BP was constructed with grasping force F, deformation D and action time t as inputs and relaxation parameters E, Ee and η as outputs. Genetic algorithm were used to optimize the initial connection weights and thresholds. The GA–BP neural network estimation model of the optimal parameters was obtained and applied to the extimation of stress relaxation parameters in the process of maniputator grasping. The results showed that the relative errors of stress relaxation parameters E, Ee and η were less than 15%, and tended to be stable. The model could be used to estimate stress relaxation parameters online.