There is a nonlinear relationship between the output of the moisture instrument weighing sensor and the weight of drying objects, and the working environment temperature change will also make sensors’ nonlinear output for the loss on drying method. After analyzing the nonlinear output mechanism of the strain gauge sensor, the paper presents a new weighing sensor’s nonlinear compensation method based on least squares support vector machine (LSSVM) and mutated modified particle swarm optimization (PSO). Firstly, this method uses noise covariance variable kalman filter algorithm to filter data to reduce the noise influence. Then, establishes a regression model based on LSSVM is established for filtered data and working temperature. Finally, a mutated PSO algorithm is applied to optimize the model parameters. The experimental results show that the accuracy of the 220 g/0.001 g moisture instrument with the proposed method is much higher than the national verification regulation standard. Moreover, the model has excellent generalization ability with small samples.