Abstract: Principle and method of correcting nonlinear error of sensors based on the support vector machine is given. The algorithm can identify and know the contrary model characteristic of sensor correctly only according to the sample, having no use for any priori knowledge about contrary model function. It also convert original problem into protruding quadratic optimizing of question. The algorithm can ensure that extreme solution is optimal and have ability of common. In the end, through using in application of wet-capacitance sensor error correction, the algorithm can make better result.