To solve the deficiencies existing in the traditional target skeleton representation based on graph structure. According to the technique of stochastic grammar used to describe the stochastic model, a new method of target skeleton representation based on stochastic grammar is proposed. The core of the method is to solve the noise-sensitive of skeleton using the anti-jamming capability of stochastic grammar, embedding skeleton weight in the stochastic grammar to catch at the main topological characteristics of the objectives. Compared to the traditional graph structure, this method has better anti-jamming capability of target skeleton to prevent the loss of the main shape for objective.