This paper proposed a 3D model feature lines extracting method guided by user's sketch.There are two main contributions of our method:on the one hand

a conditional random fields(CRF)-matching model with geometric features and shape similarity between stroke points and candidate feature points is established

and the user drawing purpose is combined with the model feature lines extraction effectively;on the other hand

a topological similarity based metrics is proposed to evaluate the rationality of extracted feature lines

and an iterative optimization process is applied to obtain global optimized model feature lines.The experimental results show that our method is efficient to extract reasonable 3D model feature lines which are similar to sketch strokes on shape and topological properties.