Abstract: This paper gives a classification of symbol fuzziness and proposes an approach based on 2-neighborhood local structures to recognize fuzzy symbols in vector drawings.By using the idea of local structures, the approach first represents the symbol to be searched and the target vector drawing as two sets of 2-neighborhood local structures, and defines the similarity distance between two local structures, then select out the similar structures according to the similarity distance.Finally, the approach clusters the similar structures based on their space distance and scale factor to find fuzzy symbols in the vector drawing.The results show that the approach can deal with fuzzy symbols with duplicate structures or stretching and gives high recognition accuracy with different kinds of fuzziness.