Abstract: Overlapping community structure can ensure the smooth and efficient network information transmission. However, the existing detection algorithm based on overlapping communities can not be effectively applied to the delay tolerant networks (DTN) with discontinuous connectivity. In order to accurately detect the community structure of DTN, This paper proposes a new algorithm based on the node similarity to detect the overlapping community. Firstly, the fuzzy relation between nodes is initially determined by using the similarity degree between nodes. Then, by computing node entropy, the communities which nodes belong to can be determined. The numerical results show that the proposed detection mechanism compared with classic community detection mechanism can significantly improve the precision of detection.