In this paper, we focus on the problem of detecting communities, where users have similar characteristics in both social relationship and check-in behavior in location based social network (LBSN). Contrast to traditional social network, LBSN not only contains users' online social relationship information but also large amounts of location information generated by users' check-in behavior, which inevitably brings challenges to community detection in LBSN. To do this, based on abundant knowledge hidden in LBSN, we first define multiple kinds of knowledge-aware similarities as well as corresponding calculation methods. Then, we propose a method called as Multi-dimensional Similarity Information Fusion for Community Detection (MFCD) on the basis of an improved K-Means algorithm. Meanwhile, we establish a set of evaluation metrics to measure community quality from different perspectives, specifically for LBSN. Finally, we conduct a series of experiments to demonstrate the excellent performance of our proposed community detection method for LBSN. [ABSTRACT FROM AUTHOR]