The co-location pattern mining discovers the subsets of spatial features which are located together frequently in geography. Instance independence has been taken as a major assumption in the co-location mining based on prevalence framework. However, in real-world spatial data sources, spatial instances are more or less correlated with each other. Prevalence-based framework can do limited work in spatial instance correlated analysis. For reducing the co-location mining results and promoting the usability of co-location patterns, this paper proposed a new framework to identify the co-location patterns with key features and extract the key features from a large collection of prevalent co-location patterns. We first give the definitions of significant co-location patterns; secondly, we design a series of metrics to evaluate significance of co-location patterns and extract the key features; thirdly, an efficient algorithm is proposed to mine the significant co-location pattern with key features. The experiments evaluate the method both on real data sets and synthetic data sets. The results show that our method can effectively identify the significant co-location patterns with key features.