Regional co-location pattern (RCP) mining is a key subfield within spatial co-location pattern mining, focuses on the identification of prevalent co-location patterns within local regions. The RCPs reflect the local relationships between spatial features, which have a wide range of practical application in human daily production and life. Existing RCP mining methods cannot effectively identify the distribution regions of RCPs driven by human activities, and face difficulties in determining a suitable prevalence threshold for mining prevalent RCPs in different instance distribution regions. To address these problems, a novel top-k RCP mining framework based on weighted Dirichlet diagram is proposed. The proposed framework first obtains the distribution regions of human activity-driven RCPs through the weighted Dirichlet diagram, and then efficiently detects the top-k prevalent RCPs in those areas. In addition, in order to solve the efficiency problem when facing large datasets, a parallel mining scheme is proposed to speed up the RCP mining process. Finally, based on the above methods, a user-friendly demonstration system is developed to promote the application of RCP mining technology in practice.