Quantifying human mobility patterns is intensively investigated by scientists from computational social science, statistical physics and complex science. In last decade, mobile phone data provide an unprecedented tool for capturing individuals’ travel activities in space and time. However, its nature of sparsity in time and imprecision in space imposes significant bias upon the derived mobility patterns. This research proposes two efficient techniques to cope with this issue. First, we implement an activity-location and travel-OD identification method to reconstruct reliable trajectories from call detailed records for mobile users. Second, we introduce the approximate entropy, which is superior to conditional entropy, for quantifying the regularity of individuals’ consecutively visited locations. With a case study in Harbin, the proposed approaches enable us to uncover meaningful patterns of urban mobility in terms of frequently and consecutively visited locations.