In many navigation application scenarios, e.g., emergency rescue, mall shopping, medical guiding, and self-guided tours, the floor plan is regarded as a prior. As a result, these navigation systems are short of scalability and robustness. Focusing on this issue, this paper proposes a dynamic landmark-based method for constructing the indoor map, which mitigates the dependence to a prior with the state-of the-art. First of all, the coarse-grained trajectories are obtained by dead reckoning, meanwhile the static indoor landmarks (such as stairs, elevators, escalators) are captured and dynamic indoor landmarks (three kinds of inflection points) are identified by inertial data. Secondly, the trajectory are constructed calibration and integration is leveraged to implement the indoor pathway mapping. Finally, constructing extensive experiments by means of the prototype to evaluate the proposed design. The analysis results demonstrate that the maximum error and average error are 1.90 m and 0.90 m respectively.