For energy saving and security in building, the information of occupant number of each zone is very important. This paper works on the estimation of the occupant number of zones in building by detecting the entrance and leaving events. In this paper, we first formulate the problem under an assumption of Markov Chain, and basing on the theoretical analysis of the model, we propose a method of occupant distribution estimation, which can be implemented distributively. The method counts occupant by detecting the entrance and leaving events of zones in real time and uses the prior information of the occupant's entrance and leaving events in each zone to reduce the estimation error, which increases the accuracy of the estimation of occupant distribution in building. Numerical experiments including simulation and field test demonstrate the performance of the method.