With the development of urbanization, the number of cars has been proliferating fast. Urban transportation provides convenience for city resident, meanwhile, it also brings many challenges, such as causing a difficulty for taking a taxi. In this paper, main work is composed of two parts: first, we propose a method to estimate the probability and waiting time for a vacant taxi at a given time and place, second, we provide an recommendations for passengers of where to wait for a taxi. In order to complete the main work, we firstly preprocess the Large-scale Taxi GPS Traces and generate the road characteristic index. Then we use NPP (Nonhomogeneous Poisson Process) to predict and recommend. In the experimental part of the paper, based on the large-scale real-world GPS traces dataset, we compare our method and other related advanced methods and the results show that our method has more advantages in accuracy and efficiency.