In this paper, a new parametric model continuous in time and space is introduced to analyze trajectory data in ecology. This model assumes that the trajectory of an individual is a solution to a stochastic differential equation. The drift of this equation is defined as the gradient of a potential map which is a mixture of an unknown number of Gaussian shaped functions. Each component of this mixture may be understood as an attractive field characterizing the propensity of the individual to move to certain regions. The parameters of this model are estimated using an Monte Carlo Expectation Maximization algorithm based on the exact algorithm proposed by \cite{beskos:papaspiliopoulos:roberts:2006} to sample trajectories exactly distributed as the solution to the stochastic differential equation. The main advantage of this discretization free method is to be efficient even when the data are obtained at irregular times and at a slow rate. The performance of the proposed model and estimation procedure is illustrated using simulated data and true GPS positions of French vessels moving in the English Channel.