Estimation of crane load parameters during tracking using expectation-maximization

In this paper we develop a method for estimating static model parameters required for marker-based visual tracking of crane loads. The static model parameters describe both the location of markers, which are used for visual tracking, and the mass center position of the crane load, which is used in the dynamical model for predicting the motion of the crane load. The proposed method is based on a particle smoothing algorithm for estimating the state variables of the crane load and an expectation-maximization algorithm for estimating the static model parameters. A benefit of the proposed method is that the static parameters can be estimated independently, which reduces the computational burden compared to estimating them together. The proposed method is verified in a simulation and in a laboratory experiment.