An evolutionary model for optimizing sensor pose in object motion estimation applications

An evolutionary control paradigm for vision-system pose planning for object motion estimation is proposed. The control of the vision system is embedded in the motion estimation process so as to adapt to the dynamic object motion behavior. A Kalman filter is employed as the motion estimator. In the Kalman filter formulation, a noise influence matrix is introduced to model the influence of vision system parameters on the measurement uncertainties. The estimation uncertainties in the Kalman filter formulation arc represented in the form of a Riccati equation. This equation describes the estimation uncertainties as an evolution process that is controlled by the vision system parameters. The control task is formulated as an optimization problem. A novel transformation of the vision system parameters is developed to relax the computational complexity of the optimization process. A hybrid genetic algorithm is proposed to search for the optimal vision system pose that is occlusion free. A series of experiments are conducted to evaluate the performance of the proposed object motion estimation model.