Automatic sensor registration using stochastic optimization methods

Sensor image registration is one of important problems in sensor fusion, it is the task of finding the correct mapping of one sensor image onto another. In this paper, the sensor image registration problem is approached as a optimization problem, then an appropriate fitness function is proposed to evaluate the mapping parameter set and several optimization methods (genetic algorithms, simulated annealing, hybrid strategy GASA) are adopted to solve this problem automatically. The experiment indicates the feasibility of the registration methods which are insensitive to noise. Among them, GASA shows good performance because of its rapid convergence.