Optical flow estimation using data fusion

CARLOS SORIA; RICARDO CARELLI; MARIO SARCINELLI FILHO

Bauru, SP, Brasil

Simposio; VI Simposio Brasileiro de Automacao Inteligente, SBAI´2003; 2003

Sociedade Brasileira de Automatica

In this work an approach is presented for the calculation of the optic flow using a data fusion method. The method searches for the value of the optical flow vector with the smallest variance. This is not a new algorithm for optical flow calculation. Rather, a well known method is used for estimating such vectors, in terms of a statistical model (mean and variance values). Then, such statistical model is used in connection to a data fusion engine, so that a better estimate is then generated for such vectors. ´l´he data fusion engine used is a Kalman filter, which guarantees the optimal data fusion. Thus, the variance of the obtained optic flow estimate will be smaller than the smallest variance entering the data fusion engine. Ths way, a very good estimate of the optic flow is obtained.searches for the value of the optical flow vector with the smallest variance. This is not a new algorithm for optical flow calculation. Rather, a well known method is used for estimating such vectors, in terms of a statistical model (mean and variance values). Then, such statistical model is used in connection to a data fusion engine, so that a better estimate is then generated for such vectors. ´l´he data fusion engine used is a Kalman filter, which guarantees the optimal data fusion. Thus, the variance of the obtained optic flow estimate will be smaller than the smallest variance entering the data fusion engine. Ths way, a very good estimate of the optic flow is obtained.