HIGH-ACCURACY MULTIVIEWPOINT STEREO MEASUREMENT USING THE MAXIMUM-LIKELIHOOD METHOD

We propose a new type of highly accurate multi-viewpoint stereo measur ement using the maximum-likelihood method. This method is used for mea suring positions of facilities, such as telegraph poles on a road. In this method, a normal distribution is assumed for a measurement error on stereo image planes. First, the probability density functions of th e measurement errors at the multi-viewpoints are combined. Next, the c ombined function is expressed using both real-space coordinates and co ordinate transformation parameters, which give relative viewpoint posi tions, and the most probable coordinates are estimated from a conditio n that minimizes the function. Then, the best viewpoint positions, whi ch minimize the area of the combined error ellipse, are investigated b y analyzing the variance included in the estimated coordinates, and th e relationship between the number of viewpoints and the variance of th e best viewpoint positions is obtained. Finally, these theoretical res ults are confirmed experimentally by measuring outdoor facilities.