Using moments to reduce object recognition to a one-dimensional search

The three-dimensional affine transformation of an object is recovered by using second- and third-order moments. Using moments eliminates the need for feature detection. This technique should be more robust than other methods using higher-order moments. The moment equations containing the parameters are solved by successively zeroing various moments. This technique requires finding the minimum of a multiple-valued function defined for angles in the interval (0, pi ). This result reduces the recognition of objects having different scales, orientations, and shears to a one-dimensional search along a finite interval. In tests, this method successfully recovers the affine transformations of objects.<<ETX>>