A high precision localization scheme for profiled extended objects based-on Generalized Hough Transform (GUT) is proposed. The proposed localization scheme utilizes the edge points' gratitude orientations of profiled extended objects to build an R-table in frame t-1. When frame t is coming, we search the R-table according to the pre-computed edge points' gratitude orientations of profiled extended objects in current frame, generating an accumulator array. The coordination corresponding to the accumulator's peak value is regarded as coarse localization point in current frame. Then, a relaxation iteration method, is proposed to refine the coarse localization point to obtain a sub-pixel localization point. Numerical experiments show that the localization scheme has following characteristics: high-precision and stability. Furthermore, we present the results of our work which involves investigation of the performance evaluation for the proposed localization scheme. (C) 2013 Elsevier GmbH. All rights reserved.