Coherent ladar range image is closely relative to physical structure property of object′s surface, and reflects essential characteristics of objects, thus it has received considerable attention in object recognition fields. Generally, it is difficult to collect a mass of range images for ladar in real application. However, under small-sample case, when the number of features increases, the recognition rate will decrease, which is called Hughes effect. In this paper, the two methods of feature selection, i. e., relief and support vector machine recursive feature elimination (SVM-RFE) are applied to solve the problem. The experimental results demonstrate that the methods of Relief and SVM-RFE are able to relieve Hughes effect that is caused by three combined moments [affine moments (AMs) and Zernike moments (ZMs), ZMs and Hu moments (HMs), and AMs, HMs and ZMs]. Moreover, the recognition rate of SVM-RFE is slightly better than that of Relief.