In this paper a new approach to recognition of single 3D curved objects is proposed. Horizontal cross-sectional slice shapes are used for 3D object representation. 3D object recognition is accomplished by matching the cross-sectional slice shapes of an input object with those of each object model. Thus, 2D image analysis techniques can be used for slice shape matching, including the use of the distance-weighted correlation for shape similarity measurement and the construction of a decision tree for decision-making. Instead of computing all the slice shapes for each input object in the recognition process, only those involved in the decision tree are computed by a data acquisition system. This increases recognition speed. Experimental results with a high recognition rate prove the feasibility of the proposed approach.