重点研究了用户对于不同风格轮毂产品的喜好度的自动判定问题。通过对海量图片的分析处理,获得图片中轮毂的造型和颜色特征的参数化信息,进而对所得特征信息进行分级聚类,特征相似图片划归同一集群,并据此构建树形轮毂风格编码库。根据用户提供的个人喜好的相关图片,利用马氏距离法判定其所属特征类似集群,确定用户喜好倾向,进而为用户提供符合用户偏好的轮毂信息。实验结果表明,本方法的判定效率更高,准确率与专家人工判定准确率一致。%An automatic determination method of users preference for car-wheel hub selection is studied in this paper. Firstly, the set of the parametric information of modeling and color features of the car-wheels hub is collected through analyzing and processing massive car-wheel images. Secondly, the images of similar feature are placed in the same cluster using hierarchical clustering of the information. Then, the tree-style car-wheel hub encoding library is built. According to favorite images provided by users, its similar clusters are matched through Mahalanobis distance. Finally, the user's favorite car-wheel hub style is selected. The results show that this method is efficient and the accuracy of this method is consistent with that of experts.
展开▼