摘要
三维形状的显著性在形状分析与处理中有不可忽视的作用.现有的三角网格显著性检测方法大多依赖某种人工设计的几何特征,缺乏灵活性.为此,提出一种基于特征融合学习的显著区域检测方法,以适应不同类别的形状.首先计算形状的多种几何特征,然后把多尺度的低层次特征输入到一维卷积神经网络中;通过优化中心正则化损失函数,得到高层次、可判别的特征向量,同时也得到显著区域检测结果.在普林斯顿网格数据集上的实验结果表明,该方法适用于不同形状的显著性检测,检测结果具有一致性,并且相比对照算法具有更好的视觉效果和定量化指标评价.
3D shape saliency plays a non-negligible role in shape analysis and processing.Nearly all of the current triangular mesh saliency detection methods rely on handcrafted geometric features,while lack of flexibility.In this paper,we develop a saliency region detection method based on the feature-fusion learning which is suitable for different kinds of shapes.Starting with computing several features of the shape,the low-level features are organized in a multi-scale way and input into a 1D convolution neural network.After optimizing the central regularized loss function,we get the high-level and discriminative feature,and the prediction of saliency.Experiments on Princeton mesh dataset show that our method is applicable to specific-targeted saliency detection,has consistent effect,and outperforms existing methods both visually and quantitatively.
作者
朱奕杰
蔺宏伟
Zhu Yijie;Lin Hongwei(State Key Laboratory of CAD&CG,Zhejiang University,Hangzhou 310058;School of Mathematical Science,Zhejiang University,Hangzhou 310027)
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2020年第2期203-212,共10页
Journal of Computer-Aided Design & Computer Graphics
基金
国家自然科学基金(61872316)
国家重点研发计划项目(2016YFB1001501)
关键词
网格显著性
卷积神经网络
中心正则化
mesh saliency detection
convolution neural network
center regularization
