.png){kind=icon}
返回
Feature representation for 3D object retrieval based on unconstrained multi-view
.png){kind=icon}
PRE
AI
DOI：10.1007/s00530-022-00939-1.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Reasonable and accurate image feature representation is the key to successful object retrieval. In this paper, we propose a 3D object feature representation method based on multiple views rather than a shape model. Unlike existing view-based methods that use pre-designed camera arrays to capture views, our method is flexible to implement by using several unconstrained views. Firstly, we generate a histogram of word frequencies to represent each view through local feature quantization. Then we integrate the histogram vectors of views belonging to the same object to generate a complete feature representation. Finally, similarity between two features is calculated for object retrieval. Several criteria are employed to evaluate the retrieval quality of the proposed method. Experimental results show that the integrated model feature is more effective and efficient than a set of individual image features and our approach is also competitive among several state-of-the-art methods.
Keyword:
3D object retrieval
Multi-view
Feature representation
Bag-of-words
Histogram vector
期刊
IF:
3.1
论文数:
2.8K
被引数:
2.7K
机构
引用论文
.png){kind=icon}
分享收藏
分享收藏分享收藏Adaptive Fusion and Category-Level Dictionary Learning Model for Multiview Human Action Recognition
PREAI
分享收藏分享收藏
查看更多内容