中国科学院机构知识库网格
Chinese Academy of Sciences Institutional Repositories Grid
Detecting ground control points via convolutional neural network for stereo matching.

文献类型:期刊论文

作者 Cao, Donglin ;  Lv, Zhihan;  Li, Shaozi ; Zhong, Zhun ;  Su, Songzhi 
刊名MULTIMEDIA TOOLS AND APPLICATIONS
出版日期2017
文献子类期刊论文
英文摘要In this paper, we present a novel approach to detect ground control points (GCPs) for stereo matching problem. First of all, we train a convolutional neuralnetwork (CNN) on a large stereo set, and compute the matching confidence of each pixel by using the trained CNN model. Secondly, we present a groundcontrol points selection scheme according to the maximum matching confidence of each pixel. Finally, the selected GCPs are used to refine the matchingcosts, then we apply the new matching costs to perform optimization with semi-global matching algorithm for improving the final disparity maps. We evaluate our approach on the KITTI 2012 stereo benchmark dataset. Our experiments show that the proposed approach significantly improves the accuracy of disparity maps.
URL标识查看原文
语种英语
源URL[http://ir.siat.ac.cn:8080/handle/172644/12467]  
专题深圳先进技术研究院_数字所
作者单位MULTIMEDIA TOOLS AND APPLICATIONS
推荐引用方式
GB/T 7714
 Cao, Donglin , Lv, Zhihan, Li, Shaozi ,et al. Detecting ground control points via convolutional neural network for stereo matching.[J]. MULTIMEDIA TOOLS AND APPLICATIONS,2017.
APA  Cao, Donglin , Lv, Zhihan, Li, Shaozi ,Zhong, Zhun ,& Su, Songzhi .(2017).Detecting ground control points via convolutional neural network for stereo matching..MULTIMEDIA TOOLS AND APPLICATIONS.
MLA  Cao, Donglin ,et al."Detecting ground control points via convolutional neural network for stereo matching.".MULTIMEDIA TOOLS AND APPLICATIONS (2017).

入库方式: OAI收割

来源:深圳先进技术研究院

浏览0
下载0
收藏0
其他版本

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。