[1]魏 东,何 雪*.基于引导信息的双目立体匹配算法[J].计算机技术与发展,2022,32(12):159-164.[doi:10. 3969 / j. issn. 1673-629X. 2022. 12. 024]
　WEI Dong,HE Xue*.Binocular Stereo Matching Algorithm Based on Guidance Information[J].,2022,32(12):159-164.[doi:10. 3969 / j. issn. 1673-629X. 2022. 12. 024]
点击复制

基于引导信息的双目立体匹配算法()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
32
期数:
2022年12期
页码:
159-164
栏目:
人工智能
出版日期:
2022-12-10

文章信息/Info

Title:
Binocular Stereo Matching Algorithm Based on Guidance Information
文章编号:
1673-629X(2022)12-0159-06
作者:
魏 东; 何 雪*
沈阳工业大学 信息科学与工程学院,辽宁 沈阳 110870
Author(s):
WEI Dong; HE Xue*
School of Information Science and Engineering,Shenyang University of Technology,Shenyang 110870,China
关键词:
立体匹配; 双目视觉; 边缘信息; 语义信息; 视差注意力机制
Keywords:
stereo matching; binocular vision; edge information; semantic information; parallax-attention mechanism
分类号:
TP391. 41
DOI:
10. 3969 / j. issn. 1673-629X. 2022. 12. 024
摘要:
针对现有立体匹配算法在边缘、遮挡、视差不连续、弱纹理等区域匹配误差较大的问题,提出一种在利用视差注意力机制的基础上引入边缘和语义信息的立体匹配算法。 在利用视差注意力机制进行代价计算和代价聚合中引入边缘细节信息改善边缘和遮挡区域匹配误差较大的问题,并对引入边缘信息时与特征提取过程中得到的不同尺度特征图融合的时机进行了讨论,确定浅层大尺度特征图引入边缘信息可以提高匹配精度;在视差优化中引入语义信息改善视差不连续和弱纹理区域匹配精度不高的问题,并对不同尺度特征图求取的语义信息对匹配精度的影响进行讨论,利用深层小尺度特征图提取语义信息可以提高匹配精度。 提出的方法在 SceneFlow 数据集上进行了测试,将基准网络 PASMNet 的误差降低了 49. 05% ,并与其他算法进行对比分析。 实验结果表明,边缘和语义等引导信息的引入有针对性地改善了现有算法在边缘、遮挡、视差不连续和弱纹理区域的视差精度,从而提高了整体预测精度。
Abstract:
Aiming at the problem that the existing stereo matching algorithms have large matching errors in areas such as edge,occlusion,parallax discontinuity and weak texture,a stereo matching algorithm based on parallax attention mechanism and introducing edge andsemantic information is proposed. In the cost calculation and cost aggregation using parallax attention mechanism, the edge detailinformation is introduced to improve the problem of large matching errors in edge and occluded regions. The timing of the fusion of theedge information and the different scale feature maps obtained in the process of feature extraction is discussed,and it is determined that theintroduction of edge information in the shallow large-scale feature map can improve the matching accuracy. The semantic information is　introduced to improve the parallax discontinuous and weak texture regions in parallax optimization, and the influence of semanticinformation extracted from feature maps of different scales on the matching accuracy is discussed,and semantic information extracted fromdeep small - scale feature map can improve the matching accuracy. The proposed method is evaluated on the SceneFlow dataset andcompared with other algorithms,and the error of the benchmark network PASMNet is reduced by 49. 05% . Experiments show that the introduction of edge and semantic information improves the disparity solution of existing algorithms in edge,　occlusion and weak texture,soas to improve the overall prediction accuracy.

相似文献/References:

[1]于勇 张晖 林茂松.基于双目立体视觉三维重建系统的研究与设计[J].计算机技术与发展,2009,(06):127.
　YU Yong,ZHANG Hui,LIN Mao-song.Research and Design of the 3D Reconstruction System Based on Binocular Stereo Vision[J].,2009,(12):127.
[2]方恒 吴怀宇.基于MRF和颜色空间的立体图像匹配算法[J].计算机技术与发展,2008,(12):28.
　FANG Heng,WU Huai-yu.Stereo Matching Based on MRF and Color Space[J].,2008,(12):28.
[3]印勇 张建华.基于三角形网格的浓密视差图提取[J].计算机技术与发展,2007,(11):52.
　YIN Yong,ZHANG Jian-hua.Dense Disparity Map Extraction Based on Triangulation Mesh[J].,2007,(12):52.
[4]毛雁明 杨慧玲.一种新的立体匹配算法[J].计算机技术与发展,2011,(03):105.
　MAO Yan-ming,YANG Hui-ling.A New Stereo Matching Algorithm[J].,2011,(12):105.
[5]周春燕 贾渊.基于遗传算法的图像配准研究及改进[J].计算机技术与发展,2011,(08):46.
　ZHOU Chun-yan,JIA Yuan.Research and Improvement of Image Registration Based on Genetic Algorithm[J].,2011,(12):46.
[6]王瑞 杨润泽 尹晓春.一种改进的立体像对密集点匹配算法[J].计算机技术与发展,2011,(09):70.
　WANG Rui,YANG Run-ze,YIN Xiao-chun.A Modified Image Dense Stereo Matching Algorithm[J].,2011,(12):70.
[7]袁力 张怡 丁丽君.基于加权匹配代价的初始视差估计算法[J].计算机技术与发展,2011,(10):36.
　YUAN Li,ZHANG Yi,DING Li-jun.Initial Disparity Estimation Algorithm Based on Weighted Matching Cost[J].,2011,(12):36.
[8]陈佳坤 罗谦 曾玉林.一种用于立体图像匹配的改进稀疏匹配算法[J].计算机技术与发展,2011,(10):63.
　CHEN Jia-kun,LUO Qian,ZENG Yu-lin.An Improved Sparse Matching Algorithm for Stereo Matching[J].,2011,(12):63.
[9]王兴昌 李炜 刘政怡 郭星.基于双密度双树复小波的立体匹配[J].计算机技术与发展,2012,(10):91.
　WANG Xing-chang,LI Wei,LIU Zheng-yi,et al.Stereo Matching Based on Double-density Dual-tree Complex Wavelet Transform[J].,2012,(12):91.
[10]汪瑶 徐杜 蒋永平 卢传泽.双目视觉小波域SIFT匹配与极线约束算法研究[J].计算机技术与发展,2012,(11):81.
　WANG Yao,XU Du,JIANG Yong-ping,et al.Study on Combined Wavelet-SIFT Matching and Epipolar Constraint Algorithm for Binocular Stereo Vision[J].,2012,(12):81.
[11]王建文 王敏.基于双目立体视觉的鞋楦三维建模[J].计算机技术与发展,2009,(04):224.
　WANG Jian-wen,WANG Min.3D Reconstruction of Shoe- last Based on Binocular Stereo Vision[J].,2009,(12):224.

更新日期/Last Update: 2022-12-10