.png){kind=icon}
返回
Omnidirectional spatio-temporal matching based on machine learning
.png){kind=icon}
PRE
AI
DOI：10.1007/s00500-022-07629-0.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
This work presents an edge-based approach to match omnidirectional stereo sequences. To estimate the stereo correspondence results, the method exploits the temporal information and machine learning. In other words, a support vector machine (SVM) classifier is trained using the temporally preceding disparity maps in order to be used as predictor for the stereo correspondence of the current pair. The algorithm starts by extracting edge points from the omnidirectional images using a spherical edge points detector. The spherical rectification is then performed on three omnidirectional pairs, which would include the current pair, the left image and its preceding one, the right image and its preceding one. The temporally preceding disparity maps are used to generate disparity ranges for the current ones as well as to train an SVM classifier. The trained SVM along with a dynamic programming method is then used on the three stereo pairs resulting in three disparity maps. Finally, the three disparity maps are combined into one 360 degrees final disparity map. The proposed approach has been evaluated on real omnidirectional sequences and the results provided are satisfactory.
Keyword:
Stereo vision
Omnidirectional images
Disparity map
Catadioptric systems
期刊
IF:
2.5
论文数:
1.0W
被引数:
2.1W
机构
引用论文
On the size distribution function of diamond nano-crystals grown on surfaces关于表面生长的金刚石纳米晶体尺寸分布函数
.png){kind=icon}
PREAI
分享收藏分享收藏
分享收藏
查看更多内容