基于双目立体视觉的多分辨率图像匹配方法研究  

Research on multi-resolution image matching method based on binocular stereo vision

在线阅读下载全文

作  者:刘华春[1] 吴广文[1] 闫静莉[1] LIU Huachun;WU Guangwen;YAN Jingli(Inner Mongolia Normal University,Hohhot 010022,China)

机构地区:[1]内蒙古师范大学,内蒙古呼和浩特010022

出  处:《现代电子技术》2025年第1期29-32,共4页Modern Electronics Technique

摘  要:在双目立体视觉系统中,面对复杂场景时噪声会损害图像特征,增加提取难度,导致匹配精度和鲁棒性下降。因此,文中提出基于双目立体视觉的多分辨率图像匹配方法,旨在从不同尺度图像中有效获取信息并实现高精度匹配。该方法利用双目立体视觉模型的双目旋转相机扫描目标并进行成像,根据内、外空间标定提升双目旋转相机的位置精度,保证目标的多分辨率成像效果;将其输入金字塔立体匹配网络中,通过网络中的类金字塔多空洞卷积操作提取双目图像特征,在此基础上,基于可变卷积增强其纹理特征细节;结合细粒度特征和互注意力机制完成双目图像匹配。测试结果显示,空间标定后,左、右两个相机的成像误差最小值分别为0.6 Pixel和0.4 Pixel;匹配点坐标偏差均值和坐标偏差方差值分别低于0.012和0.011,匹配效果良好。In binocular stereo vision systems,noise can damage image features and increase the difficulty of image feature extraction in complex scenes,leading to a decrease in matching accuracy and robustness.Therefore,a multi-resolution image matching method based on binocular stereo vision is proposed.This method aims to effectively obtain information from images of different scales and achieve high-precision matching.In this method,binocular stereo vision model with a binocular rotating camera is utilized to scan and image the objects,and internal and external space calibration is used to improve the positional accuracy of the binocular rotating camera and ensure the multi-resolution imaging effect of the objects.The imaging effect is input into the pyramid stereo matching network,and the binocular image features are extracted by the pyramid-like multiple atrous convolution operation in the network.And then,their texture feature details are enhanced based on variable convolution.Finally,the binocular image matching is achieved in combination with fine-grained features and mutual attention mechanism.The test results show that after spatial calibration,the minimum imaging errors of the left and right cameras are 0.6 Pixel and 0.4 Pixel,respectively;both of the mean value and the variance of the coordinates deviation of the matching points are lower than 0.012 and 0.011,respectively,and the matching effect is good.

关 键 词:双目立体视觉 多分辨率 图像匹配 空间标定 双目旋转相机 特征提取 特征增强 细粒度 

分 类 号:TN911.73-34[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象