摘要
对视觉图像目标匹配问题的研究,能够有效提高图像视觉效果,增强人体感官质量。首先利用双目视觉图像特征点描述向量得到图像视差值,进而计算图像匹配代价,完成双目视觉图像目标精准匹配。传统方法根据参考图与实时图之间的单应性矩阵,对实时图进行视角变换,但忽略了对图像匹配代价的计算,导致匹配精度较低。提出一种双目立体视觉图像目标精准匹配方法,通过图像平滑和插值运算降低失帧和噪声对图像质量的影响,对双目立体视觉图像进行预处理。通过对图像特征点主方向进行计算,得到图像特征点描述向量。根据特征点描述向量得到双目立体视觉图像的视差值,采用SSD度量函数和双目立体视觉图像的视差值对图像的匹配代价进行计算,完成双目立体视觉图像目标的匹配。实验结果表明,采用所提方法对双目立体视觉图像进行匹配时,抗干扰性好,匹配结果精准度较高。
The research on visual image matching problem can effectively improve the visual effect and enhance the sensory quality of the human body. First,the image parallax value is obtained by using the feature point description vector of the binocular vision image,and then the image matching cost is calculated to complete the accurate matching of the binocular vision image. The traditional method is based on the homography matrix between the reference map and the real time graph,and the real time map is transformed from the angle of view,but it neglects the calculation of the cost of the image matching,which leads to the lower matching precision. A target accurate matching method of binocular stereo vision image is proposed. Through the image smoothing and interpolation operation,the effect of the loss of frame and noise on the image quality were reduced,and the binocular stereo vision image was preprocessed. The feature point description vector was obtained by calculating the main direction of the image feature points,and the feature points were described by the feature point. The parallax value of the binocular stereo vision image was measured. The matching cost of the image was calculated by using the SSD measure function and the parallax value of the binocular stereo vision image to match the target of the binocular stereo vision image. The experimental results show that the proposed method has good anti-jamming performance and high accuracy in matching the binocular stereo vision images.
作者
孙深圳
孙洒
SUN Shen-zhen;SUN Sa(College of Electrical Materials,Hefei University of Technology,Fuyang Anhui 230009,China;School of Mechamieal Emgimeerimg,Anhui University of Technology,Fuyang,Anhui 243032,China)
出处
《计算机仿真》
北大核心
2018年第11期413-416,共4页
Computer Simulation
关键词
双目立体视觉
图像目标
匹配方法
Binocular stereo vision
Image target
Matching method
