计算机应用与软件
計算機應用與軟件
계산궤응용여연건
COMPUTER APPLICATIONS AND SOFTWARE
2014年
4期
145-148
,共4页
双目视觉%SIFT%特征匹配%区域匹配%视差梯度
雙目視覺%SIFT%特徵匹配%區域匹配%視差梯度
쌍목시각%SIFT%특정필배%구역필배%시차제도
Binocular vision%SIFT%Feature matching%Area matching%Disparity gradient
以 SIFT 特征对旋转、尺度缩放、亮度变化保持不变性的特点，使用双目视觉系统从两个不同的视觉角度采集立体图像对，提出基于 SIFT 特征匹配和改进的区域匹配相结合的匹配算法。该方法确定符合 SIFT 特征的边缘为可靠特征点，并确定其视差；根据视差梯度原理确定其他点的视差，最后生成稠密的视差图。实验结果表明，SIFT 特征与区域相结合的匹配算法的引入，提高了特征点视差的准确性，一些弱纹理区的匹配也有所改善。
以 SIFT 特徵對鏇轉、呎度縮放、亮度變化保持不變性的特點，使用雙目視覺繫統從兩箇不同的視覺角度採集立體圖像對，提齣基于 SIFT 特徵匹配和改進的區域匹配相結閤的匹配算法。該方法確定符閤 SIFT 特徵的邊緣為可靠特徵點，併確定其視差；根據視差梯度原理確定其他點的視差，最後生成稠密的視差圖。實驗結果錶明，SIFT 特徵與區域相結閤的匹配算法的引入，提高瞭特徵點視差的準確性，一些弱紋理區的匹配也有所改善。
이 SIFT 특정대선전、척도축방、량도변화보지불변성적특점，사용쌍목시각계통종량개불동적시각각도채집입체도상대，제출기우 SIFT 특정필배화개진적구역필배상결합적필배산법。해방법학정부합 SIFT 특정적변연위가고특정점，병학정기시차；근거시차제도원리학정기타점적시차，최후생성주밀적시차도。실험결과표명，SIFT 특정여구역상결합적필배산법적인입，제고료특정점시차적준학성，일사약문리구적필배야유소개선。
SIFT feature has features of keeping invariance against the variations in rotation,size scaling and brightness,considering these features,we use binocular vision system to collect stereo image pairs from two different vision angles,and propose a matching algorithm which is based on the combination of SIFT feature matching and improved area matching.The method determines the edge conforming to SIFT features as the reliable feature point and determines its parallax;the parallaxes of other points are determined according to parallax gradient principle,and finally the dense disparity map is then generated.Experimental results show that the introduction of matching algorithm combining the SIFT features and area improves the accuracy of the parallax of feature point,the matching of some weak texture areas is also improved.
