CAJ | 학술논문

提出了一种基于尺度不变特征变换和随机抽样一致算法的立体匹配方法。首先采用SIFT算法提取左右草莓图像的特征向量，再运用RANSAC和极线约束法对两幅图像进行特征点匹配，以降低特征点对之间的错配率，提高图像匹配精度。最后利用双目立体视觉原理对图像各目标进行匹配试验。结果表明，该算法较传统方法平均匹配精度提高50％，算法时间缩短66％，可满足草莓收获机器人视觉系统在自然采摘作业环境下的工作要求。
제출료일충기우척도불변특정변환화수궤추양일치산법적입체필배방법。수선채용SIFT산법제취좌우초매도상적특정향량，재운용RANSAC화겁선약속법대량폭도상진행특정점필배，이강저특정점대지간적착배솔，제고도상필배정도。최후이용쌍목입체시각원리대도상각목표진행필배시험。결과표명，해산법교전통방법평균필배정도제고50％，산법시간축단66％，가만족초매수획궤기인시각계통재자연채적작업배경하적공작요구。
A stereo matching method based on Scale Invariant Feature Transform and Random Sample Consensus algorithm was proposed in this paper. Firstly,the eigenvector of left and right strawberry images were extracted by using SIFT algorithm. Sec-ondly,image feature points between two images were matched based on RANSAC and epipolar geometric constraint algorithm, to reduce the mismatching rate and enhance the images matching precision. Finally, the matching test of each targets were recov-ered using binocular stereo visual measurement. The experimental results indicated that the average matching precision of the matching algorithm could improve by 50%,and the algorithm time is reduced by 66% than that of the conventional method, which could meet the work need of vision system for strawberry harvesting robot in natural picking operation environments.