CAJ | 학술논문

在相似区域较多的图像匹配时，SIFT（Scale Invariant Feature Transform）算法的匹配计算（KDtree-BBF）较复杂，耗时长，很难满足实时性要求。提出一种改进的匹配算法，将特征点的周围邻域的主方向梯度作为特征之一，采用主方向梯度和欧式距离相结合的计算方法进行特征点的匹配。实验结果表明：改进的算法不仅简单易行，且对图像的旋转、缩放、光照变换均具有良好的鲁棒性，比较原OpenSIFT算法还发现，改进算法的加速比范围为1.046~9.065。
재상사구역교다적도상필배시，SIFT（Scale Invariant Feature Transform）산법적필배계산（KDtree-BBF）교복잡，모시장，흔난만족실시성요구。제출일충개진적필배산법，장특정점적주위린역적주방향제도작위특정지일，채용주방향제도화구식거리상결합적계산방법진행특정점적필배。실험결과표명：개진적산법불부간단역행，차대도상적선전、축방、광조변환균구유량호적로봉성，비교원OpenSIFT산법환발현，개진산법적가속비범위위1.046~9.065。
In matching image with many similar regions, the original image matching algorithm(KDtree-BBF)based on SIFT(Scale Invariant Feature Transform)is complex, time-consuming, it is difficult to meet the real-time requirement. To overcome the shortcomings above, an improved algorithm is proposed. The method identifies the main gradient of direction of neighbor feature points as one of the features, which is combined with the distance similarity matching for matching. Experimental results show that the proposed algorithm not only is simple but also has a good robustness on the conditions of image rotating, zooming and lighting transformation. Compared with the original OpenSIFT algorithm, the improved algorithm speedup ratio is in the range of 1.046~9.065.