CAJ | 학술논문

针对SIFT局部特征在图像发生大视角变化时匹配失效的问题，提出了一种基于仿射变换模型的图像局部特征检测和匹配算法。通过建立图像仿射变换模型，对相机光轴视角变换的纬度角θ和经度角Φ进行模拟，并采用SIFT 算法进行特征检测，基于最邻近法进行特征匹配，使算法除了具有 SIFT 本身的抗图像缩放、旋转和亮度变化等特点外，还具有较强的抗仿射变换能力。试验结果表明：当图像发生相对斜率较大的仿射变换时，该图像特征检测和匹配算法仍有较强的抵抗能力。算法可用于图像末端寻的制导弹丸和空中侦察设备拍摄的图像进行有效匹配。
침대SIFT국부특정재도상발생대시각변화시필배실효적문제，제출료일충기우방사변환모형적도상국부특정검측화필배산법。통과건립도상방사변환모형，대상궤광축시각변환적위도각θ화경도각Φ진행모의，병채용SIFT 산법진행특정검측，기우최린근법진행특정필배，사산법제료구유 SIFT 본신적항도상축방、선전화량도변화등특점외，환구유교강적항방사변환능력。시험결과표명：당도상발생상대사솔교대적방사변환시，해도상특정검측화필배산법잉유교강적저항능력。산법가용우도상말단심적제도탄환화공중정찰설비박섭적도상진행유효필배。
As for the invalidation of SIFT (scale invariant feature transform) algorithm in the situation of large-angle affine transform, this paper studies on the image local feature detecting and matching algorithm based on affine transform. Firstly, the image tilt transform is simulated caused by the latitudeθand longitudeΦof camera axis movement. Then, the SIFT algorithm is used to detect the feature and match the image couples with nearest neighbor algorithm. Thus, the algorithm has the capabilities of resisting image scaling, rotation and luminance which SIFT has besides affine invariant in image matching. Experiment result shows that the proposed algorithm has nicer matching performance in larger-angle affine transform, which can be used in the image matching for image-guiding projectile and the air reconnaissance equipment.