CAJ | 학술논문

不同视角下具有一定变形的高分辨率大尺寸影像之间的匹配是遥感、摄影测量和计算机视觉等领域的难点。提出了抗仿射形变异构金字塔复合描述点特征匹配算法(RAIPy MuDePoF匹配算法)：构建了基于sinc函数卷积变换的多尺度异构金字塔影像结构，提出采用变换影像的sinc梯度、主方向和变形程度拟合仿射协变区域，在特征点的仿射归一化区域中，提出新的抗旋转投影累积量描述子和加权直方图辅助描述子进行复合描述，最后在大尺度匹配特征拟合变化参数和可信度的引导下实现尺度域的点特征匹配。大量试验表明，算法对尺度变化、旋转、噪声和一定程度的视角变换和变形具有很强的适应性，性能优于当前很好的匹配算法。
불동시각하구유일정변형적고분변솔대척촌영상지간적필배시요감、섭영측량화계산궤시각등영역적난점。제출료항방사형변이구금자탑복합묘술점특정필배산법(RAIPy MuDePoF필배산법)：구건료기우sinc함수권적변환적다척도이구금자탑영상결구，제출채용변환영상적sinc제도、주방향화변형정도의합방사협변구역，재특정점적방사귀일화구역중，제출신적항선전투영루적량묘술자화가권직방도보조묘술자진행복합묘술，최후재대척도필배특정의합변화삼수화가신도적인도하실현척도역적점특정필배。대량시험표명，산법대척도변화、선전、조성화일정정도적시각변환화변형구유흔강적괄응성，성능우우당전흔호적필배산법。
Matching for high resolution image pairs with different viewpoints and distortions is a difficult work in remote sensing, photographing and computer vision etc. Robust Affine-Invariant Isomerous Pyramid Feature and Multi-Description for Point Feature Matching algorithm was proposed. Isomerous image pyramid was constructed by sinc convolution function series, the sinc convoluted gradient, the main direction and the strength of changes were devised for determining the normalized affine-invariant area around the key point, and the rotation-invariant projective accumulated amount and the weighted histograms were given for describing the multi-changes from the isomerous image pyramid at a special position and scale, and then, the matching was implemented based on the distribution parameters and reliability calculated by the distinctive corresponding points with big scale. Experiments show that, the new algorithm is robust for scale change, rotation, noisy, a certain degree of viewpoint difference and distortion, and the match scores are better than the state of the art matching algorithms.