For Harris-Laplace registration, this paper analyses scale-invariance of characteristic scales in detail,proposes scale space projection to unify multi-scales to single scale and constructs normalized feature space as SIFT descriptors which makes feature space scale, rotation and illumination invariant.
在基于Harris-Laplace特征的配准方面,本文详细分析了其特征尺度的不变性,提出了尺度空间投影法,统一多尺度为单一尺度,并仿照SIFT特征描述子建立了归一化的特征空间,使得特征空间具有了尺度、旋转、光照不变的特性。
参考来源 - 遥感影像配准与融合技术研究·2,447,543篇论文数据,部分数据来源于NoteExpress
提出了一种基于子空间投影技术的多普勒调频率估计方法。
A Doppler rate estimation method based on sub-space projection technique was proposed.
求解大型稀疏问题最流行的方法是建立在子空间投影技术之上的。
The most popular methods to solve large sparse problems are based on projection techniques on appropriate subspaces.
训练样本与测试样本分别朝融合特征空间投影,从而得到识别特征。
After training samples and test samples are respectively projected towards the fusion feature space, recognition features are accordingly gained.
应用推荐