多模态图像之间存在显著的非线性强度差异,并且图像会因为噪声而退化,因此,多模态图像自动配准是一项具有挑战性的任务。为了解决这两个问题,本文提出一种多模态图像自动配准方法,该方法分为预配准和精配准两个阶段。在预配准阶段,通过...多模态图像之间存在显著的非线性强度差异,并且图像会因为噪声而退化,因此,多模态图像自动配准是一项具有挑战性的任务。为了解决这两个问题,本文提出一种多模态图像自动配准方法,该方法分为预配准和精配准两个阶段。在预配准阶段,通过改进SIFT算法来大致对齐多模态图像。在精配准阶段,首先,利用块Harris检测器在预配准后的参考图上提取均匀分布的特征点。然后,通过各向异性结构张量捕捉多模态图像中的结构信息来构建特征描述符,该特征描述符对噪声具有稳健性。更进一步,本文结合张量方向平行度和梯度互信息提出了一种相似度准则(tensor orientation and mutual information,TOMI)。最后,本文用多种模态图像(包括Optical,LiDAR,SAR和Map)来评估提出的方法。试验结果表明,本文提出的方法对非线性强度变化和噪声具有较好的稳健性,并且匹配效果优越。展开更多
This article puts forward a kind of parallel and distributed static augmented scene system structure to improve the performance of real time augmented simulation system. Based on static registration technique, several...This article puts forward a kind of parallel and distributed static augmented scene system structure to improve the performance of real time augmented simulation system. Based on static registration technique, several groups of processing nodes do parallel scene pictures taking, 3D registration and virtual-real merging. Process on different nodes is controlled by uniform synchronization mechanism and network transmitting. Wide field of view image can be obtained from image mosaic operation and displayed by wide view display system. Detailed system architecture, registration algorithm, method how to determine camera position and synchronization mechanism between each process node are introduced. The experiment result can validate the good performance of the designed system.展开更多
文摘多模态图像之间存在显著的非线性强度差异,并且图像会因为噪声而退化,因此,多模态图像自动配准是一项具有挑战性的任务。为了解决这两个问题,本文提出一种多模态图像自动配准方法,该方法分为预配准和精配准两个阶段。在预配准阶段,通过改进SIFT算法来大致对齐多模态图像。在精配准阶段,首先,利用块Harris检测器在预配准后的参考图上提取均匀分布的特征点。然后,通过各向异性结构张量捕捉多模态图像中的结构信息来构建特征描述符,该特征描述符对噪声具有稳健性。更进一步,本文结合张量方向平行度和梯度互信息提出了一种相似度准则(tensor orientation and mutual information,TOMI)。最后,本文用多种模态图像(包括Optical,LiDAR,SAR和Map)来评估提出的方法。试验结果表明,本文提出的方法对非线性强度变化和噪声具有较好的稳健性,并且匹配效果优越。
基金supported by the Fundamental Research Funds for the Central Universitiesthe Research Funds of Renmin University of China (10XNF039)
文摘This article puts forward a kind of parallel and distributed static augmented scene system structure to improve the performance of real time augmented simulation system. Based on static registration technique, several groups of processing nodes do parallel scene pictures taking, 3D registration and virtual-real merging. Process on different nodes is controlled by uniform synchronization mechanism and network transmitting. Wide field of view image can be obtained from image mosaic operation and displayed by wide view display system. Detailed system architecture, registration algorithm, method how to determine camera position and synchronization mechanism between each process node are introduced. The experiment result can validate the good performance of the designed system.
