Motion Detection and Segmentation in the Spatio-Temporal Space

本研究的目的是开发一个强大的 动态图像的运动检测与分割算法 场景 使用局部运动检测的初步结果 超曲面在时空域中的结构是 很有希望 它们将被扩展到动态场景中获取 使用固定的以及移动的摄像机。 拟议 方法将基于表面近似的组合 和区域生长用于动态场景的分割。 的 所提出的算法的序列中的适用性 还将研究自运动复对数空间。试验室 室外场景将用于实验验证 的算法。

The objective of this research is to develop a robust algorithm for motion detection and segmentation of dynamic scenes. Preliminary results in motion detection using local structures of hypersurfaces in the spatio-temporal domain are promising. They will be extended to dynamic scenes acquired using stationary as well as moving cameras. The proposed approach will be based on a combination of surface approximation and region growing for the segmentation of a dynamic scene. The applicability of the proposed algorithn to sequences in the egomotion complex log space will also be studied. Both laboratory and outdoor scenes will be used for the experimental verification of the algorithms.