Abstract

A modified approach to recovering the trajectories of dynamic objects in a scene from stereo images is proposed. This approach is based on the use of an extended point representation of objects, visual odometry, and a novel set of algorithms. Object identification algorithms and the block diagram of the step-by-step data processing are described. Results of numerical experiments for synthetic scenes are presented.

中文翻译：

---

从立体图像中识别和恢复动态物体的轨迹

摘要

提出了一种从立体图像中恢复场景中动态对象轨迹的改进方法。该方法基于对象的扩展点表示，视觉里程表和一组新颖的算法。描述了对象识别算法和逐步数据处理的框图。给出了合成场景的数值实验结果。