日常观点记录

自动驾驶-配准

This repositoy is part of my thesis and proposes a way to register point clouds in a scene with some autonomous vehicles using an efficient and effective way of representing the environment in which each autonomous vehicle is located. More specifically, each autonomous vehicle constructs a graph whose vertices are some features of its environment that were produced after semantic segmentation of the scene it is in. These features may consist of other vehicles, pedestrians, or even fixed objects such as light poles or road signs. Finally, the last stage follows which is the matching of the graphs of the vehicles involved in a scene. Through this matching, the relative position of one autonomous vehicle in relation to another is derived, and thus a new scene can be created containing information from both vehicles filling in any gaps.

这个重新定位是我论文的一部分，它提出了一种将场景中的点云与一些自动驾驶汽车进行配准的方法，使用一种高效有效的方式来表示每个自动驾驶汽车所处的环境。更具体地说，每辆自动驾驶汽车都会构建一个图，其顶点是对其所在场景进行语义分割后产生的环境特征。这些特征可能包括其他车辆、行人，甚至是灯杆或路标等固定物体。最后，接下来是最后一个阶段，即场景中涉及的车辆的图形的匹配。通过这种匹配，导出了一辆自动驾驶汽车相对于另一辆的相对位置，因此可以创建一个新的场景，其中包含来自两辆汽车的信息，以填补任何空白。