光栅图Raster Map也叫做位图、点阵图、像素图,简单的说,就是最小单位由像素构成的图,只有点的信息.缩放时会失真。每个像素有自己的颜色,类似电脑里的图片都是像素图,你把它放很大就会看到点变成小色块了。
与光栅图相对的是, 矢量图Vector Map也叫做向量图,记录的是点、线、面的位置和颜色信息的描述,矢量图没有直接是点的信息,还有线,面,基本图形等信息。
Master Thesis:
As important information for accurate prediction, scene context information like lane geometries
and their connectivity can be captured by HD maps. Rasterization based map representations
have been extensively applied in prediction tasks. However, rasterized maps, which create image-like raster inputs, can cause information loss and have limited receptive fields. Dealing with that, vectorized map features are leveraged. Such that lane features can be encoded through various machine learning models such as CNN and LSTM. Several state-of-the-art GNN models which consider scene context information are discussed subsequently.