Methods, systems, and computer-readable storage media for vehicles

Methods, systems, and computer-readable storage media for a vehicle are provided. The disclosed technology is implemented using an architecture that accepts point clouds and images as inputs and estimates a directional 3D bounding box of all related object classes. In an embodiment, a method include...

Full description

Saved in:
Bibliographic Details
Main Authors LANG ALEX H, VORA SHAUNAK, BEIJBOM OSCAR OLOF, HEROU BERTRAND
Format Patent
LanguageChinese
English
Published 31.01.2023
Subjects
Online AccessGet full text

Cover

Loading…
More Information
Summary:Methods, systems, and computer-readable storage media for a vehicle are provided. The disclosed technology is implemented using an architecture that accepts point clouds and images as inputs and estimates a directional 3D bounding box of all related object classes. In an embodiment, a method includes temporally matching points in a three-dimensional (3D) point cloud with an image using one or more processors of a vehicle; generating semantic data of the image by using an image-based neural network; modifying, using the one or more processors, points in the 3D point cloud with the semantic data; and estimating a directional 3D bounding box of the one or more objects using a 3D object detector with the modified points as inputs. 提供了用于运载工具的方法、系统和计算机可读存储介质。所公开的技术是使用接受点云和图像作为输入并且估计所有相关对象类的方向3D包围盒的架构来实现的。在实施例中,一种方法,包括:使用运载工具的一个或多个处理器,在时间上将三维点云即3D点云中的点与图像进行匹配;使用基于图像的神经网络,生成所述图像的语义数据;使用所述一个或多个处理器,利用所述语义数据修饰所述3D点云中的点;以及使用以经修饰的点作为输入的3D对象检测器,估计一个或多个对象的方向3D包围盒。
Bibliography:Application Number: CN202211271699