Dynamic target automatic labeling method and system in automatic driving scene

The invention discloses a method and a system for automatically labeling a dynamic target in an automatic driving scene. The method comprises the following steps of: acquiring vehicle-mounted sensor data and preprocessing the vehicle-mounted sensor data; a first detection model and a second detectio...

Full description

Saved in:
Bibliographic Details
Main Authors TANG KANG, LI DI, WU RUI, LUO KEHUA, LIU YUWEI
Format Patent
LanguageChinese
English
Published 17.05.2024
Subjects
Online AccessGet full text

Cover

Loading…
More Information
Summary:The invention discloses a method and a system for automatically labeling a dynamic target in an automatic driving scene. The method comprises the following steps of: acquiring vehicle-mounted sensor data and preprocessing the vehicle-mounted sensor data; a first detection model and a second detection model are used for processing the preprocessed vehicle-mounted sensor data, dynamic target information is obtained, and the dynamic target information comprises 2D target information and 3D target information; according to the 3D target information and in combination with the vehicle-mounted sensor data, obtaining a driving track of a vehicle; and obtaining a vehicle driving track, 2D target information and 3D target information within a preset time length, and carrying out offline multi-target tracking and post-processing on the target to obtain an automatic labeling result. 本发明公开了一种自动驾驶场景下动态目标自动标注方法及系统,方法包括以下步骤:获取车载传感器数据并进行预处理;分别利用第一检测模型和第二检测模型分别对预处理后的的车载传感器数据进行处理,获取动态目标信息,所述动态目标信息包括2D目标信息和3D目标信息;根据所述3D目标信息结合所述
Bibliography:Application Number: CN202410223064