In various examples, a three-dimensional (3D) intersection structure may be predicted using a deep neural network (DNN) based on processing two-dimensional (2D) input data. In order to train the DNN to accurately predict the 3D intersection structure from the 2D input, a first loss function that compares the 3D output of the DNN (after conversion to 2D space) with 2D ground live data and a second loss function that analyzes the 3D prediction of the DNN in consideration of one or more geometric constraints may be used to train the DNN-. Geometric knowledge of intersections may be used to penalize DNN predictions that are unaligned with known intersections and/or road structure geometry. Accordingly, the DNN may use real-time perception of autonomous or semi-autonomous vehicles to detect the 3D position of the intersection structure from the 2D input.
在各种示例中,可以使用基于处理二维(2D)输入数据的深度神经网络(DNN)来预测三维(3D)路口结构。为了训练DNN从2D输入准确地预测3D路口结构,可以使用将DNN的3D输出(在转换为2D空间后)与2D地面实况数据进行比较的第一损失函数和考虑一个或更多个几何约束来分析DNN的3D预测的第二损失函数来训练DNN——例如,路口的几何知识可用于惩罚与已知路口和/或道路结构几何形状不对齐的DNN预测。因此,DNN可以使用自主或半自主车辆的实时感知来从2D输入中检测路口结构的3D位置。
Three-dimensional intersection structure prediction for autonomous driving applications
用于自主驾驶应用的三维路口结构预测
2022-08-12
Patent
Elektronische Ressource
Chinesisch
IPC: | G06V / B60W CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION , Gemeinsame Steuerung oder Regelung von Fahrzeug-Unteraggregaten verschiedenen Typs oder verschiedener Funktion / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen |
THREE-DIMENSIONAL INTERSECTION STRUCTURE PREDICTION FOR AUTONOMOUS DRIVING APPLICATIONS
Europäisches Patentamt | 2021
|Europäisches Patentamt | 2023
|Determination Method for Intersection Driving Risk of Autonomous Vehicle
Europäisches Patentamt | 2023
|Intersection pose detection in autonomous machine applications
Europäisches Patentamt | 2024
|INTERSECTION DRIVING CONTROL METHOD AND SYSTEM FOR AUTONOMOUS VEHICLE
Europäisches Patentamt | 2023