The invention discloses a simulation traffic scene generation method and device based on natural language description. The method comprises the following steps: collecting historical traffic scene data; formulating a classification standard of the traffic scene data and corresponding natural language description; classifying and marking the collected historical traffic scene data based on the classification standard of the traffic scene data; training a traffic scene generation model by adopting historical traffic scene data which is described and labeled by a natural language; and inputting the natural language of the traffic scene into the trained traffic scene generation model to generate a corresponding simulated traffic scene. According to the method, the simulated traffic scene can be generated according to the natural language description, and the difference of the simulated traffic scene is adjusted by changing the natural language description, so that the diversity of the simulated traffic scene is ensured, and the generation efficiency of the simulated traffic scene is improved.
本发明公开一种基于自然语言描述的仿真交通场景生成方法及设备,包括:采集历史交通场景数据;制定交通场景数据的分类标准及相应的自然语言描述;基于交通场景数据的分类标准对采集的历史交通场景数据进行分类标注;采用自然语言描述标注的历史交通场景数据训练交通场景生成模型;将交通场景的自然语言输入训练好的交通场景生成模型,生成相应的仿真交通场景。本发明能够根据自然语言描述生成仿真交通场景,并通过改变自然语言描述来调整仿真交通场景的差异性,从而保证仿真交通场景的多样性,并提升仿真交通场景生成效率。
Simulation traffic scene generation method and device based on natural language description
基于自然语言描述的仿真交通场景生成方法及设备
2024-10-22
Patent
Electronic Resource
Chinese
Simulation traffic scene file generation method and device
European Patent Office | 2021
|Language-Guided Traffic Simulation via Scene-Level Diffusion
DataCite | 2024
|Traffic scene radar device and traffic scene monitoring method
European Patent Office | 2023
|