The invention discloses a multi-task sensing method and system based on automatic driving and an intelligent terminal. The method comprises the following steps: respectively acquiring video stream data of a plurality of cameras; respectively inputting each video stream data into a pre-stored feature preprocessing model to obtain a plurality of feature spaces output by each feature preprocessing model so as to obtain a sequence feature space formed by each feature space; based on a preset space combination principle and constraint conditions, performing space combination on the feature spaces to obtain a combined sequence feature space; respectively inputting the combined sequence feature spaces into a plurality of pre-stored task models so as to respectively obtain identification results output by the task models; and inputting each identification result into a perception fusion module to obtain a fusion result. The problems of high system complexity and low operation efficiency caused by parallelism of a plurality of perception models in a multi-task concurrent scene are solved.
本发明公开了一种基于自动驾驶的多任务感知方法、系统和智能终端,所述方法包括:分别获取多个相机的视频流数据;将各所述视频流数据分别输入预存的特征预处理模型,以得到各所述特征预处理模型输出的多个特征空间,以得到各所述特征空间构成的序列特征空间;基于预设的空间合并原则和约束条件,对各所述特征空间进行空间组合,以得到组合后的序列特征空间;将组合后的序列特征空间分别输入预先存储的多个任务模型,以分别得到各所述任务模型输出的识别结果;将各所述识别结果输入感知融合模块,以得到融合结果。解决了多任务并发场景下,多个感知模型并行导致的系统复杂度高,运行效率低下的问题。
Multi-task sensing method and system based on automatic driving and intelligent terminal
基于自动驾驶的多任务感知方法、系统和智能终端
2024-07-16
Patent
Electronic Resource
Chinese
Multi-task sensing method and device for automatic driving
European Patent Office | 2024
|Multi-terminal forwarding equipment, automatic driving system and automatic driving vehicle
European Patent Office | 2023
|Multi-task processing method and device for intelligent driving
European Patent Office | 2022
|Road driving environment intelligent sensing method and system, terminal and storage medium
European Patent Office | 2024
|