The invention provides a multi-task sensing method and device for automatic driving, and the method comprises the steps: carrying out the feature extraction and feature fusion of a to-be-recognized image through a feature extraction and fusion model, and obtaining an original feature map set of at least two preset sizes; the target detection model adopts a regression rule to carry out convolution branch prediction of at least two dimensions on the original feature map set to obtain a target detection object; a lane line detection model processes the original feature map set to obtain a lane line in the to-be-recognized image; a drivable region segmentation model processes the original feature map set to obtain a drivable region in the to-be-recognized image, and the lane line detection model and the drivable region segmentation model belong to semantic segmentation models.
本申请提供的一种自动驾驶的多任务感知方法和装置,该方法包括:特征提取融合模型对于待识别图像进行特征提取和特征融合,得到至少两个预设尺寸的原始特征图集合;目标检测模型采用回归规则对于所述原始特征图集合进行至少两个维度的卷积分支预测,得到目标检测对象;车道线检测模型对于所述原始特征图集合进行处理,得到所述待识别图像中的车道线;可行驶区域分割模型对于所述原始特征图集合进行处理,得到所述待识别图像中的可行驶区域,所述车道线检测模型和可行驶区域分割模型属于语义分割模型。
Multi-task sensing method and device for automatic driving
自动驾驶的多任务感知方法和装置
2024-04-09
Patent
Electronic Resource
Chinese
Multi-task sensing method and system based on automatic driving and intelligent terminal
European Patent Office | 2024
|European Patent Office | 2024
|Task scheduling strategy generation method and device for automatic driving system
European Patent Office | 2023
|Multi-task processing method and device for intelligent driving
European Patent Office | 2022
|European Patent Office | 2023
|