【作者】 瓢正泉； 赵保军； 唐林波；

【Author】 PIAO Zhengquan;ZHAO Baojun;TANG Linbo;Radar Research Lab, School of Information and Electronics, Beijing Institute of Technology;Beijing Key Laboratory of Embedded Real-Time Information Processing Technology;

【机构】 北京理工大学信息与电子学院雷达技术研究所； 嵌入式实时信息处理技术北京市重点实验室；

【摘要】 视觉感知是自动驾驶技术中的核心要素,单目测距和单目目标检测是视觉感知任务中的两大关键技术。然而当前单目测距与目标检测主要处于两个分离的研究领域,特别地,当前的目标检测技术没有考虑与单目测距技术之间的耦合。从单目测距的角度出发提出目标检测的新需求,并构建一种结合单目测距需求的目标检测模型。在公开的自动驾驶数据集上展开实验,其结果验证了优化后的检测模型可以更好地满足单目测距需求。更多还原

【Abstract】 Visual perception is the core element of autonomous driving technology, and monocular ranging and monocular object detection are two key technologies in visual perception tasks, however, currently monocular ranging and object detection are mainly in two separate research fields. In particular, the current object detection technology does not consider the coupling with the monocular ranging technology. In this paper, the new requirements for target detection from the perspective of monocular ranging are proposed, and an object detection model that combines the requirements of monocular ranging is constructed. In addition, experiments are carried out on the public dataset for autonomous driving, and the results verify that the optimized detection model can better meet the needs of monocular ranging.更多还原