• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>IEEE International Conference on Intelligent Transportation Systems >Cooperative starting intention detection of cyclists based on smart devices and infrastructure
【24h】

Cooperative starting intention detection of cyclists based on smart devices and infrastructure

机译：基于智能设备和基础设施的自行车手的协作开始意图检测

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

In future traffic scenarios, vehicles and other traffic participants will be interconnected and equipped with various types of sensors, allowing for cooperation on different levels, such as situation prediction or intention detection. This article presents an approach to cooperative intention detection of starting cyclists using smart devices and infrastructure-based sensors. A smart device is carried by the cyclists and the infrastructure is equipped with a wide angle stereo camera system. The approach is based on a two-stage cooperative intention detection process consisting of a movement primitive detection in the first stage, used to recognize the current movement type, and a trajectory forecast in the second stage. In the first stage, cooperation is conducted by means of a stacking ensemble on the level of detected movement primitives. These cooperatively detected movement primitives are used in the second stage in an adaptive gating function that weights between multiple specialized forecasting models. Our cooperative method yields an earlier detection of starting motions compared to a non cooperative approach while retaining robustness. Moreover, we also show that the cooperative method using multiple models is able to reduce the effective forecasting error, reaching the same performance as the system with perfect movement primitive classification.
机译：在未来的交通场景中，车辆和其他交通参与者将相互连接并配备各种类型的传感器，从而可以在不同级别进行协作，例如情况预测或意图检测。本文介绍了一种使用智能设备和基于基础设施的传感器对初次骑车的人进行合作意图检测的方法。骑自行车的人携带智能设备，基础设施配备广角立体摄像头系统。该方法基于两阶段协作意图检测过程，该过程由第一阶段的运动原始检测（用于识别当前运动类型）和第二阶段的轨迹预测组成。在第一阶段，通过在检测到的运动原语级别上的堆叠合奏来进行协作。这些协作检测到的运动原语在第二阶段以自适应门控功能使用，该功能在多个专业预测模型之间加权。与非协作方法相比，我们的协作方法在保持鲁棒性的同时，可以较早地检测到起始运动。此外，我们还表明，使用多种模型的协作方法能够减少有效的预测误差，达到与具有完美运动原始分类的系统相同的性能。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号