首页> 外国专利> METHOD AND APPARATUS FOR VEHICLE ODOMETRY USING MONOCULAR CAMERA IMAGES AND ON-BOARD DIAGNOSTICS

METHOD AND APPARATUS FOR VEHICLE ODOMETRY USING MONOCULAR CAMERA IMAGES AND ON-BOARD DIAGNOSTICS

机译:利用单分子相机图像和车载诊断技术进行车辆卵子检测的方法和装置

摘要

A vehicle driving record measuring method using a monocular camera image and an OBD according to an embodiment of the present invention includes a data synthesis step of combining the continuous image and the OBD information of the monocular camera; A deep neural network learning step of learning the entire deep neural network through a data synthesis step; And measuring a vehicle''s driving record through a deep neural network learning step, wherein the data synthesis step takes a continuous image of the camera and creates a sub-feature through the image processing module, and imports the OBD information to obtain the width of the image. , Duplicate to fit height, and stacking and combining the sub-features of the image and the duplicated OBD information on the channel axis, and the deep neural network learning step, obtaining the own motion by passing the synthetic data through the middle and upper steps. , And learning the entire deep neural network using the synthesized data, and the driving record measurement step includes the self-motion estimation step of the learned deep neural network, and the step of measuring the vehicle driving record using the self-motion estimation. do.
机译:根据本发明的实施例的使用单眼照相机图像和OBD的车辆行驶记录测量方法包括:将连续图像和单眼照相机的OBD信息组合的数据合成步骤;以及将单眼照相机图像和OBD信息组合在一起的数据合成步骤。深度神经网络学习步骤,通过数据合成步骤学习整个深度神经网络;并通过深度神经网络学习步骤测量车辆的驾驶记录,其中数据合成步骤拍摄相机的连续图像,并通过图像处理模块创建子特征,并导入OBD信息以获得宽度图片。 ,进行复制以适合身高,并在通道轴上堆叠并组合图像的子特征和复制的OBD信息,以及深度神经网络学习步骤,通过将合成数据传递到中层和上层步骤来获得自己的运动。并且,使用合成数据学习整个深度神经网络,并且驾驶记录测量步骤包括所学习的深度神经网络的自运动估计步骤,以及使用自运动估计来测量车辆驾驶记录的步骤。做。

著录项

相似文献

  • 专利
  • 外文文献
获取专利

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