Apollo3.5感知模块是什么

77次阅读
没有评论

共计 3013 个字符,预计需要花费 8 分钟才能阅读完成。

这篇文章主要介绍了 Apollo3.5 感知模块是什么的相关知识,内容详细易懂,操作简单快捷,具有一定借鉴价值,相信大家阅读完这篇 Apollo3.5 感知模块是什么文章都会有所收获,下面我们一起来看看吧。

简介

Apollo 感知模块具有以下新的特性:

    支持 VLS-128 线激光雷达

    使用多摄像头方法检测障碍物

    更加先进的交通灯监测模块

  可配置的传感器融合方法

感知模块综合使用 5 个摄像头 (2 个前置、2 个侧面和 1 个后置),2 台毫米波雷达 (车头和车尾) 以及 3 个 16 线激光雷达 (2 个在车尾和 1 车头) 和 1 个 128 线激光雷达来识别障碍物,并融合各个传感器的信息产生最终跟踪列表。

障碍物检测子模块对障碍物进行检测、分类和跟踪。同时预测障碍物的运动和位置信息 (例如:方向和速度)。对于车道线, 通过后期处理车道解析像素,结合计算得到的与参考车辆(如 L0, L1, R0, R1) 的相对位置信息,可以构造出车道。

体系结构

感知模块的基本结构如下所示:

感知模块的基本结构图

详细的感知模块工作流程如下所示。

感知模块的工作流程图

输入

感知模块的输入包括以下几个部分:

128 线激光雷达数据(cyber channel/apollo/sensor
/velodyne128) 
 16 线激光雷达数据(cyber channel/apollo/sensor
/lidar_front,lidar_rear_left,
lidar_rear_right)
毫米波雷达数据(cyber channel/apollo/sensor
/radar_front,radar_rear)
  图像数据(cyber channel/apollo/sensor
/camera/front_6mm,front_12mm)
  标定好的雷达传感器外参  (以 YAML  文件格式提供)
  标定好的前置摄像头的外参和内参(以 YAML  文件格式提供)
车辆的速度和角速度  (cyber channel/apollo/
localization/pose)

输出

感知模块的输出如下所示:

带有方向、速度和分类信息的 3D 障碍物轨迹(cyber channel/apollo/
perception/obstacles)
交通灯检测和识别子模块的输出(cyber channel/apollo/
perception/traffic_light)

设置指令

在配置文件 modules/perception/conf/
perception_lowcost.conf 中进行通用设置。运行命令:./scripts/bootstrap.sh  来启动 web 图形界面。在图形用户界面中选择汽车模型。使用以下命令启动感知模块:./scripts/perception
_lowcost_vis.sh start  或者在图形用户界面的 Module Controller 视图页面点击  perception  按钮。停止感知模块的命令如下:./scripts/perception
_lowcost_vis.sh stop 注意:请不要试图通过图形用户界面启动感知模块而使用脚本命令停止该模块,反之亦然。从 Apollo 开放数据平台(Open Data Platform.)下载 demo  数据。

注意:

如果你被重定向到百度云的登陆页面,请完成登陆之后重复上述第 5 步(点击开放数据平台链接)

注意事项

为了运行基于 Caffe 的感知模块,需要 Nvidia GPU 和 CUDA 环境。Apollo 在发布的 Docker 镜像中提供了 CUDA 和 Caffe 库。但是,在 Docker 镜像中没有安装 Nvidia GPU 的驱动程序。

为了运行基于 CUDA 加速的感知模块,请务必在 Docker 镜像中安装与主机电脑同样版本的 Nvidia 驱动,然后使用 GPU 选项编译 Apollo(例如, 使用./apollo.sh build_gpu  或者./apollo.sh build_opt_gpu 命令)。

请参考  How to Run Perception Module on Your Local Computer。

该模块包含一个以二进制文件格式发行的修改后的 Caffe 深度学习框架。Caffe 的版权说明如下:

1 COPYRIGHT
 2
 3All contributions by the University of California:
 4Copyright (c) 2014-2017 The Regents of the University of California (Regents)
 5All rights reserved.
 6
 7All other contributions:
 8Copyright (c) 2014-2017, the respective contributors
 9All rights reserved.
11Caffe uses a shared copyright model: each contributor holds copyright over their contributions to Caffe. The project versioning records all such contribution and copyright details. If a contributor wants to further mark their specific copyright on a particular contribution, they should indicate their copyright solely in the commit message of the change when it is committed.
13LICENSE
15Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
17 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
18 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
20 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS  AS IS

关于“Apollo3.5 感知模块是什么”这篇文章的内容就介绍到这里,感谢各位的阅读!相信大家对“Apollo3.5 感知模块是什么”知识都有一定的了解,大家如果还想学习更多知识,欢迎关注丸趣 TV 行业资讯频道。

正文完
 
丸趣
版权声明:本站原创文章,由 丸趣 2023-08-03发表,共计3013字。
转载说明:除特殊说明外本站除技术相关以外文章皆由网络搜集发布,转载请注明出处。
评论(没有评论)