环境
JETPACK 3.2
Cuda 9.0
Cudnn 7.2.1
ROS kinetic
Opencv 3.3.1
1.Jetson TX2刷机及安装JetPack 3.2
https://blog.csdn.net/QLULIBIN/article/details/78629305
注意在刷机前将系统语言设置为英文。
2.查看当前cuda及Opencv版本
Jetpack刷机时会自动下载cuda和opencv，按以下指令查看版本
cat /usr/local/cuda/version.txt
pkg-config --modversion opencv3.下载cudnn并安装
https://docs.nvidia.com/deeplearning/sdk/cudnn-install/index.html
下载对应版本，如cuDNN v7.2.1 Library for Linux
之后按官网步骤操作即可。
安装成功后检测版本
cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2显示如下
#define CUDNN_MAJOR 7
#define CUDNN_MINOR 2
#define CUDNN_PATCHLEVEL 1
--
#define CUDNN_VERSION (CUDNN_MAJOR * 1000 + CUDNN_MINOR * 100 + CUDNN_PATCHLEVEL)
#include "driver_types.h"
4.下载ZED SDK及ROS wrapper
https://docs.stereolabs.com/integrations/ros/getting-started/
5.SDK安装
chmod +x ZED_SDK_Linux_Ubuntu16_v2.0.0.run
./ZED_SDK_Linux_Ubuntu16_v2.0.0.run6.测试
cd ~/catkin_ws/
catkin_make
source ./devel/setup.bash
roslaunch zed_wrapper zed.launch注意，ZED必须接到USB3.0的口上！
7.效果
rqt选择一个topic
