Ubuntu 16.04 下 PVIO 的编译与运行

[TOC]

Overview

• code: https://github.com/zju3dv/PVIO

• paper

  @inproceedings{PRCV-LiYHZB2019,
    author={Jinyu Li and Bangbang Yang and Kai Huang and Guofeng Zhang and Hujun Bao},
    title     = {Robust and Efficient Visual-Inertial Odometry with Multi-plane Priors},
    booktitle = {Pattern Recognition and Computer Vision - Second Chinese Conference,
                 {PRCV} 2019, Xi'an, China, November 8-11, 2019, Proceedings, Part {III}},
    series    = {Lecture Notes in Computer Science},
    volume    = {11859},
    pages     = {283--295},
    publisher = {Springer},
    year      = {2019}
  }

Note: 本博客主要是记录下在本人Ubuntu 16.04的PC平台上编译时遇到的问题和解决办法。

Build

Dependencies

• cmake 3.18.0
• GCC 9.3.0
• Eigen 3.3.7
• Ceres Solver 1.14.0
• OpenCV-3.3.1 (from ROS kinetic)
• FFMPEG 4.3.2

Build

• download

  git clone https://github.com/zju3dv/PVIO.git

• cmake

  cd PVIO && mkdir build && cd build
  cmake -DCMAKE_BUILD_TYPE=Release ..

  Tips: 执行 cmake 过程中会下载 GIT_REPOSITORY，因网络原因可能导致部分下载失败，可改善网络或者多执行几次直到全部下载下来。

• make

  make -j1

Errors when make

• error: ‘::av_packet_alloc’ has not been declared

  

  解决办法：升级 FFMPEG（之前是FFMPEG2） 和 对应的 libavcodec

  sudo add-apt-repository ppa:jonathonf/ffmpeg-4
  sudo apt update
  sudo apt install ffmpeg
  ffmpeg -version
  
  sudo apt install libavcodec-dev

• /usr/lib/x86_64-linux-gnu/libavutil.so: undefined reference to clEnqueueWriteImage@OPENCL_1.0

  

  解决办法：

  因我之前装了cuda8，libavutil.so 链接到了 cuda 中的 libOpenCL.so.1

  

  使其链接到 /usr/lib/x86_64-linux-gnu/ 中的 libOpenCL.so.1 即可。

  cd /usr/local/cuda-8.0/lib64
  sudo unlink libOpenCL.so.1
  sudo ln -s /usr/lib/x86_64-linux-gnu/libOpenCL.so.1.0.0 libOpenCL.so.1

Run

• run with dataset

  ./pvio-pc/pvio-pc euroc:///home/cg/dev_sdb/datasets/ETH_ASL_Datasets/EuRoC_MAV/V1_01_easy/mav0 ../config/euroc.yaml