famoreno / srba-stereo-slam

SRBA-based Stereo Vision SLAM System

SRBA-Stereo-SLAM

Note: Preliminary version

Library that performs stereo visual SLAM within a Sparser Relative Bundle Adjustment (SRBA) framework. More info about SRBA can be found here

Building from sources

Prerequisites

• CMake (>=2.4.5)
• OpenCV (>=2.4.8, >=3.0.0 not supported due to DBoW2 not supporting it)
• DBoW2 (1.0.2) (Check dependencies for this library)
• MRPT (>=1.3.0)
• SRBA (Header-only library, must be downloaded from GitHub for now)
• stereo-vo

Compiling

This software can be compiled in Windows and GNU/Linux with gcc or clang. It should also work on OSX but it is untested.

1. Install all prerequisites above. Many of those can be installed in Ubuntu as follows:

• Only for versions older than Ubuntu Wily (15.10):

    sudo add-apt-repository ppa:joseluisblancoc/mrpt
    sudo apt-get update
  

• After that, for all Debian / Ubuntu versions:

    sudo apt-get install build-essential cmake libmrpt-dev libopencv-dev
  

• Download and build [DBoW2] (http://webdiis.unizar.es/~dorian/index.php?p=32) library following instruction therein. This library needs [boost] (http://www.boost.org) library to run (tested with 1.55).

• Clone stereo-vo and build following instructions therein. Optionally, run make test

• Clone the header-only library SRBA. Configure and generate the project with cmake so SRBAConfig.cmake is generated and, optionally, run make test to ensure everything works ok.

2. Create an empty build directory, invoke cmake and build as usual.

    mkdir build
    cd build
    cmake .. -DSRBA_DIR=[PATH_TO_SRBA_BUILD_DIR]