openmht

Open Multi-Hypothesis Tracking

Prerequisites

1. Using Ubuntu package manager:

   $ sudo apt-get install libeigen3-dev libboost-all-dev gnuplot gnuplot-qt pkg-config

2. Obtain, build, and install Munkres Hungarian Assignment Algorithm:

   $ git clone https://github.com/saebyn/munkres-cpp.git

   Make change to /path/to/munkres-cpp/CMakeLists.txt: Add the following to line 27:

   ${PROJECT_SOURCE_DIR}/src/adapters/boostmatrixadapter.h

   $ mkdir build && cd build
   $ cmake .. $ make $ sudo make install $ sudo ldconfig

3. Build openMHT:

   $ cd /path/to/openmht $ ./build.sh $ sudo ./build.sh install

   By default, this will only install a pkg-config file that points to your local git repository. This is necessary because openMHT is currently under heavy development.

4. Copy the cmake file located at /path/to/openmht/cmake/Modules/Findopenmht.cmake to your /cmake/Modules path in your cmake project to find and use openMHT easily. You will need to include openmht_INCLUDE_DIRS and link to openmht_LIBRARIES.

   include_directories(${openmht_INCLUDE_DIRS})

   target_link_libraries( your_app_name ${openmht_LIBRARIES} )

5. Look at the example in /path/to/openmht/share/multi-track-example/main.cpp for how to use the openMHT API. The important calls are:

   openmht::MHT mht; // start an instance of openMHT mht.set_dt(dt); // set the sensor update period (seconds).

   // Convert your measurement structure to openmht::Measurement openmht::Measurement m; m.set_position(Eigen::Vector2d(3.4,1.8);

   // Add the measurement to a list: std::listopenmht::Measurement m_list; m_list.push_back(m);

   // Process the list of measurements with MHT mht.process_measurements(m_list);

   // Get a list of the fused entities / contacts: std::listopenmht::Entity ents = mht.entities();

   // Each entity has a position() method that provides the // current estimated 2D position as an Eigen::Vector2d