MAP-Tk is an open source C++ collection of libraries and tools for making measurements from aerial video. Initial capability focuses on estimating the camera flight trajectory and a sparse 3D point cloud of the scene. These products are jointly optimized via sparse bundle adjustment and are geo-localized if given additional control points or GPS metadata.

This project has similar goals as projects like Bundler and VisualSFM. However, the focus here in on efficiently processing aerial video rather than community photo collections. Special attention has been given to the case where the variation in depth of the 3D scene is small compared to distance to the camera. In these cases, planar homographies can be used to assist feature tracking, stabilize the video, and aid in solving loop closure problems.

MAP-Tk uses the KWIVER software architecture. Originally developed for MAP-Tk, KWIVER is highly modular and provides an algorithm abstraction layer that allows seamless interchange and run-time selection of algorithms from various other open source projects like OpenCV, VXL, Ceres Solver, and PROJ4. The core library and tools are light-weight with minimal dependencies (C++ standard library, KWIVER vital, and Eigen). The tools are written to depend only on the MAP-Tk and KWIVER vital libraries. Additional capabilities are provided by KWIVER arrows (plugin modules) that use third party libraries to implement various abstract algorithm interfaces defined in the KWIVER vital library. Earlier versions of MAP-Tk contained these core data structures, algorithms, and plugins, but these have since been moved to KWIVER for easier reuse across projects. What remains in this repository are the tools, scripts, and applications required to apply KWIVER algorithms to photogrammetry problems.

A Qt GUI application is provided to assist with visualization of data and results with the help of VTK. The screenshots below show the MAP-Tk GUI visualizing results of MAP-Tk run on example videos from the VIRAT Video Dataset and CLIF 2007 Dataset. More information about this example data can be found in the examples directory.

The MAP-Tk GUI now supports visualization of depth maps, but compution of depth maps is not yet supported by MAP-Tk. Instead, the cameras computed by MAP-Tk can be used with a fork of PlaneSweepLib that reads in the cameras and images and produces depthmaps that the GUI can load. We are working on extending MAP-Tk to compute depth maps directly.

While the initial software implementation relies on batch post-processing of aerial video, our intent is to move to an online video stream processing framework and optimize the algorithms to real-time performance for use onboard unmanned aerial vehicles.

MAP-Tk uses CMake (www.cmake.org) for easy cross-platform compilation. The minimum required version of CMake is 3.0, but newer versions are recommended.

Currently, a compiler with at least partial C++11 support (e.g. GCC 4.4, Visual Studio 2013) is required.

We recommend building MAP-Tk out of the source directory to prevent mixing source files with compiled products. Create a build directory in parallel with the MAP-Tk source directory. From the command line, enter the empty build directory and run:

$ ccmake /path/to/maptk/source

where the path above is the location of your MAP-Tk source tree. The ccmake tool allows for interactive selection of CMake options. Alternatively, using the CMake GUI you can set the source and build directories accordingly and press the "Configure" button.

MAP-Tk has minimal required dependencies at the core level. Enabling plugins adds additional capabilities as well as additional dependencies. Some functionality is duplicated between modules to provide choices. Feature tracking requires OpenCV or VisCL. Bundle adjustment requires Ceres Solver or VXL. Geographic transformations require PROJ4.

The only hard dependencies of MAP-Tk are on the C++ standard library, KWIVER (≥ 1.0), and Eigen (≥ 3.0; also required by KWIVER).

When building KWIVER for use in MAP-Tk there are several arrows (plugins) that should be enabled to provide maximum capability to MAP-Tk. The KWIVER arrows are not a build-time dependency of MAP-Tk, but are required at run-time to provide algorithm implementations to run. The following KWIVER arrows provide algorithms which are optionally used by MAP-Tk:

• Core

  algorithm implementations with no additional dependencies

• Ceres

  supplies bundle adjustment using Ceres Solver http://ceres-solver.org/

• OpenCV

  supplies feature detectors, descriptors, matcher; homography and fundamental matrix estimators; image I/O, and more. http://opencv.org/

• PROJ

  provides geographic transforms (e.g. Lat/Lon to UTM) http://trac.osgeo.org/proj/:

• VisCL

  experimental code for OpenCL acceleration (currently not recommend for most users) https://github.com/Kitware/VisCL

• VXL

  supplies a simple bundle adjuster, image I/O, homgraphy and fundamental matrix estimation, and more. (note: requires unreleased version, use Fletch to build) http://vxl.sourceforge.net/

The visualization application (GUI) is optional, and has additional dependencies. To build the GUI, you need:

• Qt 4

  version 4.8.0 or greater (4.8.6 or greater recommended) http://www.qt.io/

• qtExtensions

  no versioned releases as of writing; use master branch http://www.github.com/kitware/qtextensions

• VTK

  version 6.2 http://www.vtk.org/

Most of the dependencies for KWIVER and MAP-Tk can be provided by a meta-project called Fletch. Fletch uses CMake to fetch, configure, and build various third party packages such that they work together in a consistent way across platforms. We recommend that you use Fletch to build Ceres, Eigen, OpenCV, PROJ, Qt, VTK, and VXL and their dependencies. Next build KWIVER and set "fletch_DIR" in CMake to point to your Fletch build. Enable the arrows recommended above in the KWIVER build. Finally, build MAP-Tk and set "kwiver_DIR" in CMake to point to your KWIVER build.

Documentation generation is another optional component that brings in additional dependencies. To build the API documentation, you need:

• Doxygen

  version 1.7 or greater http://www.stack.nl/~dimitri/doxygen/

To build the user manual(s), you need:

• Python

  version 2.6 or greater http://www.python.org/

• Docutils

  version 0.11 or greater http://docutils.sourceforge.net/

(At present, only the GUI has a user manual. Other manuals may be added in the future.)

Nightly builds of the Doxygen documentation for the primary branches are here:

Doxygen documentation for released versions are here:

If MAPTK_ENABLE_DOCS is enabled, and CMake finds, or is provided with, a path to the Doxygen tool, then the HTML documentation is built as part of the normal build process under the target "doxygen". Open ${MAPTK_BUILD_DIR}/docs/index.html in your browser to view the documentation.

If MAPTK_ENABLE_MANUALS is enabled, and CMake finds, or is provided with, a path to the Python executable which is able to import docutils, then the user manuals are built as part of the normal build process under the target "manuals". The GUI manual can be viewed from inside the GUI by choosing the "MapGUI User Manual" action from the "Help" menu.

Continuous integration testing is provided by CDash. Our MAP-Tk dashboard hosts nightly build and test results across multiple platforms including Windows, Mac, and Linux.

Anyone can contribute a build to this dashboard using the dashboard script provided. Follow the instructions in the comments.

Travis CI is also used for continued integration testing. Travis CI is limited to a single platform (Ubuntu Linux), but provides automated testing of all topic branches and pull requests whenever they are created.

Travis CI master branch:
Travis CI release branch:

MAP-Tk command line tools are placed in the bin directory of the build or install path. These tools are described below.

The primary tools are maptk_track_features and maptk_bundle_adjust_tracks. Together these form the sparse bundle adjustment pipeline. The other tools are for debugging and analysis purposes.

maptk_track_featues

Takes a list of images and produces a feature tracks file.

maptk_bundle_adjust_tracks

Takes feature tracks and produces cameras (KRTD files) and 3D points (PLY file). Can also take input POS files or geo-reference points and produce optimized POS files.

maptk_pos2krtd

Takes POS files and directly produces KRTD.

maptk_analyze_tracks

Takes images and feature tracks and produces tracking statistics or images with tracks overlaid.

maptk_estimate_homography

Estimates a homography transformation between two images, outputting a file containing the matrices.

Each MAP-Tk tool has the same interface and accepts three command line arguments:

• -c to specify an input configuration file
• -o to output the current configuration to a file
• -h for help (lists these options)

Each tool has all of its options, including paths to input and output files, specified in the configuration file. To get started, run one of the tools like this:

$ maptk_track_features -o config_file.conf

This will produce an initial set of configuration options. You can then edit config_file.conf to specify input/output files, choices of algorithms, and algorithm parameters. Just as in CMake, configuring some parameters will enable new sub-parameters and you need to re-run the tool to get the updated list of parameters. For example:

$ maptk_track_features -c config_file.conf -o config_file.conf

The above command will overwrite the existing config file with a new file. Ordering of entries and comments are not preserved. Use a different output file name to prevent overwriting the original. Continue to adjust parameters and re-run the above command until the tool no longer reports the message:

ERROR: Configuration not valid.

Note that the config file itself contains detail comments documenting each parameter. For each abstract algorithm you must specify the name of variant to use, but the list of valid names (based on which modules are compiled) is provided directly in the comment for easy reference. When the config file is complete and valid, run the tool one final time as:

$ maptk_track_features -c config_file.conf

An easier way to get started is to use the sample configuration files for each tool that are provided in the examples directory. These examples use recommended default settings that are known to produce useful results on some selected public data samples. The example configuration files include the default configuration files for each algorithm in the config directory.

MAP-Tk is a component of Kitware's collection of open source computer vision tools known as KWIVER. Please join the kwiver-users mailing list to discuss MAP-Tk or to ask for help with using MAP-Tk. For less frequent announcements about MAP-Tk and other KWIVER components, please join the kwiver-announce mailing list.

The authors would like to thank AFRL/Sensors Directorate for their support of this work via SBIR Contract FA8650-14-C-1820. This document is approved for public release via 88ABW-2015-2555.