LidarK20.rar_R树索引_lidar_lidar数据处理_索引 树

GML LIDARK LIBRARY 2.0 - README.TXT FILE =============================================================== Copyright (c) 2007-2010 MSU Graphics & Media Laboratory. mailto:vision@graphics.cs.msu.su =============================================================== INTRODUCTION: =============================================================== The LidarK library is an open-source library that allows performing different operations on multidimentional point data such as 3D LIDAR scans. Currently, it allows building a spatial index. Though it is intended to be used on LIDAR scans, it can be helpful for a wide range of problems which require spatial data processing. The latest version of the library can be downloaded from http://graphics.cs.msu.ru/en/science/research/3dpoint/lidark THEORETICAL FOUNDATIONS: =============================================================== The library core implements the original notion of Mass-Tree. Mass-Tree is the modification of R-Tree, originally introduced by Antonin Guttman in 1984. Mass-Tree is more suitable than R-Tree for building dynamic index on the real data such as LIDAR scans. It yields more compact blocks and could be considered as a heuristic way to perform cloud over-segmentation. The key differences are: * node split uses the k-means clustering algorithm * tuple insertion uses the k-pseudocentroids algorithm The kNN retrieval is implemented according to the original PruneDescent algorithm which works extremely fast on the chosen data structure. For getting more detailed information see http://graphics.cs.msu.ru/en/science/research/3dpoint/lidark LIBRARY API: =============================================================== Library user has the following abilities: * create tuple that contains point and to create his/her own tuple implementation that contains any meta-data as well * create Seg-Tree by a tuple collection or a single seminal tuple. Empty tree is not allowed * add tuples to the created tree * check if a tree node is a leaf and then get either children tuples or children nodes * get information about node bounds * find k nearest neighbours of a point * find all tuples in the given radius * find all the tuples/leafnodes intersecting the given figure * save the created tree to disk and load it from disk There are some examples explaining how to use the API. ARCHIVE CONTENTS: =============================================================== bin -- the compiled release code under MS Windows * RTree.lib -- the library kernel * RTreeKNN.lib -- the classes for the nearest neighbors retrieval cppSource -- C++ source code of the library and unit tests * RTree -- library kernel classes * RTreeKNN -- library classes for the nearest neighbors retrieval * MatlabDll -- library code for integration with MATLAB * Tests -- C++ unit tests that provide examples of using API doc -- doxygen generated documentation * html/index.html -- dicumentation index page Matlabtools -- MATLAB wrapper for the library * runTestGUI.m -- launches the GUI TestData -- examples of point clouds License.txt -- text of the license agreement ReadMe.txt -- this file LICENSE TERMS: =============================================================== See the accopanying license.txt file. IMPLEMENTATION: =============================================================== The code provides a set of classes which allows building Seg-Tree, get nodes, get children of the node, get k nearest neighbors of a point etc. Also we implemented a MATLAB wrapper for the library and some unit tests that show how to work with C++ and MATLAB library versions. The library was developed on Microsoft Windows XP + Microsoft Visual Studio 2005 but it was also tested on Ubuntu 9.10 Karmic Koala + gcc-4.4.4. The MATLAB interface was tested on MATLAB R2007b. You can find the .sln and Makefile right in the cppSource folder. MATLAB WRAPPER: =============================================================== The wrapper is now implemented as MATLAB MEX file. The binary file (for Windows) is already in the Matlabtools directory. The rest of the section is intended to those who want to rebuild it. To build it in MSVS, just build the LidarKdll project within the solution. You will need to change the path to your MATLAB directory. Namely change the following project properties: * C/C++ -> General -> Additional include directory * C/C++ -> Linker -> Additional dependencies Note that MATLAB MEX file compilation was tested only under Windows (within the MSVS project). There is the corresponding target for compiling MEX in the Makefile, however it was not tested. You also need to specify the path to Matlab dir there. After building, copy the binary (e.g. .mexw32 file) to Matlabtools and rename it to LidarKIndex, keep the extention. CHANGE LOG: =============================================================== 2.0 (10.09.2010) The second major release. There are subtle changes in the API. * One can create a tree with as few as one point. * Tree root now implements the node interface, so in most cases no special handling for the root is needed any more. * One can check if a node is leaf and get unmodifiable view to the children tuples or the children nodes respectively. * 3D Point arithmetic became more powerful. * kNN retrieval is now possible for any point, not only the ones belonging to the tree. * A bunch of tree iterators are added. They allow finding tuples in the given radius (spherical iterator), or all the tuples that fall into the given region of space (topological iterator). * Most of classes are parametrized with template parameters now. The parameter defines the class of tuples used in the container. You don't need to cast tuples to their concrete implementations any more. It is a controversive decision, since the code became more ponderous. However, the old-style class names are aliased, so some degree of backward compatibility should hold. * MATLAB integration is now done via MEX files. * Some extremely seldom reproduced bugs are fixed. * Hand-made unit tests are added. * Doxygen-generated documentation is prepared. 1.2 (05.12.2008) Some memory leaks were fixed. 1.1 (09.09.2008) Some minor changes were performed. * The growing tree bug was fixed. 1.0 (01.07.2008) The first library release. FEEDBACK: =============================================================== If you notice a bag in the code, or you have some ideas on how we can improve the library, please contact us: avelizhev@graphics.cs.msu.ru - Alexander Velizhev shapovalov@graphics.cs.msu.ru - Roman Shapovalov We always appreciate your assistance!