/AutoStitching

Automatic topology constructing and global stitching optimization for large-scale unordered UAV aerial images. [Pattern Recognition 2017]

Primary LanguageC

Introduction

This project is a generic framework for globally consistent alignment of images captured from approximately planar scenes via topology analysis. Specifically, it can resist the perspective distortion meanwhile preserving the local alignment accuracy. To guanrantee the alignment accuracy, global topological relations of images are searched firstly, and then a global optimization on alignment parameters are performed.

This C++ implemented algorithm is described in
"Globally Consistent Alignment for Planar Mosaicking via Topology Analysis", Pattern Recognition (PR), Jan. 2017.
Notice: This program is free for personal, non-profit and academic use. All right reserved to CVRS: http://cvrs.whu.edu.cn/. If you have any question, please contact: menghanxyz@gmail.com (Menghan Xia)

Here is an example for demonstration below (image topological graph and alignment result):

Usage

Dependent Libarary [compulsory]:

OpenCV 2.4.9 is recommended.

1. Project Configure:

This procedure is developed on Visual Studio 2010 under Windows 8.1 system environment, where the source code is organized with CMakeLists. So, before opening it in Visual Studio, you need to configure the project with CMake.

2. Running and Test:

2.1 Default folders in "data":

  • "Images" : put your source images in it.
  • "Cache" : used to store those intermediate results (feature points files, matching point files, topological matrix, etc) that are required by the final alignmnet optimization.

2.2 Running parameter settings:

  • Set your reference image for alignment --> Variable 'refNo' in Function main() in "main.cpp". e.g. refNo=-1 means that the program will automatically selects a reference via topological analysis.
  • Set whether your image set is sequential order or not --> Variable 'isInorder' in Function imageStitcherbyGroup() in "alignment.cpp"
  • Set whether your model need global optimization --> Variable 'needRefine' in Function imageStitcherbyGroup() in "alignment.cpp"

Besides, to use our preset working directory successfully, do not forget to UPDATE the path variable 'baseDir' as the absolute path of your directory "data" in the source file "source/Utils/util.h" [line 16]

So far, you can run the procedure and check the alignment/stitching results now.