International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 04 Issue: 07 | July -2017
p-ISSN: 2395-0072
www.irjet.net
Performance Analysis of Iterative Closest Point (ICP) Algorithm using Modified Hausdorff Distance Arunava Seal1, Abhijit Bhowmick2 1School
of Electronics Engineering, VIT University, Vellore, India Professor (Selection Grade), VIT University, Vellore, India ---------------------------------------------------------------------***--------------------------------------------------------------------2Assistant
Abstract - Registration, tracking and reconstruction of 3-D
of the meshes. This system has been used for registering the outputs of 3D scanners, which typically only scan an object from one direction at a time. ICP starts with two meshes and an initial guess for their relative rigid-body transform, and iteratively refines the transform by repeatedly generating pairs of corresponding points on the meshes and minimizing an error metric. Many different variants of ICP have been utilized over the years for different purposes. We aim to analyze these variants and their performance and accuracy parameters in terms of Modified Hausdorff Distance which gives us a measure of the similarity between two-point sets.
models in real time has many applications in the fields of pose estimation, alignment, motion estimation, object recognition, handheld scanning and augmented reality. Many varied techniques and approaches have been used to tackle the problem of registration and reconstruction of 3-D images. Our work aims to design a method for accurate and computationally efficient registration of 3-D shapes, through comparison and application of various alignment techniques. The expected outcome is a system able to align and reconstruct geometrically precise complex 3-D models in a fast and efficient manner. We mainly focus on the Iterative Closest Point Algorithm, which is the most popular method used in this field through many different forms and variations. An effort has been made to analyze these various forms which are used for object registration through the use of Modified Hausdorff Distance (MHD). Usage of outlier removal methods such as Random Sampling Consensus (RANSAC) and accelerating techniques such as kd-Tree search are also studied. This system could act as a foundation for implementing the ICP algorithm along with depth images or Simultaneous Localization and Mapping (SLAM).
The initial part of the work entails the alignment of 3D models using the various ICP algorithms and optimization. We utilize the basic ICP model for different point data sets and analyze the shortcomings and scope for improvement. Based on that, we optimize the algorithm for efficiency and performance.
1.1 Related Work Most of the early work performed in the field of object tracking, object recognition and registration revolved around global shape matching or registration having very limited classes of shapes. O. D. Faugeras and M. Hebert were among the first to perform free-form shape matching using 3-D data using a Renault auto part in the early 1980s. [1] The works of P. J. Besl, Brian Curless and Marc Levoy laid down a mathematical background upon which further work in the field would be carried out.[2][3] The first proper method for registration of 3-D shapes was performed by P. J. Besl and Neil Mckay, when they introduced the Iterative Closest Point (ICP) algorithm in 1992, which is still used to this day in various optimized forms.[4][5] The ICP algorithm provides a solution to the free-from surface matching problem and provides a general unified solution to the point-set matching problem without correspondences and the free-form curve matching problem. Besl and Mckay’s implementation required no extracted features, no surface or curve derivatives, and no pre-processing of data. The main application was to register digitized data from unfixtured rigid objects with an idealized geometric model prior to shape inspection. Many variants of the ICP algorithm have been used in diverse applications and with various parameters, some of which were thoroughly compared by Rusinkiewicz and Levoy in 2001. [6] Some of the most popular variants of ICP include Comprehensive ICP (CICP), Trimmed ICP (TICP), Picky ICP(PICP), Non-rigid ICP, Scaling
Key Words: Iterative Closest Point(ICP), Object Modeling, Object Alignment, Pose Estimation, 3-D Registration, Modified Hausdorff Distance
1. INTRODUCTION In recent years digitization of physical objects has become easy with the growing popularity of 3D scanners. Scanners use laser, light or x-rays to form a point cloud defining the shape of the object being scanned. Registration is useful in comparing two scans taken at different time/conditions and is also helpful in combining the information in two scans into a single one. This paper aims to design a method for accurate and computationally efficient registration of 3-D shapes; the expected outcome being a system able to align and reconstruct geometrically precise complex 3-D models as efficiently as possible. The Iterative Closest Point Algorithm of Besl and Mckay is one of the most popular method used for rigid transformation of roughly aligned data sets. It is widely used for the registration of free form surfaces where dense data is assumed, and a good initial estimate is available or can be easily estimated. The ICP algorithm has become the dominant method for aligning three dimensional models based purely on the geometry, and sometimes color,
Š 2017, IRJET
|
Impact Factor value: 5.181
|
ISO 9001:2008 Certified Journal
| Page 2794