Professor Kenichi Kanatani

Short Biography

Kenichi Kanatani was born on August 12, 1947 in Okayama, Japan. He received his B.S., M.S, and Ph.D. in applied mathematics from the University of Tokyo, Japan, in 1972, 1974, and 1979, respectively. He joined the Department of Computer Science, Gunma University, Kiryu, Japan, in April 1979 as Assistant Professor. He became Associate Professor and Professor there in April 1983 and April 1988, respectively. From April 2001, he is Professor of Computer Science, Okayama University, Okayama, Japan. He was a visiting researcher at the University of Maryland, U.S.A., the University of Copenhagen, Denmark, the University of Oxford, U.K., and INIRA at Rhone Alpes, France. He is the author of ``Group-Theoretical Methods in Image Understanding'' (Springer, 1990), ``Geometric Computation for Machine Vision'' (Oxford University Press, 1993) and ``Statistical Optimization for Geometric Computation: Theory and Practice'' (Elsevier Science, 1996). His research career started with studies of theoretical continuum mechanics (elasticity, plasticity, and fluid) and its application to mechanics of granular materials such as powder and soil, but his research interested has shifted to mathematical analysis of images and 3-D reconstruction from images. Currently, he is devoted to mathematical analysis of statistical reliability of computer vision and optimization procedures.

He has received many awards, including:

· Best Paper Award of IPSJ (Information Processing Society of Japan) in 1987

· Telecommunication System Technology Award of the Telecommunication Advancement Foundation in 1999

· Information Technology Promotion Award of Funai Foundation for Information Technology in 2005

· Best Paper Award of IEICE (Institute of Electronic, Information and Communication Engineers) in 2005.

· Information and System Society Activity Service Award of IEICE (Institute of Electronic, Information and Communication Engineers) in 2005.

He was elected IEEE Fellow in 2002.

Title: Statistical Optimization for Geometric Fitting: Theoretical Accuracy Bound and High Order Error Analysis

Abstract

A rigorous accuracy analysis is given to various techniques for estimating parameters of geometric models from noisy data for computer vision applications.  First, it is pointed out that parameter estimation for vision applications is very different in nature from traditional statistical analysis and hence a different mathematical framework is necessary in such a domain.  After general theories on estimation and accuracy are given, typical existing techniques are selected, and their accuracy is evaluated up to higher order terms. This leads to a ``hyper accurate'' method that outperforms all existing methods.

Robert Valkenburg :

Biography

Robert Valkenburg received a Master of Engineering degree at the University of Auckland in 1989 and joined the DSIR computer vision group in Auckland, New Zealand. This group became a part of Industrial Research Limited in 1992, where he is currently employed. His initial work with the DSIR was in the area of high speed computing for computer vision and he developed a parallel computer architecture based on transputers.  In 1993 he began research into the application of Videometrics/photogrammetry to industrial problems.  He received a Diploma in Science (Mathematics) in 1994, and a Postgraduate Diploma in Science (Mathematics) in 1996, both from the University of Auckland.  His PhD dissertation, which he received from the University of Auckland in 2002, was on estimation theory and information inequalities on the Stiefel manifold.  Dr Valkenburg's current research interests include flexible 3D scanning, the application of Clifford algebra to computer vision, and photogrammetry.

Title: Interactive Hand-held 3D Scanning

Abstract: This paper describes a hand-held three-dimensional (3D) scanner for static objects and scenes ranging in size from less than a metre to tens of metres, indoors or outdoors. The scanner's pose is optically tracked relative to a constellation of active targets placed around the scene at the start of the survey. The system auto-calibrates the target locations and defines a scene coordinate system in which all scan data is subsequently represented. The scanner can capture 3D structure of almost arbitrary complexity very rapidly. Real-time visual feedback to the operator coupled with manual control of data filtering can result in artifact-free 3D point clouds or meshes. The scan data typically contains no or very few holes because the scanner can be manoeuvred to observe occluded surfaces, and oriented optimally for obtaining ranges to difficult surfaces.

 
Brendan McCane
 

 
 
Biography:

My background is in Computer Science and I completed my undergraduate studies and PhD at James Cook University of North Queensland in Australia. My PhD was entitled "Learning to Recognise 3D Objects from 2D Intensity Images", which I completed in February 1996. I then held a temporary position as a lecturer at James Cook University, before taking up a position in February 1997 as a lecturer with the Computer Science Department here at Otago University. I am now a Senior Lecturer. My research interests include computer vision, pattern recognition and machine learning. My research focus at the moment involves developing real-time eye and hand tracking techniques for use in a virtual sculpting environment. One of the major challenges is to develop robust systems that work in a variety of environments with a variety of users, and I am looking at utilising machine learning to enhance the robustness of these systems. I also have an interest in computer graphics and participate in the computer graphics group here at the University of Otago.

 
Title: Optimizing Cascade Classifiers
 
Abstract:
Cascade classifiers have received some attention recently for their ability to significantly speed up rare event detection problems. In particular, cascade classifiers have been used to produce real-time face detectors. However, in the past training parameters have been set using ad-hoc procedures. In this talk, I will show how we have built an empirical model which can be used to find optimal or near-optimal cascade classifiers. The method is demonstrated and validated on a wide range of classification problems including rare event and non-rare event problems. The method is shown to provide significant speed-ups for all problems and is therefore a viable technique for improving the efficiency of many classification algorithms.