Resume of Guoling Shen
Current Research Projects
1. Surface Reconstruction Using Laser Range Scanner
With the advent of triangulation-based laser range scanners which are capable
of quickly capturing a large amount of range data with high accuracy, surface
reconstruction systems using laser scanners find applications in various
areas such as reverse engineering, virtual reality, and inspection of products,
etc. In recent years, lots of excellent research work have been done on
this topic, and a number of surface reconstruction systems have been developed
and entered into application, especially in manufacturing inspection and
entertainment industry. However, most of the current systems are application-oriented
(designed for specific laser scanners and system configurations), and therefore
hard to be generalized for other applications. Some key issues, such as
the robustness of data registration, the topological correctness of data
integration, and a general next-best-view algorithm, are not well addressed,
especially for large workspace.
The ultimate objective of this project is to develop an automated surface
acquisition and reconstruction system for complex objects in a large workspace,
using laser range scanner. Complex objects mean no constraint on the geometry
and the topological features of objects and neither of them is known a priori. Large workspace means that the viewing volume of the system is large compared
to the field of view of the scanner, and the driving devices have more degrees
of freedom to position the laser scanner and/or the object.
We will study the sufficient conditions of scanning for a successful registration
for a specific data registration method, e.g. the iterative closest point algorithm. Meanwhile, we will develop a next-best-view algorithm which is
collision-free and feasible in the sense of computation expense.
2. Inspection of Shiphull Using Laser Scanner
In the shipbuilding industry, at the stage of block assembly at the erection
site, adjustment work (trimming, hammering...) accounts for half of the
total fitting time. In addition, some of the substantial welding (metal
melting) time, which accounts for 1/3 of the total welding time, is expended
on depositing extra metal into the wide gaps left by inaccurate fitting.
Therefore, accuracy control at all stages of shiphull construction is essential
to eliminate extra rework during assembly. A successful accuracy control
system depends on a reliable, accurate and time-saving measurement method.
Current measurement techniques used at shipyards include photogrammetry,
theodolites, digitizers, lasers and CMM. None of these techniques are automated.
Some of them need more than one operator and thus are time-intensive, and
the measuring result is not real time (need post-processing). Most of them
measure a set of points of interest, which are marked on the manufactured
structure by retroreflective or some other kinds of labels, and therefore,
the measurement is not complete. None of them can be used to inspect a curved
shape.
With its high accuracy and data acquisition speed, laser scanner provides
the possibility to quickly measure a manufactured structure with a large
amount of data (not just a bunch of points). With the use of robots and
the development of software, it is also possible to automate the measuring
process.
Previous Research Projects
1. Approximation Measured Data with Interval B-Splines
The advent of accurate 3D laser ranging sensors that provide the capability
of producing great amount of precise data sets in extremely short time,
has created interest in automated methods for reverse engineering. A typical
application of reverse engineering is to build a CAD model from a manufactured
object which has no accurate electronic model. However, since measured data
always possess uncertainty arising from sensor precision and measurement
registration, most algorithms for surface reconstruction may produce an
ambiguously defined object and cause a solid modeling system fail due to
possible numerical instabilities when manipulating the geometric and topological
representation with complex operations. Even in the absence of measurement
uncertainty, such a failure could still happen due to the limitation of
computation precision. This project attempts to provide an approach towards
the robustness of surface reconstruction for reverse engineering. In the
approach, measured data are represented as interval data points by involving
all sources of errors, and interval surface fitting methods are developed
to make it possible to construct interval solid models which with robust
solid modeling operations based on rounded interval arithmetic, will build
up a robust solid modeling system. The creation of interval surfaces is
accomplished by either interpolation or approximation. A linear programming
process is used to minimize ranges of the resulting surfaces.
2. Numerical and Geometric Properties of Interval B-splines
As a newly developed geometric representation introduced for increased robustness
in solid modeling systems and for uncertainty representation in data fitting
problems, interval B-splines have their unique numerical and geometric properties
different from classical B-splines.
In order to preserve conservative enclosures, the computation of interval
B-splines is implemented in rounded interval arithmetic (RIA). The error propagation in the RIA evaluation of B-spline basis functions
is studied and an upper bound of the width of the function values is given.
The width of an interval B-spline curve or surface is mainly contributed
to the widths of its control points.
Unlike classical B-splines, interval B-spline curves and surfaces are no
longer invariant under transformation, if the computation is in RIA. However,
both translation and rotation are still conservative, or in other words,
in the new coordinate system, the exactly transformed interval curve or
surface is contained in the one defined by the transformed (using RIA) control
points.
The hodograph curve of an interval curve is defined as its derivative. An
interval point on the hodograph curve bounds the range within which the
tangent vector can vary. Generally, the hodograph curve intends to inflate
in width, and will become meaningless in predicting the tangent direction
if the original interval curve has a wide range.
An interval B-spline curve can also be represented as a single valued central
curve with a vector-valued error function defined as a piecewise polynomial
function in the B-spline basis. This project also studies the extraction
of the boundary of the area or volume defined by an interval geometry.
Other properties studied in this project include variation diminishing property,
knot insertion, etc.
Education
- Bachelor of Science in Marine Engineering and Naval Architecture, 1988
- Huazhong University of Science and Technology, P.R.China
- Thesis: Software Development for Signal Processing
Research Interests
- CAD/CAM
- Computational Geometry
Publications
- Approximation of measured data with interval B-splines. S. T. Tuohy, T. Maekawa, G. Shen and N. M. Patrikalakis. To appear on Computer-Aided Design.
- Numerical and Geometric Properties of Interval B-Splines. G. Shen and N. M. Patrikalakis. Submitted to International Journal of Solid Modeling for publication.
Professional Experience
- Research Assistant, Department of Ocean Engineering, MIT
- September 1994 to present
- work in the areas of CAD/CAM and computational Geometry; develop new techniques
for various applications; Develop software to implement new methods.
- Construction Engineer, Shanghai Shipyard, Shanghai, P.R.China
- September 1988 to July 1994
- Schedule and supervise the installation and test of main engine and various
equipments both in engine room and on decks.
Guoling Shen
Massachusetts Institute of Technology
Department of Ocean Engineering
77 Massachusetts Avenue, Room 5-423a
617-258-7813
glshen@mit.edu
URL: http://deslab.mit.edu/DesignLab/people/Shen.html
Last modified: September 24, 1997