UWE Bristol Engineering showcase 2015

Barry Clarbull
MEng Aerospace Manufacturing Engineering
Project Supervisor
Dr Gary Atkinson
Polarised Light in defect detection – Part B
Shape from Polarisation
The polarisation of light is a
phenomenon that can not be
detected by the human eye, for this
reason it is still a relatively new and
exciting technique within computer
vision.
Polarised light by reflection and
refraction
Light can be polarised in three ways,
the image above shows how light is
polarised by reflection, refraction
caused by a direct reflection, known
as specular polarisation and is shown
above. Polarisation can also be
caused by subsurface scatter, known
as diffuse polarisation and shown
below.
These different types of polarisation
have slightly different theory and
produced different amounts of
polarisation.
In this process a series of images
were captured using a linear polariser
at various angles. As the polariser is
rotated the intensity of the light
transmitted through it increases and
decreases and thus can be thought of
as acting like a Sine wave.
Calculations were completed to
obtain a the phase angle of the
polarised light, the degree of
polarisation on the object, and the
zenith angle of the surface normal.
From this data it is then possible to
reconstruct the object in3D.
A Snooker ball to be reconstructed
Phase angle image of a hemisphere
The Zenith angle at various points on
the sphere
The zenith angle is the angle from the
perpendicular, around to the surface
normal. This is shown in the diagram
below.
The data shown can then be
combined to show a vector at every
pixel. However as the phase angle is
only know up to 180° there is two
directions where this is true, meaning
that there is an ambiguity which
needs solving. This is a concave/
convex ambiguity and for this project
is solved by manually telling the
image which direction the vectors are
meant to be pointing, because of this
a ridge can be seen on the top of the
reconstruction.
3D Reconstruction of sphere
The 3D reconstruction above is
produced purely from diffuse
polarisation. A few anomalies are
visible on the art on the sides, as well
as a small flat on the top caused by
image saturation. However the
surface finish is smooth, and the
shape is quite well represented.
Project summary
The aim of this project was to investigate the use of
polarised light as a means of shape recovery and
quantify whether it could be adequately used for
detecting shape defects. As the project is very
complex much more emphasis was placed on
optimising shape recovery and 3D reconstruction.
Project Objectives
• Investigate the use of diffuse polarised light as a
method of shape reconstruction.
• Investigate specular polarisation and the
difference in the degree of polarisation that can be
obtained.
• Investigate pure specular reflection.
• Assess the use of polarised light as a method of
defect detection.
Project Conclusion
In conclusion good data was obtained to quantify a
shapes geometry using shape from polarisation was
very good. A 3D reconstruction was produced using
diffuse polarisation theory, however flaws in this
reconstruction are still present so the quality is not
good enough to use to asses for defects.
Further study would drive this project to combine it
with other techniques such as shape from shading to
automatically solve the concave/convex ambiguity
and to produce a much more consistent part.
Over all the project was a success and provided lot of
interesting data.