CE6023

Lighting
Instructor: Patrick Denny
Lighting
• Computer Vision Systems depend on the availability of light to obtain information about an
environment under consideration.
• When light sources are outside the control of a computer vision system, we refer to them as
passive illumination
• Examples include relying on daylight as the illumination for a self-driving car system
• Thermal imaging based weapons systems also rely on the passive thermal emissions of
targets
• When the computer vision system itself controls the provision of the illumination, we refer to
them as active illumination
• Examples include computer vision systems on production assembly lines where the
computer vision system controls the lighting.
• Precision laser-guided bombs (“smart weapons”) are guided weapons that use a laser to
“paint” a target so that weapon can be guided to it. Note that laser is not part of the
guided weapon itself, but is part of the larger computer vision system that contains both
the laser and the guided weapon, so in that sense the laser is active illumination.

(c) Patrick Denny 2024

Why active illumination sources?
• Passive illumination sources are satisfactory for systems of low-
consequence and that do not rely on high-performance and high-availability
• If there isn’t enough light for your mobile phone camera, at worst you’ll
get a bad photograph
• However, if you have a system containing a computer vision subsystem,
which requires
• high-availability and high-reliability,
• and that system
• does not contain redundancy that mitigates a dependency on the
operation of the computer vision system
• then active lighting is required

(c) Patrick Denny 2024

Why active illumination sources?
• Active illumination usually requires a higher specification than ambient room lighting
• Ambient lighting from overhead lighting fixtures or daylight are usually insufficient to illuminate
targets for a machine vision system
• Usually, the lights are not bright enough or close enough to the application environment to provide
the focused light necessary for favourable image properties such as good contrast.
• Ambient lighting changes throughout the day too; the sun may be shining through a window in the
morning but not in the late afternoon or a light may be sometimes on or sometimes off.
• This variation presents a different environment to the computer vision systems that are outside the
control or knowledge of that system, which gives an undeterministic variation in the performance of
the computer vision system.
• In order to optimize image quality, a dedicated light source should be used in any vision application.
• Dedicated lighting
• Optimizes contrast between the target object or feature and its background
• Provides uniform lighting conditions that allow image capture to be unaffected by ambient
lighting in the factory environment and this
• Simplifies image analysis by creating high contrast between “good” and the “bad” features of
interest, thereby improving the discrimination of the system.

(c) Patrick Denny 2024

Choosing the right lighting technique
• The right lighting technique can vary greatly depending on the application and factors like the target’s optical
properties such as
• Shape
• Reflectivity (at wavelengths of interest)
• Translucence
• Texture
• Target size
• Camera resolution
• Flicker
• Distance from camera
• Level of detail required
• Uniformity of lighting across an object
• Facilitation of green screening / color keying
• We’ll consider 5 common computer vision lighting techniques from the Quality Magazine article How to Choose a
Vision Lighting Technique

(c) Patrick Denny 2024

Ring Lighting

• Ring lighting is the tried-and-tested

standard in vision lighting, used most
often in general vision inspections
• Standard ring lights mount directly to
the camera or vision sensor and both
parts can be mounted as one piece
for convenience.
• Ring lights brightly illuminate the area
directly in front of the camera and are
especially useful for inspecting small
parts.

(c) Patrick Denny 2024

 Backlighting

• With backlighting, the light source – normally a specialized

backlight that provides even, high-intensity light – is positioned
behind the object and shines directly into the camera.
• Backlighting is the highest contrast lighting technique available,
producing a binary image that is easy to process.
• For opaque objects, backlighting creates an image that shows a
dark shadow of the part (a silhouette)
• The silhouette can then be inspected for proper size and shape.
• The high-contrast images created using a backlight are ideal for
high-accuracy applications like precision measurement, gauging
small parts, detecting the presence/absence of holes, and
verifying part placement and orientation
• For translucent objects, backlighting is a common technique
when looking for defects or detecting degree of translucency.
• Backlighting is not a suitable technique for detecting surface
colour or texture

(c) Patrick Denny 2024

 Directional Lighting

• Directional lighting creates shadows to detect changes in depth,

illuminate specific surface angles, and avoid glare of reflective
surfaces when directed at an angle away from lens. Directional
lighting can be either bright-field or dark-field.
• In bright-field directional lighting, the light source is positioned
so that it shines more or less perpendicular to the object, making
the object appear bright.
• The features on the object’s surface are visible but not highly
contrasted.
• However, bright-field lighting can produce glare and hotspots on
highly reflective surfaces and may not be able to show surface
texture.

(c) Patrick Denny 2024

 Directional Lighting

• In dark-field directional lighting, the light

source is positioned at a low angle, so that its
light reflects off the object, away from the
camera.
• Dark-field lighting results in a darker overall
image, but surface defects, raised features, and
textures appear bright and stand out from the
rest of the target.
• Dark-field lighting reduces glare, making it ideal
for inspecting shiny objects, and this technique
can also be used to measure height changes
by creating shadows that can easily be
measured.

(c) Patrick Denny 2024

 Directional Lighting

• Area lights, spotlights, and bar lights are common

light sources that provide directional lighting.
• Area lights and spotlights provide even illumination in a
concentrated area and can be mounted within your
existing inspection, independent of the vision sensor’s
optical axis.
• Users can change the direction or distance of the light
for improved optical contrast without adding fixtures or
realigning the inspected parts.
• Meanwhile, bar lights, also known as linear array lights,
offer higher-intensity lighting to illuminate large areas,
while cutting through dust, mist, and grime. Bar lights
provide clean, bright, and even illumination to create
the optical contrast necessary for reliable accurate
inspections of larger objects at longer distances.

(c) Patrick Denny 2024

Low Angle Ring Lighting

• A form of dark-field lighting, a low-angle

ring light has its LEDs mounted in a ring,
pointing toward the part almost
perpendicular to the camera’s direction of
view.
• Low angle ring lights create shadows and
bright spots to detect changes in depth and
surface texture.
• Low-angle ring lighting is ideal for
inspecting surface defects, etching, and
embossed marks.

(c) Patrick Denny 2024

 On-axis Lighting
• A form of bright-field lighting, an on-axis light uses
a beam splitter to focus light along the camera’s
optical axis, providing direct and diffuse light that
renders reflective surfaces bright without glare.
• The camera looks down from the top, through the
on-axis light, to the target part below.
• This technique can be used to eliminate
shadows, inspect shiny object, or detect surface
texture differences.
• The on-axis light and the low angle light produce
opposite images in terms of which features
appear dark and bright.

(c) Patrick Denny 2024

Other lighting types

• There are several other lighting

types that are hybrids which
provide a balance of
performance for the application
requirements as can be seen in
the example figure

(c) Patrick Denny 2024

Choosing the right wavelength of light
• In addition to choosing the right lighting technique, different wavelengths of light can also be used to create additional
contrast, draw out features of interest, or reduce the visibility of insignificant features.
• Colours of Light
• Different colors of light can be used to create additional contrast to either obscure or draw out specific features.
In general, when a light source is a similar color to the part, it will make the part appear brighter. For example, a
red part illuminated with a red ring light will make the red part look bright on the image.
• Conversely, when a light source color is different from the color of the part, the part will appear darker. An
example is when the same red part is illuminated by a green light, the red part will appear dark.
• Ultraviolet (UV) Light
• Ultraviolet (UV) light can be used to draw out features of interest when there is very little contrast from the
background, such as beads of clear adhesive on a part.
• Shining UV light on the adhesive may create a glowing effect (fluorescence) that makes the adhesive stand out
clearly from the background. This makes it easier for the vision system to reliably identify when adhesive is
missing or applied incorrectly.
• Infrared (IR) Light
• Infrared (IR) light can be used to hide insignificant features by reducing the contrast of certain objects. For
example, some types of ink that appear dark in the visible spectrum reflect large amounts of IR light. In these
cases, ink may disappear in a grayscale image—an effect that can be used to hide certain insignificant features.
IR light can also be used to reduce glare and reflections when inspecting shiny objects.
• Lidar cameras in self-driving vehicle environments use IR lasers that people can’t see

(c) Patrick Denny 2024

 Get started with vision
lighting – questions to ask

• Now that you understand the most

common vision lighting techniques, it is
important to understand how these apply
to your specific application.
• When choosing a vision lighting solution,
the first step is to thoroughly evaluate
your application requirements.
• The following are five key questions to help
you identify the right lighting technique and
wavelength for your application

(c) Patrick Denny 2024

What are the optical properties of the target?

• Understand your target and how the

camera sees it.
• For example, is the target highly reflective?
Under bright-field lighting, very shiny
objects will reflect most of the light back to
the camera, creating glare and making the
image difficult to process accurately.
• Meanwhile, dark-field lighting is a form of
indirect lighting that reduces glare when
inspecting reflective targets.
• Consider how the lighting leverages the
optical properties of the target to make the
desired image features more
distinguishable
• (Image taken from here)
(c) Patrick Denny 2024
 • Articulate exactly what the camera needs to see.
• Inspecting an object for surface imperfections will
What are you require a different lighting technique than inspecting
an object to verify the presence/absence of specific

looking for?
features.
• Several examples can be found in the article from
Machine Vision Basics – Vision Examples in the
Electronic Device Industry

(c) Patrick Denny 2024

What size is the target?
• What size are the features being detected/measured?
• Clarify both the size of the target and the size of the features being inspected.
• For example, ring lighting and backlighting are both common techniques for inspecting small parts.
• However, when inspecting surface texture or small raised features, dark-field directional lighting is often a
better choice.
• Also, when considering the size of the target, consider how large the target will appear in the image and
in particular how large the features will need to appear in the image in order to be resolved. This will have
you consider the size of the target in the camera’s field of view and the size of the features of the target in
the image.

(c) Patrick Denny 2024

 • This question is similar to the previous one but includes considerations like camera
focussing, the amount and homogeneity of the lighting and ultimately how large the
object and features appear in the images that are created.

How far away • What are the mounting constraints? Know the space you are working with and what
mounting options are available.

is the target? • For example, in applications where mounting options are limited, a ring light that
attaches directly to the camera means that there is only one device to be mounted.
On the other hand, large targets may require larger lights mounted farther away to
properly illuminate the entire surface.
• (Image from article on Intel’s RealSense facial recognition)

(c) Patrick Denny 2024

What is the speed of
the application?
• Know how fast your target will be travelling when inspections are in
progress.
• The actual speed of the application can significantly impact
inspection results compared to tests on slow-moving or stationary
targets.
• For example, a stationary target might be easy for a camera to
capture; however, the same target in the same environment moving
at high speeds can result in a blurred image.
• The right lighting technique can reduce the impact of application
speed to produce a better image that is easier for the camera to
process accurately.
• In general, brighter, more powerful lighting is often necessary for
high-speed applications so that a very high shutter speed can be
used to capture an image in a relatively short amount of time.
• However, to avoid LED flicker behaviours, it may be necessary to
co-ordinate high speed lighting with the camera exposure.

(c) Patrick Denny 2024

Test and validate the lighting
• After answering these questions, the next step is to test out a
solution.
• Every vision application is unique, and it is best to try the
solution in real-world conditions to make sure it works for your
application.
• Consider how the vision system needs to be maintained and
how it can be checked during use
• Ultimately, every computer vision system has a lifetime and
the system design and validation must reflect this.

(c) Patrick Denny 2024