Control Engineering Europe – March 2019

Control Engineering Europe http://www.controlengeurope.com March 2019 23

MACHINE VISION

synchronisation of multi-camera systems
by utilising the IEEE 1588 Precision Time
Protocol to allow each camera on the
network to be simultaneously triggered.

The latest version
Version 2.1, announced in 2018 now
features multi-part transmission. This
allows the sending of more complex
data structures used in 3D imaging or
any application which would benefit
from a three-coordinate data structure.
The continuing popularity of the GigE
Vision standard is undoubtedly related
both to the distances over which data
can be transmitted and the ease of
integration using industry standard
components. It therefore comes as
no surprise that increasing the data
transmission rate within this framework
in order to accommodate greater data
volumes is a continuing goal.
Different approaches have been
applied to boost standard GigE
Vision capabilities, ranging from link
aggregation to proprietary software
solutions. Nevertheless developing more
wide-ranging Ethernet solutions is a

preferable approach. The networking

stack used is divided into a number

of different layers. The ‘Ethernet’

layer is unaware of the protocols

and connections of the layers either

above or below it which means the

technology is future proofed. The

latest development is the use of

NBASE-T to allow users to get the

benefit of significantly increased data

throughput without the expense

and difficulty of replacing existing

Ethernet cabling. Fully compatible with

GigE Vision, NBASE-T is an extension

to the IEEE 802.3 Ethernet standard

and increases data transmission

using industry standard CAT 5e cable

to speeds of 2.5 and 5 Gb/sec for

2.5BASE-T and 5BASE-T respectively

and to 10 Gb/sec using CAT 6A cable

for 10BASE-T. 5BASE-T exceeds the

data transfer speeds of both USB 3.1

(Gen1) and USB 3.0, while 10BASE-T

also exceeds rates for CameraLink

(Full/80-bit) and CoaXPress (single lane

CXP-6). They all maintain the original

GigE Vision data transmission distances

of up to 100 metres. A number of

5BASE-T and 10BASE-T cameras are

now available from UKIVA members,

to take advantage of these improved

data transfer rates.

No popularity decline

Stemmer Imaging has supplied and

supported a wide range of GigE

Vision cameras since the standard was

introduced. Mark Williamson, managing

director at Stemmer Imaging UK, has

seen no decline in their popularity.

He said: “We are particularly excited

by a new range of 5BASE-T cameras

that combine the higher inherent data

transfer rates with in-built proprietary

technology utilising sophisticated pixel

analysis and processing to deliver an

overall bandwidth of 985 MB/sec. That

is over double the speed of USB3 and

approaching the speeds of 10BASE-T,

but it requires only standard Cat5e

cables that operate up to 100m cable

lengths. They are also significantly

physically smaller and consume less

power than 10 GigE Vision cameras,

allowing deployment in a wider range

of factory applications.”!

The capabilities of a given machine vision or embedded

vision system rely on more than just the hardware of the

camera. The central hardware components of a camera

often are viewed as solely responsible for the system’s

performance – components such as the sensor, the lens, the

interface and more are given all the credit, but software can

be just as impactful.

The software used in machine vision and embedded

vision systems has been making major advances. From the

food and beverage industry to the electronics industry,

improving machine vision software is having an impact.

While advances have happened across the board, there are

four areas in which major gains have been made.

High dynamic range inspection: The electronics industry is

running lines faster and with higher resolution, introducing

higher signal-to-noise ratios (SNRs). Companies are using

machine vision software to reduce SNRs for more accurate

inspections. Depending on the manufacturer’s need, this

software can be highly customised.

Optical character recognition software: Optical character

recognition (OCR) historically has been limited to black

text on white backgrounds, but this is rarely the case in

real-world manufacturing. New OCR software can detect

characters on busy backgrounds, and some of the newest

software even can recognise characters when new fonts are

introduced.

3D imaging: A number of manufacturers have been

developing photometric 3D algorithms to read 3-D text,

among other things. Some forms of 3D vision software

combine multiple images of the same part to enhance

contrast in 3D images for more accurate imaging results.

Deep learning: Some companies are making strides in

deep learning, and this technology is one of the latest

buzzwords. A few of the largest software developers have

image classification tools based on deep learning that

detect hard-to-define defects and product variations.

Machine vision software plays a critical role in the

performance of a vision system. Today, identifying the right

software for a machine vision application is as important as

finding the right hardware.

This article originally appeared on http://www.controleng.com

Machine vision software advances