Real Time Vision Guided Roboti System Featuring FPGA Technology For Ima E Processing
Loading...
Date
2006-02
Authors
Akbar, Bharmal Muhammedali
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Since the early 80s computer vision h s been extensively applied in the
field of food processing, quality evaluation, and packaging so as to release
human labourers from this tedious task and th reby increasing factory throughput
and consistency. Complete robotic automati n for handling and packaging of
food products improves food hygiene while s isfying operation safety standards
such as Hazard Analysis and Critical Co trol Points (HACCP). The most
important aspect for an efficient automated ystem in the food industry is its
vision feed-back system. Two essential aspe ts for an efficient real time vision
system are, firstly it should accurately detect he position of single and isolated
objects as well as overlapping or occluding od objects so as to provide the
robot with information to position itself accu tely in order to manipulate that
object. Secondly the image processing result should be computed at real-time
and made available to the robot.
This project integrates the modified of the Hough transform
algorithm along with contour centroid detectio method as the main recognition
engine. This method accurately detects the osition of top-most object even
under severely overlapping cases, thus provid g the robot with accurate object
position details. Vision programs consist of mplex mathematical calculations
and number of repetitive instructions for each image pixel. Implementing such
programs on the general purpose processor PP) used in the desktop computer
for this project took an average of 5 seco ds to provide the results. This is
primarily because the GPP execute instructi s sequentially in multiplexed timesharing
manner. Dedicated hardware such s FPGA chip carry out real-time
concurrent processing on customized hardw e architecture to achieve real-time
computation speeds. For the image processi g algorithm, the data-independent
modules were identified for implementing hem in concurrent processes to
enhance computations speeds. The embed ed hardware is interfaced to the
PC's parallel port.
The image processing algorithm's me ods and procedures were tested
on commercially available burgers. The al orithm was successfully able to
identify the top-most minimally overlapped nd maximally exposed specie of
burger from the heap of burgers with an accur cy of 98%. The implementation of
the vision program on the FPGA chip helped i achieving computation speeds of
0.157 milliseconds for a 640X480 image with clock cycle of 12 nano-seconds
with the customised FPGA design. Thus th dedicated hardware assisted in
achieving real-time computation speeds as compared to the average of 5
seconds on the GPP used in this project.
Description
Keywords
Complete robotic automation for , handling and packaging of food products