Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Real Time Image Segmentation using watershed algorithm on FPGA
#1

Abstract :
Watershed transformation is a powerful technique that can be efficiently used for image segmentation. Its use
for classifying and grading rice kernels has been discussed followed by its implementation on reconfigurable
devices. In this paper, we implement a watershed based segmentation algorithm on a Virtex-6 platform. The
main contribution of this work is the low execution time and minimal internal hardware resource occupation.
Watershed has been written in C-code and also in Verilog hardware descriptive language. The Verilog code is
synthesized on Virtex-6 Xilinx platform. Execution timing in both these approaches have been analyzed and the
hardware approach has been found to be much faster than its software counterpart.
Keywords: Image segmentation, Watershed, Reconfigurable Systems, FPGA, Computer vision, Embedded
imaging
1. Introduction:
A recent survey on modern methods and tools for reconfigurable embedded systems reveals that the area of
reconfigurable systems is very promising and offers quite new opportunities for development [1]. It is further
observed that these reconfigurable systems are enhancing capabilities of image processing and computer vision
domains as well. Applications such as high speed automated visual inspection; intelligent robotics, surveillance,
and transportation all require real-time complex vision systems. The nature of these applications and the hard
timing constraints associated with them makes the use of reconfigurable architectures such as FPGA (Field
Programmable Gate Arrays) a suitable implementation platform [2]. One such embedded system has been
reported for high speed inspection of grains recently [3]. Keeping in view these developments, it was felt to
implement the watershed based image segmentation algorithm used in classification and grading of rice [4].
Rice classification and grading as per standards defined by USDA has been done using Matrox imaging library
on a personal computer (PC) having 128MB RAM. Rice kernels are defined as broken when they are smaller
than three-fourths of whole kernel by USDA, 1995 [5]. Broken rice kernels are separated during milling of rice
(removal of the hull and bran layers from the rough rice) and re-added later depending on the desired quality

Download full report
http://googleurl?sa=t&source=web&cd=1&ve...06-232.pdf&ei=UgMxTujYFILosQL-1bGbCw&usg=AFQjCNF2Tisk931OwmtWyw4HkiArQTNKTw&sig2=RgVE5BIfv5tuVP7jzB-7ew
Reply



Forum Jump:


Users browsing this thread:
2 Guest(s)

Powered By MyBB, © 2002-2024 iAndrew & Melroy van den Berg.