Implementation of mean shift based on SOPC

Cong Cong Chen, Qing Liu, Lin Bo Tang, Bao Jun Zhao, Dong Shun Cui

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Mean Shift algorithm based on floating point arithmetic calculation is very difficult to be implemented on real-time systems. Another problem is that the tracking window of Mean Shift is not adaptive. In order to solve the problems above, an improved Mean Shift which can be implemented on FPGA is presented. Firstly, Hardware/software partitioning based on the complexity of the algorithm is determined the implementation of SOPC. Secondly, A pipeline structure is employed for Mean Shift algorithm to calculate and accumulate kernel function, shift vector and target area in a single clock cycle, and accuracy is above 99 percent. Finally, as floating point arithmetic is more time-consuming, custom floating-point instruction is added into CPU to improve the algorithm operation speed. Experimental results show that the presented method can adaptively update target windows and track target in 4ms stably, which improve the efficiency of program by 5 times.

Original languageEnglish
Title of host publicationNew Technologies for Engineering Research and Design in Industry
PublisherTrans Tech Publications Ltd.
Pages534-539
Number of pages6
ISBN (Print)9783038351412
DOIs
Publication statusPublished - 2014
Event2014 International Conference on Mechatronics and Intelligent Materials, MIM 2014 - Lijiang, China
Duration: 18 May 201419 May 2014

Publication series

NameAdvanced Materials Research
Volume971-973
ISSN (Print)1022-6680
ISSN (Electronic)1662-8985

Conference

Conference2014 International Conference on Mechatronics and Intelligent Materials, MIM 2014
Country/TerritoryChina
CityLijiang
Period18/05/1419/05/14

Keywords

  • Hardware acceleration
  • Mean shift
  • System on programmable chip(SOPC)

Fingerprint

Dive into the research topics of 'Implementation of mean shift based on SOPC'. Together they form a unique fingerprint.

Cite this