The following information was submitted:
Transactions: INTERNATIONAL JOURNAL of CIRCUITS, SYSTEMS and SIGNAL PROCESSING
Transactions ID Number: 19-138
Full Name: Masoud Farzam
Position: Ph.D. Candidate
Age: ON
Sex: Male
Address: 350 Victoria st., , Toronto, Ontario
Country: CANADA
Tel: 905-655-5375
Tel prefix:
Fax:
E-mail address: mfarzam@ee.ryerson.ca
Other E-mails: msfarzam@yahoo.ca
Title of the Paper: Endmember transformation and replacement in real time hyperspectral unmixing
Authors as they appear in the Paper: Masoud Farzam, Soosan Beheshti
Email addresses of all the authors: mfarzam@ee.ryerson.ca, soosan@ee.ryerson.ca
Number of paper pages: 9
Abstract: For much of the past decade Hyperspectral Imaging (HSI) systems have gained considerable attention among researchers. Recent improvements in Optics have expanded the applications of HSI systems. Real time processing of extensive volumes of Hyperspectral data calls for more efficient and accurate real time algorithms. In current algorithms, speed comes at the expense of accuracy. Nevertheless, our proposed Ultra Fast Transition and Replacement (UFTR) approach shows a substantial improvement to the processing speed while also increasing the accuracy of the present methods. In the UFTR algorithm, Hyperspectral components' signatures, known as Endmembers, are estimated in an iterative approach. In each iteration, a linear transformation of data into the abundance vectors is calculated. This iterative process causes the speed of the algorithm to be extraordinarily fast. To improve the accuracy, a correlation based approach is used to project the estimated Endmembers int!
o the library spectrum. Accurate abundance vectors are calculated using the final transition matrix and the chosen Endmembers from the library. UFTR simulation results show that in low-SNR applications, the accuracy can be improved up to 15% and the speed is 10 to 50 times faster compared to the existing methods for a data cube of 4096 pixel with 224 bands. Furthermore, unlike many existing approaches, UFTR processing time dependency on the noise level is quite low. UFTR is definitely a departure from the trade-off between speed and accuracy and has a great potential for applications in the real time Hyperspectral imaging.
Keywords: Hyperspectral imaging, Hyperspectral unmixing, Real
EXTENSION of the file: .pdf
Special (Invited) Session: A fast transition and replacement (FTR) algorithm for real time Hyperspectral imaging applications
Organizer of the Session: 603-298
How Did you learn about congress:
IP ADDRESS: 192.197.54.30
