|Table of Contents|

Noise Removal in Spectrum Above Water Surface Using Kernel Regression Smoothing(PDF)

《南京师大学报（自然科学版）》[ISSN:1001-4616/CN:32-1239/N]

Issue:
2010年03期
Page:
97-102
Research Field:
地理学
Publishing date:

Info

Title:
Noise Removal in Spectrum Above Water Surface Using Kernel Regression Smoothing
Author(s):
Wei Yuchun; Wang Guoxiang; Cheng Chunmei
Key Lab of Virtual Geographic Environment,Ministry of Education,Nanjing Normal University,Nanjing 210046,China
Keywords:
kerne l regress ion;  sm oo th;  signa l to interference ratio;  monte-car lo s imu la tion;  spectrum above w ater surface;  remo te sensing
PACS:
TP79
DOI:
-
Abstract:
Spec trum above w ater surface w ith h igh signa l to inte rference ra tio ( SIR) is the key to estima tew ate r qua lity param eters by rem ote sensing. To decrease inte rference and increase SIR of spectrum is the important content o f spectrum ana lysis. In th is paper, tw o spectrum above wa ter surface, wh ich ch lorophy l-l a concentra tion is sam e and suspended substance concentration is different, was taken as exam ples to ana lysis the de-no ising e ffect o f the kernel regression sm ooth ing. The paper uses theM onte-Car lo simu la tion m ethod to estim ate the average o f S IR by 500 rounds, g iven four interference disturbance type, .i e. Norm a l d istr ibu te, Ray leigh distribute, Exponentia l d istr ibute and Po isson d istr ibute, and four interference intensity. The SIR of ke rnel reg ression sm oothing were also com pa red w ith that o f Sav itzky-Golay smoo th filter, m ov ing av erage, loca l regression and robust lo ca l regress ion. The resu lt shows tha t kerne l regression sm ooth ing not on ly increases the SIR of spectrum, but also has the h ighest SIR than o ther fourme thods whether the interference intens ity is h igher or low er. SIR is the h ighest when interference is o f the norm a l distribute. Com pa re w ith Savitzky-Go lay sm ooth filter, the spectrum by kerne l regress ion sm ooth ing w asm ore smoo ther and keptm ore inform ation on spectrum  s peak and va lley position. The paper conc ludes tha t kerne l regress ion sm ooth ing is a be tterm e thod to dec rease in terference influence in the spectrum abovew ate r sur face.

References:

[ 1]ArstH. Optical Properties and Remo te Sensing ofMu lticomponentalW ater Bod ies[M ]. Chichester: Springer, Prax is Pub lishing, 2003.
[ 2]Analytical Spectra lDev ices Inc. Fie ldSpec Pro Use r?? s Guide[M ]. Boulder: Analytical Spectra lDev ices Inc, 2002.
[ 3]郑咏梅, 张铁强. 平滑、导数、基线校正对近红外光谱PLS定量分析的影响研究[ J]. 光谱学与光谱分析, 2004, 24 ( 12): 1 546-1 548.
[ 4]倪永年. 化学计量学在分析化学中的应用[M ]. 北京: 科学出版社, 2005.
[ 5]Bowm an A W, Azza lini A. Applied Smoo th ing Techn iques fo rData Ana lysis[M ] . New York: Ox ford Un iversity Press, 1997.
[ 6]郭波涛, 易东, 王文昌. 核平滑半参数回归模型在重复测量资料中的应用[ J]. 中国卫生统计, 2007, 24( 5) : 456-458.
[ 7]唐军武, 田国良, 汪小勇, 等. 水体光谱测量与分析Ⅰ : 水面以上测量法[ J] . 遥感学报, 2004, 8( 1): 37-44.
[ 8]Bosco lo, Pan R H, Roychowdhury V P. Independent Com ponen t Ana lys is Based on Nonpa ram e tric Dens ity Estim ation[ J]. IEEE Trans on Neura lNe tw orks, 2004, 15( 1): 55-65.
[ 9]Y i Cao. K ernel Sm oothing Regression, 2008[ EB /OL]. [ 2009-03-28 ]. http: / /www. m athw orks. com /ma tlabcentral/ fileexchange /19195-kerne-l sm ooth ing- reg ression.
[ 10]Nadaraya E A. On estim ating reg ression[ J]. Theo ry o f Probability and its App lications, 1964, 9( 1): 141-142.

Memo

Memo:
-
Last Update: 2013-04-08