摘要
水面光谱是利用遥感反演水体水质参数的数据基础,去除光谱中的噪声干扰,提高光谱的信号干扰比有助于改进水质参数的遥感反演的精度.本文选择叶绿素a浓度相同而悬浮泥沙浓度差异较大的两个水面光谱为代表,分析了核回归平滑方法对干扰的去除效果.假定干扰类型为4类,分别是正态分布、瑞利分布、指数分布和泊松分布,设定的干扰强度分为4级.利用蒙特卡洛模拟方法,通过500轮次的模拟计算了核回归平滑前后水面光谱的信号干扰比,并与多项式平滑、移动平均、局部回归和鲁棒性的局部回归平滑方法进行了比较.结果表明,不论干扰强度高或低,核回归平滑后的光谱均具有最高的信号干扰比.在四类干扰中,核回归平滑对于正态分布的干扰去除效果较好.与常用的多项式平滑方法相比,核回归平滑方法比较完整地保留了水面光谱中的峰谷位置信息,是一种值得推荐的提高水面光谱信号干扰比的方法.
Spectrum above water surface with high signal to interference ratio（SIR） is the key to estimate water quality parameters by remote sensing.To decrease interference and increase SIR of spectrum is the important content of spectrum analysis.In this paper,two spectrum above water surface,which chlorophyll-a concentration is same and suspended substance concentration is different,was taken as examples to analysis the de-noising effect of the kernel regression smoothing.The paper uses the Monte-Carlo simulation method to estimate the average of SIR by 500 rounds,given four interference disturbance type,i.e.Normal distribute,Rayleigh distribute,Exponential distribute and Poisson distribute,and four interference intensity.The SIR of kernel regression smoothing were also compared with that of Savitzky-Golay smooth filter,moving average,local regression and robust local regression.The result shows that kernel regression smoothing not only increases the SIR of spectrum,but also has the highest SIR than other four methods whether the interference intensity is higher or lower.SIR is the highest when interference is of the normal distribute.Compare with Savitzky-Golay smooth filter,the spectrum by kernel regression smoothing was more smoother and kept more information on spectrum＇s peak and valley position.The paper concludes that kernel regression smoothing is a better method to decrease interference influence in the spectrum above water surface.
出处
《南京师大学报（自然科学版）》
CAS
CSCD
北大核心
2010年第3期97-102,共6页
Journal of Nanjing Normal University(Natural Science Edition)
基金
国家自然科学基金(40771152)
江苏省普通高校自然科学研究计划资助项目(07KJB420062)
关键词
核回归
平滑
信扰比
蒙特卡洛模拟
水面光谱
遥感
kernel regression
smooth
signal to interference ratio
monte-carlo simulation
spectrum above water surface
remote sensing
