应用小波收缩方法剔除MODIS热红外波段数据条带噪声

De-striping for MODIS Infrared Band Data via Wavelet Shrinkage

采用多元并扫方式 (1km分辨率 1 0元并扫 ,5 0 0m分辨率 2 0元并扫 ,2 5 0m分辨率 4 0元并扫 )的MODIS传感器由于各探测单元在轨响应差异而引起的条带噪声对MODIS定量产品的反演计算精度造成一定影响。这种影响在MODIS的热红外波段尤其明显。为了尽可能减少这种影响 ,提高MODIS定量产品反演精度 ,提出应用小波收缩方法剔除MODIS数据条带噪声。研究首先使用连续小波变换方法分析MODIS条带数据 ,确定MODIS数据条带噪声在小波系数域中的尺度。其次 ,在分析条带噪声模式的基础上 ,使用小波收缩方法对MODIS 1B数据进行噪声剔除计算。最后比较了分别使用噪声剔除前后的两组MODIS 1B数据反演得到的部分云和大气定量遥感产品 ,结果表明 ,使用剔除噪声后的MODIS

MODIS method is a sensor to get the data by multi detector observation, among that 1km resolution waveband needs 10 detectors to scan at the same time, 500m resolution waveband needs 20 detectors, and 250m resolution waveband needs 40 detectors. The adoption of this technology has improved the scanning and data collection efficiency. However, there is a problem of the consistency between so many detectors, especially after the satellite is launched into the space the unexpected electromagnetism environment, it is very difficult to make so many detectors completely consistent. Therefore, the response discrepancy between the detectors will cause certain stripe in MODIS sensor, and the most serious stripe is in MODIS waveband 33-36. Some detectors of waveband 36 and 35 cannot reach the original defined NE Δ T . More stripe appears in the 6th, 7th and 8th detectors of waveband 34, and the 1st detector of waveband 33. Stripe also exists in other infrared waveband and visible waveband; there the stripe of window waveband is usually lower, while the stripe of atmospheric detecting waveband is more serious. This kind of stripe will cause much serious influence to the retrieving precision of MODIS quantitative products. In order to reduce the influence of MODIS stripe and improve the retrieving precision of remote sensing quantified products as soon as possible, we developed a wavelet analysis method to de stripe MODIS data. With this method, we remove the stripe of some different wave bands. These are waveband 36, 35, 34 and 33 detecting the phsical character of the cloud top, waveband 30 detecting the ozone, waveband 29, 28 and 27 detecting the water vapor, and the waveband 25, 24 and 23 detecting the atmospheric temperature. We describe a wavelet method for recovery of MODIS data from its stripe signals. Our work is organized into four broad sections. Section 1 will introduce wavelet shrinkage method for de noising noisy data; compare the character of the wavelet method and the FFT method in de noising processing. The objective of section 2 to find out the scale of MODIS stripe by the wavelet analysis for MODIS stripe data using continues wavelet transforms. Section 3 analyses stripe noise data pattern for the MODIS level 1B stripe data, present the wavelet shrinkage method for MODIS level 1B data. Section 4 will provide a comparson for MODIS cloud product and atmospheric profile product between the original data and de striped data. We can find that there is an improvement in MODIS cloud product and atmopheric profile product after de striping. And we can get more understanding for the stripe regular pattern.