Effect of atmospheric interference and sensor noise in retrieval of optically active materials in the ocean by hyperspectral remote sensing

Abstract

We present a method to construct the best linear estimate of optically active material concentration from ocean radiance spectra measured through an arbitrary atmosphere layer by a hyperspectral sensor. The algorithm accounts for sensor noise. Optical models of seawater and maritime atmosphere were used to obtain the joint distribution of spectra and concentrations required for the algorithm. The accuracy of phytoplankton retrieval is shown to be substantially lower than that of sediment and dissolved matter. In all cases, the sensor noise noticeably reduces the retrieval accuracy. Additional errors due to atmospheric interference are analyzed, and possible ways to increase the accuracy of retrieval are suggested, such as changing sensor parameters and including a priori information about observation conditions.

A Survey on Hyperspectral Image Restoration: From the View of Low-Rank Tensor Approximation

“…During imaging, the reflective or radiative signal suffers from radiometric disturbance while passing through the atmospheric environments (e.g., Stage 3 in Figure 1) and sensor distortion when spectrometer receives the incident energy of radiation (e.g., Stage 4 in Figure 1). The radiometric disturbance in atmospheric transmission process is mainly caused by three factors [31,51]: general interference of environment noise, absorption/scattering due to mixed atmospheric composition and presence of poor weather conditions, such as clouds, fogs. The sensor distortion in spectrometer inevitably occurs in collimating mirror/diffraction grating, lens optics and detector array [31,52].…”

Computational approaches for sub-meter ocean color remote sensing

“…The real-world SNR is even lower than the theoretical value on hyperspectral cameras because they simultaneously image the entire spectrum, so the lowest intensity spectra will have a lower SNR than the highest intensity pixel . The low SNR can significantly alter the accuracy of the computed ocean color products, either through the ocean color algorithms directly or through the environmental noise correction algorithms [Levin et al, 2005;Levin and Levina, 2007;Moses et al, 2012a. Therefore, the impact of optoelectronic and photonic noise needs to be considered in the most sensitive ocean color applications before using lightweight hyperspectral cameras to map optical constituents.…”

“…The optical and electronics noise sources directly affect the estimated water-leaving radiance, the noise can then propagate through software algorithms to produce imprecise estimates of OACs and inaccurate correction of surface reflected light (glint) [Levin et al, 2005;Levin and Levina, 2007]. OAC algorithms typically utilize band ratios of the remote sensing reflectance, which is the camera measured radiance divided by the irradiometer measured irradiance.…”