Noise Removal Method for Unmanned Aerial Vehicle Data to Estimate Water Quality of Miharu Dam Reservoir, Japan

Abstract

Lake Sakurako is a reservoir of the Miharu Dam in Fukushima Prefecture, Japan. The water quality of the small lake becomes significantly worse during the summer owing to the occurrence of blue-green algae. Therefore, water quality management is a serious problem. Because the primary method of water quality analysis is direct collection from the target water area, the analysis range is limited, and the analysis of the entire water area is very difficult. Therefore, performing a wider range of analyses by remote sensing is a possible solution. In this study, we analyze near infrared (NIR) data acquired by unmanned aerial vehicles (UAVs). A fuzzy regression analysis is conducted on the UAV data and water measurements. Based on the experimental results of data from August 2015, the NIR data is confirmed to be useful in estimating the water quality conditions in Lake Sakurako. Furthermore, we investigate the noise removal process using a nonlocal mean filter and demonstrate that the process provides more detailed information regarding the lake’s water quality.

Computational approaches for sub-meter ocean color remote sensing

“…CT1) [Eqs. (B2) and B3 [Su, 2017;Larson et al, 2018;Totsuka et al, 2019;Kwon et al, 2020;Kislik et al, 2018], 3,4 [Kislik et al, 2018;, 4 [Honkavaara et al, 2013]). This dissertation ameliorates the highest impact factors within chapters 2-4. .…”

“…After the algal pigment concentration maps produced by CZCS were proven to match with surface measured values [Gordon et al, 1980], the maps could then be used for research, Table 1.1: The dominant factors that limit sub-meter ocean color remote sensing, where measurements are made from low altitudes with low quality optics, are dramatically different from those at km spatial resolutions from high altitudes with state-of-the-art optics ( 1,2 [Su, 2017;Larson et al, 2018;Totsuka et al, 2019;Kwon et al, 2020;Kislik et al, 2018], 3,4 [Kislik et al, 2018, 4 [Honkavaara et al, 2013]). This dissertation ameliorates the highest impact factors within chapters 2-4.…”

“…The researchers created an environmental noise (e.g., surface reflected light) removal algorithm which they applied during the regression training process to achieve the highest accuracy estimates. The finescale fluctuation in the data were present in other data sets, which resulted in the creation of spatial filtering algorithms to eliminate the sub-meter fluctuations in the the field estimated SPM, chlorophyll, and turbidity [Totsuka et al, 2019]. In summary, the lightweight cameras that can be flown on drones are able to estimate specific optical constituents in near ideal conditions, despite the sub-meter environmental noise sources and low SNRs by leveraging spatial filtering.…”

“…A few glint removal algorithms were created to correct for this finescale glint issue in multispectral images collected from drones, but each relied upon spatial filtering. One algorithm was based upon median filtering, innately trading spatial resolution for ocean color product estimation accuracy [Totsuka et al, 2019]. Another algorithm, the matching pixel by pixel algorithm, removed glint fluctuations during empirical algorithm training by iterating through combinations of pixels in the training images to find the most accurate regression [Su, 2017].…”

“…Novel image-processing approaches have been explored to reduce the glint induced spatial variation in WQP estimates made from low-altitude drone imagery [Totsuka et al, 2019;Su, 2017], but they each exhibit some limitations. The non-local mean filter used by Totsuka et al [Totsuka et al, 2019] reduces the maximum achievable spatial resolution.…”