An Integrative Remote Sensing Application of Stacked Autoencoder for Atmospheric Correction and Cyanobacteria Estimation Using Hyperspectral Imagery

Abstract

Hyperspectral image sensing can be used to effectively detect the distribution of harmful cyanobacteria. To accomplish this, physical- and/or model-based simulations have been conducted to perform an atmospheric correction (AC) and an estimation of pigments, including phycocyanin (PC) and chlorophyll-a (Chl-a), in cyanobacteria. However, such simulations were undesirable in certain cases, due to the difficulty of representing dynamically changing aerosol and water vapor in the atmosphere and the optical complexity of inland water. Thus, this study was focused on the development of a deep neural network model for AC and cyanobacteria estimation, without considering the physical formulation. The stacked autoencoder (SAE) network was adopted for the feature extraction and dimensionality reduction of hyperspectral imagery. The artificial neural network (ANN) and support vector regression (SVR) were sequentially applied to achieve AC and estimate cyanobacteria concentrations (i.e., SAE-ANN and SAE-SVR). Further, the ANN and SVR models without SAE were compared with SAE-ANN and SAE-SVR models for the performance evaluations. In terms of AC performance, both SAE-ANN and SAE-SVR displayed reasonable accuracy with the Nash–Sutcliffe efficiency (NSE) > 0.7. For PC and Chl-a estimation, the SAE-ANN model showed the best performance, by yielding NSE values > 0.79 and > 0.77, respectively. SAE, with fine tuning operators, improved the accuracy of the original ANN and SVR estimations, in terms of both AC and cyanobacteria estimation. This is primarily attributed to the high-level feature extraction of SAE, which can represent the spatial features of cyanobacteria. Therefore, this study demonstrated that the deep neural network has a strong potential to realize an integrative remote sensing application.

Monitoring Water Quality of the Haihe River Based on Ground-Based Hyperspectral Remote Sensing

“…An overall accuracy of over 90% for our result of TP (R 2 = 0.99, RMSE = 0.0015 mg/L), Sun's result of TP (R 2 = 0.93, RMSE = 0.02 mg/L), and Song's result of TP (R 2 = 0.91, RMSE = 0.017 mg/L), was found. Among our results, the lowest fitting degree was PC (R 2 = 0.87, RMSE = 0.0011 mg/L), with results a little more than Pyo et al 's SAE-ANN model (R 2 = 0.82, RMSE = 9.32 mg/L) [13]. The retrieval of the Chl-a concentration has always been the focus of water color remote sensing research.…”

“…The substances in the water determine the spectral characteristics of the water body, and the substances affecting the spectral distribution and light intensity in the inland surface water can be roughly divided into three categories: phytoplankton pigments, yellow substances, and suspended substances [9]. Hyperspectral technology has been widely applied to water quality parameters with optical activity, such as Chl-a [10,11] PC [12,13], TSS [14,15], chromophoric dissolved organic matter (CDOM) [16], transparency [17], and turbidity [18]. Although some scholars have studied water quality parameters such as TN, TP, NH 4 -N, and NO 3 -N [19,20], it is also a great challenge to estimate the concentrations of total phosphorus, total nitrogen, ammonia nitrogen, and nitrate nitrogen in inland waters because the above parameters do not have optical activity in the perception wavelength.…”

Impact of Spectral Resolution on Quantifying Cyanobacteria in Lakes and Reservoirs: A Machine-Learning Assessment

“…Sun et al [5] employed SVR to estimate PC from hyperspectral data collected in large cyanobacteria-dominated, turbid lakes in China. Pyo et al [44] applied SVR, as well as a feed-forward artificial neural network (ANN), to the hyperspectral data collected from the Baekje reservoir located at the Geum River in South Korea to achieve atmospheric correction and retrieve PC and Chla. Pahlevan et al [45] introduced a mixture density network (MDN) to estimate Chla across different bio-optical regimes in inland and coastal waters.…”

Vision Based Real Time Monitoring System for Elderly Fall Event Detection Using Deep Learning

“…The AE is a NN for unsupervised feature learning [24]. The illustrative layers of the AE are consisting of a decoder and encoder, which consist of succeeding non-linear AE functions:…”