Machine Learning Analytic-Based Two-Staged Data Management Framework for Internet of Things

In applications of the Internet of Things (IoT), where many devices are connected for a specific purpose, data is continuously collected, communicated, processed, and stored between the nodes. However, all connected nodes have strict constraints, such as battery usage, communication throughput, processing power, processing business, and storage limitations. The high number of constraints and nodes makes the standard methods to regulate them useless. Hence, using machine learning approaches to manage them better is attractive. In this study, a new framework for data management of IoT applications is designed and implemented. The framework is called MLADCF (Machine Learning Analytics-based Data Classification Framework). It is a two-stage framework that combines a regression model and a Hybrid Resource Constrained KNN (HRCKNN). It learns from the analytics of real scenarios of the IoT application. The description of the Framework parameters, the training procedure, and the application in real scenarios are detailed. MLADCF has shown proven efficiency by testing on four different datasets compared to existing approaches. Moreover, it reduced the global energy consumption of the network, leading to an extended battery life of the connected nodes.

Use Remote Sensing and Machine Learning to Study the Changes of Broad-Leaved Forest Biomass and Their Climate Driving Forces in Nature Reserves of Northern Subtropics

Forest is the largest vegetation carbon pool in the global terrestrial ecosystem. The spatial distribution and change of forest biomass are of importance to reveal the surface spatial variation and driving factors, to analyze and evaluate forest productivity, and to evaluate ecological function of forest. In this study, broad-leaved forests located in a typical state nature reserve in northern subtropics were selected as the study area. Based on ground survey data and high-resolution remote sensing images, three machine learning models were used to identify the best remote sensing quantitative inversion model of forest biomass. The biomass of broad-leaved forest with 30-m resolution in the study area from 1998 to 2016 was estimated by using the best model about every two years. With the estimated biomass, multiple leading factors to cause biomass temporal change were then identified from dozens of remote sensing factors by investigating their nonlinear correlations. Our results showed that the artificial neural network (ANN) model was the best (R2 = 0.8742) among the three, and its accuracy was also much higher than that of the traditional linear or nonlinear models. The mean biomass of the broad-leaved forest in the study area from 1998 to 2016 ranged from 90 to 145 Mg ha−1, showing an obvious temporal variation. Instead of biomass, biomass change (BC) was studied further in this research. Significant correlations were found between BC in broad-leaved forest and three climate factors, including average daily maximum surface temperature, maximum precipitation, and maximum mean temperature. It was also found that BC has a strong correlation with the biomass at the previous time (i.e., two years ago). Those quantitative correlations were used to construct a linear model of BC with high accuracy (R2 = 0.8873), providing a new way to estimate the biomass change of two years later based on the observations of current biomass and the three climate factors.

Comparative Analysis of Machine Learning Algorithms in Automatic Identification and Extraction of Water Boundaries

Monitoring open water bodies accurately is important for assessing the role of ecosystem services in the context of human survival and climate change. There are many methods available for water body extraction based on remote sensing images, such as the normalized difference water index (NDWI), modified NDWI (MNDWI), and machine learning algorithms. Based on Landsat-8 remote sensing images, this study focuses on the effects of six machine learning algorithms and three threshold methods used to extract water bodies, evaluates the transfer performance of models applied to remote sensing images in different periods, and compares the differences among these models. The results are as follows. (1) Various algorithms require different numbers of samples to reach their optimal consequence. The logistic regression algorithm requires a minimum of 110 samples. As the number of samples increases, the order of the optimal model is support vector machine, neural network, random forest, decision tree, and XGBoost. (2) The accuracy evaluation performance of each machine learning on the test set cannot represent the local area performance. (3) When these models are directly applied to remote sensing images in different periods, the AUC indicators of each machine learning algorithm for three regions all show a significant decline, with a decrease range of 0.33–66.52%, and the differences among the different algorithm performances in the three areas are obvious. Generally, the decision tree algorithm has good transfer performance among the machine learning algorithms with area under curve (AUC) indexes of 0.790, 0.518, and 0.697 in the three areas, respectively, and the average value is 0.668. The Otsu threshold algorithm is the optimal among threshold methods, with AUC indexes of 0.970, 0.617, and 0.908 in the three regions respectively and an average AUC of 0.832.

Automatic Binary Data Classification Using a Modified Allen–Cahn Equation

In this paper, we propose an automatic binary data classification method using a modified Allen–Cahn (AC) equation. The modified AC equation was originally developed for image segmentation. The equation consists of the AC equation with a fidelity term which enforces the solution to be the given data. In the proposed method, we start from a coarse grid and refine the grid until the accuracy of the data classification reaches a given tolerance. Therefore, we can avoid a laborious trial and error procedure. For a numerical method for the modified AC equation, we use a recently developed explicit hybrid scheme. We perform several 2D and 3D computational tests to demonstrate the performance of the proposed method. The computational results confirm that the proposed algorithm is automatic.

Water Level Prediction of Rainwater Pipe Network Using an SVM-Based Machine Learning Method

Model accuracy and running speed are the two key issues for flood warning in urban areas. Traditional hydrodynamic models, which have a rigorous physical mechanism for flood routine, have been widely adopted for water level prediction of rainwater pipe network. However, with the amount of pipes increasing, both the running speed and data availability of hydrodynamic models would be decreased rapidly. To achieve a real-time prediction for the water level of the rainwater pipe network, a new framework based on a machine learning method was proposed in this paper. The spatial and temporal autocorrelation of water levels for adjacent manholes was revealed through theoretical analysis, and then a support vector machine (SVM)-based machine learning model was developed, in which the water levels of adjacent manholes and rivers-near-by-outlets at the last time step were chosen as the independent variables, and then the water levels at the current time step can be computed by the proposed machine learning model with calibrated parameters. The proposed framework was applied in Fuzhou city, China. It turns out that the proposed machine learning method can forecast the water level of the rainwater pipe network with good accuracy and running speed.

Improved Position Error Model for Airborne Hyperspectral Imaging Systems

An improved position error model for airborne hyperspectral imaging systems is proposed to quantify the position errors caused by the deficiencies of instruments and the manual installation of integrated systems. The model is based on a thorough analysis of position error sources, which include uncertainties caused by camera position and altitude, interior camera parameters, gyro stabilized mount, position setting error of the global positioning system antenna, and boresight angle error. An improved position error model that describes these error sources is established based on collinear equations. This model is then expanded into the first-order Taylor series. Finally, the error propagation law is applied to estimate the position errors. The performance of the proposed method is evaluated based on real hyperspectral images of Hymap. Results show that the mean square error of the position error is better than 2.1 pixels, which is almost the same as that of ground truth.

Spectral Spatio-Temporal Fire Model for Video Fire Detection

Fire detection technology aroused people’s attention increasingly. The main challenge of the fire detection systems is how to reduce false alarms caused by objects like fire’s colors. Most existing algorithms used only features of fire in visual field. In this work, we put forward a new algorithm to detect dynamic fire from the surveillance video based on the combination of radiation domain features model. First, a fire color model is used to extract flame-like pixels as candidate areas in YCbCr space. Second, we convert the candidate regions from the traditional color space into radiation domain in advance by camera calibration. And we use seven features to model the spectral spatio-temporal model of the fire to more accurately characterize the physical and optical properties of the fire. Finally, we choose a two-class SVM classifier to identify the fire from the candidate areas and use a radial basis function kernel to improve the accuracy of the recognition. Two different sets of data are used to validate the algorithm we proposed. And the experimental results indicate that our method performs well in video fire surveillance.

Physical evolution of the Three Gorges Reservoir using advanced SVM on Landsat images and SRTM DEM data

The Three Gorges Reservoir (TGR) is one of the largest hydropower reservoirs in the world. However, changes of the important physical characteristics of the reservoir covering pre-, during-, and post- dam have not been well studied. This study analyzed the lengths and water surface areas of the TGR using advanced support vector machine method (SVM) combined Landsat images with the Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM), which showed an increasing trend of lengths and surface areas with variable growth rates from pre-dam period to post-dam period. The highest water level (ca. 171.5 m) was reported in 1st Jan, 2015, with the longest length of 687.8 km and largest water surface area of 1106.2 km² during the study period. The lowest increasing magnitude of the reservoir length occurred in the first stage (2000-2005) but with the fastest magnitude of water surface area increase. The third stage (2010-2015) showed highest increase magnitude of length and lowest increase magnitude of water surface area. Meanwhile, the increased reservoir areas were mainly from cultivated land, forest land, and building land, with the biggest increase rate of cultivated land regardless of periods. Specifically, cultivated land contributed 39.1-46.0% to increased reservoir water area; the proportions were 22.6-29.6%, 22.1-24.1%, and 5.6-9.4% for forest, building land, and grassland, respectively. The study provides important data for the TGR physical evolution in the Holocene.