Development of Algorithm for Analysis of Vegetation Index Data from Earth Observation Satellites

Abstract

Now-a-days, various advanced sensors from satellites are utilized to generate multiple newlinetime series data with the help of temporal resolution and high spatial resolution. newlineLong time series of vegetation index data can be used for monitoring vegetation seasons and newlinefor exploring and extracting seasonality parameters. The TIMESAT program package available newlinein MATLAB is used for extracting seasonal parameters. Performance parameters like accuracy, newlineprecision are calculated using machine learning algorithms from extracted phenological newlineparameters. The suggested Naïve bayes algorithms give an enhanced performance rate. Which newlineavoids farmer interference further to know the land better. To know the land better and detect newlineagricultural droughts, pre-classifiers like Normalized difference vegetation index (NDVI) are newlinecombined with other parameters to increase its pertinency and to ensure more accuracy. A newlinestrong negative correlation exists between Land surface temperature (LST) and NDVI. newlineAlso, different approaches are initiated to perform clustering as well as validation in time series newlinedata related to the environmental eco-system platform. Various models such as the hidden newlineMarkov approach, regression-related model, genetic programming, and neural networks are newlineadvised for developing data analysis or data related to time sequence according to the frequency newlinedomain. Such approaches are utilized to forecast the current water range in the nearby river for newlineflood safeguard and also to predict different weather conditions such as fog, rainfall, and wind. newlineTemporal data observation is mainly related to land usage, land cover and etc. The abovementioned newlinemethods are experienced different kind of challenges such as offering a restricted newlineamount of interpretability, using various parameters and prior distribution or misplaced values. newlineSo, the above-mentioned challenges of the conventional vegetation classification model led to newlinedesign a new approach on the basis of deep learning models. The major aim of this proposed newlineframework is to develop a new vegetation classification model based on two stages. In the newlineinitial stage, a new model related to sequence observation to classify the vegetation type in newlinefarm region of Ujani dam, which is located near Solapur district, Maharashtra. The suggested newlinemodel enhanced ATWDTW to perform effective analysis in time sequence with the help of newlinesatellite images. Various satellite pictures collected from multiple fields are analysed at the newlineinitial stage, and they are offered observation phase by utilizing ATWDTW. Parameters of newlineconventional TWDTW are optimized by utilizing a novel approach MF-BSA to improve the newlineperformance rate of classification. After various observations are performed in the developed newlinemodel and researchers verified that the suggested model has very deep insight in vegetation newlineclassification. Further, these approaches are widely used practically in different environment newlinelike ecological and environmental time-variant data information. In the second phase, the main newlinegoal of the work is to develop a novel model for differentiating various vegetation kinds newlinepresented in the fields near Ujani dam region with the help of time series observation. Analysis newlineof time series are mainly used satellite images for the observation, and also the developed newlinemodel provides ATWDTW for effective validation. Satellite images of different fields are newlineoffered as the input then, pre-processing is performed initially in the image and subjected to newlineATWDTW for further validation. Effective classification performance rate is attained by newlinetuning the parameters of TWDTW with the help of a heuristic approach named ACCSO. Thus, newlinethe developed model has the capability to offer an effective robustness rate and also to generate newlinea low sensitivity rate at the time training is performed in the samples of TWDTW while newlineapplying in mountain vegetation kind of categorization. Finally, the developed model newlineshowcased an enhanced performance rate when contrasted with conventional approaches and newlineattained an effective classification rate than existing vegetation classification models. In the newlinefinal phase, the comparison is performed over the developed ATWDTW-MF-BSA-based newlineapproach and ATWDTW-ACCSO-based vegetation classification approach and achieved an newlineenhanced performance rate with respect to time series classification. newline