Scholarly article on topic 'Application Of Soft Computing Techniques In Coastal Study – A Review'

Application Of Soft Computing Techniques In Coastal Study – A Review Academic research paper on "Computer and information sciences"

CC BY-NC-ND
0
0
Share paper
Keywords
{"Artificial neural network" / "Coastal inundation" / "Remote sensing" / Shoreline / "Soft computing" / "Support vector machine"}

Abstract of research paper on Computer and information sciences, author of scientific article — G.S. Dwarakish, B. Nithyapriya

Abstract Coastal zone is the triple interface of air, water and land and it is so dynamic in nature which requires expeditious management for its protection. Impulsive change in shoreline and submergence of low lying areas due to sea level rise are the solemn issues that need to be addressed. Indian coastline of about 7516km is under threat due to global warming and related human interventions. Remote sensing data products provide synoptic and repetitive view of the earth in various spatial, spectral, temporal and radiometric resolutions. Hence, it can be used in monitoring coastal areas on a temporal scale. Critical Erosion hotspots have to be given proper protection measures to avoid further damages. Satellite images serve in delineating shoreline and extracting the hotspots to plan the mitigation works. Coastal inundation maps can be created using remote sensing and geospatial technologies by assuming different sea level rises. Those maps can serve as a base for planning management activities. Soft computing techniques like Fuzzy Logic, Artificial Neural Network, Genetic Algorithm and Support Vector Machine are upcoming soft computing algorithms that find its application in classification, regression, pattern recognition, etc., across multi-disciplinary sciences. They can be used in classifying remote sensing images which in turn can be used for studying the coastal vulnerability. The present paper reviews the works carried out for coastal study using conventional remote sensing techniques and the pertinency of soft computing techniques for the same.

Academic research paper on topic "Application Of Soft Computing Techniques In Coastal Study – A Review"

Accepted Manuscript

Application Of Soft Computing Techniques In Coastal Study - A Review

G.S. Dwarakish , B. Nithyapriya

PII: DOI:

Reference:

S2468-0133(16)30010-9 10.1016/j.joes.2016.06.004 JOES 22

To appear in:

Journal of Ocean Engineering and Science

Received date: Revised date: Accepted date:

21 March 2016 16 June 2016

22 June 2016

Please cite this article as: G.S. Dwarakish , B. Nithyapriya , Application Of Soft Computing Techniques In Coastal Study - A Review, Journal of Ocean Engineering and Science (2016), doi: 10.1016/j.joes.2016.06.004

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Highlights

• This paper reviews soft computing techniques and coastal remote sensing.

• Applicability of ANN in remote sensing is also dealt.

• Basics of SVM is covered.

• Evolution of types of SVM is briefly outlined.

• Advantages of SVM over other techniques is discussed.

APPLICATION OF SOFT COMPUTING TECHNIQUES IN COASTAL STUDY - A REVIEW

Dwarakish G S a and Nithyapriya B a*

a National Institute of Technology Karnataka, Surathkal * E-mail: nithyapriya.bnp@gmail.com, contact:+918015162533

ABSTRACT

Coastal zone is the triple interface of air, water and land and it is so dynamic in nature which requires expeditious management for its protection. Impulsive change in shoreline and submergence of low lying areas due to sea level rise are the solemn issues that need to be addressed. Indian coastline of about 7516 km is under threat due to global warming and related human interventions. Remote sensing data products provide synoptic and repetitive view of the earth in various spatial, spectral, temporal and radiometric resolutions. Hence, it can be used in monitoring coastal areas on a temporal scale. Critical Erosion hotspots have to be given proper protection measures to avoid further damages. Satellite images serve in delineating shoreline and extracting the hotspots to plan the mitigation works. Coastal inundation maps can be created using remote sensing and geospatial technologies by assuming different sea level rises. Those maps can serve as a base for planning management activities. Soft computing techniques like Fuzzy Logic, Artificial Neural Network, Genetic Algorithm and Support Vector Machine are upcoming soft computing algorithms that find its application in classification, regression, pattern recognition, etc., across multi-disciplinary sciences. They can be used in classifying remote sensing images which in turn can be used for studying the coastal vulnerability. The present paper reviews the works carried out for coastal study using conventional remote sensing techniques and the pertinency of soft computing techniques for the same.

Keywords

Artificial Neural Network, Coastal inundation, Remote sensing, Shoreline, Soft computing, Support vector machine

1. INTRODUCTION

Coast is ever-changing in nature and is one of the significant environmental, biological and ecological zones of India. It is the intersection of three landforms namely land, water and air, extending between 200 m bathymetric contour to 200 m elevation contour (Cracknell A.P. 1999). It gains attention due to its multiple uses like tourism, productivity, trade and commerce etc., Shoreline is considered as the incomparable part of the earth surface and is recommended to be one of the 27 features acknowledged by International Geographic Data Committee (IGDC). Coastal zones face both natural and man-made hazards namely Sea Level Rise, shoreline change, tropical cyclones, storm surge, salt water intrusion and coastal inundation. The above endangerments lead to severe socio-economic losses.

Remote Sensing (RS) is a multi-disciplinary science and which can be effectively used in the coastal studies because of its repetitive and synoptic coverage. Sensors mounted at different heights provide data in various spatial, spectral, radiometric and temporal resolutions. The temporal characteristic of the sensor helps in change detection studies. Hence, shoreline mapping which depends on a temporal scale to find out the erosion and accretion hotspots can be carried out with RS images. Classification of RS images aids in converting a raw data into a meaningful set of information. It can be of two types namely supervised or unsupervised classification. Conventional methods of supervised classification include Parallelopiped, Minimum distance to mean and Maximum likelihood classifiers. Soft classifiers are also being developed namely Fuzzy Logic (FL), Genetic Algorithm (GA), Particle Swarm Optimisation (PSO), Artificial Neural Network (ANN), Support Vector Machine (SVM), Self-Organising Maps (SOM) etc., One of the above classification strategies can be used in preparing coastal inundation maps that will help in planning the mitigation measures. The principle of traditional classification methods is Experimental Risk Minimization (ERM) where it can obtain good result when huge training data is available. On the other hand, soft methods work on Structural Risk Minimization (SRM) wherein with minimal training data, one can achieve good

results (Dai et. al., 2007).

The present paper is organised in six sections as follows: Section 1. Introduction; Section 2. Shoreline mapping - conventional method; Section 3. Coastal inundation - conventional method; Section 4. Soft computing techniques in wind turbine study; Section 5. Soft computing in classification; Sub-section 5.1 ANN in classification; Sub- section 5.2. SVM in classification; Section 6. Discussion and concluding remarks.

2. SHORELINE MAPPING - CONVENTIONAL METHOD

The occurrence of shoreline change may be due to natural or anthropogenic processes. The natural processes includes the effects of waves, currents, tides and winds while the latter includes sand mining, offshore dredging or building of coastal structures. Shoreline changes draw more attention since they are the most important environmental indicants that directly impact on the economic development and coastal land management. The nine maritime states of India are Gujarat, Maharastra, Goa, Karnataka, Kerala, TamilNadu, Andhra Pradesh, Orissa and West Bengal as shown in fig. 1.

Maritime Slates oflndia

«0 «W 11»

have been carr

Fig 1 Maritime states of India

Various studies have been carried out in all the above mentioned states by many researchers. Murali et. al., (2009) have studied the coastline of Paradip in the state of Orissa from 1998 - 2005. Paradip region is located at the mouth of Mahanadi river and hence the change in coastline due to the seasonality of the river had been well understood. Natesan et. al., (2015) have done an extensive study on the most vulnerable coast of TamilNadu by using geospatial technologies and Digital Shoreline Analysis System (DSAS) for the period of 1978 - 2014. For ease of analysis, TamilNadu coast had been divided into four zones, each stretch covering the coastal areas of Bay of Bengal, Palk Strait, Gulf of Mannar and Indian Ocean. Natesan et. al., (2013) and Mujabar et. al., (2011) have also done the study on Vedaranyam coast and Kanyakumari coast of TamilNadu respectively and concluded that proper nourishment had to be given alongside the shoreline to prevent accretion and erosion.

Research work by Rao et. al., (2007) on the narrow Kerala coast had identified the actuating factors, processes and frequency of dynamic changes in coastline stretches. It has been found that shoreline change is maximum in the stretch underlain by soft sediments and also at the intersection of lineaments. Karnataka coast has one major port and ten minor ports and is one of the most indented areas with a number of river mouths, bays, spits, sand dunes, lagoons, creeks, cliffs and long beaches. This clearly unveil the problems

associated with this coast. A numerous researchers have studied this coast either as a complete stretch or by considering a particular river mouth. ChenthamilSelvan et.al. (2014) and Hegde et. al., (2015) have analysed the entire stretch covering three coastal districts namely Dakshina Kannada, Udupi and Uttar Kannada for the years 1973 - 2006 and 1991 - 2014 respectively by utilizing DSAS. Dakshin Kannada coast has two crucial river mouths namely Netravathi - Gurpur river mouth (with breakwater) and Mulky- Pavanje river mouth (without breakwater). Thus this stretch of Karnataka coast is ideal in studying the effect of breakwater on coastal erosion. Shetty et. al., (2015) have worked on the shoreline change along Mangalore coast having the above said river mouths for the period of (2005 - 2013) without employing DSAS. In addition to the studies on shoreline change over past years, there are also researchers working to predict the position of shoreline from historical data (Maiti et. al., 2008). For the entire coast of India, long-term shoreline change for a period of 38 years from 1972-2010 has been done by Institute of Ocean Management, Anna University, Chennai and the database had been created and presented in the public domain.

A numerous studies have reported in the recent past stating the significance of satellite images in studying the retrieval of shoreline after the damage had happened. Katchal island in Andaman and Nicobar where the erosion is more prevalent has been studied for aftermath of 2004 Tsunami (Yunus et. al., 2016). The impact of coastal structures such as sea wall, groins and breakwater on shoreline change has been studied in the Puducherry region (Selvan et. al., 2016).

The summary of data used in the above studies are presented in this section. Survey of India (SOI) toposheets, Indian Remote Sensing (IRS) - Linear Imaging Self Scanner - III (LISS III) data product, Shuttle Radar Topographic Mission's (SRTM) Digital Elevation Model (DEM) and/or LandSat data products have been used. The general methodology followed is geometric correction of the satellite images, delineating the shoreline as vector maps using ArcGIS software and analysing it with or without employing DSAS. Table 1 shows the works done by various researchers related to shoreline mapping.

Table 1. Study on Shoreline Mapping

Researchers' Name Year Study area

Rao et. al., 2007 Kerala Coast

Maiti et. al., 2008 Balasore, Orissa and Midnapur, West Bengal

Murali et. al. 2009 Paradip Coast, Orissa

Mujabar et. al., 2011 Kanyakumari Coast

Natesan et. al., 2013 Vedaranyam Coast

ChenthamilSelvan et.al. 2014 Karnataka Coast

Natesan et. al., 2015 TamilNadu Coast

Hedge et. al., 2015 Karnataka Coast

Shetty et. al., 2015 Mangalore Coast, Karnataka

Yunus et. al., 2016 Katchal island, Andaman

Selvan et. al., 2016 Puducherry

3. COASTAL INUNDATION - CONVENTIONAL METHOD

Sea Level Rise (SLR) is the one of consequences of global warming. This, in turn, causes the inundation of coastal low lying areas that bring on potential habitat and economic losses. Hence, it is primary to quantify

the perils associated with the SLR to guide decision makers in future coastal infrastructure development. Land Use/ Land Cover (LU/LC) of the coastal area and its dynamics have to be considered in prior to contemplating the SLR owing to inundation losses. Coastal LU/LC undergoes change due to the dynamic nature of their surrounding environment.

Change detection studies are becoming important in the coastal areas. A review about all the types of change detection techniques for various LU/LC had been done in detail (Lu et. al., 2004). Image differencing, Change Vector Analysis (CVA), Gramm Schmidt (GS) transformation, Principal Component Analysis (PCA), Spectral mixture model, Li Strahler reflectance model and Tasselled Cap (TC) Transformation are some of the techniques on change detection. CVA method had been applied in monitoring changes in marine environments (Michalek et. al., 1993). CVA method along with Transformed Normalized Difference Vegetation Index (TNDVI) have been used to study the recovery potential of Tsunami affected area along the Andaman sea coast of Thailand (Romer et. al., 2012). The change of spatial use of the coastal area can be estimated by calculating the exposure value which is the sum of Land Use Intensity (LUI) and Shoreline Use Intensity (SUI). LUI can be calculated by using the ratio of various LU/LC types. Similarly SUI can be calculated by using the ratio of various types of shoreline (Huang et. al., 2010).

Murali et. al., (2014 and 2015) have examined the implications of SLR on LU/LC classes of coastal zones of Orissa and Cochin respectively. Effect had been studied on the major LU/LC classes like Agriculture, Built-up, Water body etc., by preparing inundation maps. They are useful in assessing the extent of SLR and potential impact on inundation. LU/LC map generated with conventional maximum likelihood classifier is merged with Digital Elevation Model for quantifying inundation. They considered the SLR of 1m and 2m to calculate the habitat loss. Dwarakish et. al., (2009) had done an extensive study along the Udupi coast of Karnataka, by calculating Coastal Vulnerability Index (CVI) along the coast and by preparing LU/LC map and inundation map of the study area. Nayak et. al., (2013) have divided the whole Indian coast into four regions namely coasts along Mumbai, Kochi, Chennai and Vishakapatnam and estimated that 34.906 sq. km. of Indian coast had been inundated due to SLR. Similar study had been done by Marfai et. al., (2008) over Semarang city, Indonesia using IKONOS data. This study gained significance because there are about 17500 islands covering 88000 km of coastline in the Indonesian region.

Flood extent maps have been produced for Punjab province of Pakistan using Moderate Resolution Imaging Spectroradiometer (MODIS). In-situ data coupled with remote sensed data have been used to develop flood damage index (Atif et.al., 2016). Coastal Vulnerability Index (CVI) has been assessed for 298 km of Karnataka coast by dividing it into eight Taluks namely Karwar, Ankola, Kumta, Honnavar, Bhatkal, Kundapura, Udupi and Mangaluru. It had been reported that 68.65 km of the coast is under very high vulnerable category (Belur et.al., 2016). The repercussion of a cyclone event on the rice crop has been studied using Remote Sensing over the region of Odisha. Aforementioned study has been done using Synthetic Aperture Radar (SAR) data where backscattered radiation is recorded (Haldar et. al., 2016). Saline water intrusion pathways has been found with the help of land forms and lineaments extracted from LISS III and SRTM (Dhakate et. al., 2016). Table 2 shows the works done by various researchers related to coastal inundation.

Table 2. Study on Coastal inundation

Researchers' Name Year Remarks

Michalek et. al., 1993 CVA in Marine region

Lu et. al., 2004 Studied various change detection techniques for LU/LC

Marfai et. al., 2008 Estimated inundation due to SLR

Dwarakish et. al., 2009 Studied CVI of the Coast

Huang et. al., 2010 LUI and SUI to study coastal change detection

Romer et. al., 2012 CVA and TNDVI to

study Andaman coast

Nayak et. al., 2013 Estimated inundation due to SLR

Murali et. al., 2014 Implication of SLR

Murali et. al., 2015 Implication of SLR

Atif et.al., 2016 Flood extent using MODIS

Belur et.al., 2016 CVI of Karnataka Coast

Haldar et. al., 2016 Study on aftermath of cyclone

Dhakate et. al., 2016 Saline water intrusion

4. SOFT COMPUTING TECHNIQUES IN WIND TURBINE STUDY

Coastal and offshore area serve as a potential region for the construction of wind farms to produce electricity. The wind speed over the offshore region is higher when compared to land and so fewer offshore wind turbines cab be able to produce more energy. Wind energy from a turbine encounters problem such as curtailment in the wind speed due to wake effect by other turbines. Wake effect is caused when a turbine is placed within the area of turbulence caused by another turbine or the area behind another turbine. This in turn triggered the reduction in wind speed leading to lessened power production. So as to increase the efficiency of wind farm, evaluating the parameters that influence on the wake effect, is one of the important research areas. One of the soft computing techniques, employed to study the various aspects of wind turbines and wake effect is, an Adaptive Neuro-Fuzzy Inference System (ANFIS). The prediction capacity of ANFIS has also been studied in the estimation of wind speed distribution. The output has been validated against most commonly used two parameter Weibull distribution. It is seen that ANFIS is preferable for assessing Weibull parameters for wind energy applications (Petkovic, D. 2015). Similar research work has been done by Shamshirband S. et. al. (2015) by using Extreme Learning Machine (ELM) and its efficiency had been compared with Support Vector Machine (SVM), Artificial Neural Network (ANN) and Genetic Programming (GP) for estimating the Weibull parameters (shape and scale parameter). A new class of Diffuser Augmented Wind Turbine (DAWT) is in use wherein the diffuser draws more wind towards low pressure regions and thus increases the energy production. Nikolic V. et. al., (2015) have studied the effect of diffuser on the performance of rotor through estimating power output, torque output and rotational speed of the rotor using ANFIS and Simulink. Support Vector Regression (SVR) has also been used for envisaging wind turbine reaction torque (Shamshirband S. et. al. 2014). Some more researchers have also contributed to the above area of research (Shamshirband S. et. al. 2014, Petkovic, D. 2014 and Nikolic V. 2015). An attempt has been made by Pai et. al., (2009) to estimate the wind and wave parameters using SAR imagery. A numerous studies are made on wave and tidal forecasting that aid in proficient coastal management (Gowthaman et. al., 2013 and Rakshith et. al., 2014).

5. SOFT COMPUTING TECHNIQUES IN REMOTE SENSING CLASSIFICATION

Soft computing techniques can be used capably for Remote sensing image classification. In conventional classification technique, training and testing are done based on one pixel - one class method. So there arises a conflict in classifying mixed pixels, which most often end up in mis-classification. Fuzzy Logic (FL) which takes into account of the belongingness of an element to a particular class through membership functions had been used for this purpose as an alternative method. (Wang 1990 and Binaghi 1997). Zadeh proposed the

concept of fuzzy logic that dates back to 1960's. Crispy sets, where an abrupt boundary exist were replaced by fuzzy sets where a tolerance is given to the boundary based on the degree of belongingness. The outcome of fuzzy classification depends on the predefined fuzzy rules. Genetic Algorithm (GA) which depicts the natural evolution can be used for optimising the fuzzy rules which in turn provides good classification result (Yuan et. al., 1996). Particle Swarm Optimisation (PSO), one of the effective data mining technique was proposed by Eberhart and Kennedy. It had also been used in framing classification rules for remote sensing images (Wang et. al., 2007).

Artificial Neural Network (ANN) mimics the human brain through a number of elements, similar to neurons which are interconnected to do processing. This processing depend on weighted connections, that are similar to synapses in the human brain. Neural network has to be trained to establish a relation between the input and output given during training. A good data set is required to draw better classification accuracy.

Machine learning examines the study and building of algorithms that can efficiently learn from and make predictions on data. SVM is one such powerful tool to classify the data into two categories, which implies that SVM is a binary classifier. The principle is to optimize the decision boundary that would maximise the margin around it. Margin is the region around each training sample through which decision boundary cannot pass. There exists two problems in classification. (i) When a problem is easy to classify and the boundary function is complicated more than it needs to be, the boundary is likely overfit. (ii) When a problem is hard and the classifier is not powerful enough, the boundary becomes underfit. SVMs are excellent examples of supervised learning that tries to maximize the generalization by maximizing the margin. It also support nonlinear separation using advanced kernels, by which SVMs try to avoid over-fitting and under-fitting (Yu et.al., 2003). Multi- class SVM is achievable by concatenating multiple binary classifiers. In this paper, ANN and SVM are chosen to review their applications in the classification of remote sensing data and coastal study.

5.1. ANN IN REMOTE SENSING DATA CLASSIFICATION

Each of the pixel in the remote sensing images have its own characteristic value. The spatial organisation of these pixels will be used for LU/LC classification (Fung and Chan 1994). The variation of the above mentioned characteristic values of the pixel give rise to the concept of texture. Several studies had been carried out based on texture based image classification by applying several parameters (Haralick et. al., 1973). These techniques had been applied to Landsat images and an accuracy upto 80% had been obtained.

ANN research had experienced three periods of extensive activity. The first peak in 1940s, when McCulloch and Pitts (1943) proposed the first mathematical neuron model. This was followed by Rosenblatt's perceptron theory in 1960s. Since the late 1980s, there has been a dramatic growth in the level of research activity in the neural network field. The major development was the introduction of back- propagation learning algorithm for the multilayer perceptron. These advantages make ANNs an accepted alternative to traditional statistical methods for improving classification accuracy. Studies carried out by Hilbert and Ostendorf (2001), Blaschke et al. (2004) and Kavzoglu (2009) prove that in the last decade, neural network classifiers have been widely used in the fields of pattern recognition, image interpretation and clustering/categorization. Hepner (1990) explains about how neural network can be applied to land cover classification of Thematic Mapper images. The neural network model had several training site examples to 'learn' the association between these sites and the desired land- cover classes. The obtained results were compared with the conventional supervised classification method and it was found that ANN is superior to conventional classification methods. Zhou and Yang (2011) studied about developing some guides for parameterizing the multi-layer perceptron (MLP) feed-forward back-propagation neural networks. It is found that the internal parameters significantly affect the classification accuracy. Madhubala et al. (2010) classified LISS-III satellite images into different classes as agriculture, urban and water body by using pixel based classification and Back propagation technique in neural network Model. Accuracy was found to be increased as the size of the training dataset is increased. Gray scale aerial images can be classified into various types of classes like forest, water, city, agricultural field etc., by using ANN based classification (Ashish 2002). With the advent of new technology, ANN was combined with various other methods like Fuzzy Logic, Genetic algorithms etc., Foody (1996) carried out a combination study of neural networks and fuzzy classification and concluded that fuzzy classification techniques are only a partial solution to the mixed pixel problem, as they only enable an appropriate representation of the land cover composition of pixels. Milan et al. (1990) studied about the comparison of fuzzy classification and neural network classification with each

other and both of them with Minimum distance method. Fuzzy and Neural network showed significant increase in classification accuracy than the conventional method. Yongzhu Xiong et al. (2010) used both ANN and Decision Trees (DT) classifiers to extract LU/LC from remote sensing images and compared their accuracies with the statistical Minimum Distance (MD) classifier. Results indicate that ANN has better accuracy than the DT and MD classifiers. This shows that ANN is a more effective method for remote sensing image classification of mountainous areas because of its higher accuracy and superior performance than DT and MD classifiers. Investigating the sensitivity of neural networks with respect to various parameter settings has been the subject in an increasing number of studies, since this knowledge is critical to the design of efficient neural network models for improved performance (Yoshida and Omatu 1994; Kanellopoulos and Wilkinson 1997).

ANN is suitable for problems when a large diversity in the dataset exist. It can also be used for segmentation and classification purposes effectively (Schalkoff 1992). ANN had been used in image classification and high accuracy had been obtained (Ritter and Hepner 1990). Some of the classical statistical classification methods were compared with ANN to prove its superiority (Bischof et. al., 1990, Schowengerdt 1995 and Zhang et. al., 2001). The use of artificial neural networks is, however, challenged by the difficulty of network design and parameterization. Many factors affect the performance of ANN that include network topology, training algorithm and parameters setting, and network architecture (Kavzoglu and Mather 2003; Zhou and Yang 2010; Zhou and Yang 2011). Table 3 shows the works done by various researchers related to ANN in Remote Sensing

Table 3. Study on ANN in Remote

Researchers' Name Year Remarks

McCulloch and Pitts 1943 First Neuron model was proposed

Haralick et. al., 1973 Texture based Image classification

Hepner 1990 Application of ANN to LU/LC classification

Milan et. al., 1990 /X Comparison of Fuzzy classification and ANN classification

Bischof et. al., 1990 Comparison of ANN and statistical classification techniques

Schalkoff 1992 ANN for segmentation and classification

Fung and Chan 1994 Pixel based LU/LC classification

Yoshida and Omatu 1994 Analysis of Sensitivity parameters of NN

Schowengerdt 1995 Comparison of

ANN and

statistical

classification

techniques

ANN in

Foody 1996 combination with Fuzzy technique

Wilkinson 1997 Analysis of Sensitivity parameters of NN

Comparison of ANN and

Zhang et. al., 2001 statistical classification techniques

Classification of

Ashish 2002 Aerial images using ANN

Comparing Pixed based

Madhubala et. 2010 classification

al., and Back propagation technique

Yongzhu Xiong et. al., 2010 Comparing ANN and Decision

Tree algorithm

Zhou and Yang 2011 Development of MLP

on thm

5.2. SVM IN REMOTE SENSING DATA CLASSIFICATION

Image classification and pattern recognition have become important in many fields of engineering and medicine. Many researchers are working towards development of algorithms so as to improve the accuracy in classifying and recognizing the patterns of image. SVM is a supervised and non- parametric machine learning algorithm and it was invented as a linear binary classifier by Vapnik and Chervonenkis in 1963. Most of the real world data is non - linear in nature. So, in 1992 some of the researchers recommended nonlinear classifiers by applying the kernel trick to study the real world data by maximizing the margin between hyperplanes (Boser et. al., 1992). SVM classifier can optimally locate hyperplanes with minimal training data. Widely used soft margin based SVM was proposed by Cortes and Vapnik in 1993 and published in 1995 (Cortes et.al., 1995). SVM based classification had been known well to reach the right balance between accuracy attained on a given minimal amount of training data and the capacity to generalize unseen data (Mountrakis et. al., 2011). SVM had been used to classify various type of data such as heart data, diabetes data, satellite data and shuttle data (Durgesh et.al., 2010).

Some of the softwares that are being used to execute the SVM algorithm are listed out in this section. SVM llght, SVM struct, mySVM, JmySVM, LIBSVM (Library for Support Vector Machines), looms, BSVM, SVMTorch, Weka, SVM in R, M-SVM (Multi - class SVM), Gist, MATLAB SVM Toolbox, SmartLab, Gini-SVM, GPDT, HeroSvm, SVMsequel, LSVM (Lagrangian support Vector Machine), ASVM (Active Support Vector Machine), PSVM (Proximal Support Vector Machine), etc.,

The application of remote sensing data in various fields is evolving day by day. The high spectral capacity of

the sensors had given rise to multispectral and hyperspectral imagery. The usefulness of those images lies in, how well the data is converted into information. Image classification is the process of data mining technique that involves extracting interesting patterns representing knowledge from the satellite data products.

Hyperspectral data having a large number of narrow bandwidths helps us in detail study about the earth features. SVM had been applied in classifying AVIRIS (Airborne Visible Infra-red Imaging Spectrometer) data as it do not suffer from the limitations of data dimensionality and limited samples. Kernels in the SVM algorithm can make use of the high dimensionality of hyperspectral data to transform original input space to higher dimensional feature space (Zhang et. al., 2001). A broad study had been done on the kernels in SVM that will be suitable for image classification. It had been seen that Radial Basis Function (RBF) outperformed polynomial kernel in the same context (Kavzoglu et. al., 2009). To avoid the confusion in selection of kernels, Support Vector Selection and Adaption (SVSA) method had been used. The SVSA method consists of selection of the support vectors that contribute most to the classification accuracy and their adaptation based on the class distributions of the data without using the kernels (Kaya et. al., 2009). Techniques like Transducitve SVM (TSVM) and Contextual SVM had been used to make study on hyperspectral data (Plaza et. al., 2009).

The complexity of the classifier in working with huge amount of data can be reduced by using Clustering-Based SVM (CB-SVM). CB-SVM utilizes a hierarchical micro-clustering algorithm. It examines the entire data set only once to provide an SVM with high quality samples that carry the statistical summaries of the data such that the summaries maximize the benefit of learning the SVM (Yu et. al., 2003). This in turn, assists in SVM fast implementations, approximations, selective sampling and SVM for dynamic environments.

Neural Network had been in use for image classification in late 1960s. Researchers have compared the efficiency of neural network classifiers with other statistical based algorithms (Benediktsson et.al., 1990). Pal et. al., (2005) have compared the promising development of SVM with Neural network classifiers and Maximum likelihood classifier using multispectral Landsat ETM+ and hyperspectral Digital Airborne Imaging Spectrometer (DAIS) data. Studies have also been made to prove the time efficiency of SVM when compared with neural networks (Dixon et. al., (2007)). SVM also outperforms traditional classifier like Maximum likelihood classifier (MLC) and Decision Tree Classifier (DTC) (Huang et. al., 2002) .

Hierarchical classification approach combined with SVM had been verified to increase the accuracy and also reduces the computational time which is achieved through multilevel wavelet decomposition of each band of hyperspectral image. Pixels of the highest level (lowest resolution) are classified using all support vectors, while pixels at sub-sequent lower levels (higher resolution) are classified using neighbourhood SVM classification, that only uses support vectors of classes to which the corresponding neighbours of the higher level belong to (Demir et. al., 2009).

LU/LC analysis is the primary work which is being done by most of the Remote Sensing and GIS analysts. LU/LC of a particular area will be heterogeneous in nature. As SVM is well known as a linear binary classifier, researchers have worked on it to extend to multi-level classification. There are two general approaches: One against One (1A1) and One against All (1AA) approach wherein, both requires multiple implementations of binary classifiers and cascading the outcomes to achieve a single multi-class classification (Hsu et. al., 2002). 1A1 and 1AA approaches were compared by incorporating linear, quadratic, Polynomial and Radial Basis Function kernels. (Anthony et.al., 2007). The choice of 1A1 or 1AA becomes a matter of preference as both yield similar results. There also exists Directed Acyclic Graph SVM (DAGSVM) algorithm that combines the outcomes of 1A1 SVMs to get multi-class output (Chen at. al., 2009). Multi -class SVM with a single classifier is also possible, but with larger optimization problem to arrive at very accurate classification (Foody et. al., 2004). An improvement had been brought to the cascade SVM by Parallel SVM training algorithm for large scale classification problems that works efficiently (Zhang et. al., 2005).

Most often studies are based on a particular type of LU/LC like Agricultural land, wetlands, Urban area, etc., (Zhu et.al., 2009). SVM is a best suitable algorithm for such studies. The mapping of Mangrove fringes had been done to high accuracy by merging Object Oriented Image Analysis (OBIA) and SVM (Heumann 2011). Remote sensing of tea plantations suffer from the reflectance due to the surrounding orchards and bushes. SVM had been used to map such tea plantations using Modified Normalized Difference Vegetation Index

(MNDVI) as a supporting parameter (Dihkan et. al., 2013). Sanchez-Hernandez et. al., (2007) have mapped the coastal salt marsh habitats using Support Vector Data Description (SVDD). It only uses training data from the class of interest and targets to calculate an optimal hyperplane boundary that closes around the target class and separates it from any other possible classes. In the context of study of urban area using remote sensing, there exists some confusion between buildings and roads. SVM has been successfully employed in differentiating the above spectrally similar classes (Rafiee et. al., 2008). They have also made a mention that Maximum Likelihood classifier fails in such a case, because it considers the mean of training data as representative of each class and not the individual values of training data. Spatial correlation analysis like Principal Component Analysis (PCA), Discrete Wavelet Transform (DWT), Segmentation and Semivariance plays an important role in remote sensing data analysis. SVM had been showed to be effective in studying the images with one or combination of the above spatially correlated images (Hwang et. al., 2010). Coastlines are of many types namely Bedrock coast, Sandy coast, Muddy coast, Mangrove coast and Artificial coast. Different image processing techniques are being applied on each type of shoreline to enhance the same. Adaptive filtering for sandy coast, image sharpening for muddy coast and edge enhancement for bedrock and artificial coast had been effectively used as image enhancement techniques. Enhanced images can then be classified by applying sigmoid kernel in SVM (Hanny et. al.,2013). Modification of Normalized difference Water Index (MNDWI) in combination with SVM yields good results in classifying or mapping shoreline (Yin et.al., 2011).Table 4 shows the works done by various researchers related to SVM in Remote Sensing.

Table 4. Study on SVM in Remote Sensin

Researchers' Name Year Remarks

Vladimir N. Vapnik and Alxey Ya. Chervonenkis 1963 Invention of SVM as linear binary classifier

Boser et. al., 1992 Non-linear SVM with Kernel trick

Zhang et. al., 2001 Classification of hyperspectral data using SVM

Hsu et. al., 2002 Multi- class SVM

Yu et. al., 2003 Cluster Based SVM

Foody et. al., 2004 Multi- class SVM with single SVM

Pal et. al., 2005 Comparison of SVM and NN classifiers

Zhang et. al., 2005 Parallel SVM training

Dixon et. al., 2007 Time efficiency of SVM

Anthony et.al., 2007 Comparison of 1A1 and 1AA techniques of Multi-class SVM

Sanchez-Hernandez et. al., 2007 SVM to map Coastal salt marsh habitats

Kavzoglu et. al., 2009 Study on

suitability of Kernels

Kaya et. al., 2009 Support Vector Selection and Adaption method

Transductive

Plaza et. al., 2009 SVM and Contextual SVM

Hierarchical

Demir et. al., 2009 classification

combined with SVM

Durgesh et.al., 2010 SVM based data classification

Hwang et. al., 2010 SVM with PCA and DWT

SVM with

Heumann 2011 Object Oriented Image Analysis

SVM for

Yin et.al., 2011 mapping shoreline

SVM with

Dihkan et. al., 2013 MNDVI to map tea plantations

6. DISCUSSION AND CONCLUDING REMARKS

• The present review focused on the application of remote sensing techniques for the coastal studies namely shoreline mapping, coastal LU/LC and coastal inundation mapping. Also the application of soft computing techniques for evaluating the wind turbine performance had also been dealt.

The conventional techniques that had been discussed show the effectiveness of using satellite imagery to study the earth features, in particular, coastal area which requires continuous monitoring.

• The main aim of coastal management planning is to control erosion and flooding. Now a days, microwave radiometers, infrared radiometers, altimeters, ocean color sensors and synthetic aperture radars are used in satellite oceanographic studies.

• However, satellites operating in the optical range like Landsat and Indian Remote Sensing satellites can also be used.

• Conventional image processing techniques are useful to understand the process involved.

• Considering shoreline mapping, which started in the earlier days by manually surveying the beach profile, then developed to the process of digitization using ArcGIS and manually calculating the rate of change. Advancement in software development led to DSAS which can automatically calculate the rate of change of shoreline provided the parameters are properly set.

• Automatic shoreline delineation is also becoming possible due to the algorithmic advancement.

• Considering LU/LC mapping, unsupervised classification technique though not accurate, were in use to study about the earth features at reconnaissance level. Then supervised techniques are developed to improvise the accuracy in mapping the land use and land cover.

• Among the known supervised classification algorithms, Maximum Likelihood Classifier (MLC) gives the better classification accuracy. However, the conventional methods are data intensive.

• Soft computing techniques which find its foot print in computer science application has grown across multi - disciplinary sciences. It take into account of the uncertainty that exist in the real world data. Hence, it can be constructively applied to remote sensing data where huge amount of data is involved.

• Fuzzy Logic reflects the degree of belongingness, Genetic Algorithm reflects the theory of evolution,

Particle Swarm Optimization reflects the representation of movement of organisms, Artificial Neural Network mimics the functions of human brain and Support Vector Machine considers the machine learning algorithm.

ANN and SVM had been widely used in remote sensing field. They had gained a lot of improvement in their configuration.

The performance of ANN in remote sensing data classification lies mainly on deciding the parameters like number of hidden layers, learning rate of the network, activation function that controls the internal weights of the neurons, momentum coefficient of the neural architecture and number of iterations control.

However, with an advent of SVM and the concept of kernels for non linear problems, it occupied higher position in classification algorithm. Selection of kernels play a crucial role in SVM classifier. In short, it can be concluded that, high learning capacity with limited amount of available data, adapting itself to newly seen data and kernel tricks on nonlinear data make SVM a promising technique for remote sensing data classification.

REFERENCES

1. Anthony, G., Greg, H. and Tshilidzi, M. (2007). "Classificat 'images using support vector machines." arXiv preprint arXiv:0709.3967.

2. Ashish, D. (2002). "Land-use classification of aerial using artificial neural networks.' International Journal of Remote Sensing. 30(8): 1989-20

3. Atif, I., Mahboob, M. A., & Waheed, A. (2016). "S)atio-Temporal Mapping and Multi-Sector Damage Assessment of 2014 Flood in Pakistan using I emote Sensing and GIS." Indian Journal of Science and Technology, 9(1).

4. Beluru Jana, A., & Hegde, A. V. (2016). "GIS Based Approach for Vulnerability Assessment of the Karnataka Coast, India." Advances in Civil Engineering, 2016.

5. Benediktsson, J., Swain, P. H. an d Ersoy, O. K. (1990). "Neural network approaches versus statistical methods in classification of multisource remote sensing data." Geoscience and Remote Sensing, IEEE Transactions on, 28(4), 540-552.

6. Binaghi, E., Madella, P., Montesano, M. G. and Rampini, A. (1997). "Fuzzy contextual classification of multisource remote sensing images." Geoscience and Remote Sensing, IEEE Transactions on,

35(2), 326-3

7. Bischof, H., Schneider, W. and Pinz, A.J., (1990). "Multispectral classification of Landsat-images using neural networks." IEEE Transactions on Geoscience and Remote Sensing. 30 (3): 482-490.

Lke, T. (2010). "Object based image analysis for remote sensing." Journal of Photogrammetry emote Sensing. 65 (1): 2-16.

9. Boser, B. E., Guyon, I. M. and Vapnik, V. N. (1992, July). "A training algorithm for optimal margin classifiers." In Proceedings of the fifth annual workshop on Computational learning theory . ACM. (pp. 144-152).

10. Chen, P. and Liu, S. (2009, August). "An improved dag-svm for multi-class classification." In

Natural Computation, 2009. ICNC'09. Fifth International Conference on (Vol. 1, pp. 460-462). IEEE.

11. ChenthamilSelvan, S., Kankara, R. S., and Rajan, B. (2014). "Assessment of shoreline changes along Karnataka coast, India using GIS & Remote sensing techniques." Indian Journal of Marine Sciences,

43(7).

12. Cortes, C. and Vapnik, V. (1995). "Support-vector networks."Machine learning,.20(3), 273-297.

13. Cracknell, A. P. (1999). "Remote sensing techniques in estuaries and coastal zones an update." International Journal of Remote Sensing, 20(3), 485-496.

14. Dai, C. G., Huang, X. B., and Dong, G. J. (2007, August). "Support vector machine for classification of hyperspectral remote sensing imagery." In Fuzzy Systems and Knowledge Discovery, 2007. FSKD 2007. Fourth International Conference on, Vol. 4, pp. 77-80. IEEE.

15. Demir, B. and Erturk, S. (2009, November). Improving SVM classification accuracy using a hierarchical approach for hyperspectral images. In Image Processing (ICIP), 2009 16th IEEE International Conference on (pp. 2849-2852). IEEE.

16. Dhakate, R., Sankaran, S., Kumar, V. S., Amarender, B., Harikumar, P., & Subramanian, S. K. (2016). "Demarcating saline water intrusion pathways using remote sensing, GIS and geophysical techniques in structurally controlled coastal aquifers in Southern India." Environmental Earth Sciences,75 (5), 1-19.

17. Dihkan, M., Guneroglu, N., Karsli, F. and Guneroglu, A. (2013). "Remote sensing of tea plantations using an SVM classifier and pattern-based accuracy assessment technique." International Journal of Remote Sensing, 34(23), 8549-8565.

18. Dixon, B. and Candade, N. (2008). "Multispectral landuse classification using neural networks and support vector machines: one or the other, or both?." International Journal of Remote Sensing, 29(4), 1185-1206.

19. Durgesh, K. S. and Lekha, B. (2010). "Data classification using support vector machine." Journal of Theoretical and Applied Information Technology, 12(1), 1-7.

20. Dwarakish, G. S., Vinay, S. A., Natesan, U., Asano, T., Kakinuma, T., Venkataramana, K., Jagadeesha Pai, B., and Babita, M. K. (2009). "Coastal vulnerability assessment of the future sea level rise in Udupi coastal zone of Karnataka state, west coast of India." Ocean & Coastal Management, 52(9), 467-478.

21. Foody, G. M. And Mathur, A. (2004). "A relative evaluation of multiclass image classification by support vector machines." Geoscience and Remote Sensing, IEEE Transactions on, 42(6), 1335-1343.

22. Foody, G.M. and Strahler, A.H. (1996). "Artificial Neural Network Response To Mixed Pixels In Coarse-Resolution Satellite Data." Remote Sensing Of Environment. 58: 329-343.

23. Fung, T. and Chan, K. (1994). "Spatial composition of spectral classes: A Structural approach for Image Analysis of Heterogeneous Land-Use and Land-Cover Types." Photogrammetric Engineering and Remote Sensing. 60 (2): 173-180.

24. Gowthaman, R., SanilKumar, V., Dwarakish, G. S., Mohan, S. S., Singh, J., & AshokKumar, K. (2013). "Waves in Gulf of Mannar and Palk bay around Dhanushkodi, Tamil Nadu, India.

25. Haldar, D., Nigam, R., Patnaik, C., Dutta, S., & Bhattacharya, B. (2016). "Remote sensing-based assessment of impact of Phailin cyclone on rice in Odisha, India." Paddy and Water Environment, 111.

26. Hannv, Z., Qigang, J. and Jiang, X. (2013). "Coastline Extraction Using Support Vector Machine from Remote Sensing Image." Journal of Multimedia, 8(2), 175-182.

27. Haralick and Shanmugan. (1973). "Textural Features for Image Classification." IEEE Transactions on Systems, Man and Cybernatics. 3(6): 610-621.

28. Hegde, A. V., and Akshaya, B. J. (2015). "Shoreline Transformation Study of Karnataka Coast: Geospatial Approach." Aquatic Procedia, 4, 151-156.

29. Hepner, G.F. (1990). "Artificial Neural Network classification using minimal training set: Comparison to conventional supervised classification." Photogrammetric Engineering and Remote Sensing. 56: 469-473.

30. Heumann, B. W. (2011). "An object-based classification of mangroves using a hybrid decision tree— Support vector machine approach." Remote Sensing, 3(11), 2440-2460.

31. Hilbert, D.W. and Ostendorf, B. (2001). "The utility of artificial neural networks for modelling the distribution of vegetation in past, present and future climates." Ecological Modelling. 14(6): 311327.

32. Hsu, C. W. and Lin, C. J. (2002). "A comparison of methods for multiclass support vector machines." Neural Networks, IEEE Transactions on, 13(2), 415-425.

33. Huang, C., Davis, L. S. and Townshend, J. R. G. (2002). "An assessment of support vector machines for land cover classification." International Journal of Remote Sensing, 23(4), 725-749.

34. Huang, J., Cui, S., Qiu, Q., Shi, L., and Ma, K. (2010). "Estimates of exposure of a coastal city to spatial use changes-a case study in Xiamen." International Journal of Sustainable Development & World Ecology, 17(4), 292-298.

35. Hwang, J. T. and Chiang, H. C. (2010, June). "The study of high resolution satellite image classification based on Support Vector Machine." In Geoinformatics, 2010 18th International Conference on, (pp. 1-6). IEEE.

36. Kanellopoulos, I. and Wilkinson, G.G. (1997). "Strategies and best practice for neural network image classification." International Journal of Remote Sensing. 18(4): 711-725.

37. Kavzoglu, T. (2009). "Increasing the accuracy of neural network classification using refined training data." Environmental Modeling & Software. 24: 850-858.

38. Kavzoglu, T. and Colkesen, I. (2009). "A kernel functions analysis for support vector machines for land cover classification." International Journal of Applied Earth Observation and Geoinformation, 11(5), 352-359.

39. Kavzoglu, T. and Mather, P.M. (2003)."The Use of Backpropagating Artificial Neural Networks in Land Cover Classification." International Journal Of Remote Sensing. 24(23): 4907-4938.

40. Kaya, G. T., Ersoy, O. K. and Kama§ak, M. E. (2009, June). "Support vector selection and adaptation and its application in remote sensing." In Recent Advances in Space Technologies, 2009. RAST'09. 4th International Conference on (pp. 408-412). IEEE.

41. Lu, D., Mausel, P., Brondizio, E., and Moran, E. (2004). "Change detection techniques."

International Journal of Remote Sensing, 25(12), 2365-2401.

42. Madhubala, M., Mohan, S.K., and Ravindra, B.G. (2010). "Classification of Liss III images by using Artificial Neural Networks." IJCA special issue on 'Recent trends in image processing and pattern recognition'. 152-157.

43. Maiti, S., and Bhattacharya, A. K. (2009). "Shoreline change analysis and its application to prediction: a remote sensing and statistics based approach."Marine Geology, 257(1), 11-23.

44. Marfai, M. A. and King, L. (2008). "Potential vulnerability implications of coastal inundation due to sea level rise for the coastal zone of Semarang city, Indonesia." Environmental Geology, 54(6), 1235-1245.

45. McCulloch, W. and Pitts, W. (1943). " A logical calculus of the ideas immanent in nervous activity." Bull. Math. Biophys. 5: 115-133.

46. Michalek, J. L., Wagner, T. W., Luczkovich, J. J., and Stoffle, R. W. (1993). "Multispectral change vector analysis for monitoring coastal marine environments." Photogrammetric Engineering & Remote Sensing, vol. 59, No.3, pp. 381 - 384.

47. Milan, I.A., Hamrah, M. and Majdabadi. (1990). "Classification of Multi Spectral Images by Fuzzy and Neural Network Approaches." IEEE Transactions on Geosciences and Remote Sensing. 25(3): 14-22.

48. Mountrakis, G., Im, J. and Ogole, C. (2011). "Support vector machines in remote sensing: A review." ISPRS Journal of Photogrammetry and Remote Sensing,66(3), 247-259.

49. Mujabar Sheik and Chandrasekar. (2011). "A shoreline change analysis along the coast between Kanyakumari and Tuticorin, India, using digital shoreline analysis system." Geo-spatial Information Science, 14(4), 282-293.

50. Murali Mani, R., Shrivastava, D. and Veth; imony, P. (2009). ' Monitoring shoreline environment of Paradip, east coast of India using remote sensing." Current science, 97(1), 79-84.

51. Murali, R. M. (2014). "Application of geo-spatial technologies in coastal vulnerability studies due to sea level rise (SLR) along the Central Orissa Coast, India." Geospatial Technologies and Climate Change, 187-199.

52. Murali, R. M. and Kumar, P. D. (2015). "Implications of sea level rise scenarios on land use/land cover classes of the coastal zones of Cochin, India." J. Environ. Manage., 148, 124-133.

53. Natesan, U., Parthasarathy, A., Vishnunath, R., Kumar, G. E. J., and Ferrer, V. A. (2015). "Monitoring Longterm Shoreline Changes along Tamil Nadu, India Using Geospatial Techniques."

Aquatic Procedia, 4, 325-332.

54. Natesan, U., Thulasiraman, N., Deepthi, K., and Kathiravan, K. (2013). "Shoreline change analysis of Vedaranyam coast, Tamil Nadu, India." Environmental monitoring and assessment, 185(6), 50995109.

55. Nayak, S., Mandal, M., Adhikari, A. and Bhatla, A. (2013). "Estimation of Indian coastal areas inundated into the sea due to sea-level rise during the 20th century." Current Science, 104(5), 583585.

56. Nikolic, V., Petkovic, D., Shamshirband, S., & Cojbasic, Z. (2015). "Adaptive neuro-fuzzy estimation of diffuser effects on wind turbine performance." Energy, 89, 324-333.

57. Nikolic, V., Shamshirband, S., Petkovic, D., Mohammadi, K., Cojbasic, Z., Altameem, T. A., and Gani, A. (2015). "Wind wake influence estimation on energy production of wind farm by adaptive neuro-fuzzy methodology." Energy, 80, 361-372.

58. Pai, Jagadeesha B and Kumar, Raj and Sarkar, Abhijith and Bhowmick, Sunchandra Aich and Hegde, Vittal and Dwarakish, G S (2009). " Wind and Wave Parameters of West Coast of India Using Synthetic Aperture Radar Data." International Journal of Earth Sciences and Engineering, 2 (2). pp. 125134.

59. Pal, M. and Mather, P. M. (2005). "Support vector machines for classification in remote sensing." International Journal of Remote Sensing, 26(5), 1007-1011.

60. Paola, J. and Schowengerdt, R.A. (1995). "A review and analysis of backpropagation neural networks for classification of remotely-sensed multispectral imagery." International Journal of Remote Sensing. 16: 3033-3058.

61. Petkovic, D. (2015). "Adaptive neuro-fuzzy approach for estimation of wind speed distribution." International Journal of Electrical Power & Energy Systems, 73, 389-392.

62. Petkovic, D., Cojbasic, Z., Nikolic, V., Shamshirband, S., Kiah, M. L. M., Anuar, N. B., and Wahab, A. W. A. (2014). "Adaptive neuro-fuzzy maximal power extraction of wind turbine with continuously variable transmission." Energy, 64, 868-874.

63. Plaza, A., et. al., (2009). "Recent advances in techniques for hyperspectral image processing."

Remote Sensing of Environment, 113 (1), S110-S122.

64. Rafiee, A. and Sarajian, M. R. (2008, December). "Classification of Buildings and Roads Using Support Vector Machine. In Digital Image Computing: Techniques and Applications (DICTA)." (pp. 111-116). IEEE.

65. Rakshith, S., Dwarakish, G., & Natesan, U. (2014). "Tidal-Level Forecasting Using Artificial Neural Networks along the West Coast Of India." Journal of JSCE, 2(1), 176-187.

66. Rao Prabhakara, P., Nair, M. M., and Raju, D. V. (1985). "Assessment of the role of remote sensing techniques in monitoring shoreline changes: a case study of the Kerala coast." International Journal of Remote Sensing, 6(3-4), 549-

67. Ritter, N.D. and Hepner, G.F. (1990). "Application of an artificial neural network to land-cover classification of thematic mapper imagery." Computers and Geosciences. 16 (6): 873-880.

68. Romer, H., Jeewarongkakul, J., Kaiser, G., Ludwig, R., and Sterr, H. (2012). "Monitoring posttsunami vegetation recovery in Phang-Nga province, Thailand, based on IKONOS imagery and field investigations-a contribution to the analysis of tsunami vulnerability of coastal ecosystems. "

International Journal of Remote Sensing, 33(10), 3090-3121.

69. Rosenblatt, F. (1958). "The Perceptron: A Probabilistic Model For Information Storage And Organization In The Brain." Psychological Review. 65: 386-408.

70. Sanchez-Hernandez, C., Boyd, D. S. and Foody, G. M. (2007). "Mapping specific habitats from remotely sensed imagery: support vector machine and support vector data description based classification of coastal saltmarsh habitats." Ecological informatics, 2(2), 83-88.

71. Schalkoff, R. (1992). "Pattern Recognition: Statistical, Structural and Neural Approaches", Wiley, New York.

72. Selvan, S. C., Kankara, R. S., Markose, V. J., Rajan, B., & Prabhu, K. (2016). "Shoreline change and impacts of coastal protection structures on Puducherry, SE coast of India." Natural Hazards, 1-16.

73. Shamshirband, S., Mohammadi, K., Tong, C. W., Petkovic, D., Porcu, E., Mostafaeipour, A., ... & Sedaghat, A. (2015). "Application of extreme learning machine for estimation of wind speed

distribution." Climate Dynamics, 1-15.

74. Shamshirband, S., Petkovic, D., Amini, A., Anuar, N. B., Nikolic, V., Cojbasic, Z. and Gam, A. (2014). "Support vector regression methodology for wind turbine reaction torque prediction with power-split hydrostatic continuous variable transmission." Energy, 67, 623-630.

75. Shamshirband, S., Petkovic, D., Anuar, N. B. and Gani, A. (2014). "Adaptive neuro-fuzzy generalization of wind turbine wake added turbulence models." Renewable and Sustainable Energy Reviews, 36, 270-276.

76. Shetty, A., Jayappa, K. S., and Mitra, D. (2015). "Shoreline Change Analysis of Mangalore Coast and Morphometric Analysis of Netravathi-Gurupur and Mulky-pavanje Spits". Aquatic Procedia, 4, 182-189.

77. Wang, F. (1990). "Fuzzy supervised classification of remote sensing images." Geoscience and Remote Sensing, IEEE Transactions On, 28(2), 194-201.

78. Wang, Z., Sun, X. and Zhang, D. (2007). "A PSO-based classification rule mining algorithm." In

Advanced intelligent computing theories and applications. With aspects of artificial intelligence (pp. 377-384).

79. Yin, J. and He, F. (2011, June). "Researching the method of coastline extracted by remote sensing image." In Remote Sensing, Environment and Transportation Engineering (RSETE), 2011 International Conference on, (pp. 3441-3444). IEEE.

80. Yongzhu, Xiong., Run, Wang. and Zhi, Li. (2010). "Extracting Land Use/Cover of Mountainous Area from Remote Sensing Images Using Artificial Neural Network and Decision Tree Classifications." International Symposium on Intelligence Information Processing and Trusted Computing.

81. Yoshida, T. and Omatu, S. (1994). "Neural Network Approach to Land Cover Mapping." IEEE Transactions on Geoscience and Remote Sensing. 32(5):- 1103-1109.

82. Yu, H., Yang, J. and Han, J. (2003, August). "Classifying large data sets using SVMs with hierarchical clusters." In Proceedings of the ninth ACM SIGKDD international conference on Knowledge discovery and data mining, ACM. (pp. 306-315).

83. Yuan, Y. and Zhuang, H. (1996). "A genetic algorithm for generating fuzzy classification rules."

Fuzzy sets and systems, 84(1), 1-19.

84. Yunus, A. P., Dou, J., Avtar, R., & Narayana, A. C. (2016). "Shoreline and Coastal Morphological Changes Induced by the 2004 Indian Ocean Tsunami in the Katchal Island, Andaman and Nicobar-A Study Using Archived Satellite Images." In Tsunamis and Earthquakes in Coastal Environments (pp. 65-77). Springer International Publishing.

85. Zhang, J. and Foody, G. M. (2001). "Fully-fuzzy supervised classification of sub-urban land cover from remotely sensed imagery: statistical and artificial neural network approaches", International Journal of Remote Sensing. 22 (4): 615-628.

86. Zhang, J. P., Li, Z. W. and Yang, J. (2005, August). "A parallel SVM training algorithm on large-scale classification problems." In Machine Learning and Cybernetics, 2005. Proceedings of 2005 International Conference on, (Vol. 3, pp. 1637-1641). IEEE.

87. Zhang, J., Zhang, Y. and Zhou, T. (2001). "Classification of hyperspectral data using support vector machine." In Image Processing, 2001. Proceedings. 2001 International Conference on, Vol. 1, pp.

882-885). IEEE.

88. Zhou, L. and X. Yang. (2010). "Training algorithm performance for image classification by neural networks." Photogrammetric Engineering and Remote Sensing. 76(8): 945-951.

89. Zhou, L. and X. Yang. (2010). "Training algorithm performance for image classification by neural networks." Photogrammetric Engineering and Remote Sensing. 76(8): 945-951.

90. Zhou, L. and X. Yang. (2011). "An assessment of internal neural network parameters affecting image classification accuracy." Photogrammetric Engineering and Remote Sensing. 77(12): 1233-1240.

91. Zhu, H., Yang, X. and Luo, Y. (2009, August). "Classification of urban remote sensing image based on support vector machines." In Geoinformatics, 2009 17th International Conference on, (pp. 1-6).

Graphical Abstract