Professional Experience & Contribution

Associate Professor /Tenured

Assistant Professor (Tenure Track)

Department of CIS and Technology

University of West Alabama

Aug 2016 -Present

Jan. 2012 – July 2018 (6.5 years)

Courses currently taught:- ( 150 hours of classroom teaching)

C programming
Object Oriented Programming with C++ and Python
Electric Motor Controls with Lab
AC-DC Circuit Theory with Lab
Programmable Logic Circuits with Lab
Advanced Statistics
Solid State and Digital Electronics with Lab
Industrial Electro-Mechanical Controls with Troubleshooting
Advanced Programmable Logic Controller Programming
Educational Statistics for Graduate Students (Adjunct Professor for College of Education)
Robotics & Micro-Controller Programming
Special Topics in Mechatronics/Advanced PLC programming
Special Topics in Micro-Controller Programming using PIC 24F, ATmega and ARM Processor.

Post Doctoral Associate

Geosystems Research Institute,

HPCC, Mississippi State University

Aug. 2010 – Jan 2011 (1.6 years)

Project 1: My role as a pattern recognition engineer is to devise novel landslide detection approaches on earthen levees of Mississippi river, using airborne synthetic aperture radar data from NASA’s UAVSAR sensor. This project is funded by Southeast Region Research Initiative (SERRI ).

Ø A hybrid approach of multi-polarized feature extraction and machine learning based feature classification is developed for fine tuning of the complex and generalized landlside detection model [4]

Project 2: An exploitation study of recently available Worldview-2 sensor data for urban remote sensing is investigated though the data-fusion contest conducted by Geoscience and Remote Sensing Society (GRSS). Its unique multi-angular and very high resolution capabilities are explored for very accurate classification of urban man-made objects.

Ø A multi-level automatic building extraction approach was developed using the milti-view optical imagery. A template matching algorithm is formulated for estimating the relative building height in the process. The building detection accuracy of the proposed approach is improved to 88% compared to 83% without using multi-angular information.

Proposal 1: Hybrid CPU/GPU parallelization of pseudo-spectral solver for diret numerical simulations of transitional flows. The objective of the study is to: (a) accelerate the performance of a pseudo-spectral solver using hybrid Central Processing Units (CPUs)/Graphics Processing Units (GPUs) parallelization and scale up to 50K cores; (b) apply the solver for temporally developing bypass transition flows over a flat plate to create a direct numerical simulation (DNS) database for a range of turbulence intensities. This proposal is submitted to Oakridge Research National Laboratory (ORNL) for director discretionary awards towards code development and possible INCITE allocation [5].
Proposal 2: for possible collaboration and membership with National Center for Computational Sciences at ORNL through NSF Center for Autonomic Computing at MSU. Several meetings have been conducted between the chief collaborators regarding this membership and scope for development of scientific computing applications.

Graduate Research Assistant

Center for Advanced Vehicular Systems,

HPCC, Mississippi State University

Jan. 2007 – Aug. 2010 (3.6 years)

Development of Spatio-Temporal based Machine Learning (STML) techniques and Approaches for automatic target recognition on Multi-spectral and Hyper-spectral sensor datasets. The above techniques ranged from new boundary based segmentations, several feature selection procedures including up to Genetic Algorithms, Fast-Fourier-Transform (FFT) based template matching techniques. The kernel based approaches were experimented rigorously ranging from support vector machines, support vector regression, neural networks, ensemble techniques, data fusion techniques etc.

a) Developed XML Schema for Image Information Mining Standardization

In the period of my Graduate Research Assistantship and Doctoral Dissertation, I developed and implemented the new XML schemas based standardization for sensor-web data mining and also developed methods and approaches for spatio-temporal machine learning to improve the classification of the target classes in the multi-spectral (MS) satellite image datasets. My dissertation based publications has been cited significantly with 27 citations in the last 4 years. We initiated the discussion within the image information mining (IIM) community to evolve and develop specifications for image information and description [22]. The proposal of XML schemas based standardization is based on the current Services Oriented Architectures (SOA’s) which provides loosely coupled services that enable cross domain information integration and querying. Web services decouple objects that are platform specific and facilitate interactions among platform independent objects, which are able to access data from anywhere on the Web. They rely on loose, rather than tight coupling among the web components which enables a flexible and dynamic interchange in open, distributed web environments. The Open Geospatial Consortium (OGC) provides a platform for government organizations, academia, and industry to come to a consensus and standardization of geospatial technologies for IIM.

Architecture of the proposed IIM standardization

b) spatio-temporal modelling of Harmful Algal Bloom events in Gulf of Mexico

In my dissertation research, I developed a Machine Learning based Spatio-Temporal (STML) data mining approach for the detection of Harmful Algal Blooms (HABs) events in the region of the Gulf of Mexico. HABs (also called Red Tide in Gulf of Mexico) pose an enormous threat to the U.S. marine habitation and economy in the coastal waters. Federal and state coastal administrators have been devising a state-of-the-art monitoring and forecasting system for these HAB events [1], [6-8], and [18]. The efficacy of a monitoring and forecasting system relies on the performance of HAB detection.

In this study, a “spatio-temporal cubical neighborhood” around the training sample is formulated to retrieve relevant spectral information of both HAB and non-HAB classes. The feature relevance is studied through mutual information criterion to understand the important features in classifying HABs from non-HABs. Kernel based support vector machine is used as a classifier in the detection of HABs. This STML approach gave a significant performance improvement by reducing the false alarm rate. Further, with the achieved classification accuracy, the seasonal variations and sequential occurrence of algal blooms are predicted from spatio-temporal datasets. A New variability visualization is introduced to illustrate the dynamic behavior of HABs across space and time.

Red Tide event in Gulf of Mexico

The Spatio-Temporal Evolution of HAB for 10 Days in early fall season

In further studies, we investigated the evolutionary feature subset selection using wrapper based genetic algorithms on Multi-temporal datasets [38]. Feature subset selection helps in reducing the original feature dimension and also yields high performance. The evolutionary strategy attains a global optimum by reducing the computations iteratively and by traversing intelligently in the entire feature space. This method gave a very high performance improvement up to 0.97 kappa accuracy with a best reduced feature dimension for harmful algal bloom detection (see Fig. 5).

Fig.5 shows the performance of the STML-HAB model using Genetic Algorithm wrapper based feature subset selection.

c) Ensemble Methodology:--

The available empirical remote sensing techniques for target recognition in coastal datasets are reliant on prior observations and thresholds. These techniques tend to give high false alarm rate, as they are limited in spatiotemporal contextual information and decision combination techniques. We propose a multistage learning based ensemble methodology addressing the above constraints for performance improvement of target detection [2], and [6]. Machine learning-based spatiotemporal data mining approach, along with empirical relationships, is used for target detection in the first stage of the ensemble, to exploit the potential benefits of each individual detection technique. The decision outputs from these detection techniques are fused in the second stage using nonlinear modeling-based combination techniques unlike conventional weighted averages. The proposed ensemble methodology outperforms all of the individual members and gave a significant overall performance improvement up to 0.8632 kappa accuracy. The performance is evaluated over tenfold cross validation average and compared against various ensemble methods and combination techniques.

The process diagram of ensemble methods which combines multiple classifiers decision with a Neural Network to produce high prediction performance.

d). Hyper-Spectral remote sensing datasets: - For Crop Stress Infestation Recognition

Rotylenchulus reniformis is a newly emerging nematode species affecting the cotton crop and quickly spreading throughout the southeastern United States. Effective use of nematicides at a variable rate is the only economic counter measure. It requires the nematode population in the field to be known, which in turn depends on the collection of soil samples from the field and analyzing them in the laboratory. This process is economically prohibitive.

Hence there is a need to develop alternative methods through which the actual numbers of reniform nematode present in the field can be determined. For the above study, we proposed a methodology in which a canopy reflectance model (PROSAIL) is inverted using machine learning approaches to retrieve the biophysical parameters, and relate the key variables to the nematode levels, so that it is possible to quantify the nematode infestation, at all multi-temporal intervals in geographically distributed fields [25], and [40]. A Support Vector Machine (SVM) Regression method is used for the inversion and retrieval of key biophysical parameters which help to understand and quantify the nature of the nematode infested vegetation. The performance of this approach is analyzed by the accuracy measures of root mean square error (RMSE) and N-fold cross validation average on a considerable data set. Finally, a graphical web portal is being developed for facilitating the end users, to use their field collected data to determine the extent of the nematode infestation in their crop and retrieve other spatiotemporal statistics.

Reniform Nematodes Infestation in Cotton Crop

Female Reniform Nematodes (pink) parasitizing a cotton root

e.). Digital system Design for vehicle collision warning

Here an Infrared Sensor is used to detect the distance of an object. This sensor takes a continuous distance reading and reports the distance as an analog voltage with a distance range of 4cm (~1.5") to 30cm (~12"). When the range decreases to the allowable distance between the two objects the buzzer will be alarmed with a high beep sound and the LEDs give warning by flickering. This project is titled Collision Warning System integrating all the following modules; Infrared sensor, A-D (8-bit) Converter, Spartan -3 board, external power supply, Piezo buzzer, and connecting wires [17].

Graduate Research Assistant

National Remote Sensing Center,

Indian Space Research Organization, INDIA

June. 2005 – Nov. 2006 (1.5 years )

During my tenure at National Remote Sensing Center, Department of Space, Govt. of INDIA, in 2005, my research and scholarly interests were stimulated in the fields of remote sensing (RS) and geospatial information system (GIS) applications. I worked as a research scholar for one and half year period in developing the boundary based segmentation approach for extraction of buildings and roads in urban areas from high resolution CARTOSAT (2.1m) and IKONOS (1m) satellite data [27] and [28]. My contribution was in the pioneering lines of using the CARTOSAT-1 satellite sensor data and validating its high spatial resolution capabilities for Urban Land-Use/Land-Cover (LU/LC), and urban sprawl applications. The pattern-recognition approach successfully delineated the man-made objects such as roads and buildings in an urban scene against the water bodies, waste land, barren land, vehicles, and agriculture lands in the urban satellite image scenes. The limitations in my work were shadows of the adjacent buildings and view-angle of the satellite sensor

Awards: - Best student paper presentation certificate with cash award for the below listed conference presentation.

Reference:- Gokaraju B, Jayanthi S.C, Neelakantan K, Raghavswamy V, Bapiraju S, and Nathawat M. S, “Automatic extraction of man-made objects from high-resolution satellite data in urban areas,” Proceedings of 25th Indian Society of Remote Sensing Annual Symposium, 6-9 DEC, 2005. ISRS, India.
Reference:- Gokaraju B, “Digital Watermarking Technique in protection of Map data or Rasterized data,” Proceedings of National Seminar on Role of Geoinformatics in Decentralized Planning for Better Governance. 24-26 OCT, 2004, Ranchi, India.

CARTOSAT-1a dataset of Mumbai study area with total number of buildings in the scenes 108

Automatic Detection Output of the study area with total number of buildings extracted =92

Controls Engineer

Indian Heavy Vessels and Plates Manufacturing Ltd. INDIA

June. 2003 – May. 2004 (1 year)

Maintained and wrote PLC programs, for the PLC based manufacturing lines of for heavy metal bending, cutting and lathes. Specialized in both Allen Bradley PLCs and SIEMENS PLCs.

Google Sites

Report abuse