Research projects
The Geospatial Technology Laboratory(GTL) remains frontier in various research domain like Agricultural, Drought, IoT, Smart city, Mineral and soil. The expertise and research scholar has developed a unique methodology and advanced analytics which is based on recent development in remote sensing and GIS domain.
[toggle title=”I. Agricultural Drought Assessment System and Tool Based on Geospatial Data” state=”open”]
Traditionally, the climatic observation using weather stations were the primary input for the drought monitoring and assessment system. The conventional methods for agricultural drought monitoring and assessment is not effective due to change in climate and agricultural ecosystem. The advancement in satellite remote sensing technology provides an opportunity to assess the impact of drought on the agricultural and ecosystem. Moreover, satellite-based vegetation indices along with ground observation are reliable to measure agri-ecosystem response to changing the weather pattern. The Internet of Things (IoT) based agro-metrological and Soil moisture system has developed for evaluation and calibration of drought modelling. Moreover, smartphone application is also developed to collect the ancillary data such as crop attributes, Ground Truth (GT) points and agricultural attributes. The satellite-based climatic parameter along with ground observation is reliable to measure agri-ecosystem response to climate variability. Total 5 villages of Vaijapur tehsil such as Ghaigaon, Jambargaon, Kolhi, Janephal, Khandala has selected for the pilot study. In the current research study, the case study has completed which is based on identification and classification of drought severity using Hyperspectral EO-1 Hyperion data along with ancillary data. The developed drought assessment system will be used to generate a thematic map of drought severity, crop health, soil moisture. Additionally, the system will be helpful in predicting the early warning of droughts which will be used by the decision makers for planning, monitoring and management of the drought.
Future scope of Research
Our focus is on development of multivariate drought index, automatic drought prone area mapping system and Early warning drought prediction system.
[/toggle]
[toggle title=”II. Design and Development of Soil Analysis System using Geospatial technology” state=”close”]
An accurate and reliable detection of soil properties is a difficult and complicated issue in soil science. The soil properties may be varied by spatially and temporally with the complexity of nature. In the past, soil properties detection has been obtained through routine soil physicochemical laboratory analysis. However, these laboratory methods do not fulfil the rapid requirements. On the other hand, the digital assessment of surface soil types, spatial distribution and its mapping is a somewhat formidable task due to various soil attributes and assorted effect of various features of planet surface that can affect spectral and spatial features of soils. The soil classification is essential for farming practices to grow the food production for future. The present study highlights the use of high spectral and spatial resolution remote sensing datasets to identify and classify the surface soil with its condition. Accordingly, DRS can be used to nondestructively detect and characterize soil materials with a superior solution. In the present research, we report study done through spectral curves in Visible (350-700nm) and Near-Infrared (700-2500nm) region of seventy-four soil specimens which were agglomerated by farming sectors of Phulambri Tehsil of Aurangabad region of Maharashtra, India. The quantitative analysis of VNIR spectrum was done. The spectra of agglomerated farming soils were acquired by the ASD Field spec 4 spectroradiometer. The soil spectra of VNIR region were preprocessed to get pure spectra which were the input for regression modelling. The PLSR model was developed to construct the calibration models, which were individually validated for the prediction of soil properties from the soil spectrum. The developed model was based on a correlation study between reflected spectra and detected soil properties. The detected soil properties were soil organic carbon, Nitrogen, soil organic matter, pH values, electrical conductivity, phosphorus, potassium, iron, sand, silt and clay. The results are significant for soil analysis and its mapping of the complex region. The outcome of the present research will be apt for precision farming and decision making.
Future Scope of the Research
In the future scope, more soil samples will be considered for developing better predictive models and developing the spectral information of surface soil types which will be used as reference (endmember) spectra. The fusion approach of high spectral and spatial resolution remote sensing datasets will be carried out to overcome the limitations of using single datasets only. The detected soil properties will be quantitatively mapped through satellite or airborne hyperspectral imagery for the large area.
[/toggle]
[toggle title=”II. Design and Development of Agricultural Geospatial Information for Decision Making” state=”close”]
The identification of agricultural crops is a tedious task using conventional methods which is essential in condition and growth monitoring of the crop. The objectives of the ongoing project were to develop and evaluate spectral signatures of crops to improve nutritional diagnosis of the crop. The present research study describes the enlargement of crop spectral signature using multiband frequencies for crop discrimination. Additionally, crop identification based on spectral signature, botanical pigments extraction methods and morphological parameters estimation for various applications. The proposed methodology can be used for identified spectral signatures of crops will be used as an end member extraction for hyperspectral imagery. Amalgamation of diverse methods greatly improves the accuracy of targeted yield acreage to increase the crop productivity. The outcome of this research study is major inputs for crop insurance policy schemes like Pradhan Mantri Fasal Bhima Yojana (PFBY). Furthermore, the research will be benefitted to the agricultural researcher, food policy makers, students, government and non-government organization. The present invention provides a facile method for identification and discrimination of crop types with an inexpensive manner using geospatial data.
Future Scope of the Research
We envision my future research to span across some interrelated sub-areas of hyperspectral remote sensing. The unifying theme of the research will be the development of spectral signatures of crops using non-imaging spectroradiometer and imaging Pika-L for precision farming and agriculture monitoring.
[/toggle]
[toggle title=”IV. Identification of Medicinal Plant Species using Hyperspectral Spectroradiometer” state=”close”]
Very few efforts have been made around the world for designing such medicinal plants identification system. Although the significant amount of research work has been done by studying various aspects of plant semi-automated systems, time-consuming and requires more chemical extraction procedure. A state-of-the-art system which is fully automatic, non-chemical content and requires least human interaction is yet to be developed. The proposed system is based on non-imaging and imagining hyperspectral remote sensing for identification and classification of medicinal plant and its ingredient. The database of plant species was collected from the university campus area of Dr Babasaheb Ambedkar Marathwada University Aurangabad, Maharashtra, India. The variations of leaf chlorophyll content, Chl a, Chl b and Carotenoid along with Xanthophyll, Anthocyanin were calculated of healthy leaves of plant based on spectral indices. The linear regression models were developed for the calculation of correlations between spectral indices and pigment contents using the developed tool. In the present study, the water content in four types of plants species (Synygium Cumini, Azadirachta Indica, Adhatola Zeylanica, and Aloe Vera) have been measured by using non-imagining ASD Field Spec4 spectrometer, whereas the 150 SM soil moisture sensor has used for the moisture detector. The EO-1 Hyperion hyperspectral satellite images of 10nm spectral resolution of the study area were used for analysis. The outcome of the research study, the spectral indices and designed regression model has given an accurate result for estimating chemical properties and biophysical properties of medicinal plants. The research study will be benefitted from drug discovery, pharmaceutical industries, forest department for identification of medicinal plant powder, researchers, and students.
Future scope of the research
We are developing a fully automatic system based on the spectral signature of the medicinal plant which able to identify the adulteration in the powder.
[/toggle]
[toggle title=”V. Land Use Classification of Urban Areas From Hyperspectral Data Based on Spectral and Spatial Feature Extraction” state=”close”]
India is urbanizing, therefore, cities and towns are at the Centre of India’s development trajectory. In the coming decades, the urban sector will play a critical role in the structural transformation of the Indian economy. Because of dynamic urban development and high mapping costs, municipal authorities are interested in effective urban surface mapping that can be used for development of Smart City. Determination and Identification of Land-Use-Land-Cover (LULC) of urban area have become a very challenging issue in planning a city development. Although urban area represents a little portion of Earth’s surface area, it brings an unbalanced impact on its surrounding areas. An urban region is a complex ecosystem and made up of a heterogeneous material. In worldwide an urban expansion is done by occupying a cultivated land which causes a serious problem in our ecosystem. Now there is a need to identify an urban area and its growth pattern. A recent Hyperspectral Remote sensing technology can be used to monitor built-up areas and also it can detect the growth and spatial distribution of urban built-up. In this study, both Hyperspectral and Multispectral data were used for LULC identification of Aurangabad Municipal Corporation area. First LISS-III image data set of February 2015 and January 2009, obtained from NRSC Hyderabad, India, for the region of Aurangabad city (India) has been used. We have applied four classification techniques by assuming six types of objects in the extracted the images of Aurangabad City. The present work highlights the Geospatial Technology for classification, identification of LULC using multispectral and Hyperspectral datasets. Specifically, it introduces novel techniques for remote sensing application for LULC Change Detection. The research outcome has been carried out as per revealed literature and it has accepted by the scientific community which works on this type of study. The developed methodological model can be used for spectral and spatial feature extraction of urban features with better results. The land use land cover information of Twelve years and sixteen different Anderson’s LULC Level 2 classes will be helpful for decision makers
Future Scope of the Research
In the future, this work will be elaborate using very high spatial and spectral Drone-based Hyperspectral images and LIDAR to identify the minute features of urban sectors. Using these high spatial and spectral resolutions Hyperspectral images a different urban surface material can be identified.
[/toggle]
[toggle title=”VI. Hyperspectral data analysis for endmember extraction using machine and deep learning approach” state=”close”]
The advancements in hyperspectral remote sensing are increasing continuously and recording a wealth of spatial as well as spectral information about an object, but resulting the high volume of data. Analysis of hyperspectral data needs a ground truth data or spectral library or image-based endmembers which assist to unmix the mixed pixels and map their spatial distribution. Till date, though several hyperspectral endmember extraction algorithms have been proposed (PPI, NFINDR, VCA, CCA, FIPPI, SGA and ATGP), every algorithm has its own limitations. The perfect endmember extraction algorithm would find unique spectral signatures directly from the image data without any prior knowledge. Automatic identification of endmembers is a crucial task in hyperspectral data exploitation because they may appear as anomalies since their population is relatively small. In most of the cases, sub-scene represented by each observed pixel is not homogenous. As a result, the hyperspectral signature collected by the sensor at each pixel is formed by an integration of signatures, associated with the purest portion of sub-scene. Correctly extracted pure spectral signatures or endmembers from hyperspectral images will significantly improve the supervised classification, mixture modelling and spectral unmixing in hyperspectral images Proposed work aims to extend the machine and deep learning capabilities to identify the hyperspectral endmembers without any prior knowledge about the scene automatically. The outcome of the proposed research will be the spectral unique signatures (i.e. endmembers) in the form of spectral signatures, spatial coordinates of the corresponding pixels and asci files containing the endmember information (i.e. wavelength vs reflectance) for the scene under observation. The outcomes will be used as image derived ground truth information for supervised classification and Spectral Unmixing.
Future Scope of the Research
Once, the precise and accurate hyperspectral endmembers are extracted automatically, the further work can be extended to improve the per-pixel as well as within-pixel supervised classification along with spectral unmixing.
[/toggle]
[toggle title=”VII. Dimensionality Reduction for Remote Sensing Data Using Advance Linear Approach” state=”close”]
Hyperspectral satellites collect imagery in hundreds of continuous spectral bands simultaneously, covering wavelengths from the near-UV to the short-wave-IR regions. This continuous nature of acquisition produces a very high volume of data. This data can provide direct identification of surface materials and are used in a wide variety of remote-sensing applications. Most of the spectral channels of the hyperspectral images are characterized by redundant information. The processing and analysis of these images with original dimensionality results into higher storage and computational burden. There are some techniques evolved to solve this issue namely PCA, MNF and ICA. These techniques are dependent on the user-defined parameters and give inconsistent output varying with input parameters. Therefore there is a need to automatize dimensionality reduction issue using hybrid mechanism. The outcome of the proposed research work will reduce the redundancy in hyperspectral data and preserve the significant information. It will be characterized with the automatic nature of dimensionality reduction
Future Scope of the Research
Dimensionality reduced data can be further used to improve the accuracy of hyperspectral endmember extraction, supervised and unsupervised classification with lower time and space complexity.
[/toggle]