• Analysis of Urban Heat Islands by Using Multi-Sensor and Multi-Temporal Remote Sensing Images

      Umamaheshwaran, Rajasekar (2011-09-20)
      This doctoral dissertation research has developed models to facilitate in characterization,analysis and monitoring of urban heat islands (UHI). Over the past few years there has been evidence of mass migration of the population towards urban areas which has led to the increase in the number of mega cities (cities with more than 10 million in population) around the world. According to the UN in 2007 around 60% (from 40% in 2000) of world populations was living in urban areas. This increase in population density in and around cities has lead to several problems related to environment such as air quality, water quality, development of Urban Heat Islands (UHI), etc. The purpose of this doctoral dissertation research was to develop a synergetic merger of remote sensing with advancements in data mining techniques to address modeling and monitoring of UHI in space and in time. The effect of urban heat islands in space and over time was analyzed within this research using exploratory and quantitative models. Visualization techniques including animation were experimented with developing a mechanism to view and understand the UHI over a city. Association rule mining models were implemented to analyze the relationship between remote sensing images and geographic information system (GIS) data. This model was implemented using three different remote sensing images i.e., Advanced Space borne Thermal Emission and Reflection Radiometer (ASTER), Landsat and Moderate Resolution Imaging Spectroradiometer (MODIS). The effect of the spatial resolution on the model and the phenomenon were analyzed in detail to determine variables which strongly associate with land use land cover (LULC) in space and in time.A non-parametric process convolution model was developed and was used to characterize UHI from MODIS time series images. The resulting characterized images were used to study the relationship between LULC and UHI. The behavior of UHI including its movement and magnitude was analyzed in space and time.The intellectual merits of these methods are two-fold; first, they will be a forerunner in the development and implementation of association rule mining algorithm within remote sensing image analysis framework. Second, since most of the existing UHI models are parametric in nature; the non-parametric approach is expected to overcome the existing problems within characterization and analysis. Parametric models pose problems (in terms of efficiency, since the implementation of such models are time consuming and need human intervention) while analyzing UHI effect from multiple imageries. These proposed models are expected to aid in effective spatial characterization and facilitate in temporal analysis and monitoring of UHI phenomenon.
    • Classification of Urban features using Airborne Hyperspectral Data

      Babu, Bharath Ganesh (2010-05-11)
      Accurate mapping and modeling of urban environments are critical for their efficient and successful management. Superior understanding of complex urban environments is made possible by using modern geospatial technologies. This research focuses on thematic classification of urban land use and land cover (LULC) using 248 bands of 2.0 meter resolution hyperspectral data acquired from an airborne imaging spectrometer (AISA+) on 24th July 2006 in and near Terre Haute, Indiana. Three distinct study areas including two commercial classes, two residential classes, and two urban parks/recreational classes were selected for classification and analysis. Four commonly used classification methods – maximum likelihood (ML), extraction and classification of homogeneous objects (ECHO), spectral angle mapper (SAM), and iterative self organizing data analysis (ISODATA) - were applied to each data set. Accuracy assessment was conducted and overall accuracies were compared between the twenty four resulting thematic maps. With the exception of SAM and ISODATA in a complex commercial area, all methods employed classified the designated urban features with more than 80% accuracy. The thematic classification from ECHO showed the best agreement with ground reference samples. The residential area with relatively homogeneous composition was classified consistently with highest accuracy by all four of the classification methods used. The average accuracy amongst the classifiers was 93.60% for this area. When individually observed, the complex recreational area (Deming Park) was classified with the highest accuracy by ECHO, with an accuracy of 96.80% and 96.10% Kappa. The average accuracy amongst all the classifiers was 92.07%. The commercial area with relatively high complexity was classified with the least accuracy by all classifiers. The lowest accuracy was achieved by SAM at 63.90% with 59.20% Kappa. This was also the lowest accuracy in the entire analysis. This study demonstrates the potential for using the visible and near infrared (VNIR) bands from AISA+ hyperspectral data in urban LULC classification. Based on their performance, the need for further research using ECHO and SAM is underscored. The importance incorporating imaging spectrometer data in high resolution urban feature mapping is emphasized.
    • Neural Network Classification of Hyperspectral Imagery for Urban Environments: a Case Study

      Lulla, Vijay (2010-09-22)
      Urban environments are complex because many different artificial and natural objects occur in close proximity. Being able to understand the processes and workings of these environments requires the ability to observe and record data with high spatial and spectral resolution. Hyperspectral sensors have been gaining popularity for this task as they are becoming more affordable. In this research, a commonly used maximum likelihood (ML) classifier and artificial neural network (ANN) classifier have been compared for classifying urban land use and land cover (LULC) using AISA+ hyperspectral data. Further, the best set of bands were identified for classification of urban areas for use in ANN classification. Optimum bands based on a spectral separability measure were used with a neural network classifier to compare its performance with maximum likelihood classifier. It was found that both the classifiers had an overall classification accuracy of more than 80% and the neural network classifier with optimum band selection performed better in all of the study sites.