Theses and Dissertations from UMD
Permanent URI for this communityhttp://hdl.handle.net/1903/2
New submissions to the thesis/dissertation collections are added automatically as they are received from the Graduate School. Currently, the Graduate School deposits all theses and dissertations from a given semester after the official graduation date. This means that there may be up to a 4 month delay in the appearance of a give thesis/dissertation in DRUM
More information is available at Theses and Dissertations at University of Maryland Libraries.
Browse
3 results
Search Results
Item Integrated use of Landsat and Corona data for long-term monitoring of forest cover change and improved representation of its patch size distribution(2016) Song, Danxia; Townshend, John R; Huang, Chengquan; Geography; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Forest cover change has profound impact on global carbon cycle, hydrological processes, energy balance, and biodiversity. The primary goal of this dissertation is to improve forest cover change characterization by filling a number of knowledge gaps in forest change studies. These include use of Corona data to extend satellite based forest cover change mapping back to pre-Landsat years in the 1960s, quantification of forest cover change over four decades (1960s – 2005) for a major forested province in China using Corona and Landsat data, and development of more accurate patch size-frequency modeling methods for improved representation of forest disturbances in ecosystem and other spatially explicit models. With comprehensive data coverages in the 1960s, Corona data can be used to extend Landsat-based forest change analysis by up to a decade. The usefulness of such data, however, is hindered by poor geolocation accuracy and lack of multi-spectral bands. In this study, it was demonstrated that combined use of texture features and the advanced support vector machines allowed forest mapping with accuracies of up to 95% using Corona data. Further, a semi-automated method was developed for rapid registration of Corona images with residual errors as low as 100 m. These methods were used to assess the forest cover in the 1960s in Sichuan, a major forest province in China. Together with global forest cover change products derived using Landsat data, these results revealed that the forest cover in Sichuan Province was reduced from 45.19% in the 1960s to 38.98% by 1975 and further down to 28.91% by 1990. It then stayed relatively stable between 1990 and 2005, which contradicted trends reported by inventory data. The turning point between sharp decreases before 1990 and the stable period after 1990 likely reflected transitions in forest policies from focuses on timber production to forest conservation. Representation of forest disturbances in spatially explicit ecosystem models typically relies on patch size-frequency models to allocate an appropriate amount of disturbances to each patch size level. Existing patch size-frequency models, however, do not provide accurate representation of the total disturbance area nor the patch sizes at each frequency level. In this study, a hierarchical method was developed for modeling patch size-frequency distribution. Evaluation of this method over China revealed that it greatly improved the accuracy in representing the patch size at different frequency levels and reduced error in total disturbance area estimation over existing methods from around 40% to less than 10%. The significance of this dissertation is the contribution to improve the characterization of forest cover change by extending the satellite-based forest cover change monitoring back to the 1960s and developing a more accurate patch size distribution model to represent the forest disturbance in ecosystem models. The work in the dissertation has a broader impact beyond developing methods and models, as they provide essential basis to understand the relationship between the long-term change of forest and the socioeconomic transitions. They also improve the capacities of ecosystem and other spatially explicit models to simulate the vegetation dynamics and the resultant biodiversity and carbon dynamics.Item Temporal Tracking Urban Areas using Google Street View(2016) Najafizadeh, Ladan; Froehlich, Jon E; Computer Science; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Tracking the evolution of built environments is a challenging problem in computer vision due to the intrinsic complexity of urban scenes, as well as the dearth of temporal visual information from urban areas. Emerging technologies such as street view cars, provide massive amounts of high quality imagery data of urban environments at street-level (e.g., sidewalks, buildings, and aesthetics of streets). Such datasets are consistent with respect to space and time; hence, they could be a potential source for exploring the temporal changes transpiring in built environments. However, using street view images to detect temporal changes in urban scenes induces new challenges such as variation in illumination, camera pose, and appearance/disappearance of objects. In this thesis, we leverage Google Street View’s new feature, “time machine”, to track and label the temporal changes of built environments, specifically accessibility features (e.g., existence of curb-ramps, condition of sidewalks). The main contributions of this thesis are: (i) initial proof-of-concept automated method for tracking accessibility features through panorama images across time, (ii) a framework for processing and analyzing time series panoramas at scale, and (iii) a geo-temporal dataset including different types of accessibility features for the task of detection.Item Change Detection in Stochastic Shape Dynamical Models with Applications in Activity Modeling and Abnormality Detection(2004-08-04) Vaswani, Namrata; Chellappa, Rama; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)The goal of this research is to model an ``activity" performed by a group of moving and interacting objects (which can be people or cars or robots or different rigid components of the human body) and use these models for abnormal activity detection, tracking and segmentation. Previous approaches to modeling group activity include co-occurrence statistics (individual and joint histograms) and Dynamic Bayesian Networks, neither of which is applicable when the number of interacting objects is large. We treat the objects as point objects (referred to as ``landmarks'') and propose to model their changing configuration as a moving and deforming ``shape" using ideas from Kendall's shape theory for discrete landmarks. A continuous state HMM is defined for landmark shape dynamics in an ``activity". The configuration of landmarks at a given time forms the observation vector and the corresponding shape and scaled Euclidean motion parameters form the hidden state vector. The dynamical model for shape is a linear Gauss-Markov model on shape ``velocity". The ``shape velocity" at a point on the shape manifold is defined in the tangent space to the manifold at that point. Particle filters are used to track the HMM, i.e. estimate the hidden state given observations. An abnormal activity is defined as a change in the shape activity model, which could be slow or drastic and whose parameters are unknown. Drastic changes can be easily detected using the increase in tracking error or the negative log of the likelihood of current observation given past (OL). But slow changes usually get missed. We have proposed a statistic for slow change detection called ELL (which is the Expectation of negative Log Likelihood of state given past observations) and shown analytically and experimentally the complementary behavior of ELL and OL for slow and drastic changes. We have established the stability (monotonic decrease) of the errors in approximating the ELL for changed observations using a particle filter that is optimal for the unchanged system. Asymptotic stability is shown under stronger assumptions. Finally, it is shown that the upper bound on ELL error is an increasing function of the ``rate of change" with increasing derivatives of all orders, and its implications are discussed. Another contribution of the thesis is a linear subspace algorithm for pattern classification, which we call Principal Components' Null Space Analysis (PCNSA). PCNSA was motivated by Principal Components' Analysis (PCA) and it approximates the optimal Bayes classifier for Gaussian distributions with unequal covariance matrices. We have derived classification error probability expressions for PCNSA and compared its performance with that of subspace Linear Discriminant Analysis (LDA) both analytically and experimentally. Applications to abnormal activity detection, human action retrieval, object/face recognition are discussed.% with experimental results.