http://hdl.handle.net/10204/3826 2013-05-19T15:19:56Z http://hdl.handle.net/10204/6727 Title: Relevance of new multispectral imagery for assessing tropical forest disturbance: RapidEye and WorldView-2 Authors: Cho, MA; Ramoelo, AI; Mutanga, O; Van Deventer, H; Debba, P; Mathieu, R Abstract: The aims of this study were to assess utility of RapidEye imagery for predicting leaf nitrogen concentration and evaluate the effects of forest fragmentation on leaf nitrogen distribution in the Dukuduku forest, KwaZulu Natal, South Africa. RapidEye and WorldView-2 images were acquired for the study area. Leaf nitrogen concentration was accurately (R2 = 0.52, p < 0.05) estimated using the MERIS terrestrial vegetation index (MTCI) derived from the RapidEye image. Land cover types were accurately classified (overall accuracy = 85%) using WorldView-2 imagery. Differences in leaf nitrogen concentration between land cover types were then analysed. Remnant forest patches showed higher leaf nitrogen than grassland patches in the degraded landscape. In conclusion, foliar nitrogen can be mapped at peak productivity using RapidEye sensor. Forest fragmentation significantly affects leaf nitrogen concentration. Description: 9th International Conference of the African Association of Remote Sensing and the Environment, El Jadida, Morocco, 29 October-2 November 2012 2012-11-01T00:00:00Z http://hdl.handle.net/10204/6715 Title: Comparing parametric and non-parametric classifiers for remote sensing of tree species across a land use gradient in a Savanna landscape Authors: Cho, MA; Naidoo, L; Mathieu, R; Asner, GP; Ramoelo, A Abstract: Several classification techniques have been used to map vegetation communities or land cover types using remote sensing data including maximum likelihood (ML), discriminant analysis and spectral angle mapper classifiers. ML classifier is a commonly used supervised classification method with conventional multispectral data that considers both first order variations (e.g. mean values) and second order variations (e.g. covariance matrices). However, there is a limitation with the application of the ML classifier in situations of high within-species variability. The objective of this study is to ascertain which classification techniques are suitable for classification of savanna tree species across a land-use gradient. Eight savanna tree species were classified for two sites in the vicinity of the Kruger National Park, South Africa using two parametric (ML and Mahalanobis distance classifiers) and three non-parametric classifiers (spectral angle mapper (SAM), artificial neural networks (ANN) and Random Forest (RF)). The spectral data used consisted of 8 WorldView-2 multispectral bands simulated from 72 VNIR bands image acquired over the study areas using the Carnegie Airborne Observatory (CAO) system. With the exception of SAM, the nonparametric classifiers provided higher classification accuracies (RF = 78%, ANN = 75%, SAM = 36%) when compared to the parametric classifiers (ML = 65%, Mahalanobis distance = 68). This study moves remote sensing closer towards classification of savanna tree species over large areas. Description: 9th International Conference of the African Association of Remote Sensing and the Environment, El Jadida Morocco, October 29 to November 2, 2012 2012-11-01T00:00:00Z http://hdl.handle.net/10204/6693 Title: Fingerprint pores extractor Authors: Mngenge, NA; Nelufule, NN; Nelwamondo, FV; Msimang, M Abstract: Automatic Fingerprint Recognition Systems (AFRSs) rely on minutiae position and orientation within the fingerprint image for matching. Minutiae information is highly accurate provided that the fingerprint image matched is of high quality. However, this is not always the case because of diseases and hash working conditions that affect fingerprints. In order to maintain high level of security independent of varying fingerprint image quality research suggests the use of other fingerprint features to compliment minutiae. These are things like ridge contours, sweat pores, dots, and incipient ridges. Sweat pores have been proven as one of the most distinctive among these feature. Thus in order to improve accuracy of AFRSs pores can be fused with minutiae or used alone. Sweat pores have been less utilized in the past due to constraints imposed by fingerprint scanning devices and resolution standards. Recently, progress has been made on both scanning devices and resolution standards to support the use of pores in AFRSs. However, very few techniques exist for extracting, matching and fusing them with minutiae. Matching and fusion can only be possible if pores are available. Some techniques have been proposed to reliable extract pores. However, existing techniques can only work on one resolution i.e. an algorithm proposed and tested on 500dpi cannot work on 1000dpi without minor modifications because pores size change if resolution changes. In addition, existing pore extraction techniques are computationally expensive. In this paper an algorithm to extract feature level 3 (pores) is proposed. The algorithm uses Laplacian of Gaussian (LoG) in Fourier domain in order to reduce computation. The performance of the proposed algorithm is tested on two distinct databases with different resolutions in order to validate its accuracy. The accuracy of the proposed algorithm is further measured using false detection rate (FDR) and true detection rate (TDR). Results show that FDR ranges from 10-35% while TDR ranges from 65-90%. Description: 2012 National Conference on Computing and Communication Systems, Durgapur, West Bengal, India, 21- 22 November 2012. To be published in IEEE Xplore 2012-11-01T00:00:00Z http://hdl.handle.net/10204/6655 Title: Spherical stochastic neighbor embedding of hyperspectral data Authors: Lunga, D; Ersoy, O Abstract: In hyperspectral imagery, low-dimensional representations are sought in order to explain well the nonlinear characteristics that are hidden in high-dimensional spectral channels. While many algorithms have been proposed for dimension reduction and manifold learning in Euclidean spaces, very few attempts have focused on non-Euclidean spaces. Here, we propose a novel approach that embeds hyperspectral data, transformed into bilateral probability similarities, onto a nonlinear unit norm coordinate system. By seeking a unit l2-norm nonlinear manifold, we encode similarity representations onto a space in which important regularities in data are easily captured. In its general application, the technique addresses problems related to dimension reduction and visualization of hyperspectral images. Unlike methods such as multidimensional scaling and spherical embeddings, which are based on the notion of pairwise distance computations, our approach is based on a stochastic objective function of spherical coordinates. This allows the use of an Exit probability distribution to discover the nonlinear characteristics that are inherent in hyperspectral data. In addition, the method directly learns the probability distribution over neighboring pixel maps while computing for the optimal embedding coordinates. As part of evaluation, classification experiments were conducted on the manifold spaces for hyperspectral data acquired by multiple sensors at various spatial resolutions over different types of land cover. Various visualization and classification comparisons to five existing techniques demonstrated the strength of the proposed approach while its algorithmic nature is guaranteed to converge to meaningful factors underlying the data. Description: Copyright: 2012 IEEE Xplore. This is an ABSTRACT ONLY. The definitive version is published in IEEE Transactions on Geoscience and Remote Sensing, vol. 51(2), pp 857- 871 2012-07-01T00:00:00Z