2. Thesis and Dissertations

Now showing 1 - 2 of 2

Advanced Spectral Spatial Approaches for Dimensionality Reduction of Hyperspectral Data
(National Institute of Technology Karnataka, Surathkal, 2024) C, DEEPA; SHETTY, AMBA; NARASIMHADHAN, A.V.
Recent advances in sensor technology have enabled the collection of large data in hyperspectral remote sensing. Although rich spectral information is captured in hundreds of narrow contiguous bands, the hyperspectral data possess several limitations such as mixed pixels, high intraclass variability, interclass similarity, and the curse of dimensionality which restricts the potential of conventional machine learning classifiers. Dimensionality reduction (DR) and incorporation of spatial information can be taken into account to increase the interpretability of hyperspectral data. The thesis mainly focuses on the implementation of different approaches for DR of hyperspectral data to address the curse of dimensionality, limited samples and labelled data issues inherent in hyperspectral data. First, a quality measure based on the co-ranking matrix has been proposed for the performance evaluation of 15 DR techniques for mineral exploration. The selection of appropriate techniques for a particular task is challenging due to the diversity and ever-increasing number of DR techniques. A few important aspects in this regard have been explored in detail. Clustering is performed using the K-means algorithm and the relationship between the quality index and clustering accuracy has been examined concurrently for the first time in hyperspectral remote sensing. Furthermore, the loss of quality in the process of DR has also been analyzed which provides sufficient input for the end-user to select an appropriate DR technique. Second, the ability of the Convolutional Neural Network (CNN) for supervised learning of hyperspectral data is explored. A fast and compact hybrid CNN which combines the strengths of 3D and 2D convolutions to extract joint spectral-spatial information has been proposed to analyze the impact of different feature extraction techniques on classification performance. The effect of input patch size on final results has been well demonstrated. A detailed investigation of classification accuracy, execution time, and comparison with nine state-of-the-art approaches has been demonstrated. ii Next, a novel deep feature selection strategy using autoencoders inspired by knowledge distillation has been implemented for the model compression and selection of informative bands. The potential of convolutional autoencoders has been well explored in selecting discriminative bands. Sensitivity analysis tests and different applications have been considered to verify the generalization capability of the proposed model. The potential of unsupervised learning schemes has been discussed in detail. Finally, a generator model based on Generative Adversarial Networks (GAN) has been proposed for virtual sample generation and compact representation of hyperspectral data. The training instability issue in Vanilla GAN has been addressed by the effective implementation of deep convolutional GANs. By comparing the spectra of the generated hyperspectral images to the corresponding real ones, the quality of the images is assessed. The potential of augmented data for improvement in classification accuracy has also been investigated.
Gait Features-Based Human Recognition Approaches
(National Institute of Technology Karnataka, Surathkal, 2022) R, Anusha; C D, Jaidhar
Research into biometrics is an on-going and open research challenge to achieve robust human identification in a visual surveillance environment. Compared to other biometrics, gait has gained considerable attention in current years due to the unique benefits that other biometrics may not offer. Most significantly, it can be used with video feeds captured at a distance without alerting the subject and with low-resolution video. Interest in gait has increased appreciably because of the promising recognition results achieved from research in this area under controlled environments. Recent research is focused more on improving the recognition rate in realistic environments, where it is necessary to address the effects of changes in view, resolution, and fluctuation of gait patterns, due to carrying goods, walking speed variations, different footwear or clothes. The influence of various gait challenging conditions makes the real-world gait recog- nition system struggle for better performance. The development of appropriate features by using the information source available is the only solution to deal with these chal- lenges. In this work, solutions that can enhance the performance of a gait recognition system are proposed to assist security applications. One of the significant challenges of the appearance-based gait recognition system is to augment its performance by using a distinctive low-dimensional feature vector. Therefore, this study presents the low- dimensional feature vector that is capable of capturing the spatial, gradient, and texture information. These features are obtained by the computation of Histogram of Oriented Gradients (HOG), followed by the sum variance Haralick texture descriptor from nine cells of Gait Gradient Magnitude Image (GGMI). The improved recognition rate is achieved on the five publicly available gait datasets. The clothing variance is one of the most common covariate influences which can in- fluence the performance of the gait recognition approach in real-world scenarios. This study presents a gait recognition approach proficient in choosing information charac- teristics for individual identification under different clothing conditions. The proposed approach deals with the feature extraction technique by introducing a binary descriptor called Modified Local Optimal Oriented Pattern (MLOOP). Furthermore, the proposed approach is assessed on the OU-ISIR B and CASIA B gait datasets, and it achieves improvement in recognition performance over other binary descriptors. One of the difficulties of the appearance-based gait approach is to enhance the per- formance of frontal gait recognition, as it carries less spatial and temporal data when compared with other view variations. As a result, to increase the performance of the frontal gait recognition, this study presents a method that uses a two-step procedure; the Hierarchical Centroid Shape Descriptor (HCSD) and the similarity measurement. One more method is proposed, which uses the contour image and contour vertices to extract three discriminative feature vectors from the Gait Energy Image (GEI). Thus, it captures the spatial dynamics of frontal gait efficiently to improve gait recognition performance. These two methods are assessed on the broadly used CASIA A, CASIA B, and CMU MoBo gait datasets. The experimental outcomes show that the proposed methods yield the promising results and outperform certain state-of-the-art methods in terms of recognition accuracy. In this work, effective approaches are proposed to remove the effect of walking speed in a gait detection system. The first approach uses the Region of Interest (ROI) extracted from GEI to classify a probe sample into a gallery sample. The Mutual Infor- mation (MI) obtained from a probe and gallery sample, followed by their classification, efficiently improves the gait recognition performance. The proposed method shows an improved performance for two datasets when compared to other methods reported in this thesis. The next method identifies the most similar parts of the probe and each gallery sample independently and uses these parts to obtain a similarity/dissimilarity measure through three metrics. This method represents the spatial dynamics of GEI efficiently to improve gait recognition performance. Further, the proposed methods are evaluated on CASIA C and OU-ISIR A gait datasets. Experimental results demonstrate the capability of the proposed approaches in comparison with the existing gait recog- nition methods. This approach shows an increased performance for two datasets when compared to other methods reported in this thesis. Finally, the possibility of identifying individuals by using their running video is mostly unexplored. Hence, this study proposes a method that extends the feature-based approach to recognize people by the way they run. Here, the statistical, texture-based, and area-based features are extracted from each image of a gait cycle. The experiments are carried out on the KTH and Weizmann dataset. The several feature extraction algorithms proposed in this thesis are focused pre- dominantly on appearance-based methods because of their exceptional performance and simplicity. Overall, the aim of this research work is to increase the gait recognition sys- tem performance by contributing to areas such as dimensionality reduction of a feature vector, identification of an individual is attempted by using running patterns, to accom- plish frontal gait recognition, speed invariant gait recognition and clothing invariant gait recognition. The proposed solutions in this work contribute to improving gait recogni- tion performance in various practical scenarios that further enable the adoption of gait recognition into various applications.

2. Thesis and Dissertations

Browse

Filters

Settings

Sort By

Results per page

Search Results