This thesis work is an exploration into the application of machine learning to a cur- rent problem relating to streams of aerial images, with application to the imaging systems within the Texas A&M GEOSAT group an...
详细信息
This thesis work is an exploration into the application of machine learning to a cur- rent problem relating to streams of aerial images, with application to the imaging systems within the Texas A&M GEOSAT group and to the Texas A&M AgriLife Extension Ser- vice. The streams of images referred to are taken by a unmanned aerial vehicle (UA v) over some space of land. Our problem relates to flight inconsistencies in the UA v. In an ideal scenario, our UA v captures a set of images from a specified height, pointed normally at the ground, with a predetermined amount of overlap between images. However, due to inconsistencies inherent in most flights, the UA v will occasionally tilt or swing in such a way that the images captured are not in line with adjacent images, i. e., they do not have the required amount of overlap, and their information may be blurred. This creates a problem when attempting image stitching afterwards, as such anomalous images will be irreconcilable with adjacent images. A prime goal then consists of designing a system which can, in pseudo-real-time, detect images which are out of line, through the use of feature extraction and machine learning, so that they can be discarded and possibly recaptured. Our research compares and evaluates a number of image features and distance metrics, along with supervised classification learning algorithms, on the basis of their performance in this task. After feature evaluation, pixel intensity distribution, binary robust independent elementary features (BRIEF), and blob detection were selected as image features. The classification models chosen include logistic regression, neural networks, decision trees, and support vector machines (SvM). After training and testing, the logistic regression model yields the highest performance, with a detection rate of 95.6% and a false alarm rate of 7.7%, making this a viable option for this task. Neural networks and SvM yield lower levels of performance in detection and false alarm
The generalized negative binomial distribution (GNB) is a new exible family of dis- crete distributions that are mixed Poisson laws with the mixing generalized gamma (GG) distributions. This family of discrete distrib...
详细信息
Purpose of Review: This review covers the basic principles of machine learning (ML) and current applications in the subspecialty of cardiac imaging at computed tomography in diagnostic radiology. Recent Findings: This...
详细信息
The 2D non-separable linear canonical transform (2D-NS-LCT) can model a range of various paraxial optical systems. Digital algorithms to evaluate the 2D-NS-LCTs are important in modeling the light field propagations a...
详细信息
ISBN:
(数字)9781510609686
ISBN:
(纸本)9781510609679;9781510609686
The 2D non-separable linear canonical transform (2D-NS-LCT) can model a range of various paraxial optical systems. Digital algorithms to evaluate the 2D-NS-LCTs are important in modeling the light field propagations and also of interest in many digital signal processing applications. In [Zhao 14] we have reported that a given 2D input image with rectangular shape/boundary, in general, results in a parallelogram output sampling grid (generally in an affine coordinates rather than in a Cartesian coordinates) thus limiting the further calculations, e.g. inverse transform. One possible solution is to use the interpolation techniques;however, it reduces the speed and accuracy of the numerical approximations. To alleviate this problem, in this paper, some constraints are derived under which the output samples are located in the Cartesian coordinates. Therefore, no interpolation operation is required and thus the calculation error can be significantly eliminated.
This work focuses on estimating an accurate 3D transformation in real time, which is used to register images acquired from different viewpoints. The main challenges are significant image appearance differences, which ...
详细信息
The paper proposes a modification of the pyramid method for constructing algorithms for the difference solution of the d'Alembert equation on a graphics processor in the event of a shortage of video memory. The au...
详细信息
The paper proposes a modification of the pyramid method for constructing algorithms for the difference solution of the d'Alembert equation on a graphics processor in the event of a shortage of video memory. The authors demonstrate the effectiveness of the method on the practical example of dividing the grid area into two sub domains. Acceleration reaches the characteristic for the case of a domain entirely located in the video memory. In the article investigated the effectiveness of using the author's approach depending on the height of the pyramid and showed the boundaries of applicability of the proposed modification.
The goal of this paper is to develop a fast algorithm for local peak filtering of two-dimensional arrays. The ambiguity of the concept of a local peak is demonstrated and additional conditions introduced that resolve ...
详细信息
The goal of this paper is to develop a fast algorithm for local peak filtering of two-dimensional arrays. The ambiguity of the concept of a local peak is demonstrated and additional conditions introduced that resolve it. A correct peak filter that takes into account the above conditions is developed. To evaluate effectiveness of the proposed algorithm two known algorithms for finding local maxima are described. Estimates of the computational complexity of algorithms for the best and worst cases are given. Analysis of dependency of the algorithm execution time from image size, sliding window size and a number of local maxima, is made. The results of experimental research showed that performance of the correct peak filter is higher than its incorrect counterparts.
Remote visual inspection using video endoscopes equipped with stereoscopic prism-based optical systems requires operative and robust evaluation of measurement uncertainty. Known solutions are based on significant assu...
详细信息
Multichannel imageprocessing, in particular, supervised classification, requires designing novel time effective algorithms, because in the most of the cases slightly dimension increase leads to significant processing...
详细信息
Multichannel imageprocessing, in particular, supervised classification, requires designing novel time effective algorithms, because in the most of the cases slightly dimension increase leads to significant processing time growth. In this paper we describe our supervised multichannel image classification algorithm based on a hierarchical representation of multivariate histograms. The algorithm estimates the joint sample set distribution, the particular distributions of each class and the decision rule by means of specific data structure called histogram-tree. Proposed algorithm provides faster learning and classification of multidimensional input data. The experimental evaluation of the algorithm has been conducted for the hyperspectral remote sensing images. The results demonstrate that proposed algorithm is faster than the commonly used C4.5 classifier.
The study of substances with a crystal structure is a complex multi-step process. The key step in the crystalline substance analysis is the unit cell parameter estimation. The estimation of the crystal lattice unit ce...
The study of substances with a crystal structure is a complex multi-step process. The key step in the crystalline substance analysis is the unit cell parameter estimation. The estimation of the crystal lattice unit cell parameters is a particular problem that involves the search of the crystal lattice model's parameters according to the information which can be extracted from the substance. In these recent times, the most accurate information about the substance structure can be obtained with the electron microscope whose linear resolution is high enough to observe the atomic structure of a substance. The problem of parameter estimation in this case means the reconstruction of the three-dimensional crystal lattice with 2-dimentional images received by an electron microscope, and the estimation of the crystal lattice unit cell parameters by reconstructed lattice. In the previous papers the crystal lattice parametric identification algorithms based on solving the local optimization problem were presented. However, the analysis of a large crystal lattice database requires a lot of computations. In this paper, a high-performance crystal lattices parametric identification algorithm using the CUDA technology is proposed. The investigation of the algorithm effectiveness is carried out on the GPU GeForce Nvidia GTX 1070 Ti. With data dimension more than 32 translations the acceleration is higher than 70. The algorithm runs more efficiently at the use of a large number of CUDA-blocks.
暂无评论