Thesis - Open Access
Master of Science (MS)
In the fast growing field of remote sensing acquiring information through the use of cameras and related devices, such as radar and thermal infrared sensors, operated from aircraft and spacecraft - several useful developments are taking place. Several of these developments are centered around the remote sensing vehicle known as ERTS (the Earth Resources Technology Satellite), which is currently orbiting and photographing the earth at an altitude of approximately 570 miles. Experience has shown that even an expert, well versed in remote sensing, usually cannot derive adequate information from merely a single “frame” of remote sensing imagery covering the area that he is interested in. Various types of remote sensing research, however, have recently demonstrated that the ease, accuracy, and completeness with which information can be derived is likely to be greatly improved through what might be termed the “multi” approach to remote sensing, in which several frames of imagery, all covering the same general geographic area, are variously enhanced and analyzed. This concept is applicable to several different aspects of remote sensing. The goal of this project was to study in detail the mode seeking algorithm and its application in pattern recognition. The algorithm was first applied to a set of computer generated data and based on that certain observations were made. These are discussed in Chapter 3. The mode seeking algorithm was then modified slightly. It was then applied to two practical problems; 1) a three class problem in a four feature space (photographic transmission in four spectral bands), and 2) a three class problem in a single feature space. Results of these example problems are given in Chapter 4. In Chapter 5, a possible application of mode seeking algorithm in training the K-class classifier is discussed.
Library of Congress Subject Headings
South Dakota State University Theses
Number of Pages
South Dakota State University
Kaveriappa, G. K., "Unsupervised Iterative Clustering in Pattern Recognition" (1973). Electronic Theses and Dissertations. 3892.