This project proposes a basis for a flexible processing and tracking scheme based on image classification. Satellite scene images consisting of various background scenarios were classified using linear discriminant analysis and a support vector machine based on four statistical features and the bag of keypoints descriptor. The bag of keypoints descriptor did not render significantly improved results for this dataset. Increasing the number of image classes increased the number of features needed to achieve linear class separability. This methodology can be applied for imagery of any waveband. Target trackers modified to leverage this classification scheme may be allowed to track more successfully in a variety of background scenarios.

• This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.

Creator

• Banco, Daniel Pasqual

Publisher

• Worcester Polytechnic Institute

Identifier

• E-project-122214-205853

Advisor

• Iannacchione, Germano S.
• Clancy, Edward A.

Year

• 2014

Center

• MIT Lincoln Laboratory-Lexington, Massachusetts Project Center - MQP

Sponsor

• MIT Lincoln Laboratory

Date created

• 2014-12-22

Resource type

• Major Qualifying Project

Major

• Interdisciplinary

Rights statement

• In Copyright