Implementation Scheme for Recursion in Spectral Dimension
Abstract
Developing recursive detection algorithms for thermal image processing is very important for real time implementation. In his previous work for CRDEC, Warren used a first order autoregressive time series to model the background of thermal image with which he further developed a recursive algorithm in time frames. In this report, we revisit his work and present an alternative approach to developing a recursive algorithm for thermal image target detection. The new approach uses Kalman filter theory, which has proven to be very powerful in real time processing because of its recursive nature. The algorithm developed by Warren is designed for a single spectral band. In reality, however, the characteristics of the vapor cloud and background may vary from band to band in spectral domain; thus, it is practical to extend Warren's and the Kalman filtering approaches to cover multiple spectral bands. To alleviate the difficulty of processing multiple bands, two suboptimal models (separable spectral correlation and separable spectral-Markov spatial correlation) are also proposed for the background.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1992
- Accession Number
- ADA248945
Entities
People
- Chein-i. Chang
Organizations
- University of Maryland, Baltimore County