A program system for efficient multispectral classification
Abstract
The pixelwise multispectral classification for analyzing remotely sensed imagery data was implemented by a simple one-dimensional box-classifier, a class-pivotal canonical classifier, and a local neighborhood filtering algorithm to decrease the computer time. The class-pivotal canonical classifier uses linear transformations to evaluate the pixel and to compute its maximum likelihood ratio. The method for supervised classification has been implemented in the IBM 370 batch-mode system, the DEC 10 PILIP dialog system, and its interactive PDP 11/34 program.
- Publication:
-
Tokyo International Astronautical Federation Congress
- Pub Date:
- September 1980
- Bibcode:
- 1980toky.iafcQ....A
- Keywords:
-
- Image Processing;
- Imaging Techniques;
- Multispectral Band Scanners;
- Photointerpretation;
- Remote Sensing;
- Algorithms;
- Canonical Forms;
- Classifications;
- Classifiers;
- Linear Transformations;
- Maximum Likelihood Estimates;
- Optical Scanners;
- Instrumentation and Photography