AFREET: HUMAN-INSPIRED SPATIO-SPECTRAL FEATURE CONSTRUCTION FOR IMAGE CLASSIFICATION WITH SUPPORT VECTOR MACHINES
The authors examine the task of pixel-by-pixel classification of the multispectral and grayscale images typically found in remote-sensing and medical applications. Simple machine learning techniques have long been applied to remote-sensed image classification, but almost always using purely spectral information about each pixel. Humans can often outperform these systems, and make extensive use of spatial context to make classification decisions. They present AFREET: an SVM-based learning system which attempts to automatically construct and refine spatio-spectral features in a somewhat human-inspired fashion. Comparisons with traditionally used machine learning techniques show that AFREET achieves significantly higher performance. The use of spatial context is particularly useful for medical imagery, where multispectral images are still rare.
- Research Organization:
- Los Alamos National Laboratory (LANL), Los Alamos, NM (United States)
- Sponsoring Organization:
- US Department of Energy (US)
- DOE Contract Number:
- W-7405-ENG-36
- OSTI ID:
- 774619
- Report Number(s):
- LA-UR-01-891; TRN: AH200121%%88
- Resource Relation:
- Conference: Conference title not supplied, Conference location not supplied, Conference dates not supplied; Other Information: PBD: 1 Feb 2001
- Country of Publication:
- United States
- Language:
- English
Similar Records
Implementing Spatio-Temporal 3D-Convolution Neural Networks and UAV Time Series Imagery to Better Predict Lodging Damage in Sorghum
Land cover classification in multispectral imagery using clustering of sparse approximations over learned feature dictionaries