Introduction:
Over the last decades, there has been a remarkable
increase in the number of remote sensing sensors on-board various
satellite, aircraft, and land vehicle platforms. Large volumes of
panchromatic, multispectral, and hyperspectral data with high
resolutions (e.g., meter to sub-meter level for satellites) have been
collected periodically. Fusion of these multidimensional remote sensing
data along with in situ observations from multiple sensors can help us
to derive more information than is possible from a single sensor alone.
Examples include the determination of the composition of ground
vegetation, localization of mineral resources, and other traditional
application areas. However, the more detailed information such as shape
and object type cannot often be derived precisely. New automatic target
recognition methods that can distinguish critical object attribute and
geometric characteristics would allow extraction of scale and
rotationally variant objects and targets in a scene, or distinguishing
two objects made of the same material. Recent development and advances
in the biologically inspired methods involve segmenting patterns,
materials, and objects, among other capabilities. If such human
perception supported object recognition methods can be combined with
much developed geometry based object recognition techniques, a variety
of object recognition tasks in civilian, military, and intelligent
applications can be significantly improved and speed up. They can also
open up new application fields that may be otherwise impossible. In this
research, we will develop an innovative biologically and geometrically
inspired approach to target recognition for multispectral/hyperspectral
and multiplatform image analysis.
Objectives of the Research
The objectives of this research project are to:
-
Develop a biologically inspired and extended
algorithm of S-LEGION (Spatial LEGION) to quickly analyze and
extract information from multispectral/hyperspectral remote sensing
images covering large areas,
-
Develop a new vector geometric active contour (GAC)
model for target boundary extraction, refinement and shape
reconstruction in interest areas extracted by S-LEGION,
-
Research on a memory-based target recognition
method based on S-LEGION that can perform the final recognition of
interested targets considering spectral, contextual, and geometric
patterns, and
-
Establish an integrated multi-sensor (satellite,
airborne, descent, and ground) model for scale and rotationally
variant object extraction and target recognition across multiple
platforms to support the biologically and geometrically inspired
approach.
|
