Automated Target Detection and Grouping From Remotely Sensed Data
Abstract
Targets are objects that rise above the bottom surface more than a specific amount defined by IHO survey standards. Currently target detection requires time consuming analysis by the human expert. The contracted task is to design and implement an automated target detector that can be used as a tool by human experts. We tested the implementation internally and have sent some of the test results to experts at NAVO for assessment. In addition we also designed and implemented a target grouping procedure that clusters the targets according to a proximity metric. The resulting grouping can be used to produce polygon outlines that will replace selected clusters of densely spaced targets. Several issues and possible improvements were identified from our testing and analysis. They include alternative algorithms for target identification, computational optimization and parallelization of the implementation, application of machine learning algorithms for optimization of parameters for target identification, and a systematic testing regime.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 2006
- Accession Number
- ADA484526
Entities
People
- Brian Bonnlander
- Choh M. Teng
- Clark Glymour
Organizations
- Florida Institute for Human and Machine Cognition