Automated Detection of Ship Tracks in Multispectral Satellite Data

摘要

The objectives of this work are to develop robust ship track detection methods,demonstrate their utility using satellite imagery, and design the framework of an automated ship track detection system for operational use. The approach is to (1) use geographical and multispectral information to reduce the data stream greatly based on contexts for which ship tracks are physically allowed; (2) optimally enhance satellite images using multispectral signals; (3) apply state-of-the-art edge-detection, dilation and erosion operators in order to find and enhance candidate ship tracks with weak signatures; (4) determine features or parameters that best characterize the higher reflectivity and curvilinearity of ship tracks; (5) apply rule-based and cluster analysis techniques to reduce the data stream to a limited number of subscenes with potential tracks; (6) apply state-of-the-art neural net and statistical discriminant analysis methods as final detection filters; (7) assess detection success and error rates; (8) develop a prototype design of the automated system. The algorithms used here are one that have exhibited success on this or similar problems.