Integrated Planning for Automated Image Processing
Steve Chien, Helen Mortensen, Christine Ying, Shouyi Hsiao · 1995
The Multimission VICAR Planner (MVP) sys-tem is an AI planning system which constructs executable image processing programs to sup-port Operational Science Analysis (OSA) re-~juests made to the Jet Propulsion LaboratoryPL) Multimission Image Processing Subsys-tem (MIPS). MVP accepts as input: image files and a high-level specification of desired corrections, enhancements, output properties (such as for mosaics). MVP then derives: un-specified but required processing steps, rele-vant image processing library programs, and appropriate parameter settings for such pro-grams- constructing an executable image pro-cessing program to fill the image processing re-quest. MVP is currently available to analysts to fill requests and reduces the effort to fill ra-diometric correction, color triplet reconstruc-tion, and mosaicking tasks by over an order of magnitude. Problem Description In recent times, improvements in spacecraft imag-ing hardware have caused a massive increase in the amount of scientific data and variety of science data products. Simultaneously, increased sophistication of image processing algorithms has complicated the image processing task. While ensuring physical ac-cess to the vast amounts of space-related data can be achieved, it is often extremely difficult for the av-erage user to manageably prepare and process the available scientific data. One method for reducing this data access, prepa-ration, and analysis problem is the development of general purpose data processing languages and in-terfaces. These languages and interfaces allow users to access and process data within a common en-vironment. For image processing, the VICAR en-vironment (Video Image Communication and Re-trieval 1) (LaVoie t. al. 89) is a major constituent