초록 ▼

Target detection using algorithms is based on the adaptive real-time endmembers selection from hyperspectral image data. New pixels are correlated with prestored clutter endmembers to obtain subsequently detected target and clutter pixels and then suppressing the latter, leaving about 10% of the ima

Target detection using algorithms is based on the adaptive real-time endmembers selection from hyperspectral image data. New pixels are correlated with prestored clutter endmembers to obtain subsequently detected target and clutter pixels and then suppressing the latter, leaving about 10% of the image as targets. Targets are then reduced in number by size filtering.

대표청구항 ▼

[ Having thus described my invention, what I claim as new and desire to secure by Letters Patent is as follows:] [1.]1. A method for separating man-made military targets from natural background clutter while scanning a scene in a raster format with a hyperspectral pixel sensor, comprising the steps

[ Having thus described my invention, what I claim as new and desire to secure by Letters Patent is as follows:] [1.]1. A method for separating man-made military targets from natural background clutter while scanning a scene in a raster format with a hyperspectral pixel sensor, comprising the steps of:A) Prescanning said scene to obtain a set of sample pixels;B) normalizing each pixel of said set with a computer to provide a unity vector endmember with a spectral angle that represents the target signature of said sample pixel;C) deleting all substantially redundant ones of said unity endmembers that represent various types of background clutter;D) deleting all but one of the endmembers which represent man-made targets;E) storing said remaining endmembers in a library section of the memory bank of said computer;F) scanning a new scene with said sensor to obtain a set of new pixels;G) normalizing said new pixels and correlating each one of said new pixels with all of said endmembers to provide each said new pixel with a set of scores proportional to its difference in spectral angle from each endmember;H) assigning each of said new pixels a target contrast value of one;I) separating targets from clutter by reassigning each of said new pixels a clutter contrast value of zero, if its scores indicate it correlates closest to a clutter endmember;J) eliminating any group of said pixels, with contast value one that is too large or too small to be a valid target; andK) Utilizing the data thus obtained for target acquisition as it is processed.