A SIMILARITY-HISTOGRAM BASED SEMI-THRESHOLDING APPROACH FOR SUBPIXEL TARGET DETECTION IN UNKNOWN BACKGROUND

摘要

Detecting and classifying targets from an unknown background is a frequent and extremely difficult task, particularly for mixed pixel targets. In Human discrimination, a target in a natural scene may have distinct signatures between the target and neighboring interferences. In this work, we present an efficient means of isolating the specific interested material from interferences within mixed pixels using a similarity histogram-based semi-thresholding technique (SHST). Experimental results obtained from applying the SHST method to AVIRIS data demonstrate that SHST can accurately detect a subpixel target in unknown environments without the selectivity problem.