Determining target detection limits and accuracy delineation using an incremental technique

Classification of terrestrial materials using remotely sensed imagery is becoming an increasingly popular method to assess natural systems. This study uses predictions derived from remotely sensed hyperspectral images to determine minimum target detection thresholds and to statistically constrain classification accuracies. Minimum detection thresholds are determined using an iterative regression breakpoint technique, and accuracy is delineated by quantifying the variance above the correlation breakpoint. Assessing confidence and determining detection limits generates a greater understanding of the classification process and adds significant utility to the classified product.