Skip to main content
ARS Home » Research » Publications at this Location » Publication #164497
Title: Detection-threshold calibration and other factors influencing digital measurements of bare ground

Author
item Booth, D
item Cox, Samuel
item JOHNSON, D - OREGON STATE UNIVERSITY

Submitted to: Journal of Range Management
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/19/2005
Publication Date: 11/5/2005
Citation: Booth, D.T., Cox, S.E., Johnson, D.E. 2005. Detection-threshold calibration and other factors influencing digital measurements of bare ground. Journal of Range Management 58:598-604.

Interpretive Summary: Rangelands are a vast resource and monitoring for ecological health requires significant time and money. Monitoring costs might be reduced by computer software that will analyzed rangeland photographs at hundreds per minute to measure plant cover or bare ground; however, measurements are subject to human judgment, errors and inconsistencies in setting detection thresholds. We developed a calibration procedure that makes threshold adjustment less subjective and we tested our calibration by comparing manual and automated measurements. In 3 tests measurements by calibrated software did not differ from manual measurements by more than 7% -- compared to a 10 to 26% difference without calibration -- suggesting the potential for computers to significantly reduce the cost of ecological monitoring.

Technical Abstract: New methods of image acquisition and analysis are advancing rangeland assessment techniques. Current versions of image-analysis software require users to set thresholds for color / object classification -- a subjective process affected by human errors and inconsistencies. We measured the effect of subjective threshold adjustment and developed a calibration procedure that makes threshold adjustment less subjective. We tested our calibration method by comparing manual and automated image-analysis ground-cover measurements. Measurements were made using software with a calibrated threshold (a method that allows for batch-processing of hundreds of images) against a manual method similar to the image point sampling advanced by researchers in 1966 and 1971. Cover measurements derived from batch processing using calibrated software had much less deviation from manual measurements than would be expected from subjective adjustment of a color-recognition threshold.