Error sources in a digital image analysis system

Abstract Experience with a computer vision system used for measurements on sweet potatoes has revealed a number of sources of error, some from equipment and some from procedure. Nine sources - reflection, edge pixels, lens quality, camera warm-up, auto black circuits, pixel aspect ratios, variation with screen location, gray scale range, and object size versus gray scale range - are identified and discussed along with results of tests to reveal the magnitude of the errors introduced.