Constructing a Search Model to Predict Percent Correct Counts for Each Sample

Abstract

Drop in reliability of counting inspection is due to the miscounting of objects being inspected. A series of experimental runs was taken to determine the causes of this miscounting. In the experiment, the eye movements of the subject were measured with an eye camera. The analysis of the measured eye movements showed that there are two causes of miscounting: making a smaller number of eye movements than the number of objects to be counted and overlooking objects present in the field of peripheral vision. It was found that the former type of miscounting occurs during the counting process, one of the two human information processes, and that the latter occurs during the immediate apprehension process. A model to predict miscounts was constructed from the obtained findings.