EYE MOVEMENT BEHAVIOR AS A PREDICTOR OF PERFORMANCE ON PACED VISUAL INSPECTION TASKS
Abstract
In many areas of industry, inspection stations on assembly lines require that inspectors view products moving on a conveyer belt. They must determine whether the product passing on the conveyor belt conforms to a set of standards and must take appropriate action. This research was designed to assess the relationships between the demands placed on the inspector by such a dynamic visual inspection task, the visual abilities required to perform the task and the inspector's performance, measured by his error rate. A two part experiment was conducted to evaluate both visual abilities and performance. In Part I, dynamic visual acuity (DVA) threshold was measured as the subjects viewed Landolt targets passing before them from left to right across a curved screen. Three levels of angular target velocity and three levels of exposure time were varied. Part II of the experiment measured errors on a simulated dynamic visual inspection task. A 14 foot conveyor was used and three levels of target spacing and three levels of belt velocity were varied. Targets were viewed on a one at a time basis. Eye movements were measured on both parts of the experiment using Electro-oculography. Each part of the experiment was configured as a completely repeated measures (CRM) design with each of the 12 subjects being randomly assigned to all treatment combination in both experiments. Eye movements were broken down into three time categories including target acquisition time (t(,1)), target search time (t(,2)), and time remaining in the viewing window (t(,3)). Errors were broken down in 13 categories. Results of the error analysis indicated that the best performance (lowest percent error) was achieved with the slowest belt velocity and largest target spacing. A comparison was made between error rates and eye movements for the same task condition. This comparison showed that the conditions which provided the lowest error rates also provided the longest available time for target acquisition and search. A high correlation between t(,2) and total errors (r = -.91) supports the conclusion that eye movement behavior, particularly target search time t(,2) is highly related to performance on dynamic visual inspection. Dynamic visual acuity (DVA) threshold was found to have a lower correlation with errors on dynamic visual inspection (r = .71). Good dynamic visual acuity is necessary to achieve high levels of performance on DVI, however, DVA threshold appears to be best suited to determine those inspectors who cannot perceive the necessary detail size presented by DVI task conditions. Once inspectors have been chosen, DVA threshold does not appear sufficiently sensitive to distinguish between differences in performance for these inspectors. The results of this research indicate that visual abilities and performance on DVI tasks are highly related with slower belt velocities and larger target spacings providing the best conditions for improvement in performance, longer available time for target acquisition and search, and improved levels of dynamic visual acuity.
Degree
Ph.D.
Subject Area
Industrial engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.