Xiaochun Jiang

Theoretical issues in the design of visual inspection systems

The objective of this review essay is to both chronicle and analyse literature in the area of visual inspection. Classical as well as contemporary papers are included to describe both the historical development and the state of the art of visual inspection theories and technologies. Human operators, despite well-documented problems, often perform visual inspection. While supervized machine systems obviate some of the problems associated with human inspectors, other problems still exist. In particular, accounting for a supervisors perception of a machines performance (as gauged, e.g. by trust) and consequent actions. The difficulties associated with these two alternatives have led to the emergence of a third alternative: collaborative human/machine or hybrid systems, which combine the advantages of both alternatives—in theory. However, in practice, how to best distribute the functions between a human and machine, in a dynamic environment in real time, is problematic. Moreover, a humans perceptions of its counterparts performance remain an issue. These unresolved problems are subjects for future research. In the interim, the alternatives are critiqued to create a basis for establishing guidelines to select the alternative that is best suited for a given situation.

EFFECT OF ERROR SEVERITY ON HUMAN TRUST IN HYBRID SYSTEMS

Computerized systems are often employed to support control and decision-making tasks in complex and dynamic environments. Trust or mistrust in these systems has been demonstrated to significantly affect operator performance. Consequently, errors of trust or mistrust may compromise system performance, with potentially disastrous results. Accordingly, trust should be considered in both the design and operation of human/machine systems. In order to do so, metrics and methods for the measurement of trust must be developed along with models of human performance that incorporate trust and related system variables. Current approaches to trust measurement rely solely on subjective metrics, which are based on different theoretical concepts of trust between humans that may not necessarily be as relevant to machines. Although researchers have been able to establish a relationship between trust and behavior, these models lack an analytic foundation. Therefore, the purpose of this research is to develop a quantitative approach that relates trust to changes in system parameters and severity of errors.

AN EVALUATION OF A JOB-AIDING TOOL FOR AN INSPECTION TASK

Various job aids have been developed to improve visual-inspection performance. Many of the job aids have been developed with the purpose of training inspectors to adopt systematic search strategies. While various research has shown that systematic search strategies are superior to random search strategies, exactly what systematic search strategy constitutes the most efficient search is still unclear. The purpose of this paper is to examine the effectiveness of a job-aiding tool that allows inspectors to inspect a stimulus with a natural search strategy while informing inspectors of areas that have already been searched. In general, the job-aiding tool significantly increased accuracy while decreasing search times and stopping times.

Impact of Task Complexity on Hybrid Inspection System Performance

Recently, 100% inspection using automated systems has seen more frequent application than traditional sampling inspection using human inspectors. Nevertheless, humans still outperform machines for most attribute inspection tasks. Since neither humans nor automation can achieve superior inspection system performance, hybrid inspection systems where humans work cooperatively with machines merit study. In response to this situation, this research was conducted to evaluate the impact of task complexity on inspection performance in a hybrid environment. Results from the study showed that task complexity has a significant effect on inspection performance. Furthermore, the study looked at the impact of human intervention in such a hybrid inspection system and indicated that human inspector will continue to play a vital role in future hybrid systems.

Modeling Errors in Sampling Inspection: Effect of Degraded Performance

Studies have shown that inspection is not error-free and is systematically influenced by such factors as time available, the payoffs, complexity of the task, the decision goal, defect rate, and so on. This study shows how models of decision-making can be used to obtain better designs of sampling plans in the presence of human inspection error. This study also outlines a methodology to develop sampling plans under different levels of degraded human performance. Results revealed that sampling plans are sensitive to decision goals and the level of degradation.

A Preliminary Study of the Measurement of Trust in a Hybrid Inspection System

Since human trust in automation can directly impact inspection quality and overall inspection performance, it is critical to study the issue of trust in hybrid inspection systems. After developing a trust questionnaire for hybrid inspection systems, we conducted a preliminary study to measure humans trust in the hybrid inspection systems using the questionnaire developed. Two hybrid inspection systems were used in the study. Results showed that the trust questionnaire was sensitive to different systems. A stepwise regression procedure selected competency and reliability as the better predictors.