Predicting Topographical Defect Severity and Repair Requests from Simple Physical Measures

Charles Lloyd and Yunjian He

From an August 1996 technical report written for the Lighting Research Center, Troy, New York.


This evaluation was conducted in two parts. In the first, eight observers were asked to rate the severity of small-area topographic defects in an automobile paint finish when viewing the paint outdoors, and indoors using a lighting installation having the essential features of the system proposed for the paint inspection tunnels used by the manufacturers. The eight observers had all purchased a new car for more than $12,000 in the last three years. The defect severity ratings for the same defect made indoors and outdoors were found to be highly correlated. Further, the proportion of observers who would request a repair was consistently related to the defect severity rating.

In the second part of the evaluation, a simple photometric method for objectively measuring the area and luminance modulation (contrast) of a small-area, topographical defect was developed. The measurements of the area and modulation of the defects used in the first part were combined with the severity ratings to construct two predictive models. Using these models, the defect severity and the probability that a customer will request a repair can be closely related to the area and modulation of the defect.

The results obtained are limited to small-area topographical defects of the type that occur when dirt is embedded in the paint. They are also limited to paint defects occurring on flat surfaces and by the small number of observers used. However, the results are sufficiently promising to recommend an attempt at validation with a larger set of participants who have recently purchased automobiles. Such an attempt would require the construction of a larger set of defect samples seen by a much larger and more representative sample of new vehicle buyers. It is also recommended that an instrument for measuring defect area and modulation, suitable for field use and based on the method reported here, be developed and tested. Finally, it is recommended a wider range of outdoor lighting conditions be used as part of the next evaluation.