Classification Performance
Statistical performance curves plot the trade-off between the true positive rate and the false positive rate for a diagnostic or classification system across different discrimination thresholds. In electronics inspection, the receiver operating characteristic defines the ability of automated optical inspection systems to separate true solder defects from false calls. This graphical evaluation helps engineers select threshold settings that minimize both escapes and false alarms.
It is used as a standard tool to quantify the detection capability of testing hardware.
Threshold Optimization
Adjusting the sensitivity of an automated test system shifts the operating point along the curve, changing the ratio of true defects caught to false boards flagged. If the threshold is set too high, the system catches all defects but generates too many false positives, which increases the labor cost of manual review. Conversely, setting the threshold too low avoids false alarms but allows defective boards to pass through to the next production step undetected.
By analyzing the area under the curve, quality engineers can compare the effectiveness of different inspection algorithms and choose the optimal setting. This analysis ensures the line operates efficiently without compromising product quality.
System Evaluation
Testing protocols run benchmark boards through inspection systems to generate data points for plotting these performance curves. These reference boards contain known defect distributions to ensure the accuracy of the resulting metrics. Using the curves allows manufacturers to verify that newly installed inspection systems meet factory performance specifications before line release.
The resulting data guides the calibration of both visual and X-ray test methods.
