Multi-model integrated line detection robust to environmental variations using Bayesian optimization

As industrial automation and vision-based inspection technologies become increasingly widespread, the demand for reliable linedetection across diverse manufacturing processes continues to grow. However, real industrial environments involve illuminationchanges, metallic surface reflections, and texture variability, which cause substantial performance fluctuations in single-model linedetectors. Moreover, when data collection is limited, learning-based approaches alone may not provide sufficient robustness tosuch environmental variations. In this study, we propose a multi-model integrated line detection approach that runs two parallelAirLine detectors operating in different parameter spaces and combines their complementary outputs using weighted RANSAC. Wefurther apply risk-aware Bayesian optimization with Conditional Value-at-Risk as the objective to automatically derive parameterconfigurations that are robust to environmental variations. The proposed method is validated on 779 real industrial images collectedfrom a shipyard mid-assembly welding environment, achieving up to a 33.5% performance improvement over the initial configuration