Exploratory image data analysis for quality improvement

• Exploratory Image Data Analysis (EIDA) is an extension of exploratory data analysis (EDA), focusing on leveraging image data to identify patterns, generate hypotheses, and improve quality processes.
• EIDA integrates steps such as image processing, quantitative data analysis, feature identification, and interpretation to uncover causal relationships and support decision-making.
• Case studies in welding, automotive, and pipeline quality demonstrate EIDA’s practical applications in detecting defects, prioritizing improvements, and enhancing quality management.

Exploratory Image Data Analysis (EIDA) builds on the principles of exploratory data analysis (EDA) by adapting its framework for image data, allowing for hypothesis generation and quality improvement. The EIDA framework involves four key steps: image processing to enhance and prepare images for analysis, quantitative data analysis to derive actionable insights, identification of salient features that reveal patterns or deviations, and interpretation to develop hypotheses about causes of quality issues. EIDA helps translate complex image data into meaningful insights by focusing on transparency and reproducibility.

The method is illustrated through real-world examples in welding inspection, automotive body-in-white (BIW) dimensional measurement, and pipeline defect analysis. For instance, in a welding study, EIDA identified key conformance issues like undercut and stick out by analyzing image clusters and linking them to quality standards, helping prioritize power level adjustments. Similarly, in automotive assembly, spatial and temporal analysis of image-derived measurements revealed deviations in door gaps and flushes, enabling targeted improvements in robotic fixture calibration.

EIDA emphasizes graphical data representation to highlight patterns and facilitate hypothesis testing. Pareto charts, control charts, and spatial maps provide a clear visual context for identifying critical variables and deviations. By aligning with a structured, principle-based approach, EIDA bridges the gap between raw image data and quality improvement and offers a framework for future applications, including video analysis and integration into commonly used quality management software.

Read the full article