Dynamic-template processing for passive thermograms: Applied to automotive stamping split detection

This manuscript introduces a passive thermal-vision modality to detect splits and tears in automotive stampings in real time. This system scans 3D temperature maps of high reflective steel panels, with complex geometry in addition to the drawing die, to locate abnormal temperature readings indicative of high local wrinkling pressure (cause of metal splitting). Preprocessing utilizes a coefficient-correction map to account for the emissivity dependency on part geometries, and the stamping-die changing the thermal profile. Also, a pattern-matching algorithm is used to synchronize the thermal analysis with the die location (stamping pace) through a new concept of matching thermal histograms. A new image-processing routine named "dynamic-template" is developed to retrieve the split's shape and identify its location. The template is computed through a region of interest (ROI) approach with a dynamic threshold, while accounting for spatial deviations (mainly press-line vibration). The developed hardware and software are implemented on an automotive press-line, and the real-time thermal acquisition is processed to illustrate the system performance and success; different split shapes and sizes are detected under actual operating conditions.