Quality Control of Pixel Detector Modules

Modified Steger’s Algorithm with Kalman Filter for Robust Line Tracking

This project presents an improved line-tracking algorithm for automatic wire bond inspection, based on a modified version of Steger’s ridge detection method augmented with a Kalman filter for stabilization. The work targets failure modes observed in dense and noisy visual inspection data from ATLAS ITk pixel modules.

While Steger’s algorithm is effective at identifying line centers, it becomes unstable in regions where multiple wires cross or where image quality is degraded by reflections and noise. In such cases, the original linking procedure can make ambiguous or inconsistent decisions about track continuation.

To address this, a Kalman filter was introduced into the linking stage, allowing the algorithm to combine current measurements with predictions based on previous states. The tracked state includes position, local direction, and line strength, enabling smoother and more physically consistent tracking through intersections and low-contrast regions.

The improved tracker enables reliable identification of broken, loose, and touching wires by enforcing geometric continuity and endpoint consistency. This approach substantially increases robustness while remaining computationally efficient, making it suitable for large-scale automated inspection workflows.

Links