Phase-based crossline center detection for robust vibration monitoring under large motion and pose variations

• Robust vibration measurement under large motion and pose variations. • Crossline provides perspective-invariant features for robust center detection. • Crossline center accurately located using intersection of phase zero-crossing lines. • Enhanced pixel-wise camera imaging simulation improves sub-pixel motion validation. • Validated technique achieves image-plane tracking accuracy under 0.003 pixels. Monitoring of a vibrating target undergoing large motion has been an interest in the field of industrial informatics, aerospace testing, etc. However, the large translation and the possible pose variation of target impose difficulties to a stable and accurate vibration monitoring using vision tracking techniques. Herein, this study proposes a phase-based crossline center detection technique capable of accurate and robust absolute image-plane displacement tracking (mean absolute error under 0.003 pxl) even under large motion and severe pose changes. The two orthogonal lines of crossline pattern remain as line patterns even after perspective transformation, providing a stable feature for lateral phase extraction and center detection. Each line position is determined by searching for the phase zero-crossings after phase extraction along two directions, without the requirement of constant phase assumption and phase pyramid as the conventional phase-based optical flow. The center is accurately detected by intersecting the two lines. This study provides a derivation of the phase-based crossline center detection process, demonstrating that the marker center can be precisely located using phase zero-crossing. An improved pixel-wise camera imaging simulation method is also proposed to validate the accuracy and robustness of the proposed technique with different filter parameters and crossline poses and sizes. Laboratory experiments are conducted on a movable shaker with constant-amplitude vibration output testing the accuracy and robustness of the proposed technique. Field experiment is performed by monitoring the vibration of a moving spindle on a hybrid machine tool with comparison with accelerometers. The results confirm the effectiveness of the proposed phase-based crossline center detection and its potential for vibration monitoring tasks even with large motion and pose changes. In this context, accuracy refers to image-plane displacement, which corresponds to physical motion under planar conditions, while under significant pose changes and large motion it primarily indicates the motion’s order of magnitude, trends, and frequency content.