Integrated Real-Time Automated fiber placement Inspection System with High-Speed Point Cloud Alignment and Fusion

Automated fiber placement (AFP) demands realtime dimensional verification and defect detection. We present an integrated inspection system that fuses $\mathbf{1 0 0 ~ H z}$ line-laser profiles with CNC pose data via a linear resampling algorithm using a center-symmetric sparse matrix, ensuring uniform-density point clouds despite sampling-rate mismatches. Fusion is accelerated through sparse tensor multiplication, reducing CPU time from 39.8 ms to 6.8 ms per batch. Experimental validation demonstrates point cloud stitching accuracy within 0.9 mm across large-scale component scanning.