One assignee leads the ranked field with 117 records, well ahead of the fifth-ranked assignee at 22 and the tenth at 15. That gap indicates the leader has built breadth across the measurement pipeline rather than relying on a small cluster of related filings. The top five assignees combined hold 246 of 863 records in scope, about 28.5% of the total, so the field has a genuinely concentrated top tier even though the ranking runs to 100 companies overall.

The data shows publications falling from a 2019 peak of 83 to 30 in 2024, a -29% move from the 42 recorded in 2021. That is a real cooling in the recent complete years, but it should not be read as the field disappearing: publication normally lags filing by around 18 months, so the 2025 and 2026 figures in any such dataset are still incomplete and will rise as filings from those years get published. The safest read is that the earlier boom years of rapid growth have passed, not that activity has stopped.

The dominant class is G01B, measuring length and dimensions, which covers 59.8% of the 863 records in scope — reflecting that this is fundamentally a dimensional measurement technology. G06T (image data processing) follows at 31.1% and H04N (pictorial communication) at 20.5%, showing heavy overlap with computational imaging. A61B (diagnosis and surgery) appears on 9.4% of records, pointing to meaningful use in medical and dental scanning applications such as intraoral imaging.

This filing from Nanjing University of Science and Technology combines a multi-frequency temporal phase unwrapping method with a residual convolutional neural network trained to resolve fringe order at high frequency, using four projected pattern frequencies (1, 8, 32 and 64). It sits specifically at the intersection of deep learning and temporal phase unwrapping rather than covering fringe projection broadly. Anyone building a phase unwrapping pipeline that does not use a trained neural network for fringe-order resolution, or that uses a different frequency-combination strategy, is working outside its specific claim scope, though a full freedom-to-operate check should compare it against the surrounding patents from the same period.

Relative to the dense core around G01B measurement and calibration methods, branches like specular surface compensation, calibration-target-free scanning pipelines and real-time point cloud registration for moving targets show thinner filing density in this dataset. Co-assignment is also limited to only 10 recurring pairs, suggesting cross-company joint development is uncommon and most portfolios are built in-house. That combination — a concentrated top tier plus specific under-filed technical branches — is where new entrants have the clearest room to file defensible first claims.