Iterative automatic global registration algorithm for multi-view point cloud of underground tunnel space

Aiming at the narrow and long tunnel structure, few internal features, a large amount of point cloud data, existing registration algorithms commercial software results are not ideal, an iterative global algorithm is proposed for massive underground registration, which composed local initial pose acquisition adjustment. Firstly, feature coordinates in extracted, then station-by-station performed according to Rodrigues matrix. Finally, result considered as value parameter, adjustment all observations carried out. The observation values weighted by selection weight iteration method weights constantly modified process until threshold conditions met stops. In this paper, experimental made up 85 stations from Xiamen subway tunnel, about 1300 m long. When accumulated error station-to-station too large, several regarded partial wholes, optimal achieved through multiple adjustments, accuracy within 5 mm. Experimental confirm feasibility effectiveness algorithm, provides new space tunnel.