Abstract: The points in two scanned point clouds are not repeated and unrecognizable; all monitoring points must be computed by some methods that consider the point cloud characteristics, such as the gravity center method, fitting method, model difference method and so on. These existent methods are sometimes limited in practical applications because they either get very few deformation points or require continuous and smooth surface. This paper proposes a new method that can directly calculate the deformation based on scanned point clouds during two periods through following processing: segment point clouds of deformable object; transforming coordinates and projection, dividing the point clouds into grids; computing and sorting the shortest distance; averaging median of sorted distance; inversing transformation deformation. Simulated and field experiments show that the proposed method possesses following characteristics: non–contact measurement, no need for data preprocessing, strong anti–noise, monitoring point pairs with high–density, surface–independent and mm–level monitoring accuracy within 150 meters.

Key words: point cloud, segmentation, grid division, shortest distance, median, deformation monitoring