Hidden periodicity in LiDAR point-count rasters

Alejandro Urias 

LAS files store coordinates on a fixed step, usually 1 cm, and counting points in raster cells can add a regular pattern that comes from the storage and counting grids rather than the ground. We ask how a user can test whether a repeating pattern in a point-count raster was introduced this way. We model three mechanisms (grid mismatch, boundary rounding and uneven filling of stored positions), predict their size and, for grid mismatch, where they appear in a spectrum, and test the predictions with controlled changes on UAV, mobile and airborne point clouds and in PDAL and lidR. Stripe sizes matched the predictions (median regression slope 1.00), and moving a 10 cm grid by 5 mm reduced the RMS relative count change per cell of a UAV raster from 11.6 % to 0.5 %. In an airborne tile where the pattern was more than 30 times weaker than the counting noise of one cell, its spectral lines stood well above the spectral background, and random thinning showed the expected rise of line-to-background ratio with the square root of the number of points. Across eight cell sizes the grid-mismatch line followed its predicted ground wavelength while a reference component stayed fixed. The effects are largest for cells below a few decimetres. Because each mechanism responds to a different controlled change, a suspicious pattern can be tested, and each mechanism has its own remedy.