Each zone accumulates orthogonal projections of every neighbour point p i onto the BFP.
Vector v i connects orthogonal projections of sampling point p p and neighbour point p i onto BFP (Figure 5).
For each neighbour point p i, inner product of normal n i and vector v i is also calculated.
Distance with sign means that the sign of the value depends on the location of neighbour point p i with respect to the BFP normal vector (Figure 3).
Weight w i for each neighbour point is proportional to the Gaussian distribution of the distance s i between orthogonal projections of sampling point p p and neighbour point p i : w i ∼ e − s i 2 2 r 2, where r is the neighbourhood radius.
A set of random points (approximately 1000, or less if model is smaller than 1000 points) is selected and for each point the distance to the closest neighbour point is calculated and the average distance is found.
BFP division is made using vector v 1, which connects projections of the arbitrary neighbour point p 1 with division centre p p, and using vector v 2, which is a vector product of BFP normal and vector v 1.
The reason is that, if a too small value gau(i) of nearest-neighbour point i is summed to compute the density together with other big values, that would bring big bias, and thus the point should be gotten rid of.
This grid is composed of M radial sectors and L angular sectors, as shown in Figure 4. Every neighbour point p j within B σ r + 1 f is associated to a grid sector, and for each sector we compute two score values derived by the normal and saliency information.
The same car, one neighbour pointed out, that Cristiano Ronaldo drives.
