Table 2 M-DBSCAN algorithm

Input: A data set X containing n intersections, neighborhood radius eps, minimum number \(mi{n_{pts}}\), Mahalanobis distance matrix D.

Output: Clustering set C, noise parameter is noise.

1 Mark all intersections as untraversed

2 Randomly select an unobserved intersection p

3 Mark point p is processed

4 If the eps neighborhood of p has at least \(mi{n_{pts}}\) intersections (core intersections) based on D

5 Create a new cluster C and add p to C

6 Create a set N that contains all intersections in the eps neighborhood of p

7 for all \(q \in N\) do

8 If q is unprocessed

9 Mark q as processed

10 If the eps neighborhood of q has at least \(mi{n_{pts}}\) points, add these intersections to N

11 If q is not yet a member of any cluster, add q to C

12 end

13 Output C

14 Otherwise, mark p as noise (spurious target cluster)

15 Until there are no unobserved intersection