Algorithm 1. DBSCAN procedure

Input: a set of objects S

Output: a cluster of objects C

Set all the objects in S to be unprocessed, $C\leftarrow Ø$;

while there is an unprocessed object in S

Choose an arbitrary unprocessed object $p\in S$ and calculate $N_{\epsilon }(p)$;

if $\left|N_{\epsilon }(p)\right|\ge MinPts$

Build a new cluster E containing the only object p and add E into C;

$Child\leftarrow N_{\epsilon }(p)-p$;

while there exist unprocessed or noise objects in Child

for each unprocessed or noise object $q\in Child$

Insert q into E and compute $N_{\epsilon }(q)$;

if $\left|N_{\epsilon }(q)\right|\ge MinPts$

$Child\leftarrow Child\cup N_{\epsilon }(q)-q$;

end if

end for

end while

else

Set the object p to be noise;

end if

end while