Algorithm 1: Candidate Path Search (CPS)
Input: GPS trajectory $trj=\left\{p_1,p_2,\cdots ,p_m\right\}$; the set of edges in road network $E$; search radius $r$; heading change threshold $\beta$; Output: a set of $Path=\left\{pat{h}_1,pat{h}_2,\cdots ,pat{h}_n\right\}$
1.	for ($i=1;i\le m;i++$)
2.	$C{E}_i\leftarrow$ CandidateRoadSegmentQuery($p_i$,$E$,$r$,$\beta$)
3.	end for
4.	$Path\leftarrow C{E}_1$
5.	for ($i=2;i\le m;i++$)
6.	$Path{}^{\prime }\leftarrow \varnothing$;
7.	foreach ($p$ in $Path$)
8.	foreach ($ce$ in $C{E}_i$)
9.	$\mathsf{\sigma }\leftarrow {{\displaystyle \sum }}_{k=1}^m{e}_k.length/e_k.speedlimit$
10.	$paths\leftarrow$ PathForestQuery($p.lastedge$,$ce$,$\sigma$);
11.	$Pat{h}^{\prime }\leftarrow paths$;
12.	end foreach
13.	end foreach
14.	$Path\leftarrow \mathrm{PathConnect}\left(Path,Pat{h}^{\prime }\right);$;
15.	end for