Algorithm 2: Path Inference.

1 Sort the candidate frames by ascending order of timestamps and denote them by ${\mathrm{F}}_1,{\mathrm{F}}_2,\cdots ,{\mathrm{F}}_{\mathrm{k}}$ ; 2 for $1\le \mathrm{i}\le \mathrm{k}$ do 3   for each incoming neighbor ${\mathrm{F}}^{\prime }\to {\mathrm{F}}_{\mathrm{i}}$ do 4    for each partial route $\mathrm{P}\in \mathrm{R}\left({\mathrm{F}}^{\prime }\right)$ do 5     Extend $\mathrm{P}$ to a new path ${\mathrm{P}}^{\prime }=\mathrm{P}\to \mathrm{F}$ i; 6     Incrementally estimate the score of ${\mathrm{P}}^{\prime }$ ; 7     if score $\left({\mathrm{P}}^{\prime }\right)>\mathsf{\delta },$ then 8       $\mathrm{R}\left({\mathrm{F}}_{\mathrm{i}}\right)=R\left({\mathrm{F}}_{\mathrm{i}}\right)\cup {\mathrm{P}}^{\prime }$ ; 9      if len( $\mathrm{P}\prime$ ) $>\mathrm{len}\left({\mathrm{P}}_{\max }\right),$ then 10        ${\mathrm{P}}_{\max }={\mathrm{P}}^{\prime }$ ; 11    $\mathrm{R}\left({\mathrm{F}}_{\mathrm{i}}\right)=\mathrm{R}\left({\mathrm{F}}_{\mathrm{i}}\right)\cup \mathrm{Fi}$ ; 12 return ${\mathrm{P}}_{\max }$ ;