In　frontier-based　exploration,　the　primary　work　is　to　iteratively　detect　frontiers　and　generate　frontier　points　to　guide　the　mobile　robot　to　explore　the　unknown　environment.　However,　most　frontier　detection　algorithms　detect　the　frontiers　by　processing　the　entire　map　data.　This　is　a　time-consuming　process.　In　this　work,　we　present　a　novel　approach　to　dynamically　generate　safe　and　reachable　frontier　points.　The　proposed　approach　examines　only　the　new　coming　laser　data　to　extract　coarse　frontier　points　and　then　refines　them　to　the　global　optimal　frontier　points　by　combining　them　with　global　topological　map.　Only　the　current　laser　data　and　topological　information　are　processed,　the　computational　cost　of　the　proposed　method　has　a　significant　reduction　comparing　to　traditional　approaches.　Thus,　this　method　provides　an　opportunity　to　improve　the　efficiency　of　autonomous　exploration.　Experimental　results　verified　the　effectiveness　and　real　time　performance　of　the　proposed　approach.