Rapidly-exploring　Random　Tree　(RRT)　algorithm　is　widely　used　in　path　planning,　while　the　RRT　is　inefficient　for　robotic　exploration　in　large-scale　environments　with　multi-obstacles　and　narrow　entrances.　Here,　we　propose　a　Hybrid　Frontier　Detection　(HFD)　strategy　for　autonomous　exploration　which　incorporates　a　variable　step-size　random　tree　global　frontier　detector,　a　multi-root　nodes　random　tree　frontier　detector,　and　a　grid-based　frontier　detector　algorithm.　The　proposed　strategy　enables　a　robot　to　quickly　search　for　the　frontier　in　real-time.　Compared　with　the　traditional　RRT-based　strategy,　the　exploration　time　and　traveling　length　of　the　proposed　HFD　strategy　are　respectively　decreased　by　over　15%　and　12%　in　the　simulation　environment　and　decreased　by　over　14%　and　11%　under　the　same　experimental　conditions　in　the　experimental　environment.　The　results　indicate　that　the　HFD　strategy　effectively　solves　the　problem　of　autonomous　exploration　in　the　environment　with　multi-obstacles　and　narrow　entrances.　©　2021　IEEE.