For　the　problem　of　low　accuracy,　time-consuming　and　quantified　faults　cannot　be　provided　that　the　current　methods　have,　a　fault　recognition　method　based　on　affinity　propagation　clustering　algorithm　and　least　square　curve　fitting　is　proposed.　Firstly,　discontinuous　points　of　horizons　were　located　by　the　connected　component　labeling　method　and　used　as　control　points　of　the　faults.　Then,　the　discontinuous　points　were　clustered　by　the　affinity　propagation　clustering　algorithm　and　the　points　of　the　same　cluster　were　used　to　determine　a　fault.　The　number　of　the　faults　was　determined　by　the　number　of　classes　after　clustering.　At　the　end,　the　discontinuous　points　of　the　same　cluster　were　fitted　by　the　least　square　curve　fitting,　and　the　curves　were　the　desired　faults.　By　the　proposed　method,　the　time　consuming　cross　correlation　calculation　of　the　traditional　method　was　discarded,　the　computing　process　was　simplified　and　the　faults　were　quantified.　To　validate　the　effectiveness,　the　proposed　algorithm　was　compared　with　the　traditional　and　the　latest　methods　on　the　real　seismic　data　and　the　experimental　results　confirmed　that　the　accuracy　of　the　proposed　algorithm　is　better　and　the　time　consumption　is　reduced　greatly.　©　2016,　Scibulcom　Ltd.　All　rights　reserved.