Abstract: As　one　of 　the 　core 　ste ps　in　point 　cloud 　data　processingthe 　re gistration 　result 　has　great 　influences 　on　the 　subsequent　data　operations．Traditional 　precise 　re gistration　methods　mainly　depend　on artificial 　tar gets 　and　feature 　points．These　methods　are 　limited 　by　the 　external 　environmentinitial conditionsfeature 　points 　are 　not　eas y　to 　find 　and　so　on．To　overcome 　the 　limitation this　paper 　proposes 　an 　improved　Particle　Swarm　optimization PSOal gorithm．Using　the 　sum　of 　normal 　vectors cross 　products 　to　define 　the 　fitness 　function the 　current 　al gorithm　applies 　an　efficient Universal Searchand 　implements 　scattered 　cloud 　data 　re gistration 　based　on　the　best 　re gistration 　points．By　the experiment　with　the 　cloud 　data 　received　by　a　multi－station 　scanning　of 　a　hi gh　stee p　slo pe　rock　and comparing　the 　result　with　the 　classical 　al gorithms 　such　as 　ICPthe 　improved　PSO　al gorithm　is 　proved to　be 　feasibleefficient 　and　stable．It 　can　effectivel y　solve 　the　problem　of 　the 　tar gets 　or 　the 　feature points 　are 　not　eas y　to 　find 　in 　re gistration 　process．