In　road　engineering,　large　amounts　of　pavement　vibration　acceleration　signals　can　be　collected　by　using　acceleration　sensors.　On　the　basis,　the　ability　to　obtain　more　valuable　information　related　to　pavement　performance　by　dealing　with　these　data　has　become　a　key　problem.　In　order　to　provide　relatively　accurate　velocity　and　displacement　for　further　studies,　this　research　introduced　the　formation　principle　of　trend　items　during　integration,　and　emphatically　expounded　the　integral　processing　of　vibration　acceleration　from　two　aspects　of　indoor　and　outdoor　tests.　By　comparing　and　analyzing　the　integral　results　in　time　domain　and　frequency　domain,　it　was　found　that　it　is　necessary　to　process　the　original　data　before　integration　to　avoid　the　trend　items,　and　the　frequency　domain　integral　method　can　reduce　integral　errors　greatly　in　the　process　of　continuous　integration.　©　2018　Taylor　&　Francis　Group,　London.