The　radial　and　axial　vibration　signals　of　a　rotating　machine　are　crucial　information　to　understand　the　machine　operation　and　to　diagnose　potential　faults.　Instead　of　using　three　single　sensors　to　measure　the　horizontal,　vertical　and　axial　displacements　of　rotating　shafts,　a　novel　non-projection　vision-based　system　was　proposed　to　realize　simultaneous　measurement　of　the　radial　and　axial　displacements　with　high　accuracy　and　good　reliability　using　an　tailored　artificial　constant　density　sinusoidal　fringe　pattern　(CDSFP),　which　was　pasted　around　the　shaft　surface　and　worked　as　a　tri-axial　(i.e.　the　horizontal,　vertical　and　axial)　displacement　sensor.　The　measurement　principle　and　setup　of　the　proposed　measurement　system　were　well　established.　The　horizontal　displacement　could　be　correctly　obtained　from　the　fringe　period　density　changes　of　the　CDSFP　image　sequence　recorded　by　a　high-speed　camera.　Simultaneously,　the　vertical　displacement　could　be　acquired　by　tracking　the　centerline　of　shaft　whilst　the　axial　displacement　could　be　obtained　by　locating　the　peaks　of　the　cross-correlation　sequence　of　the　fringe　intensities.　A　sub-pixel　method　was　employed　to　improve　the　displacement　resolution　of　the　developed　system.　The　performance　of　the　proposed　system　was　demonstrated　by　the　comparison　of　the　experiments　using　eddy　current　sensors.　It　showed　that　the　proposed　method　was　an　effective　and　accurate　technique　for　real-time　tri-axial　vibration　monitoring　of　rotating　shaft.　Experimental　results　verified　the　feasibility,　effectiveness　and　good　robustness　of　the　proposed　methodology,　which　demonstrated　that　the　proposed　system　was　capable　of　achieving　accurate　tri-axial　vibration　displacements　of　rotating　shaft　compared　to　the　commercial　eddy　current　sensor　which　could　only　measure　one　dimensional　displacement　at　each　measurement.　Therefore,　the　vision-based　tri-axial　vibration　monitoring　system　could　be　recommended　for　real　engineering　applications　in　condition　monitoring　of　rotating　shafts.　©　2018　Elsevier　Ltd