Fast　and　accurate　rotational　speed　measurement　is　required　both　for　condition　monitoring　and　faults　diagnose　of　rotating　machineries.　A　vision-　and　fringe　pattern-based　rotational　speed　measurement　system　was　proposed　to　measure　the　instantaneous　rotational　speed　(IRS)　with　high　accuracy　and　reliability.　A　special　double-sine-varying-density　fringe　pattern　(DSVD-FP)　was　designed　and　pasted　around　the　shaft　surface　completely　and　worked　as　primary　angular　sensor.　The　rotational　angle　could　be　correctly　obtained　from　the　left　and　right　fringe　period　densities　(FPDs)　of　the　DSVD-FP　image　sequence　recorded　by　a　high-speed　camera.　The　instantaneous　angular　speed　(IAS)　between　two　adjacent　frames　could　be　calculated　from　the　real-time　rotational　angle　curves,　thus,　the　IRS　also　could　be　obtained　accurately　and　efficiently.　Both　the　measurement　principle　and　system　design　of　the　novel　method　have　been　presented.　The　influence　factors　on　the　sensing　characteristics　and　measurement　accuracy　of　the　novel　system,　including　the　spectral　centrobaric　correction　method　(SCCM)　on　the　FPD　calculation,　the　noise　sources　introduce　by　the　image　sensor,　the　exposure　time　and　the　vibration　of　the　shaft,　were　investigated　through　simulations　and　experiments.　The　sampling　rate　of　the　high　speed　camera　could　be　up　to　5000　Hz,　thus,　the　measurement　becomes　very　fast　and　the　change　in　rotational　speed　was　sensed　within　0.2　ms.　The　experimental　results　for　different　IRS　measurements　and　characterization　of　the　response　property　of　a　servo　motor　demonstrated　the　high　accuracy　and　fast　measurement　of　the　proposed　technique,　making　it　attractive　for　condition　monitoring　and　faults　diagnosis　of　rotating　machineries.　©　2017　Elsevier　Ltd