The　measurement　of　instantaneous　rotational　speed　for　rotating　machinery　is　essential　for　ensuring　the　operational　stability,　enabling　predictive　maintenance,　and　guaranteeing　the　safe　operation　of　rotating　equipment.　Encoders,　as　a　common　measurement　method,　face　challenges　such　as　inconvenient　installation　and　load　effects　on　the　shaft.　To　tackle　these　issues,　a　novel　vision-based　measurement　method　for　instantaneous　rotational　speed　(IRS)　of　the　shaft　using　projection　fringe　was　proposed.　The　proposed　method　only　requires　processing　a　single　row　data,　thus　offering　advantages　in　measurement　efficiency　and　data　volume.　A　pattern　with　sine　white　edge　was　firstly　preset　to　circumferential　surface　of　the　shaft.　Then　a　constant　density　fringe　from　the　projection　lamp　was　projected　onto　the　white　pattern,　in　which　the　fringe　center　will　be　sine-modulated　by　the　white　pattern　during　rotation　of　the　shaft.　Cross-correlation　was　utilized　to　track　the　coordinate　variation　of　fringe　center,　thereby　obtaining　its　variation　curve.　An　improved　short-time　Fourier　transform　with　adaptive　window　length　was　proposed　to　accurately　estimate　rotational　frequency　from　fringe　center　variation　curve,　enabling　the　determination　of　shaft　IRS.　In　this　study,　a　system　simulation　model　was　established　to　analyze　the　various　factors　influencing　its　performance.　Then　the　effectiveness　of　proposed　system　and　algorithm　was　verified　through　experiments,　proving　that　its　measurement　accuracy　was　comparable　to　that　of　encoders.　Constant　speed　measurement　of　300　rpm　showed　that　the　average　speed　error　was　0.99　rpm　with　a　relative　error　of　0.33%.　Linear-varied　speed　measurement　revealed　that　the　average　speed　error　was　3.01　rpm　with　a　relative　error　of　0.48%.　Sine-varied　speed　measurement　indicated　that　the　average　speed　error　was　5.80　rpm　with　a　relative　error　of　0.89%.　Consequently,　the　proposed　system　offers　a　promising　and　reliable　avenue　for　IRS　measurement　of　rotating　equipment.　©　2001-2012　IEEE.