It　is　of　great　significance　to　detect　slag　carbon　content　rapidly　and　accurately　for　boiler　combustion　condition　monitoring,　boiler　automated　controlling,　and　reduction　of　energy　consumption　and　emission.　We　present　a　new　method　for　detecting　carbon　content　of　boiler　slag　based　on　machine　vision　technology,　which　is　simple,　fast　and　accurate.　The　slag　images　are　acquired　and　then　the　carbon　content　of　the　slag　is　obtained　according　to　the　relationship　model　established　here.　Then　the　relation　between　slag　image　and　slag　carbon　content　is　constructed　via　the　information　entropy　difference　of　slag　image　and　reference　image.　Experiment　verifies　that　the　average　deviation　of　the　proposed　method　is　less　than　2%,　and　the　method　spend　shorter　detection　time　than　the　traditional　weight　loss　method,　so　the　new　method　meets　the　detection　requirements.　Especially,　the　detection　results　are　unaffected　by　the　illumination　environment,　enabling　real-time　testing　of　the　slag　carbon　content.　©　2018,　Editorial　Office　of　Journal　of　Xi＇an　Jiaotong　University.　All　right　reserved.