In　this　paper,　a　Morse　signal　automatic　detection　method　is　proposed　base　on　computer　vision.　Firstly,　an　energy　accumulation　preprocessing　method　is　proposed,　and　the　signal　area　is　found　by　nonlinear　transformation.　Secondly,　based　on　the　description　method　of　computer　vision,　different　types　of　signal　are　transformed　into　feature　matrix.　Finally,　the　signal　detection　classifier　is　built　and　trained　based　on　machine　learning.　Performance　of　the　classifier　is　evaluated　and　the　generalization　ability　of　the　classifier　is　proved　by　real-time　data　testing.　©2019.