To　a　mobile　robot　vision　system,　it　is　of　vital　importance　to　separate　objects　from　the　background.　This　needs　us　firstly　to　extract　edges　from　the　images　collected　by　the　mobile　robot.　It　is　known　that　traditional　edge　detection　algorithms　are　sensitive　to　noise　and　insufficient　for　complex　edge　feature.　For　the　corridor　image　of　a　building,　its　signal-to-noise　ratio　is　low　and　the　edge　complex.　If　we　use　traditional　edge　detection　algorithms　to　dispose　it　directly,　the　result　is　not　expected.　In　order　to　solve　these　problems,　a　novel　multi-scale　multi-structure　element　edge　detection　algorithm　based　on　mathematical　morphology　is　proposed.　Experiments　with　corridor　images　of　a　teaching　building　prove　that　the　proposed　algorithm　can　preserve　the　detailed　edge　information　in　the　original　image　while　smoothing　it　simultaneously.　Hence,　the　proposed　algorithm　can　meet　the　needs　of　robot　vision　system.