In　order　to　solve　the　problems　of　poor　robustness,　bad　adaptability　and　low　precision　in　motion　estimation　for　mobile　robot　by　using　single　camera　based　on　flat　ground　assumption,　a　pseudo-omnidirectional　vision　system　and　its　components　are　introduced　firstly,　and　a　novel　calibration　method　for　ground-parallel　parameters　of　camera　based　on　twostep　movement　of　robot　is　presented　considering　the　characteristics　of　the　vision　system.　Then　a　fusion　estimation　algorithm　based　on　pseudo-omnidirectional　vision　for　autonomous　mobile　robot　self-motion　is　proposed.　According　to　such　factors　as　robot　nonholonomic　constraints,　flat　ground　assumption　and　consistence　measure　among　motion　estimation　parameters,　the　algorithm　can　synthetically　evaluate　the　performance　of　each　visual　estimation　in　the　pseudo-omnidirectional　vision　system.　And　based　on　the　estimation　result,　the　motion　estimation　parameters　that　are　more　reliable　and　more　robust　can　be　fused　and　determined　finally.　The　experimental　results　validate　the　proposed　method　in　the　aspects　of　robustness,　precision　and　real-time　property.