A　novel　stereo　matching　algorithm　for　a　binocular　mobile　robot　is　presented,　which　is　based　on　image　segmentation　and　reliable　point　correspondences.　Firstly,　less　and　similar　texture　regions　of　stereo　images　are　determined　by　the　segmentation　of　the　right　and　left　image　using　the　texture　features　verifying　and　segmentation;　Then,　the　disparity　of　reliable　point　could　be　obtained　by　SIFT　features　point　matching.　Secondly,　the　preliminary　disparity　could　be　obtained　by　SAD　area-based　matching.　Finally,　the　image　is　processed　according　to　the　distribution　of　SIFT　features　matching　point,　disparity　continuous　constraint　and　minimum　distance　classifier.　Meanwhile,　the　higher　precision　disparity　map　could　be　obtained　by　removing　the　smaller　region　via　the　blob　filter.　It　can　be　drawn　from　the　results　of　experiment　that　the　proposed　algorithm　is　effective,　and　could　obtain　quicker,　accurate　obstacle　detection.