In　this　paper,　the　accuracy　improvement　of　velocity　measurement　of　intelligent　wheelchair　based　on　optical　flow　is　studied.　When　wheelchair　moves　indoor,　mechanical　jittering　occurs　slightly,　which　has　an　impact　on　the　accuracy　of　optical　flow　odometry.　To　solve　the　problem,　a　novel　restoration　method　of　optical　flow　field　is　proposed.　Firstly,　we　select　the　starting　and　end　points　of　optical　flow　vectors　as　the　matched　feature　points,　then　solve　the　initial　affine　kinematic　parameters　using　least　square　method.　In　the　following,　Kalman　Filter　(KF)　is　used　to　smooth　the　parameters　with　futile　shake.　In　the　end,　optical　flow　field　can　be　compensated　to　prepare　for　the　follow-up　optical　flow　odometry.　Our　experiments　demonstrate　the　effectiveness　of　our　method　on　velocity　measurement　of　intelligent　wheelchair.　©　2016　IEEE.