Improvement　of　the　extraction　means　of　the　ORB　(Oriented　FAST　and　Rotated　BRIEF)　feature　primarily　includes　optimization　concerning　excessive　aggregation　of　ORB　features　and　the　improvement　of　the　problem　that　the　correct　features　could　not　be　extracted　when　regional　image　illumination　is　too　bright.　First,　the　local　self-adaptive　threshold　was　calculated　on　the　basis　of　the　threshold　and　FAST　features　were　extracted　based　on　the　local　threshold　as　the　candidate　feature　points.　Then,　image　iteration　was　divided　into　disparate　regions　and　the　optimal　feature　points　of　the　local　region　were　selected　as　the　extraction　result.　The　experimental　data　showed　that　the　aggregation　level　of　the　improved　ORB　feature　lowered　more　obviously　than　that　of　the　ORB　feature,　which　adapted　to　local　illumination　and　the　threshold　value　with　high　stability;　however,　the　time　still　met　real-time　demands.