Unmanned　Aerial　Vehicle　(UAV)　Remote　Sensing　has　great　advantages　over　traditional　methods,　such　as　lower　cost,　simpler　operation,　faster　access　speed　and　higher　resolution.　In　this　paper,　after　analyzing　the　spectral　characteristics　of　vegetation　and　non-vegetation　in　UAV　images,　which　only　contains　red,　green,　and　blue　bands,　we　found　that　the　vegetation　spectral　had　the　feature　of　green　band＞red　band＞blue　bands,　which　means　vegetation　had　the　biggest　reflection　in　the　green　band　and　had　the　smallest　reflection　in　the　blue　band.　However,　non-vegetation　region　had　the　reflection　feature　of　red　band＞green　band＞blue　band　and　blue　band＞green　band＞red　band.　The　pixels　value　of　the　vegetation　region　was　smaller　than　the　non-vegetation　region.　For　overall　consideration　of　the　above　characteristics　and　the　features　of　the　healthy　green　vegetation＇s　spectral　profile,　and　in　order　to　enhance　the　vegetation　information　and　minimize　the　vegetation　signal,　we　referenced　the　form　of　NDVI　and　put　forward　a　new　vegetation　index--VDVI　(visible-band　difference　vegetation　index).　Then　we　calculated　the　vegetation　index　of　VDVI,　EXG,　NGRDI,　NGBDI,　and　RGRI.　After　calculation　of　the　vegetation　index,　we　used　the　same　AOI　region　of　the　prior　analysis　in　a　typical　spectral　characteristic　and　we　made　the　line　charts　to　analyze　the　feasibility　of　each　index.　After　observation　of　the　line　charts,　we　found　that　NGRDI　was　not　suitable　to　extract　the　vegetation　from　a　UAV　image　because　the　index　values　were　overlapping　with　each　other,　except　for　bare　soil.　On　the　contrary,　the　NGBDI　and　VDVI　was　suitable　for　the　extraction　of　vegetation　from　the　image,　because　there　was　little　overlapping.　And　the　vegetation＇s　EXG　index　value　was　greater　than　twenty　and　the　value　of　non-vegetation　was　lesser　than　twenty　except　for　the　area　of　a　building.　However,　there　was　also　some　overlapping　of　a　building　and　the　field,　which　could　cause　some　mistake　in　the　extraction　of　the　result.　After　that,　we　determined　the　threshold　value　of　each　index.　The　threshold　is　most　important　in　the　extraction　of　vegetation　from　an　image,　but　the　value　is　very　difficult　to　determine.　After　observation　of　the　histogram　of　the　vegetation　index,　we　decided　to　use　the　bimodal　histogram　and　the　histogram　entropy　threshold　method　to　determine　the　threshold　value　of　each　vegetation　index,　and　compared　each　extraction　result　of　the　two　methods　and　chose　the　value　which　had　the　higher　extraction　accuracy　and　used　it　as　the　final　value　to　extract　the　vegetation　from　the　UAV　image.　After　comparison　of　the　VDVI　extraction　results　with　EXG,　NGRDI,　NGBDI,　and　RGRI,　we　found　that　whatever　method　was　used　to　determine　threshold　value,　the　accuracy　of　VDVI　extraction　result　was　always　most　precise,　and　the　EXG　must　use　the　histogram　entropy　threshold　method　to　determine　the　threshold.　It　can　also　extract　the　most　vegetation　information.　The　NGRDI　and　RGRI　have　bad　extraction　results.　In　conclusion,　in　the　VDVI　overall　consideration　of　the　vegetation　characteristic　of　reflection　in　the　green　band　and　the　absorbent　in　the　red　and　blue　bands,　and　its　value　ranged　from-1　to　1.　And　the　extraction　accuracy　of　VDVI　is　higher　than　the　others,　and　its　threshold　is　easier　to　determine　which　is　near　to　zero.　In　order　to　verify　that　the　VDVI　was　suitable　for　another　UAV　image,　we　chose　another　UAV　image　to　calculate　its　VDVI　and　extracted　vegetation　from　it.　Then　we　generated　200　random　points　to　evaluate　the　vegetation　extraction　accuracy　from　the　UAV　image.　From　an　accuracy　report,　we　found　the　total　extraction　precision　was　91.5%,　and　the　kappa　coefficient　was　0.8256.　We　concluded　that　the　VDVI　was　broadly　suitable　to　extract　vegetation　from　N　UAV　image　which　only　contains　a　visible　band.　©,　2015,　Chinese　Society　of　Agricultural　Engineering.　All　right　reserved.