Objective:　This　study　proposes　a　method　for　estimating　the　stem　volume　based　on　UAV　images　before　and　after　cutting,　and　provides　reference　for　UAV　remote　sensing　estimation　of　forest　stem　volume.　Method:　Based　on　the　state-owned　forest　of　Jinsen　Forestry　Co.　Ltd.　in　Jiangle　county,　Sanming　city,　Fujian　Province,　the　first　step　in　this　dissertation　was　to　use　unmanned　aerial　vehicle　remote　sensing　to　get　images　whose　resolution　was　more　than　10　cm,　and　got　point　cloud　after　Pix4D　processing.　Based　on　it,　the　point　cloud　of　before　cutting　canopy　was　matched　to　the　point　cloud　of　surface　cloud　after　cutting.　Secondly,　the　forest　canopy　and　the　surface　cloud　were　separated　by　the　cloth　simulation　filtering　algorithm,　the　digital　surface　model(DSM)and　digital　elevation　model(DEM)was　generated　by　natural　neighbour　method,　canopy　height　model(CHM)was　generated　by　the　two　model　subtraction.　Then,　the　tree　height　was　extracted　by　the　improved　local　maximum　algorithm　to　searched　the　top　of　tree　in　canopy　height　model.　Finally,　according　to　the　tree　high　and　diameter　at　breast　height(DBH)of　400　Pinus　massoniana　and　Cunninghamia　lanceolata,　five　DBH　estimation　equation　in　Fujian　Province　were　established.　The　highest　correlation　coefficient　model　was　selected　to　calculate　the　DBH,　then　using　single　wood　produce　volume　formula　in　Fujian　Province　to　estimating　sub-compartment　stem　volume.　Result:　1)　The　matching　of　two-stage　UAV　point　cloud　can　better　eliminate　the　impact　of　large　terrain　slope　on　tree　height　extraction.　2)　The　improved　local　maximum　algorithm　effectively　reduces　the　errors　that　usually　happen　in　the　fixed　window　searching　for　canopy　vertex.　3)　The　estimated　number　of　tree　is　339,　the　measured　number　of　tree　is　366,　the　estimated　average　height　of　stand　is　18　m,　the　measured　average　height　of　stand　is　19　m,　the　estimated　volume　of　sub-compartment　is　182　m3　and　the　measured　volume　is　199　m3,　the　estimation　accuracy　of　the　number　of　tree　average　height　of　stand　and　volume　are　higher.　Conclusion:　With　the　technology　of　UAV　remote　sensing,　automated　estimation　of　forest　stem　volume　can　be　achieved,　thus　greatly　reducing　the　cost　of　traditional　field　investigation　and　promoting　the　rapid　investigation　and　updating　of　the　forest　resources.　©　2019,　Editorial　Department　of　Scientia　Silvae　Sinicae.　All　right　reserved.