Owing　to　the　computing-cost　calculation　of　the　form　factor　and　the　radiation　energy　in　the　traditional　radiosity,　a　new　method　of　parallel　radiosity　Based　on　CUDA　(Compute　Unified　Device　Architecture)　was　proposed　to　improve　the　performance　of　the　implementation.　Specifically,　the　implementation　processes　were　as　follows.　The　detailed　3D　plant　canopy　model　was　loaded　in　the　virtual　scene.　The　solar　altitude　angle,　solar　azimuth　and　the　radiation　intensity　were　estimated　according　to　the　given　time　and　the　geographic　location,　and　the　light　panel　was　generated　afterwards,　which　was　defined　as　radiation　source　on　the　basis　of　uniform　triangular　subdivision.　To　realize　the　parallel　computation　of　light　distribution　within　the　virtual　plant　canopy,　the　3D-discrete　View　Factor　method　was　used　to　calculate　the　view　factor　and　the　corresponding　radiation　energy　based　on　CUDA.　The　experimental　results　show　that　the　presented　algorithm　can　achieve　a　speed-up　ratio　about　20　with　a　considerable　accuracy,　compared　to　the　traditional　method　running　in　the　CPU.　©　2015,　The　Editorial　Board　of　Journal　of　System　Simulation.　All　right　reserved.