Crop　planting　structure　plays　an　important　role　in　crop　spatial　pattern,　and　it　is　the　foundation　for　the　optimal　allocation　of　regional　land　and　water　resources.In　this　paper,　taking　the　Kai-Kong　River　Basin　agricultural　area　in　Xinjiang　as　the　study　area,　a　method　for　crop　classification　that　comprehensively　utilizes　high　temporal　resolution　MODIS　images　and　high　spatial　resolution　Landsat　images　was　proposed.Due　to　large　scope　of　study　area,　it　is　difficult　to　obtain　crop　sample　points　uniformly　throughout　whole　research　area　with　traffic　and　time　constraints.MODIS　and　Landsat　data　for　2016,　combined　with　crop　phenology　data,　were　used　to　construct　experimental　sample　points;　thereby　providing　a　better　solution　for　crop　extraction　in　this　area　wherein　sample　access　is　difficult.NDVI　time-series　curves　for　different　crops　were　constructed　based　on　experimental　sample　points.Based　on　the　NDVI　time-series　curves,　the　critical　period　of　crops　during　growing　season　were　obtained.For　these　key　periods,　Landsat　8　OLI　images　were　selected.Next,　extraction　knowledge　rules　for　the　main　crops　were　constructed,　and　identification　and　classification　of　crops　were　performed　based　on　decision　tree.In　2016,　main　crop　planting　area　was　5.07×105　hm2　of　the　Kai-Kong　River　Basin　agricultural　region,　with　largest　planting　area　found　for　cotton　(1.97×105　hm2),　followed　by　those　for　corn　and　wheat.Bosten　Lake　and　Kaidu　River　agricultural　area　was　dominated　by　pepper,　corn,　and　wheat,　and　the　planting　structure　was　relatively　scattered.Planting　structure　of　the　Peacock　River　agricultural　area　was　relatively　simple,　with　cotton　and　pear　as　main　crops.A　comparative　experiment　based　on　time-series　MODIS　images　for　crop　recognition　and　classification　was　also　conducted.Results　were　verified　and　compared　with　sample　points　of　field　survey.The　accuracy　of　crop　classification　using　MODIS　and　Landsat　data　was　obviously　improved　as　compared　with　the　accuracy　of　crop　classification　using　only　time-series　MODIS　data.Overall　classification　accuracy　increased　from　62.58%　to　88.37%,　and　kappa　coefficient　increased　from　0.53　to　0.86.The　use　of　high　temporal　resolution　MODIS　data　and　high　spatial　resolution　Landsat　data　can　improve　the　accuracy　of　crop　extraction　to　a　certain　extent;　this　avoided　(1)　poor　classification　accuracy　caused　by　the　insufficient　spatial　resolution　of　MODIS　data　and　(2)　phase　selection　blindness　or　data　redundancy　caused　by　the　insufficient　resolution　of　Landsat　data.Therefore,　this　approach　has　high　potential　application　value　in　the　extraction　of　crop　planting　structure　in　arid　areas.　©　2020,　Science　Press.　All　right　reserved.