【摘要】 [目的]为农作物产量预报提供技术支持。[方法]通过分析安徽省小麦生长季MODIS多时相NDVI特征并利用该技术识别水体、城镇居民区、旱地、水田、林地研究了该技术在地物识别及农作物产量预报中的应用。[结果]一季稻和双季晚稻的NDVI最大峰值出现在6月。水体在10月至次年6月NDVI(i,j)小于0或略大于0(如浅水体),而NDVI(i,j)小于0.16像元通常为城镇等居民点,水田NDVI(i,j)往往小于0.3,一般林地和种植越冬作物的耕地NDVI(i,j)均大于0.3。以NDVI多时相特征对5大地物类型进行识别时可识别出82.7%林地,67.4%旱地,52.1%水田,15.2%居民区。水体、旱地、水田、林地的识别可信度均达60%以上,其中识别出来的水体中有近80%为现实中的水体,城镇的识别可信度不足20%。[结论]该法在农作物产量预报中具有良好的应用前景。更多还原

【Abstract】 [Objective] The study aimed to supply technical support for the prediction of crop yield.[Method] Through analyzing the multi-temporal features of MODIS/NDVI in the growing season of wheat in Anhui Province and recognizing forest land,townish residential area,upland field,paddy field and woodland by using this technique,its application in ground-object identification and prediction of crop yield was studied.[Result] The maximum peak values of NDVI of single-crop rice and late double-crop rice appeared in Jun.The NDVI(i,j) of water body was smaller or a little bigger(such as shallow water body) than 0 from Oct.to next Jun.The pixels with NDVI(i,j) below 0.16 were residential areas usually,such as town.The NDVI(i,j) of paddy field was often smaller than 0.3,but that of general woodland and cultivated land of overwinter crops were bigger than 0.3.When the 5 ground-object styles were recognized by using the multi-temporal features of NDVI,the recognized rates of forest land,upland field,paddy field and residential area were 82.7%,67.4%,52.1% and 15.2%.All the recognition confidences of water body,upland field,paddy field and woodland were above 60%,among them that of water body was close to 80% of realistic water body,and that of town was less than 20%.[Conclusion] This method had good application prospect in the prediction of crop yield.更多还原