论文部分内容阅读
精确的农作物分类信息对于农业环境评估、水资源利用规划非常重要,尤其是在干旱、半干旱地区。本文利用30 m分辨率的Landsat NDVI时间序列数据进行了新疆石河子垦区混合农作物精确区分的潜力研究。首先利用S-G滤波重构了Landsat NDVI时间序列,然后基于SVM模型对研究区域农业类型进行了精确分类。在SVM分类模型作用下,S-G重构后的时间序列有效地将该地区棉花、玉米、小麦等主要作物区分开来,精度高于0.86,Kappa系数大于0.82。结果表明,S-G滤波能够有效提高NDVI时间序列数据质量;TM影像时间序列在监测干旱、半干旱地区的作物类型和种植方式随时间的变化方面存在巨大潜力。
Accurate crop classification information is important for agricultural environmental assessment and water resource planning, especially in arid and semi-arid areas. In this paper, using the Landsat NDVI time-series data with a resolution of 30 m, the potential of hybrid crops in Shihezi reclamation area in Xinjiang was studied. Firstly, the Landsat NDVI time series was reconstructed by S-G filtering, and then the agricultural types in the study area were classified accurately based on the SVM model. Under the SVM classification model, the S-G reconstructed time series can effectively distinguish the main crops such as cotton, corn and wheat in this area. The accuracy is higher than 0.86 and the Kappa coefficient is greater than 0.82. The results show that S-G filtering can effectively improve the quality of NDVI time series data. TM image time series have great potential in monitoring the changes of crop types and planting patterns over time in arid and semiarid regions.