摘要
不同分辨率的遥感数据对农作物识别和面积估算精度的敏感程度是目前研究的热点。本文利用20m分辨率CBERS-02BCCD数据,通过构建不同分辨率遥感影像序列,采用最大似然、支持向量机、人工神经网络3种分类器,分别从像元和区域尺度讨论了空间分辨率对分类和面积估算精度的影响;同时,还进行了作物种植成数和聚集度的影响分析。结果表明,随着空间分辨率的降低,分类精度和面积估算精度均呈下降趋势;作物种植成数越高,作物种植越密集,面积估算精度随分辨率的降低下降速率越慢;对于分布十分密集的作物,当分辨率高于80m时,聚集度对其精度结果起主导作用。最后,针对不同作物种植结构区,提供了要满足一定的区域精度所需要的遥感数据和最佳分类方法。
It is necessary and valuable to study the effect of spatial resolution on crop identification and crop acreage estimation.In this paper,remote sensing images of different spatial resolution were generated from CBERS-02BCCD satellite data with 20mspatial resolution,then several crop identification tests were carried out using classifiers including(a)maximum likelihood,(b)support vector machines,(c)artificial neural network.At last,pixel and regional accuracy assessment were implemented to evaluate the impacts of spatial resolution on crop identification and acreage estimation accuracy.The following were concluded:with the decrease of spatial resolution,the overall accuracy and regional accuracy will decrease due to the increasing of mixed pixels;the higher of the crop proportion and the lower of the aggregation index,the slower decrease of the regional accuracy.The analytic results provide academic and experimental reference to resolve the problem of data and classifier selection in crop acreage estimation by remote sensing.
出处
《遥感信息》
CSCD
2014年第2期36-40,46,共6页
Remote Sensing Information
基金
国家自然科学基金(41071277)
中国科学院重点布署项目(KZZD-EW-08-05)
关键词
空间分辨率
作物识别
种植面积估算
精度
spatial resolution
crop identification
acreage estimation
accuracy
