训练样本对TM尺度小麦种植面积测量精度影响研究(Ⅱ)——样本质量对小麦测量精度的影响分析

本文以小麦为例,在构建标准训练样本数据集的基础上,引入欧式距离、马氏距离和纯度三个指标对训练样本数据进行质量评价和分等定级。然后将相同样本数量,但不同质量等级的训练样本输入到最小距离、最大似然和支持向量机三种分类器中进行影像分类和冬小麦提取,并对小麦识别结果进行分析和比较。研究结果显示:同一分类器和不同分类器对相同指标划分的不同质量等级的训练样本响应不同,对不同指标划分的相同质量等级的训练样本响应也不同;相同分类器,面积误差最小和总体像元精度最高所对应的训练样本的质量等级不一样。

The effects of training samples on the wheat planting area surveying accuracy in TM scale(II): The accuracy response of different classifiers to samples quality

A standard sample set was constructed.Then three indexes(Euclidean distance,Mahalanobis Distance and Pure index) were introduced to evaluate training samples quality.And training samples with same amount and different quality were respectively input into three different classifiers to classify image and identify wheat,which are minimum distance,maximum likelihood and support vector machine.The results indicate: The response of same classifier to different sample quality level is different.The response of different classifier to same sample quality level is also different.As concern of one classifier,the minimal area measurement error and the highest overall accuracy can't be gotten at same quality lever.