论文部分内容阅读
在遥感影像计算机自动识别与分类中,选取最佳的波段子集对地物进行分类对提高分类精度至关重要。根据统计学原理,遥感影像中属于某类别地物的特征向量服从正态分布,训练样本的正态性检验是关键,基于此理论,本文利用TM影像数据,通过检验所选取的训练区的正态性与否,让计算机自动的选取最优的波段组合,并对分类的精度进行评估。研究表明,计算机自动选取最佳波段组合后对分类精度的预先评估,较常规分类后再进行数据检验精度评估方法方便,快捷,省时,省力。
In the automatic recognition and classification of remote sensing image computer, it is very important to select the best band subsets to classify the objects to improve the classification accuracy. According to the principle of statistics, the eigenvectors belonging to a certain category in the remote sensing image obey the normal distribution and the normality test of the training samples is the key. Based on this theory, the TM image data is used to test the positive State or not, so that the computer automatically select the optimal band combination, and to assess the classification accuracy. The research shows that the computer automatically selects the best band before the classification accuracy of the pre-assessment, compared with the conventional classification of data accuracy test method is convenient, fast, time-saving and labor-saving.