基于分类回归树方法的遥感信息快速提取研究

Research on remote sensing information rapid extraction based on classification and regression tree method

遥感信息具有一定的连续变化性,这将会在一定程度上使得遥感信息快速提取存在偏差,其提取的时间也随之增加,容错率下降,为此文中提出基于分类回归树的遥感信息快速提取方法。通过噪声调整的主成分分析法(NAPCA)提取遥感信息的特征,利用复小波变换法对图像进行去噪处理,同时结合邻域值函数完成小波系数收缩。通过分类回归树方法进行样本训练,连续不间断获取遥感信息,结合Bayes判别准则完成遥感信息快速提取。实验结果表明,所提方法能够有效提升容错率,降低遥感信息快速提取偏差和时间。

Remote sensing information has a certain continuous variability,which will cause some deviation in the rapid extraction of remote sensing information to a certain extent.Its extraction time will also increase,and its fault tolerance rate will decrease.Therefore,a method of remote sensing information rapid extraction based on classification and regression tree method is proposed.In the method,the features of remote sensing information are extracted by means of the noise-adjusted principal component analysis(NAPCA),the image is denoised by complex wavelet transform,and the wavelet coefficient shrink is completed by combination of the vicinity threshold function.The classification and regression tree method is used for sample training,and the remote sensing information is continuously and uninterruptedly obtained.In combination with the Bayes discriminant criterion,the rapid extraction of remote sensing information is completed.The experimental results show that the proposed method can effectively improve the fault tolerance rate and reduce the deviation and time of rapid extraction of remote sensing information.