基于马氏距离的TM数据森林分类方法研究

THE STUDY ON MAHALANOBIS DISTANCE METHOD IN FOREST CLASSIFICATION BASED ON THE TM DATA

以呼和浩特市土默特左旗白石头沟林场为研究对象,利用ENVI4.7对研究区TM数据进行处理,建立基于马氏距离分类的主要树种的训练样本,并提取训练样本TM4、3、2波段的灰度值作为特征向量,取其均值代入基于两个总体的马氏距离判别公式,建立判别函数判定未知类别。结合研究区的森林资源二类调查数据,进行分类结果检验,得到总体分类精度为67.73%,落叶松林、油松林、白桦林的分类精度分别为39.79%、41.62%、88.17%。

Take the state forest farm in White Stone Valley Tomato Left County Hohhot as the object of study,processing the TM data of study area by using ENVI 4.7,eastablishing the training samples of main tree spices needed to be classified,extracting the gray values of TM4、3、2 band combination as eigenvector,and taking the mean into the two general mahalanobis distance discriminate formulars,thus the decision function is set up to determine the unkonwn category.Then combining with the two kinds of cruising data of study area,the classification result is tested.When training samples are selected at random to be discriminated,the precision is 67.73%,the precision of Larch、Pinus tabulaeformis Carr、Betula platyphylla respectively are 39.79%、41.62%、88.17%.