SPOT5遥感数字影像识别技术探讨

Discussion on recognition technology of SOPT 5 remote sensing digital image

利用Erdas Imagine 8.6遥感图像处理软件,对吉林省蛟河林业实验区管理局的SPOT5遥感数字影像进行监督分类,提取地物信息,并对分类后的地物信息采取实地复测方法进行检验。结果表明:主成分变换法适用于阔叶林的识别,精度达到57.14%;比值变换法适用于针叶林的识别,精度达到62.50%;乘积变换法对阔叶林和针叶林的识别正确率都不高;三种融合方法均可应用于混交林的识别,精度均达到100%。

The SOPT 5 remote sensing digital images of the Jiaohe Forestry Experimental Area Administration in Jilin Province were supervised classification according to the extraction of ground feature information by Erdas Imagine 8.6 remote sensing image software. The classification was then tested by field measurements. The results showed： the Principal Component method was suitable to the identification of the broad - leaved forest, with 57.14% accuracy; while the Brovey Transform method fit the identification of coniferous forest, with 62.5% accuracy. However, the Multiplicative method had low preciseness on identification of both broad - leaved forest and coniferous forests. All the three methods could be applied to the identification of mixed forest with precision up to 100%.