基于多种分类器组合的森林类型信息提取技术研究

Investigation of Forest Types Extraction Technology Using Multiple Classifiers Combination

遥感图像分类是遥感应用系统中的关键技术,提高遥感图像的分类精度是发展遥感技术重点,采用多分类器组合的算法对黑龙江塔河县森林类型进行分类。根据黑龙江省森林资源调查技术规定及研究区二类调查数据制定分类系统,最终的分类级别为针叶林、阔叶林和针阔混交林。通过分析TM数据的原始波段和NDVI、BI等植被指数提取各分类类型的光谱特征。选择最小距离法、最大似然法、马氏距离法对研究区进行分类,计算出各分类器的精度。在分类器组合的过程中采用信息熵方法确定组合分类器中各分类器的权重系数,利用组合后新的分类器对研究区进行分类。结果表明:多分类器组合的分类精度达75.59%,比单分类器精度提高了3.85%,对阔叶林、针阔混交林、针叶林3种分类类型的分类精度分别达82.32%、66.45%、75.49%,比单分类器进度分别提高了2.87%、4.82%。4.10%。

Remote sensing image classification is a key technology in remote sensing applications, and the improvement on the accuracy of remote sensing image classification is emphasized for remote sensing technology. The forest types in Tahe county of Hei- longjiang province were classified based on multiple classifiers combination. The classification system was designed based on the forest resource inventory technical regulations of Heilongjiang province and inventory data for study area, and the classifications include co- niferous forest, broadleaf forest, coniferous and broadleaf mixed forest level. Through the analysis on the TM data band and NDVI, BI vegetation indices, the spectral characteristics of each type of forest were extracted. The minimum distance, maximum likelihood, and the mahalanobis distance method were used to classify the forest types. The weighting factor of the classification was determined by the entry weight method based on the classification accuracy, and combined a new classifier for classification. The experiment results showed that the accuracy of multiple classifier combination achieved 75.59%, a increase of 3.85% compared with the single classifier accuracy. The classification accuracy of the three forest types were as follows： broadleaf forest 82. 32%, coniferous forest 66.45%, coniferous and broadleaf mixed forest 75.49%, respectively, which increased by 2. 87%, 4. 82%, 4. 10%, respectively, com- pared with the single classifier accuracy.