基于人工神经网络与决策树相结合模型的遥感图像自动分类研究

本文提出了一种新的基于Kohonen神经网络与决策树相结合模型的遥感图像自动分类方法。选取绍兴地区为实验区，对TM图像进行了分类实验。并将该模型分类结果与基于Kohonen网络模型的分类结果进行了比较，发现对于江南低山丘陵河网密集区的TM图像应用该模型进行分类能够得到较为满意的分类结果，其分类精度可达到85．16％，较之单纯使用Kohonen网络模型提高了20．12％。

Study on Model of ANN Combined with Decision Tree Algorithmfor Image Classification

This work carried out a new auto classification model of remote sensing image using Kohonen network combined with decision tree algorithm. Shaoxing county of Zhejiang Province was selected as a case study area. Both model of Kohonen network and model of Kohonen network combined with decision tree algorithm were validated by TM image of study area and accuracy assessments were compared . The results suggested that satisfactory results could be achieved by applying the model of Kohonen network combined with decision tree algorithm to TM image classification of hillside area with concentrated network of waterways at the south of Yangtse River. The accuracy of this trial was 85.16%, much higher than the classification result using Kohonen network alone.