作者:李 昊 ， 于 虹 ， 饶 桐 ， 周 帅 ， 沈 锋
Authors:LI Hao, YU Hong , RAO Tong , ZHOU Shuai , SHEN Feng摘要:高光谱图像包含丰富的地物信息 ,被广泛应用于许多场合 。 由于各分类模型具有不同的分类性能 ,如 何有效利用各分类模型性能的差异性是实现融合分类的重要环节 ,为此提出了 一 种基于 DS 证据理论的多模型融 合分类的高光谱图像分类方法 。 由于现有的分类模型从 HSI 数据的空间域和光谱域提取不同的特征 , 因此产生的 预测结果不同 。本融合方法采用多层感知机网络和随机森林网络进行融合分类实验 ,该网络借助各分类网络的提 取特征的差异性 ,提高了分类结果的准确性 。实验结果表明 , 当不同网络的分类精度存在 一定差异时 ,DS 融合模 型能提高分类精度 ,同时优于线性平均加权融合模型。
Abstract:Hyperspectral images contain rich information of ground objects and the technology is widely used in many occasions. Since each classification model has different classification performance, effectively utilizing the difference in the performance of each classification model is an essential step to achieve fusion classification. Therefore, a hyperspectral image classification method based on DS evidence theory is proposed. Different prediction results are generated since the existing classification models extract different features from the spatial and spectral domains of HSI data. In this fusion method, a multi-layer perceptron network and random forest network are used for fusion classification experiments. The network improves the accuracy of classification results by taking advantage of the difference in extracted features of each classification network. The experimental results show that the DS fusion model can improve the classification accuracy and is superior to the linear average weithted fusion model when there are some differences in the classification accuracy of different networks.

PDF全文下载地址:

可免费Download/下载PDF全文