【作者】 赵泉华； 宋伟东； 鲍勇；

【Author】 Zhao Quanhua, Song Weidong, Bao Yong(School of Geomatics,Liaoning Technical University, Fuxin 123000,China)

【机构】 辽宁工程技术大学测绘与地理科学学院；

【摘要】 为了解决快速、高精度的遥感影像分类问题,本文在MATLAB平台基础上,提取离散分形布朗模型的水平分维、垂直分维、对角分维、综合分维四种分形纹理特征,采用BP神经网络对遥感图像进行监督分类,首先将提取好的特征作为样本矩阵送入BP网络进行训练,然后用训练好的网络对待分类遥感影像进行分类,最后采用混淆矩阵法对分类结果进行精度评定。试验表明采用分形纹理特征的神经网络方法进行遥感影像分类具有运算速度快、分类精度高的特点。更多还原

【Abstract】 In order to improve the speed and precision of the classification of remote sensing image, horizontal fractal dimension, vertical fractal dimension, diagonal fractal dimension and integrated fractal dimension of discrete dractal brownian model were extracted from the selected samples. the article adopts BP neutral network to finish the supervisory classification based on MATLAB. Firstly the extracted features are delivered as sample matrix to the BP neural network classifier. Then we use the trained BP neural network to realize the remote sensing image classification. At last, we make an accuracy assessment by confusion matrix. The results show that the classification method based on BP neural network is a convenient and high precision method for image classification更多还原