基于BKNNSVM算法的高分辨率遥感图像分类研究（英文）

发布时间：2018-04-11 10:14

  本文选题：高分辨率遥感图像分类 + KNNSVM算法　； 参考：《中南民族大学学报(自然科学版)》2016年01期

【摘要】：为了解决局部支持向量机算法KNNSVM存在的分类时间过长不利于具有海量数据量的高分辨率遥感图像分类的不足,提高KNNSVM的算法表现,提出了改进的基于不确定性的BKNNSVM算法.该算法利用二项式分布的共轭先验分布Beta分布根据近邻的分布情况推导该未标记样本属于正类或负类的概率大小,从而计算每一个未标记样本在类属性上的不确定性大小.再通过设置不确定性阈值的大小,对不确定性低于阈值的未标记样本直接采用KNN进行分类,而对高于阈值的样本利用其近邻建立局部支持向量机分类器进行分类.对高分辨率图像分类的实验结果表明:合适的阈值能够有效降低原始KNNSVM算法的时间开销,同时能保持KNNSVM分类精度高的特点.
[Abstract]:In order to solve the problem that the classification time of local support vector machine (KNNSVM) algorithm is too long which is not conducive to the classification of high-resolution remote sensing images with large amount of data, and to improve the performance of KNNSVM algorithm, an improved BKNNSVM algorithm based on uncertainty is proposed.By using the conjugate priori Beta distribution of binomial distribution, the probability of the unlabeled sample belonging to a positive or negative class is derived according to the distribution of the nearest neighbor, so as to calculate the uncertainty of each unlabeled sample in class attributes.By setting the uncertainty threshold, the unlabeled samples with uncertainty below the threshold are classified directly by KNN, and the samples above the threshold are classified by local support vector machine (SVM) classifier using their nearest neighbors.The experimental results for high-resolution image classification show that the appropriate threshold can effectively reduce the time cost of the original KNNSVM algorithm and maintain the high accuracy of KNNSVM classification.
【作者单位】： 中南民族大学电信学院;中国地质大学地球物理与空间信息学院;
【基金】：湖北省自然科学基金资助项目(PBZY14019)
【分类号】：TP751
，

本文编号：1735550