遥感影像K-最近邻图目标分类改进算法的研究

Improved KNN Classification Algorithm Based on K-nearest Neighbor Graphfor Remote Sensing Images

针对高分辨率遥感影像数据中典型目标的判别,提出基于K-最近邻图KNN改进算法的深度学习模型。该模型采用深度学习方法研究目标的属性,充分利用数据之间的关联,建立抗变换性的目标特征,可提高目标判别的准确度。高分辨遥感影像目标检测实验表明该方法的有效性。

In this paper,we proposed an improved deep learning model based on K-nearest neighbor(KNN)algorithm for the typical target discrimination of high-resolution remote sensing images.The proposed model used the deep learning algorithm to study the target attributes,and made full use of the correlation between data to establish the target characteristics that were resistant to transformation target,which could improve the accuracy of target discrimination.Finally,we proved the effectiveness of this method by the experiments of target detection of high-resolution remote sensing images.