基于卷积神经网络的遥感图像语义分割方法研究

Semantic Segmentation of Remote Sensing Image Based on Convolutional Neural Network

高分辨率遥感图像大部分情况下包含相对来说较为复杂的语义信息以及容易混淆的目标,对高分辨率遥感图像进行语义分割是一项很重要并且具有挑战性的任务。近几年来,深度卷积神经网络(Deep Convolutional Neural Network, DCNN)为代表并结合条件随机场(Conditional Random Field, CRF)的算法在图像分割领域中有着杰出的表现。本文基于DeepLap V3+网络结构,结合DCNN,设计出了一种针对高分辨率遥感图像的语义分割网络,仿真实验结果验证了该方法的有效性和鲁棒性。

In most cases, high-resolution remote sensing images contain relatively complex semantic information and easily confused targets. Semantic segmentation of high-resolution remote sensing images is a very important and challenging task. In recent years, deep convolutional neural network (DCNN) as a representative and combined with Conditional Random Field (CRF) algorithm has out-standing performance in the field of image segmentation. Based on the DeepLap V3+ network structure and combined with the DCNN, this paper designs a semantic segmentation network for high-resolution remote sensing images. The results of simulation experiments verify the effectiveness and robustness of the method.