基于编码-解码神经网络遥感图像语义分割应用研究

图像分割的实现经历了从传统方法到神经网络方法的演变。本文从图像分割的发展过程入手,介绍了图像分割与语义分割的区别,对最近几年传统图像分割方法在遥感图像分割领域的应用进行梳理分析,总结了传统遥感图像分割方法的不足。基于此,归纳了几种经典编码-解码神经网络架构在遥感图像语义分割领域的应用,对其改进方式进行了综合性分析,并对其未来的发展趋势进行展望。

Remote sensing image semantic segmentation based on coding-decoding neural network

The realization of image segmentation has experienced the evolution from traditional methods to neural network methods.This article starts with the development process of image segmentation,introduces the difference between image segmentation and semantic segmentation,analyzes the application of traditional image segmentation methods in the field of remote sensing image segmentation in recent years,and discusses the shortcomings of traditional remote sensing image segmentation methods.Based on this,the application of several classical encoding-decoding neural network architectures is summarized in the field of remote sensing image semantic segmentation,the improvement methods of these methods are comprehensively analyzed,and their future development trend is prospected.