基于改进DeeplabV3+模型的云检测

Cloud Detection Based on Improved DeeplabV3+Model

国产卫星影像数量的快速增长对国产影像的质量控制的精度和效率提出更高的要求,而云检测是遥感影像质量检测的首要问题。针对现有云检测的深度学习模型存在误判、漏判和训练需要花费大量时间的问题,研制一套云检测算法具有重要意义。文章提出了一种基于改进DeeplabV3+模型的云检测方法,通过对Xception网络和空间金字塔池化模块(atrous spatial pyramid pooling,ASPP)进行改进,并加入迁移学习,进而提高模型的精度和效率。分析结果表明,该改进的云检测模型与传统的DeeplabV3+模型相比,准确率提高了3.34%,精确率提高了3.78%,召回率提高了4.47%,平均交并比提高了5.39%,且训练时长和预测时长也有明显的减少。

With the rapid growth of the number of images from domestic satellites,higher requirements are put forward for the accuracy and efficiency of the quality control of domestic images.Cloud detection is the primary problem of remote sensing image quality detection.In view of the problems that the existing depth learning model of cloud detection has errors,omissions,and time-consuming training,it is of great significance to develop a set of cloud detection algorithms.This paper proposes a cloud detection method based on the improved DeeplabV3+model,which improves the accuracy and efficiency of the model by improving the Xception network and atrous spatial pyramid pooling(ASPP)module,and adding migration learning.The analysis results show that compared with the traditional DeeplabV3+model,the improved cloud detection model improves the accuracy rate by 3.34%,the precision rate by 3.78%,the recall rate by 4.47%,and the average crossover/merge ratio by 5.39%,respectively,and the training time and prediction time are also significantly reduced.