对流层散射传播特性及信号谱特性分析

Target Detection of Remote Sensing Images Based on Deep Learning:Method and System Design

为解决传统的目标检测算法难以满足遥感图像数据爆发式增长需求这一问题,文章提出基于深度学习的遥感图像目标检测系统软件。首先,为给深度学习网络训练提供高质量的样本数据,在GIS平台上实现了样本标注功能和数据集兼容性转换功能,并提供图像预处理方法对样本进行扩充;其次,针对遥感图像场景分类与遥感图像特定目标检测,应用深度学习技术,分别实现了模型训练、迁移学习、目标检测等功能;最后,采用了形态学处理、矢量化、直角化约束等方法,对遥感图像场景分类的效果进行改善。实验结果表明,文章的遥感图像目标检测系统在遥感图像场景分类方面取得了85%的分类精度,在特定目标检测方面取得了95%的检测精度,明显优于传统的遥感图像处理方法。该系统软件满足目标检测应用需求,能够为遥感影像分类、信息提取、变化检测等任务提供技术支持。

In order to solve the problem that the traditional target detection algorithm is difficult to meet explosive growth demand of remote sensing image data,a target detection system software based on deep learning for remote sensing images is proposed.Firstly,in order to provide high-quality label data for the training of deep learning network,the functions of sample annotation and compatibility transformation of data set is realized on the platform of GIS,and the image preprocessing method is provided to expand the sample.Secondly,deep learning is used for scene classification and specific target detection of remote sensing images,the functions of model training,migration learning,and target detection is realized.Finally, morphological processing,raster data vectorization,right angle constraint and other methods are used to improve the effect of scene classification of remote sensing images.The experimental results show that the target detection system of remote sensing images in this paper has achieved 85% classification accuracy in scene classification of remote sensing images,and 95% detection accuracy in specific target detection of remote sensing images,obviously better than the traditional remote sensing image processing methods;the system software meet the application requirements of target detection,can provide technical support for tasks of remote sensing images such as image classification,information extraction and change detection.