摘要
为提高遥感图像的清晰程度,研究基于迁移学习的遥感测绘图像细节增强方法。定义邻域边界矩阵,计算细节损失梯度坐标,提取边界函数,重构遥感测绘图像边界;建立转换数据幅值函数,计算拉伸函数不同方向的算子,基于迁移学习校正像素参数;通入低通滤波,计算光照参数,代入光照格式函数,在拉伸处理后获取阴影部分细节纹理。实验结果显示本文的增强方法平均梯度、信息熵、均值均最大。在图像细节增强的实例检测中,可以清晰地看出该图像增强方法较好,得到的图像更清晰。
In order to improve the clarity of remote sensing images,the method of detail enhancement of remote sensing mapping images based on transfer learning is studied.Define neighborhood boundary matrix,calculate detail loss gradient coordinates,extract boundary function,and reconstruct remote sensing mapping image boundary;Establish the amplitude function of the converted data,calculate the operators in different directions of the stretching function,and correct pixel parameters based on migration learning;Introducing low-pass filter,calculating illumination parameters,substituting illumination format function,and obtaining detailed texture of shadow part after stretching processing.The experimental results show that the average gradient,information entropy and average value of this method are the largest.In the example detection of image detail enhancement,it can be clearly seen that the image enhancement method is better and the obtained image is clearer.
作者
李德伟
LI De-wei(Huaiyang District Housing and Urban Rural Development Bureau,Zhoukou 466700 China)
出处
《自动化技术与应用》
2023年第3期67-71,共5页
Techniques of Automation and Applications
关键词
迁移学习
遥感测绘技术
遥感图像
图像增强算法
transfer learning
remote sensing mapping technology
remote sensing images
image enhancement algorithm