基于机器学习的倾斜摄影影像质量改善方法

Quality Improvement Method of Oblique Photogrammetry Image Based on Machine Learning

针对倾斜摄影影像常出现的质量差、清晰度低等问题,提出了一种基于机器学习的倾斜摄影影像质量改善方法。该方法通过生成式对抗网络模型训练样本,并从颜色、纹理、内容3个方面建立损失函数以保证模型的训练质量,使得倾斜摄影影像的色调、清晰度能得到高效精确的调整,将该方法运用于广西某地的基础测绘作业中,取得了相较于传统方法更好的质量改善效果,证明该方法具有良好的推广和应用价值。

Aiming at the problems of poor quality and low definition of oblique photogrammetry,we proposed a quality improvement method of oblique photogrammetry based on machine learning.In this method,the training samples of the generative adversarial networks model were used,and the loss function was established from the three aspects of color,texture and content to ensure the training quality of the model,so that the tone and definition of the oblique photogrammetry could be adjusted efficiently and accurately.We applied this method to the basic surveying and mapping operation in a certain place in Guangxi,and obtained a better quality improvement effect compared with the traditional method,which could prove the method has good popularization and application value.