Remote sensing technology is an effective tool for sensing the earth’s surface. With the continuous improvement of remote sensing technology, remote sensing detectors can obtain more spectral and spatial information, including clear feature contours, complex texture features and spatial layout rules. This information was detected in mineral resources. Surface substance identification, water pollution information monitoring and many other aspects have played an important role. The coding algorithm and defects, storage algorithm and interference from atmospheric cloud radiation information during the imaging process lead to varying degrees of distortion and deterioration of remote sensing images during imaging, transmission and storage. This makes it difficult to process, analyze and apply remote sensing images. Therefore, the design of a reasonable remote sensing image quality evaluation method is not only conducive to the remote sensing image quality evaluation in the real-time processing system of remote sensing image, but also conducive to the optimization of remote sensing image system and image processing algorithm. The application is worthwhile. In this paper, the deteriorating features of remote sensing images will change the statistical distribution. We propose a method for evaluating the quality of remote sensing images in depth learning. Feature learning and blurring as well as noise intensity classification for image remote sensing using convolutional neural network are carried out. The evaluation model is modified by masking effect and perceptual weighting factor, and the quality evaluation results of remote sensing images are obtained according to human vision. The research shows that this method can effectively solve the problem of removing and evaluating the noise of remote sensing image, and can effectively and accurately evaluate the quality of remote sensing image. It is also consistent with subjective assessment and human perception.

中文翻译：

---

基于深度学习的遥感图像质量评价

遥感技术是感知地球表面的有效工具。随着遥感技术的不断进步，遥感探测器可以获得更多的光谱和空间信息，包括清晰的特征轮廓、复杂的纹理特征和空间布局规则。这些信息是在矿产资源中检测到的。地表物质识别、水污染信息监测等诸多方面发挥了重要作用。成像过程中的编码算法和缺陷、存储算法和大气云辐射信息的干扰导致遥感图像在成像、传输和存储过程中出现不同程度的失真和劣化。这使得处理、分析和应用遥感图像变得困难。所以，设计合理的遥感影像质量评价方法，不仅有利于遥感影像实时处理系统中的遥感影像质量评价，而且有利于遥感影像系统和影像处理算法的优化。申请是值得的。在本文中，遥感图像的恶化特征将改变统计分布。我们提出了一种在深度学习中评估遥感图像质量的方法。使用卷积神经网络对图像遥感进行特征学习和模糊以及噪声强度分类。评估模型通过掩蔽效应和感知权重因子进行修改，遥感图像质量评价结果是根据人的视觉得到的。研究表明，该方法能够有效解决遥感图像噪声的去除和评价问题，能够有效准确地评价遥感图像的质量。这也符合主观评价和人类感知。