Research on Cloud Detection System of Multi-Source Satellite Remote Sensing Image Based on Computer Deep Learning Technology

This project intends to integrate SegNet and conditional random field to study SegNet-based cloud detection technology, so as to solve the problems such as noise sensitivity and inaccurate contour extraction of existing cloud detection algorithms. An image cloud recognition method based on multi-view images is proposed. This method is used to improve the image characteristics and algorithm robustness of the cloud cluster to be detected. This method has high accuracy on the premise that the cloud boundary is kept well. The experimental results show that the deep learning network is used for remote sensing target detection, which has good accuracy and robustness, thus providing diversified technical support for the application of domestic high-resolution remote sensing satellites.