Landslide identification method and system based on deep learning

The invention discloses a landslide identification method based on deep learning, and the method comprises the steps: S100, carrying out the feature extraction of a landslide image, and enabling the extracted features to be a feature map; s200, processing the feature map through an RPN detection network to obtain a recommended target candidate region; and S300, performing feature extraction on the feature map of the target candidate region to obtain a target classification result. According to the method, a channel attention mechanism is realized by increasing few parameters, and the target detection effect in a complex environment is improved. 本发明公开了一种基于深度学习的滑坡识别方法，包括：S100：对滑坡图像进行特征提取，所提取特征为特征图；S200：通过RPN检测网络处理特征图，得到推荐的目标候选区域；S300：对目标候选区域的特征图进行特征提取，获取目标分类结果。本发明通过增加很少的参数，实现了通道注意力机制，提高复杂环境下目标检测的效果。